Content uploaded by Odile Mackett
Author content
All content in this area was uploaded by Odile Mackett on Apr 05, 2021
Content may be subject to copyright.
A gender-based investigation of the determinants of labour
market outcomes in the South African labour market
Odile Mackett
Submitted in partial fulfilment of the academic requirements for the degree of Masters in
Development Theory and Policy
Faculty of Commerce, Law and Management
School of Economic and Business Sciences
University of the Witwatersrand
Johannesburg, South Africa
Supervisor: Professor Daniela Casale
February 2016
ii
Plagiarism Declaration
I, Odile Mackett, declare that
1. The research reported, except where otherwise indicated, is my original research.
2. This report has not been submitted for any degree or examination at any other
university.
3. This report does not contain other persons’ data, pictures, graphs or other information,
unless specifically acknowledged as being sourced from other persons.
4. This report does not contain other persons' writing, unless specifically acknowledged
as being sourced from other researchers. Where other written sources have been
quoted, then:
a. Their words have been re-written but the general information attributed to them
has been referenced
b. Where their exact words have been used, then their writing has been placed in
quotation marks, and referenced.
5. This report does not contain text, graphics or tables copied and pasted from the
Internet, unless specifically acknowledged, and the source being detailed in the report
and in the references section.
_____________________________
Odile Mackett
Date:
iii
Abstract
In this report, the individual and household circumstances which influence the probability of
a person having a certain labour market outcome, and how these outcomes differ by gender,
will be investigated. While a number of similar studies have been conducted, this report
contributes to the South African literature by investigating, using more recent data from the
National Income Dynamics Study, what the determining factors are that drive women and
men to the labour market, and determine employment outcomes. Furthermore, the
investigation is extended by exploring whether these factors differ for men and women by
age cohort. The main hypothesis of the study is that the determinants, which impact labour
market outcomes and a successful transition from being not economically active or
unemployed in a given period, to becoming employed in another period, differ for males and
females; with factors such as education, labour market experience, and other household
factors like marital status and children in the home being more important for women than for
men. The results of the econometric analysis suggest that education is important for both
sexes, but is of particular importance in determining the labour force participation and
employment probabilities of women and the youth cohort. Furthermore, the location in which
an individual resides is an important determinant of the labour market outcomes of women,
with women in urban areas having the most favourable labour market outcomes. Having
pensioners in the home has an adverse effect on the employment probabilities of men, while
it is positively related to the employment probabilities of young women. Children in the
home reduce the labour force participation of both men and women, but have a negative
effect on the employment probabilities of women.
iv
Acknowledgements
The completion of this report, and ultimately my degree, would not have been possible
without the support of a number of people who have chosen to accompany me on this
challenging journey.
I am particularly grateful to the following people:
My supervisor, Professor Daniela Casale, who commenced her employment at the
University of the Witwatersrand at a time that could only suggest that she was heaven
sent. Her vast knowledge, experience and passion for research and teaching, which
shone through every piece of advice and constructive feedback she gave, not only
provided me with a great guide in writing this report, but also with a role model in my
career as an academic.
The lecturers and researchers in the School of Economic and Business Sciences, who
took their time to provide some of the most insightful lectures in the coursework
component of the Masters in Development Theory and Policy. I would like to thank
Dr. Lotta Takala-Greenish and Dr. Nicohas Pons-Vignon in particular, for their
dedication to the programme and alternative theories. I hope that they will continue to
provide other students with the knowledge which they imparted on myself and my
class mates.
My colleagues in the Department of Economics at the University of South Africa,
who were of great support, while I undertook this programme on a full-time basis.
The Nedbank Foundation for providing me with a scholarship by means of which to
complete my studies.
Finally, I am forever indebted to my parents, Gregory and Rebecca Mackett, who
have inevitably provided me with this opportunity through their hard work and
passion for education. They have provided me with unfailing support and continuous
encouragement throughout my years of studying which ultimately resulted in me
being able to pursue my calling. My accomplishments would not have been possible
without them, and so this report is dedicated to them.
v
TABLE OF CONTENTS
LIST OF TABLES ................................................................................................................. vii
CHAPTER 1: INTRODUCTION ........................................................................................... 1
CHAPTER 2: LITERATURE REVIEW .............................................................................. 5
2.1. Introduction ................................................................................................................. 5
2.2. Theoretical literature review ....................................................................................... 5
2.3. Empirical literature review ........................................................................................ 11
2.4. Conclusion ................................................................................................................. 19
CHAPTER 3: METHODOLOGY ....................................................................................... 20
3.1. Introduction ............................................................................................................... 20
3.2. The Data Set .............................................................................................................. 20
3.3. Analysis Techniques ................................................................................................. 21
3.4. Description of variables ............................................................................................ 24
3.5. Youth and non-youth sub-samples ............................................................................ 31
3.6. Limitations ................................................................................................................ 31
3.7. Conclusion ................................................................................................................. 33
CHAPTER 4: DESCRIPTIVE RESULTS .......................................................................... 34
4.1. Introduction ............................................................................................................... 34
4.2. Labour market outcomes by gender .......................................................................... 34
4.3. Labour market outcomes by marital status ............................................................... 36
4.4. Labour market outcomes by number of children in the household ........................... 38
4.5. Labour market outcomes by education level............................................................. 39
4.6. Labour market outcomes by race .............................................................................. 41
4.7. Labour market outcomes by location ........................................................................ 42
4.8. Labour market outcomes by province ....................................................................... 43
4.9. Transition matrices .................................................................................................... 44
vi
4.10. Conclusion ............................................................................................................. 45
CHAPTER 5: REGRESSION ANALYSIS ......................................................................... 46
5.1. Introduction ............................................................................................................... 46
5.2. Cross-sectional probit regression analysis on labour force participation probabilities .
................................................................................................................................... 46
5.3. Cross-sectional multinomial logistic regression analysis on probabilities of different
labour market outcomes ....................................................................................................... 52
5.4. Panel logistic regression analysis on employment probabilities ............................... 58
5.5. Conclusion ................................................................................................................. 62
CHAPTER 6: CONCLUSION.............................................................................................. 64
APPENDICES ........................................................................................................................ 68
A 1: Labour market outcomes by gender and marital status (column percentages) ............ 68
A 2: Labour market outcomes by gender and education level (column percentages) ......... 68
A 3: Labour market outcomes by gender and race (column percentages) ........................... 69
A 4: Labour market outcomes by gender and location (column percentages) ..................... 69
A 5: Labour market outcomes by gender and province (column percentages) .................... 70
A 6: Probability of employment across periods, by gender - excluding NEA individuals .. 71
A 7: Probability of employment across periods, by gender - Wave 2 to Wave 3 ................ 73
A 8: Wave 2 labour market outcomes by gender (%) .......................................................... 74
A 9: Mean characteristics of variables by gender ................................................................ 75
REFERENCES ....................................................................................................................... 76
vii
LIST OF TABLES
Table 1: Broad labour force participation and unemployment rates by gender and age cohort
(%)............................................................................................................................................ 34
Table 2: Labour market outcomes by gender and age cohort (%) ........................................... 35
Table 3: Labour market outcomes by marital status and gender (%) ...................................... 36
Table 4: Labour market outcomes by marital status and gender (%) - Youth only ................. 37
Table 5: Mean values of the number of children in the household by gender ......................... 38
Table 6: Mean values of the number of children in the household by gender – Youth only ... 39
Table 7: Labour market outcomes by education level and gender (%) ................................... 40
Table 8: Labour market outcomes by education level and gender (%) - Youth only .............. 40
Table 9: Labour market outcomes by race and gender (%) ..................................................... 41
Table 10: Labour market outcomes by location and gender (%) ............................................. 42
Table 11: Labour market outcomes by province and gender (%) ............................................ 43
Table 12: Transition matrices of labour market states between Wave 1 and Wave 2 (%) ...... 44
Table 13: Transition matrices of labour market states between Wave 1 and Wave 3 (%) ...... 45
Table 14: Probability of labour force participation, by gender and age cohort ....................... 50
Table 15: Probability of labour market outcomes, by gender and age cohort ......................... 55
Table 16: Probability of employment across periods, by gender. ............................................ 60
1
CHAPTER 1: INTRODUCTION
Why do men and women make different labour market decisions or experience different
labour market outcomes? This is a question that has been posed for decades. While, men have
traditionally held the role of breadwinner and women the role of homemaker, women started
entering the labour market gradually, and began to engage in wage employment towards the
end of the twentieth centry (Smith & Ward, 1985). This process took place in both developed
and developing countries, as well as capitalist and socialist regimes, taking place in some
more rapidly than others (Folbre, 1994). In an ideal world, this transformation would have
meant that women and men could share in housework and childcare responsibilities in order
for them to perform their duties in the formal economy, however, this has, instead, had
differing consequences for men and women in different societies. The entry of women into
formal employment thus not only had economic consequences, but also had an impact on the
social relations between the sexes.
Although countries differ in terms of the customs and traditions to which they provide a
home, having a large labour force is beneficial to the economy, and policymakers would thus
want as many people actively participating in the economy as possible. However, for
governments to induce individuals to supply their labour, it is useful for them to have
knowledge of the factors which act to encourage or hinder labour market entry, as well as to
be cognisant of the ways in which these factors affect men and women differently. Studying
these factors is an exercise in need of pursuit on a regular basis, due to the changing nature of
the labour market and social relations in society.
A distinguishing feature of the South African labour market over the years has been its
persistent gender and racial inequalities (Ranchhod, 2010). Studies have found that gender
inequalities in labour markets have a negative effect on economic growth (Kabeer, 2012;
Klasen & Lamanna, 2009) and it is thus of vital importance for policymakers to have a
strategy in place to reduce gender inequality where it occurs. To do this, it is important to
understand the factors which encourage or impede individuals from entering the labour
market and gaining employment. These factors are likely to differ for men and women.
2
In this report, the individual and household circumstances which influence the probability of
a person having a certain labour market outcome receives consideration. Although this
phenomenon has been investigated in a number of previous studies which also focused on
developing countries (see for example, Bbaale and Mpuga (2011), Bridges and Lawson
(2008) and Bridges, Lawson and Begum (2011) on Uganda and Bangladesh), this report will
build on the existing literature, by exploring this topic for the South African labour market.
Although, in South Africa, numerous authors have also contributed to this topic (see for
example Dinkelman and Pirouz (2002) and Ntuli (2007)), this report will contribute to South
African literature by investigating, using more recent data, those determining factors driving
women and men into the labour market, and which determine employment outcomes.
Furthermore, the investigation is extended by exploring whether these factors differ for men
and women by age cohort.
Purpose of the study
The report will thus explore gender differences in the factors determining labour market
outcomes
1
in South Africa. In addition to this, the study will investigate whether these factors
are different for men and women in the youth and non-youth cohorts. The purpose of the
study is to determine whether there are gender differences in the factors which determine
labour market outcomes, or whether someone is likely to transition from being not
economically active or unemployed in one period to becoming employed in the next, and
whether these gender differences are more or less pronounced amongst the youth.
Research questions
The report will seek to provide answers to the following questions:
a) Are there gender differences in the factors which determine labour force participation
(LFP)?
b) Are there gender differences in the factors which determine labour market outcomes,
namely entry into one of three states – employed, unemployed or not economically
active (NEA)?
c) Are there gender differences in the factors which determine whether someone who is
NEA or unemployed in one period is likely to become employed in a subsequent
period?
1
Labour market outcomes in this report refer to whether an individual is employed, unemployed or not
economically active (NEA). These states will be discussed in detail in the rest of the report.
3
d) Are these factors different among the youth and non-youth cohorts?
While there has been an increase in gender studies of the labour market in the post-apartheid
period in South Africa, a detailed microeconomic analysis such as the proposed study has not
been undertaken for the South African labour market, using the data employed here, namely
the National Income Dynamics Panel Study (NIDS).
2
Hypotheses of the study
Based on the findings of previous studies, which are extensively discussed in the literature
review, this study hypothesises that the determinants which impact labour market outcomes
differ for males and females, respectively. Furthermore, the factors which determine a
successful transition from being inactive or unemployed to becoming employed from one
period to another are also likely to differ by gender. It is possible that education and labour
market experience will be more important for women than for men, if employers value
signals of productivity differently by gender. In addition, household factors (like marital
status and children) might differentially affect the probability of joining the labour market
and finding employment. Given the changing gender norms over time, and improved labour
legislation in the post-apartheid period, it is possible that differences in the factors
determining labour market states between men and women are likely to be less pronounced
among the youth cohort, when compared to the non-youth cohort.
Research methods
The research questions posed will be investigated, making use of longitudinal data from the
National Income Dynamics Study (NIDS). The data will first be utilised to determine how
socio-economic and demographic characteristics of an individual impacts upon their LFP
decisions
3
or outcomes, whereafter it will then be utilised to determine how the same factors
impact the likelihood of an individual being in one of three labour market states (NEA,
unemployed or employed), and lastly, how these characteristics assist or impede the chances
of an unemployed or NEA individual obtaining employment in the future. All the regressions
in the report are disaggregated by gender so as to determine how these factors impact the
labour market outcomes of men and women differently, and where sample size allows, the
2
NIDS is a survey conducted by the Southern Africa Labour and Development Research Unit (SALDRU) at the
University of Cape Town.
3
The LFP decisions of an individual refer to whether an individual chooses to supply their labour to the labour
market or whether they choose to be NEA.
4
data is further disaggregated to explore how these differ between the youth and the non-youth
cohorts of the population.
Outline of the report
The report consists of six chapters, with Chapter One providing the introduction to the study.
This is followed by Chapter Two, which provides discussion of the literature which is
relevant to the topic. It also introduces a number of theories which are related to the
determinants of LFP, particularly for women. Chapter Three provides a description of the
data that were utilised to investigate the research questions and the analysis techniques
applied to interpret the data. This chapter also describes the variables used in the regressions
and the limitations of the research. Chapter Four consists of a discussion of the descriptive
statistics of the variables used in the analysis, while Chapter Five provides the results of the
regressions. Lastly, Chapter Six will provide a summary of the findings of the report, make
recommendations and identify areas for future research.
5
CHAPTER 2: LITERATURE REVIEW
2.1. Introduction
An individual’s labour market state has a number of implications - not just for that individual,
but also for his or her family, community, and country. These labour market states include
either being employed, unemployed or not economically active (NEA). For the individual,
being in a particular labour market state is likely to influence his or her bargaining power in
society, and have consequences for his or her family, and it might result in other members of
the family being able to choose a different labour market state, or being compelled into a
particular state.
For the community and the country, labour market states have both macroeconomic and
microeconomic consequences (Escudero & Mourelo, 2013). When it comes to the
macroeconomic, there are implications for economic growth and the employment rate (Tsani,
Paroussos, Fragiadakis, Charalambidis & Capros, 2012). Whatever the consequences may be
for the individual actor in the labour market, there is consensus that any given country would
aim to keep its labour force participation (LFP) rate as high as possible (Klasen & Lamanna,
2009). Although there are many measures that might be taken by policymakers in an effort
towards increasing the LFP rate, there are individual circumstances in which citizens may
find themselves that prevent them from entering the labour market, just as there are
circumstances which may compel them to be part of the actively participating labour force. A
theoretical literature review will be followed by an empirical literature review on those
factors likely to influence individual labour market states, the way in which these factors
influence labour market states, and how they are likely to differ by gender and age.
2.2. Theoretical literature review
The debate around women’s LFP has changed in the last few decades, and has become
particularly important as high female LFP has proven to be highly beneficial to a country’s
economic growth (Bbaale & Mpuga, 2011; Tsani et al., 2012). Neoclassical economists view
the LFP decision as a choice between work and leisure time, and this is based on the
assumption that the individual making the decision is rational and, as a result, will choose to
engage in the activity which has the lowest opportunity cost. The rationality of the individual
6
has been described as “the ability of individuals to order their preferences (their likes and
dislikes) in a manner that is logically consistent and then, given that preference structure, to
make choices that maximise their self-interest” (Barker, 1999b: 571). This decision is
influenced by a number of factors, such as an increase in the wage an individual earns, which
could lead to an increase in leisure time, as individuals are given the opportunity to work
fewer hours and earn the same wage (income effect), or an increase in working hours as the
opportunity cost of leisure time has become too high (substitution effect) (Becker, Murphy &
Tamura, 1994).
Feminist economists, however, reject the rigidity of the neoclassical model, the perceived
rationality of human beings, and the scarcity framework under which society is assumed to
live. They claim that the choices people make every day are neither value-free nor gender
neutral, where these decisions cannot simply be viewed through a utility maximising and self-
interested lens, but instead involve complex interests, which are not necessarily self-
reflective, but may involve considering the consequences of a community or a family as well
(Barker, 1999b). Women and men fulfil multiple roles in society, some of which are not
captured in the simplistic work/leisure trade-off model presented by the neoclassical school.
Mjoli-Mcube (1998: 208) refers to this phenomenon as “gender contract[s], ‘which are the
invisible social contracts within which men and women act in the belief that this is what
societies expect of them’ (Schlyter and Zhou 1995, 5)”. Furthermore, these decisions often
differ for men and women as a result of this gender contract, and the patriarchal structures
that have been perpetually enforced by society (Folbre, 1994). According to the neoclassical
assumptions of rationality and individuality, each human being thus has the ability to
maximise his or her utility, and would choose a labour market outcome that might provide
him or her with the greatest utility. This has lead to the development of the human capital
theory.
2.2.1 Human capital theory
Van Der Merwe (2010: 107) posits that “a core thesis of human capital theory is that
education renders people more productive, that is, it raises the marginal productivity of an
educated worker relative to one not so educated”. This theory, which is based on
“individualism, perfect knowledge, rationality, private property rights and market economy
(competition)” (Van Der Merwe, 2010: 108), makes no distinction between the returns which
women and men may face, but rather assumes that everyone, if given the opportunity to be
7
educated, could expect the same returns from their increased knowledge and skills. In
addition, “human capital theory assumes that individuals have perfect foresight about future
earnings for every level of education” (Lam, Leibbrandt & Mlatsheni, 2008: 3). The human
capital theory implies that differences in wages between workers, both men and women, are
attributable to differences in the level of education, the number of years of experience, and
the level of skill a person acquires. Human capital theorists have claimed that a wage gap
may exist between men and women, as women generally invest in human capital that does
not have a monetary return as high as the human capital in which men invest (Jacobsen,
1999). Where an individual is married, the partner with the ability to earn a higher salary
spends more time in the labour market, while the other partner allocates more time to taking
care of housework (Delaunay, 2010). However, in reality, there exists a wage differential
between certain groups, which cannot be explained by the postulates of the human capital
theory. Lips (2013) suggests that this earnings gap is attributable to discrimination in labour
markets.
Human capital theorists further ignore the role that customs and traditions play in the labour
market, and simply emphasise the increasing number of working opportunities for women
which accompany a growing capitalist system (Seguino, 1997). There is an assumption that
everyone can freely choose the type of human capital they will invest in, and how this human
capital will be utilised (Jacobsen, 1999). Workers who expect greater returns from the labour
market would thus be more likely to be economically active, and actively engage in searching
for a job if they were to be unemployed. By the same token, it would therefore be logical to
conclude that certain women may be less optimistic about their labour market returns, due to
the presence of labour market discrimination. While Marxists and neoclassical economists
agree that the expansion of a capitalist system provides more working opportunities for
women as more labour is demanded, there are differences in the way in which they view how
these opportunities affect women (Seguino, 1997). Neoclassical economists believe that
employers are rational beings who will employ the labour which is cheapest, while Marxists
believe that this only serves to exploit the class relations which exist between the working
class and the capitalist class. However, a criticism that feminist economists have posed to
Marxists is that “production for exchange, typically men’s work, takes precedence over
production for use or reproduction, typically women’s work” (Albelda, 1999: 539).
8
2.2.2 Reproductive labour
Reproductive labour, which is generally not included in the gross domestic product (GDP) of
countries, is defined as “the work of managing a household, cooking, cleaning, keeping
home, clothing and domestic equipment in good repair, and caring for family members and
friends and neighbours” (Bakker, 1999: 85). The division between productive and
reproductive work is known as the sexual division of labour (Barker, 1999a). Men are
typically engaged in work for which one would receive monetary remuneration, whereas
women are typically involved in reproductive work. The reason for this, according to
neoclassical theorists, is because “women simply have a greater preference for family life
than men, and are therefore willing to sacrifice more [for it]” (Folbre, 1994: 98). However,
others argue that the expectations which society has of men and women in terms of work
makes a women’s decision about whether to participate in the labour force or not a more
complex issue than it is for men (Coleman, 1999). The debate on reproductive labour has
come to the forefront, with the development of feminist theories, and the large influx of
women into the labour market. Although the International Labour Organisation (ILO) defines
economically active individuals as “all persons of either sex who furnish the supply of labour
for the production of goods and services during a specified time-reference period” (ILO,
http://www.ilo.org), reproductive work is just as important in ensuring that the ‘productive’
side of the economy functions well. A burgeoning literature has developed around the
importance of reproductive labour, especially in debates involving class and race (Lewis,
2001), however, the heavy burden placed on women by housework responsibilities maintains
segregation in productive occupations (Folbre, 1994). As a result of this ‘responsibility’
placed on women, “the labour supply decision of women is based on a complex of needs
including financial necessity, social goals for well-being, gender determined non-market
responsibilities and personal interest” (Coleman, 1999: 503). Lim (2002: 204) further adds to
this and states that “the conflict between women’s productive and reproductive roles
significantly raises the opportunity cost of having children” and the longer this conflict
persists, the more likely we are to witness a decrease in fertility rates, especially as women
become more educated.
2.2.3 Reservation wage
A reservation wage is defined by Walker (2003: 4) as “the highest wage” at which a person
will choose not to work, thus, a reservation wage will be lower for poorer individuals and will
9
tend to be higher for those who are wealthier (O’Higgins, 2001). An individual’s reservation
wage has been found to be a significant determinant of that person’s labour market state and
has a positive relationship with an individual’s unemployment duration (Walker, 2003). The
reservation wage, according to Wittenberg (2002: 4) should be “set not just to cover the cost
of leisure foregone, but also, the opportunity cost of foregoing additional search”, as
searching for a job does cost money. While the determination of wages is not a supply-side
issue, it does have a determinate effect on the individual decision to supply labour or not. The
marital status of an individual is said to be a significant factor in determining a person’s
reservation wage. Other individual characteristics, affecting the reservation wage are whether
or not an individual is the head of a household, how many employed individuals reside within
the household, as well as the level of education which an individual possesses (Walker,
2003). As these tend to differ by gender, it is expected that there would be gender differences
in the reservation wages of individuals, and thus gender differences in the decisions people
make as to whether or not they will be an active participant in the labour market.
Furthermore, a NEA housewife, for instance, who has the support of a husband’s wage
(which is sufficient to support the whole household), may have a higher reservation wage
than an unemployed male who lives in a household with inadequate income available.
Changes in the household may thus change a person’s reservation wage, and give rise to
effects such as the added worker effect.
2.2.4 Added worker effect
Another phenomenon which has had an impact on the labour market decisions individuals
make, is the ‘added worker effect’. The added worker effect is defined by Fernandes and De
Felício (2005: 887) as “the effect of a job loss of a husband on the labor supply of his wife,
which encompasses both an increased labor force participation rate (LFPR) as well as
increased hours of work for those wives who are already in the labor force”. The added
worker effect can be extended to other individuals in the household, such as children of
working-age, who may need to enter the labour market once the breadwinner is no longer
able to provide for everyone. While women tend to be NEA when married, their entrance into
the labour market after the loss of a husband’s job indicates the effect which a women’s
marital status has on her reservation wage, as a wife will only enter the labour market if the
possible income to be earned is larger than her reservation wage, which increases once she is
married (Serumaga-Zake & Kotze, 2004). The reservation wage should thus be an important
10
factor for policy-makers, especially those who wish to encourage entry into the labour force
of educated individuals.
2.2.5 Youth
Youth unemployment, which is a global issue, has, in many countries, become as pervasive
as gender inequality. While the youth generally have fewer employment prospects due to
factors such as a lack of skills and inexperience relative to older adults, there are certain
factors which may aggravate this issue. In times when economic growth is slow, employers,
who are apt to minimise costs, are likely to hire fewer workers or likely to lay off some who
are already employed. As many young workers are employed on a contractual basis, it is
easier for employers to lay off young workers than workers with permanent contracts and
many years of experience. Another factor which may impact youth employment negatively is
the number of young people entering the labour market at any given time, as the more people
searching for employment, the more employment opportunities need to be made available,
which is difficult in times of economic distress (Escudero & Mourelo, 2013; Mlatsheni &
Rospabé, 2002).
This may be particularly true for females, who are often employed under precarious
conditions, and are more likely to have less work experience and fewer skills than men, due
to intermittent absences from the labour market, especially during their peak childbearing
years (Escudero & Mourelo, 2013; Lim, 2002). This has been confirmed by O’Higgins
(2001), who found that while youth unemployment is a problem in both developed and
developing countries, the problem is even more severe for young women who have fewer
employment opportunities than their male counterparts.
Age has proven to be a significant factor in determining which labour market state
individuals find themselves in, with the youth being more likely not to engage in economic
activity when compared with adults (Escudero & Mourelo, 2013). Young people tend to be
supported by their families, and thus, the cost of being unemployed tends to be lower for a
young person with negligible responsibilities than for an adult who may be responsible for an
entire household (Knight & Kingdon, 2000). Furthermore, young people are more likely to be
furthering their education, should they have the resources to do so, thus accounting for their
concentration in the NEA part of the population. In addition, job security may be more
valuable to an older adult, who may have particular minimum requirements for wages and job
benefits, whereas young people may find it easier to obtain an entry-level job from the
11
numerous firms who are likely to be advertising such jobs. Thus, young people will have a
higher likelihood of being unemployed or NEA, or be more comfortable to move from
employment into unemployment than older individuals might, and are likely to have a lower
reservation wage than might an older individual (Knight & Kingdon, 2000).
Changing gender roles and norms may result in young men and women making decisions
differently from older men and women. As young women have parenting responsibilities,
they have traditionally been more likely to be NEA, when compared to their male
counterparts. In addition, like older women, young women are traditionally more likely to
engage in unpaid housework (Colman, 1998). However, with perceptions changing about the
value of housework and, in some instances, those who can afford it, employing domestic
labour to take care of housework and childcare responsibilities, there is an expectation that
young women may enter the labour force in larger volumes than before. Thus, although there
are still young women who choose the traditional path in terms of their responsibilities as
mothers and wives, others may be making decisions differently. Therefore, should the
changing norms have become entrenched in society, there is an expectation that the gender
differences between young men and women will be less pronounced than those between older
men and women; otherwise, we would expect to see the same pattern which has been
displayed over the years, with women leaning towards their parenting and housework
responsibilities, and engaging less in productive work.
2.3. Empirical literature review
In a number of countries, women are found to have jobs which are generally insecure and
precarious, while they also experience higher unemployment rates when compared to their
male counterparts (Lim, 2002). Thus, while the neoclassical theorists could have been correct
in saying that employers – as rational beings – hire the cheapest labour that they can find, the
Marxists predicted that the nature of the employment relationship would most likely turn out
to be exploitative, hence the precarious and insecure nature of women’s work. Many Asian
countries have increased their competitiveness by making use of female labour in their export
industries, although it has also been found that the labour conditions under which these
women operate are exploitative in some cases (Seguino, 1997). However, while many women
are being exploited globally, employment has allowed some women to gain bargaining power
within their homes, and some have even managed to mobilise efforts in the workplace in the
form of collective bargaining, overcoming the traditional barriers which have been placed on
12
them as women in their countries (Klasen & Lamanna, 2009; Seguino 1997, 2000). Though
women may find strength in their collectivism, it is their individual circumstances which
would ultimately determine whether or not they would be willing and able to supply their
labour, and whether they find work. Knowing what these individual characteristics are would
prove to be useful to policymakers if they are to create an environment in which people can
supply productive labour and manage reproductive work simultaneously. While there are
many variables that will affect an individual’s labour market decisions and outcomes, the
ones which are most likely to display gender differences are discussed below.
2.3.1 Education
Educational attainment is expected to increase the LFP rates for both sexes, as the higher the
educational attainment of an individual, the higher the opportunity cost of choosing to be
NEA (Ntuli & Wittenberg, 2013). Although it has been found that the gender pay gap for
women who are highly educated is smaller than for women who are not, the pay gap still
exists, which is contrary to the hypothesis of the human capital theory, which states that no
pay gap should exist between those with the same education levels and skills (Addabbo,
Favaro & Magrini, 2012). Educational attainment has generally been found in a number of
studies to have a positive effect on women’s LFP as well as their probability of finding a job
(see for example Bbaale and Mpuga (2011) for Uganda and Siphambe and Motswapong
(2010) for Botswana). Furthermore, an increase in educational attainment for women has
been found to be related to a decrease in fertility (Duflo, 2012), which is significant, as the
number of children in the household also affects the labour market outcomes of females
(discussed below). Having gender inequality in education not only reduces human capital
capabilities, but also means that employers are forced to choose from a less talented labour
force (Duflo, 2012; Klasen & Lamanna, 2009). While these negative consequences may stem
from gender discrimination in education, increased educational attainment has proven to
reduce fertility levels and child mortality rates, as well as to prompt an increase in women’s
bargaining power at home and in the workplace (Lim, 2002).
In South Africa, both men and women are more likely to be employed the higher their level
of education, and more likely to be NEA the lower their level of education (Ntuli, 2007;
Ranchhod, 2010; Van Der Westhuizen, Goga & Oosthuizen, 2007). Increased levels of
education, particularly amongst African women, have been a key driving factor in the rapid
13
increase of female LFP rates in the country (Casale & Posel, 2002; Maja & Nakanyane, 2006;
Ntuli & Wittenberg, 2013). High levels of educational attainment, such as the possession of a
Matric certificate or a degree from a higher education institution, are of particular importance
in driving women to enter the labour force, while unemployed individuals mainly consist of
those individuals with very low educational levels, or no education at all (Naudé &
Serumaga-Zake, 2001; Van Der Westhuizen et al., 2007). For the younger demographic, Yu
(2013) confirms that youth who are more educated have a higher likelihood of being
employed and also have a higher likelihood of being amongst the searching unemployed, than
the non-searching unemployed. However, Mlatsheni and Rospabé (2002: 10) have found that
a young male is more likely than a young female to be in employment or self-employment,
than to be unemployed.
Youth entering the labour market at an early and young age with little educational attainment
is a serious concern for the South African labour market, and significantly contributes to the
high youth unemployment rates in the country. Lam et al. (2008) state that limited access to
resources to further education or an obligation to supplement the family income, especially
where younger children are present in the household, are some of the various reasons why
many young people cease completing their education and actively search for work. They
further go on to challenge the assumption of perfect knowledge in the human capital theory,
as certain groups, such as the youth, are uncertain about their future prospects, their returns to
education, and prospective employment when they make a decision to further their education.
Education thus generally affects the youth in the same way that it affects the adult cohort of
the population, with those who have higher levels of education being significantly more
likely to find employment than those who leave school early to search for employment (Lam
et al., 2008).
2.3.2 Marital status
Single women who head households in Uganda are more likely to be in employment than
married women who live in a household headed by a male (Bbaale & Mpuga, 2011), while
similar evidence was found in Botswana (Siphambe & Motswapong, 2010). Men are
typically more likely to be employed when married, while women may be more economically
dependent on their husbands, as opposed to a woman who has never been married or who is
divorced or widowed (Ntuli, 2007). Furthermore, without a husband to provide necessities,
14
there may be more of a need for women to be in employment if they are responsible for an
entire household.
Evidence found by Delaunay (2010: 36) for cohabiting and married couples in Portugal
indicated that men spend more and women less time in productive labour. In addition, Naudé
and Serumaga-Zake (2001) found that being married increases the probability of male
employment, while Ntuli and Wittenberg (2013) found that being married significantly
reduced the probability of an African woman being employed; however they add that the
decision not to participate in economic activity, where being married may be endogenous
(confirmed also by Dinkelman and Pirouz (2011)). The reason why an individual’s marital
status could be endogenous in an analysis of labour market outcomes is that individuals may
be in a particular labour market state due to their marital status, such as a woman leaving her
job and becoming NEA once she becomes married. However, being in a certain labour
market state could also induce an individual to enter into a marital state, such as an employed
male having the financial resources to get married. In South Africa, this is of particular
importance, where African men are required to make a dowry payment (or bride wealth) if
they wish to get married, indicating that those who are employed or have the financial
resources to do so will probably be more likely to get married than those who do not (Casale
& Posel, 2010b).
2.3.3 Presence of children, pensioners and working-age adults in the household
Women and men’s labour market decisions are influenced by the composition of the
household, particularly the number of young children in the household. It is for this reason
that there is an increasing number of countries implementing legislation that compels certain
employers to provide childcare facilities to workers, along with maternal and paternal leave,
as well as flexible working hours where possible; likely addressing the ‘conflicting’
relationship between reproductive and productive work for both men and women (Lim,
2002). In Kenya, having young children in the household was not significant in determining
the likelihood of being employed or NEA for the youth, however, it had a significantly
negative impact among the non-youth cohort of the population (Escudero & Mourelo, 2013).
In Uganda, women who are more educated were found to have fewer children, thus
increasing the likelihood that they are to enter the labour market, showcasing the combined
effect which education and the presence of children can have on the LFP decision of a
woman (Bbaale & Mpuga, 2011; Duflo, 2012).
15
Using the South African October Household Survey to determine the factors which influence
the participation of married women in the labour market, Serumaga-Zake and Kotze (2004)
found that the presence of young children in the household had a negative impact on the
participation rates of married women. This is further substantiated by Ntuli & Wittenberg
(2013), who found fertility to have a negative effect on the LFP rates of women, however, the
results differed between those who were broadly
4
unemployed and those who were strictly
5
unemployed. It was consequently concluded by Ntuli & Wittenberg (2013: 367) that “the
presence of young children in the household increases the prospects of wanting to work, but
not actively searching for work”. Using a panel to analyse the effects that certain factors have
in determining whether an individual will move from regular employment to non-
employment/subsistence agriculture or vice versa, Cichello, Leibbrandt and Woolard (2014:
79) found unexpectedly that women are more likely to move from non-employment into
regular employment when there are young children in the home, and men are more likely to
stay in non-employment/subsistence agriculture.
While the decision to be in the labour market or not, with children in the household, could be
one of preference, particularly for married individuals, the presence of a child could also
mean that someone is a recipient of a childcare grant from the government. Ranchhod (2010)
finds that recipients of childcare grants are 7.4 percent less likely to be employed, whereas
other individuals living in the same household are more likely to be NEA. Having this
income stream in the household could allude to some form of income distribution in the
household, which is likely to increase the reservation wages of the individuals belonging to it.
The number of working-age adults who reside in a household may also affect the income
levels within that household, and might be expected to influence the reservation wages of its
members. It has been found that the larger the number of working adults present in the
household, the higher the reservation wage of individuals living in the household (Dinkelman
& Pirouz, 2011; Walker, 2003). Having working adults in the home allows for a distribution
of income to those individuals who are unemployed or NEA, and also provides some sort of
security for these individuals. Although having a high number of unemployed or NEA
4
The definition of broad unemployment in South Africa includes everyone who is without work and reports
wanting to work, regardless of whether they have been searching for work or not (Leibbrandt, Woolard,
McEwan & Koep, 2010).
5
The strict definition of unemployment “only considers as unemployed those who have actively searched for
work in the last 4 weeks and are able to accept a job within the next week. All other ‘discouraged’ workers (who
would like to work but are not actively seeking work) are classified as not economically active” (Leibbrandt et
al., 2010: 9).
16
working-age adults in the home could dilute the effect this income distribution is likely to
have, as there are more mouths to feed.
Having a pensioner in the household has been found to reduce the likelihood of a young
person being employed, as opposed to being a non-searching unemployed person (Yu, 2013).
Ranchhod (2010) substantiates this with findings that show households receving government
grants, such as child care grants and pensions, are more likely to include individuals who are
NEA. It was found that adults living with a pensioner are 7.9 percent less likely to be
employed, having the same effect as having an employed individual in the household. For
women, however, this meant that they were able to enter the labour market, as grandparents
could potentially provide supervision for their children, but could possibly also provide them
with the financial resources with which to undertake a job search (Aassve, Arpino & Goisis,
2012; Posel, Fairburn & Lund, 2006).
2.3.4 Location
The geographical location of an individual is equally important when considering labour
market outcomes. Individuals who reside in urban areas are expected to be more likely to be
employed than unemployed, and are more likely to be unemployed than NEA, due to the
increased number of working opportunities within urban areas where central business
districts are found, and where there is a reduced job search cost. Siphambe and Motswapong
(2010) found that residing in an urban area increased a women’s likelihood of being
employed as well as an active labour force participant in Botswana, although O’Higgins
(2001) adds that in Tanzanian urban areas, young women are more likely than young men to
be unemployed, and the author notes that this might lend itself to an inquiry into the types of
jobs which are available to the young women in these areas. This is substantiated with
evidence from India, where at first glance, the unemployment rates for young men and
women do not seem to differ much, however, when disaggregated by location young women
faced far higher unemployment rates in urban areas than did young men (O’Higgins, 2001).
Comparing employment likelihoods between urban and rural areas, Bbaale and Mpuga
(2011) found that in Uganda, women in rural areas were eight percent more likely to be
economically active than women in urban areas. Evidence found for the South African labour
market correlates with the findings of Siphambe and Motswapong (2010) and O’Higgins
(2001). Naudé and Serumaga-Zake (2001) found that living in an urban area increased the
17
likelihood of being employed, where unemployment rates tended to be lower in urban areas
than in rural areas for women.
Similar evidence was found by Ntuli (2007) and Cichello et al. (2014) for South African
women, and Ntuli and Wittenberg (2013) for black South African women; whereas the
difference in urban and rural location did not have a significant effect for men (Cichello et al.,
2014). Additionally, women living in the Western Cape had a greater probability of moving
into regular employment, whereas men in Gauteng and Mpumalanga had greater prospects of
moving into employment across two periods. Wittenberg (2002), who undertook an analysis
of the African population, found that those who reside in urban areas had a higher likelihood
of being active labour force participants, however, Ntuli and Wittenberg (2013: 366) have
cautioned that this variable could possibly be endogenous when undertaking an analysis of
labour market outcomes, as women who are unable to find work in “urban areas may return
to rural areas, where the cost of living is lower”, but where job search costs may also be
higher.
2.3.5 Youth
From the theoretical literature, there are numerous reasons why the youth may be more likely
to be unemployed or NEA, relative to individuals in the non-youth cohort. In Uganda young
people are less likely to be part of the labour force than older individuals, this might be due to
the fact that they may still be schooling or have the fall-back position of a family providing
for them (Bbaale & Mpuga, 2011). In Kenya, it has also been found that the youth are more
likely to be NEA than the non-youth cohort of the population, with the effect being greater
for the 15 to 24-year-old cohort than the 25 to 34-year-olds. These groups are also more
likely to be unemployed than are individuals in the adult cohort of the population. The
evidence from South Africa is no different, with the 20 to 24-year-old cohort making up the
largest proportion of unemployed individuals in the country, and the second largest NEA
group, after the 55 to 59-year-old cohort (Ranchhod, 2010). In more recent data, the 15 to 24-
year-old cohort had the lowest LFP rate amongst the working-age population
6
(Stats SA,
2015). The low LFP rates are most likely due to the extended periods young people use to
further their education.
6
This is based on the strict definition, thus discouraged work-seekers are not considered to be part of the labour
force.
18
In addition, an attempt to transform the South African labour market through a number of
affirmative action policies
7
which have been implemented in the interest of redressing the
social injustices of the past, may have an even greater impact on the South African youth and
their labour market decisions. Although the effectiveness of these policies has been
questioned, as it perpetuates racial identities and prejudices, and in many instances results in
decreased morale amongst employees, both those who are beneficiaries thereof and those
who are not (Thomas, 2002), “affirmative action, can only be meaningful in the context of
individuals who are similarly qualified or skilled and where those who ‘belong’ to one of the
‘designated groups
8
’ have to be given preference over the others” (Alexander, 2006: 95). In
this context, it is clear why affirmative action may not have been as effective amongst the
older cohort (from the designated groups), as they come from a history of differential access
to education, and are thus not expected to have the same skill levels as their white
counterparts.
9
Young individuals from designated groups, however, have had greater access
to opportunities to further their education than their parents and grandparents. It is for this
reason that access to labour market opportunities may be more accessible for young
individuals, especially those from the designated groups, and this is expected to increase the
opportunity cost of being NEA. These policies may have proven to be particularly beneficial
to young women from designated groups, as these policies aim not only to redress the racial
injustices of the past, but also the gender discrimination which has affected women of all
races (Naidoo & Kongolo, 2004).
With an increase in educational attainment amongst the youth population and affirmative
action policies potentially providing greater opportunities in the labour market for the youth
who are more educated than the non-youth cohort, one may expect a large supply of youth
labour after the completion of studies. However, for young females, this pattern of behaviour
may not be as simplistic, as they are in their peak childbearing years, possibly making them
more likely to be NEA. Thus, it will be the path that the youth view as most valuable that
they will choose, however, there is still an expectation that many young women will be NEA
due to parenting responsibilities, particularly those who are married.
7
Policies which have been implemented in this regard refers to the “Public Services Act, the Employment
Equity Act, the Skills development Act and the Skills Development Levy Act” (Alexander, 2006: 93).
8
Designated groups are referred to in these policies as “black people, women and people with disabilities”,
while the term ‘black people’ include “Africans, Coloureds and Indians” (Alexander, 2006: 94).
9
While there are older individuals from designated groups who have been able to obtain the same level of
education as their white counterparts, they have not been able to do so in large volumes; this is evident in the
criticisms of South Africa’s affirmative action policies, which is said to have benefitted a small portion of the
black middle class, and as a result, perpetuated class inequalities (Alexander, 2006).
19
Serumaga-Zake and Kotze (2004) found that age is a significant indicator for married
females, and that the probability of employment increases as an individual ages. These
findings suggest that one might expect to find greater differences in the labour market
outcomes of young males and females, compared to the non-youth cohort. This could be an
indication of women having a reduced likelihood of being active labour force participants,
due to decreased participation in their childbearing years, but that they can more easily obtain
employment as they exit this phase.
2.4. Conclusion
From the existing literature, we thus expect higher LFP rates amongst women and men who
have high levels of education; this is expected to be true for both the youth and the non-
youth. The marital status of individuals is also expected to render important gender
differences. Lower LFP rates for married women are expected, both among the youth and
non-youth, while higher LFP rates are expected for married men. The presence of young
children in the home is expected to result in a lower LFP rate for women, particularly among
the youth who are of peak childbearing years, although increased labour market opportunities
may cause some young people to choose differently. The presence of pensioners in the home
may have a negative effect, if the extra income increases the reservation wage of those living
in the home, making employment less attractive. While the location of a man is not expected
to matter, the location of a woman may have a different effect on LFP, depending on whether
she resides in a rural or urban area.
Although numerous studies on LFP have been undertaken in South Africa using different
methods and data sets, this study intends to provide a more comprehensive analysis of the
factors important for the LFP of men and women, using more recent data, and making use of
panel data. In addition to this, a comparison is made between the youth and the non-youth
cohorts disaggregated by gender.
20
CHAPTER 3: METHODOLOGY
3.1. Introduction
In this chapter, the methods utilised to investigate the research questions are described. The
data set used and where it can be obtained will be described first, followed by a discussion of
the analysis techniques applied. A description of the variables discussed in the literature
review will be provided in the context of this study, and a discussion of the limitations of the
study will be presented. The research questions to be answered with the methods described
below are:
a) Are there gender differences in the factors which determine LFP?
b) Are there gender differences in the factors which determine labour market outcomes,
namely entry into one of three states – employed, unemployed or NEA?
c) Are there gender differences in the factors which determine whether someone is likely
to transition from being NEA or unemployed in one period to becoming employed in
the next period?
d) Are these factors different among the youth and non-youth cohorts?
3.2. The Data Set
The data utilised in this report are from the National Income Dynamics Study (NIDS),
10
a
survey conducted by the Southern Africa Labour and Development Research Unit
(SALDRU) at the University of Cape Town. NIDS is the first South African panel study
which has, up until now, collected and released three Waves of data; Wave 1 released in
2008, Wave 2 in 2011 and Wave 3 in 2012. The majority of the regressions
11
in this report
make use of Wave 1 only, with the exception of the set of panel regressions which utilise
Wave 2 and Wave 3 as well
12
.
10
NIDS data are freely available from http://www.nids.ac.za.
11
The statistical package STATA was utilised to analyse the data.
12
While the Quarterly Labour Force Survey, published by Statistics South Africa, may also have proven useful
in investigating the questions, NIDS data were more appropriate due to the panel nature of the data set and the
fact that NIDS tracks individuals, rather than households (Posel, Casale & Vermaak, 2014).
21
In Wave 1 of NIDS, 7 300 households and 28 000 individuals were interviewed. Of those
who were interviewed in Wave 1, 21 098 individuals were successfully re-interviewed in
Wave 2 (Brown, De Villiers, Leibbrandt & Woolard, 2012) and 23 604 of those interviewed
in Wave 2 were successfully re-interviewed in Wave 3 (De Villiers, Brown, Woolard,
Daniels & Leibbrandt, 2014). Wave 1 was used for the cross-sectional analyses in this report
while all three Waves were utilised for the panel regressions. For the panel, only continuing
sample members who were successfully interviewed in both Waves were utilised. This
resulted in 21 108 individuals for the Wave 1 to 2 analysis and 21 384 individuals for the
Wave 1 to 3 analysis.
13
As NIDS is the only longitudinal study in South Africa, it provides an interesting opportunity
to investigate how changing circumstances of individuals impact their labour market
decisions over time. The report thus takes advantage of this by estimating panel regressions
to determine how certain demographic and socio-economic characteristics of unemployed
and NEA individuals can result in these individuals becoming employed in the future. For
this analysis, Wave 2 and Wave 3 is utilised together with Wave 1.
The cross-sectional analysis in this report makes use of Wave 1 post-stratified weights which
are thus used to “account for survey design and initial non-response” (Baigrie & Eyal, 2013).
These weights have been specifically designed to take into consideration the geographic data
in the 2011 Census. To analyse the panel, panel weights from Wave 2 and Wave 3 were
utilised to correct for attrition bias
14
and survey design (De Villiers, Brown, Woolard, Daniels
& Leibbrandt, 2013).
3.3. Analysis Techniques
A number of analysis techniques are employed in this study. The first is a set of probit
regressions, to determine the likelihood of LFP for males and females in Wave 1, followed by
a set of multinomial logistic regressions, which estimate the likelihood of being in a
particular labour market state in Wave 1. Lastly, a set of logistic panel regressions are
estimated to determine the factors that are likely to influence a person’s movement from
being NEA or unemployed in one period, to being in employment in the next. In addition to
13
An analysis of Wave 2 to 3 is also included in the appendix of the report, which includes observations for
20 462 individuals.
14
Attrition occurs when individuals which were interviewed in one wave is not interviewed in a subsequent
wave. This can lead to attrition bias, when those who are no longer in the sample differ from those who remain
in the sample. This phenomenon is discussed under the limitations section (section 3.6) of the report.
22
estimating separate regressions for men and women, the estimations are disaggregated by
youth and non-youth to explore whether the gender differences are more or less pronounced
amongst the youth. The youth samples are restricted to individuals between the ages of 20
and 35 years old for both sexes, while the non-youth samples are restricted to individuals
aged 36 to 64 years for males and 36 to 59 years for females.
15
The regressions, which
include the whole sample, include individuals between the ages of 20 to 64 years old for
males, and 20 to 59 years old for females.
3.3.1 Binary probit model for labour force participation
A set of probit regressions are used to determine how likely individuals of working-age are to
be labour force participants, based on a number of demographic and socio-economic
characteristics which take on the following form:
(1)
Where
is a binary dependent variable, which takes on a value of 0 if the individual is NEA
and 1 if the individual is economically active (employed or unemployed
16
), represents the
cumulative distribution function and represents the observed characteristics of individual
(Woolridge, 2010). The regressions are run separately for males and females and are further
disaggregated into youth and non-youth cohorts. Observed characteristics in this model
include age, race, location, marital status, level of education, English language proficiency,
the number children, working-age adults and pensioners in the household, perceived relative
family background and difficulty in performing basic daily tasks. These regressions are run
using data from NIDS Wave 1.
3.3.2 Multinomial logistic model for labour market outcomes
Generalised multinomial logistic regressions are utilised to determine the likelihood of an
individual of working-age being in a particular labour market state, based on various
characteristics. The outcomes of the multinomial logistic model are “disjunct and
exhaustive”, thus, in the context of labour market outcomes, someone can only be in one of
15
Different age brackets are used for men and women, as men and women were eligible for social pensions at
different ages at the time of data collection.
16
‘Unemployed’ includes searching and non-searching unemployed individuals.
23
the labour market outcomes mentioned, and there are no further categories available (Cramer,
1991: 43). The model takes on the following form:
(2)
Where is the indicator variable of choices, is equal to 0 for NEA, 1 for unemployed
(searching and non-searching) and 2 for employed, is the vector of independent variables
and is the corresponding coefficient vector (Cramer, 1991; So & Kuhfeld, 1995;
Woolridge, 2010). The regressions are run separately for males and females and are further
disaggregated for the youth and non-youth cohorts. The vector of independent variables
includes age, race, location, marital status, the level of education, English language
proficiency, the number of children, working-age adults and pensioners in the household,
perceived relative family background and difficulty in performing basic daily tasks. These
regressions also analyse NIDS Wave 1 data.
3.3.3 Binary logistic panel model for employment likelihood
Logistic regressions are then used to estimate panel models for two sets of periods; Wave 1 to
Wave 2 and Wave 1 to Wave 3.
17
The regressions are used to determine which factors are
likely to lead to the employment of an individual who was unemployed or NEA in Wave 1.
18
For example, the Wave 1 to 2 regressions estimate what the characteristics of the unemployed
or NEA in Wave 1 are, and which is likely to lead to these individuals becoming employed in
Wave 2. In this instance, Wave 1 is referred to as the “previous Wave” and Wave 2 is
referred to as the “subsequent Wave”. Both the NEA and unemployed individuals are
included in this analysis, as there was an inexplicable increase in the number of NEA
individuals in the NIDS data set between Wave 1 and 2 which is not consistent with statistics
from Stats SA, alluding to possible fieldwork errors in classifying individuals into the
searching unemployed, non-searching unemployed and NEA categories (Cichello et al.,
2014; Posel et al., 2014). Both NEA and unemployed individuals are included in the Wave 1
to 3 analysis as well.
17
These regressions are also run for the period between Wave 2 and 3, and the results are included in Appendix
A 7.
18
For the regressions which analyse Wave 2 to Wave 3, the regressions estimate which of the characteristics of
the unemployed or NEA in Wave 2 are likely to lead to becoming employed in Wave 3.
24
The logistic model takes on the following form:
(3)
Where
is a binary dependent variable, which takes on a value of 0 if the individual is
unemployed in the subsequent Wave, and the value of 1 if the individual is employed in the
subsequent Wave, represents characteristics of the individual in the previous period,
and its corresponding coefficient vector (Cramer, 1991). Individual characteristics from the
previous period include: whether an individual was searching unemployed; age; race;
location; marital status; level of education; English language proficiency; the number of
children, working-age adults, and pensioners in the household; perceived relative family
background; difficulty in performing basic daily tasks, and whether the individual has
previous work experience or not.
3.4. Description of variables
To answer the research questions posed, a number of variables will be used in the models
described above. Hereafter follows a description of the dependent variables and the
explanatory variables that will be used in the analysis, as well as their expected outcomes.
3.4.1 Dependent variables
As the study is focused on determining which variables are likely to influence the labour
market decisions and outcomes of men and women, the dependent variables consist of labour
market outcomes. As mentioned, these outcomes are active in the labour force or not for the
probit regressions, employed, unemployed, or NEA for the multinomial logistic regressions
and either employed or not employed for the logistic panel regressions. For all the
regressions, no distinction is made between searching unemployed and non-searching
unemployed individuals, as it has been found that compared to non-searching individuals,
those who are searching are not more or less likely to find employment in the future (Cichello
et al., 2014; Dinkelman & Pirouz, 2011; Knight & Kingdon, 2000; Posel et al., 2014). The
term ‘unemployment’ in this study thus refers to the broad definition of unemployment.
However, a distinction is made between those who are unemployed and those who are NEA,
as these groups differ markedly from one another (Dinkelman & Pirouz, 2011; Posel et al.,
2014).
25
3.4.2 Explanatory variables
In this section, a description of each explanatory variable is provided with the motivation for
including each variable in the set of regressions. The inclusion of each explanatory variable is
informed by a large body of literature (discussed in the literature review) on the determinants
of LFP. Given the focus on gender differences in outcomes, a discussion of potential differing
effects for men and women of each variable is provided.
Age
The data are restricted to individuals aged 20 to 64 for males and 20 to 59 for females, so as
to remove the effects which large numbers of young NEA scholars and NEA retirees will
have on the analysis. The upper age brackets are different for males and females and are
consistent with the ages at which individuals were eligible for social pensions at the time of
data collection (ILC, 2011).
An individual’s age is of utmost importance in determining what their labour market state
will be. Ranchhod (2010) has found that each additional year of age creates a greater
likelihood of someone being an active participant in the labour force (see Serumaga-Zake and
Kotze (2004) for similar evidence). One would expect the probability of employment to
increase as age increases, as an increase in the level of education and work experience is
often accompanied by an increase in age. Furthermore, while the youth may have a fall-back
position in the presence of their families, older individuals may be more likely to have a
greater obligation towards finding employment, as they often have families whom they need
to care for, and would thus have a lower probability of being NEA (Dinkelman & Pirouz,
2011).
Race
People from ethnic minorities are generally discriminated against, in both developed and
developing countries, and it has been found that ethnic minorities often tend to have lower
educational attainment, higher LFP rates and higher unemployment rates than their white
counterparts. Evidence of this has been found by O’Higgins (2001) in the United States, the
United Kingdom as well as in Hungary. However, women also face discrimination in the
labour market, and have come from a long history of oppression, although women from
26
different races and classes have different labour market experiences, and experience
discrimination differently.
South African designations of race are divided into four groups, namely, African, coloured,
Indian and white. Race has been found to be a significant determinant, in South African and
international studies, of LFP (Lim, 2002). In South Africa, Indian women, in particular, are
more likely to be NEA, while coloured and white females tend to be NEA from their mid-
twenties onwards. Black women tend to have higher LFP rates than women of other races in
South Africa (Serumaga-Zake & Kotze, 2004). There is thus an expectation that African
women and men will have greater LFP rates than men and women from other races, and that
the magnitudes of the coefficients will be greater for men. In contrast, it is expected that a
higher proportion of Africans will be unemployed than the other race groups and that the
magnitudes of these coefficients will be greatest for African females.
Location (provinces and geographical location)
Geographical variables are classified in accordance with the Census 2011 structure. The
geographical locations (urban, traditional and farms) used in this report are defined
accordingly: an urban area is “a continuous built-up area that is established through township
establishments such as cities, towns, ‘townships’, small towns, and hamlets”; traditional areas
are “communally owned land under the jurisdiction of traditional leaders. Settlements within
these areas are villages”. Lastly, farms are defined as “land allocated for and used for
commercial farming including the structures and infrastructure on it” (NIDS, 2014: 3).
The rural-urban divide plays an important role in South Africa’s labour market, as well as
internationally. A number of studies have found that individuals are more likely to be
employed if they reside in an urban area, as opposed to a rural area (Mlatsheni & Rospabé,
2002). However, from these definitions, one would expect the employment rate on the farms
to be greater than in traditional and urban areas, due to the fact that these areas have been
demarcated for productive activity. It has been found that the location where one resides is
especially important for women, with women having a higher likelihood of transitioning into
employment if they live in close proximity to urban areas (Cichello et al., 2014).
Dummy variables are included for all nine provinces, with Gauteng used as the reference
province. Studies have found that the Eastern Cape has the lowest rate of LFP (Serumaga-
Zake & Kotze, 2004), while Gauteng has the highest. This is not surprising, as Gauteng is an
economic hub in the country, with employment opportunities being far more plentiful in this
27
area than in other areas. There is thus an expectation that those who reside in the Eastern
Cape will have a lower probability of being active in the labour force and a lower probability
of being employed, while those residing in Gauteng have a greater probability of being active
labour force participants and employed. Van Der Westhuizen et al. (2007) also noted that
provinces which have areas that were ‘homelands’ under the apartheid regime would tend to
be less economically productive and have higher unemployment rates, such as Limpopo,
Mpumalanga, the North West and the Eastern Cape. While Gauteng and the Western Cape
are expected to be the most productive provinces and have the highest employment rates,
KwaZulu-Natal would be expected to have lower employment rates compared to these two
provinces, due to the vast rural areas in this province (similar evidence is found by Cichello
et al. (2014), Dinkelman and Pirouz (2011) and Yu (2013)).
Education
Education has been proven to be an important indicator of LFP for both males and females,
but especially so for African females, who have been entering the labour market in large
numbers (Casale & Posel, 2002). Education becomes an important variable for labour market
entry, when an individual possesses a higher degree, certificate or diploma, especially for
women, while possession of a Matric certificate is likely to induce men to enter the labour
market (Dinkelman & Pirouz, 2011; Mlatsheni & Rospabé, 2002; Van Der Westhuizen et al.,
2007). One would also expect education to increase the probability of employment for men
and women, and perhaps more so for women, as women tend to be crowded into jobs where
the returns to education are high.
19
English language proficiency is included in the regressions as a dummy variable, in order to
act as a proxy for educational quality. This variable indicates whether a respondent is able to
read and write very well in English, as reported by the respondent. The historical challenges
which face the South African education system are of a great concern to employers and
higher education institutions, however, English language proficiency generally acts as a
signal for the quality of education or for an individual’s qualifications (Casale & Posel,
2010a). The fluency in English variable takes on a value of 1 if the respondent writes very
well and reads very well in English, and 0 otherwise. English language proficiency is a
significant determinant when considering labour market returns from education (see Casale
19
King (1999: 508) suggests that these jobs are located in the “women’s sector”, in which returns to previous
work experience are low, but returns to education are high. Women’s jobs thus “do not seem to fall squarely into
the primary sector, which is theorized [sic] to reward both education and experience, nor into the secondary,
which is conceptualized [sic] as rewarding neither” (King, 1999: 508).
28
and Posel (2010a) and Mehtabul, Aimee and Nishith (2013) for a discussion on this). One
might, therefore, expect those with greater English language proficiency to have a higher
likelihood of being a labour force participant, or of being employed.
Marital status
The marital status of men and women has repeatedly been found to result in differential
labour market outcomes for these groups; it is thus one of the most important variables to
include in a set of regressions explaining differential gender effects. Being married increases
the probability of a male being an active labour force participant, while it has the opposite
effect for women (Dinkelman & Pirouz, 2011; Naudé & Serumaga-Zake, 2001; Ntuli &
Wittenberg, 2013). Being divorced has been found to decrease the probability of LFP for
African women (Ntuli & Wittenberg, 2013). These effects are likely due to the different
responsibilities which men and women are perceived to have towards productive and
reproductive work as well as the effect which the marital status has on the reservation wages
of individuals, as discussed in the literature review.
Number of children
The number of young children in the household has been found to be a significant
determinant in women’s labour force participation decisions, both internationally and locally.
Children in the home tend to decrease women’s LFP (see Serumaga-Zake and Kotze (2004)
for married women, Cichello et al. (2014), Dinkelman and Pirouz (2011), and Ntuli and
Wittenberg, (2013)). This effect is expected to be especially pronounced for young females,
as they are of peak childbearing age, and thus expected to be out of the labour force. Young
people have also been found to have a smaller probability of being employed, due to their
high propensity to have young children in the home, and there is a suggestion that being a
recipient of a childcare grant matters (Mlatsheni & Rospabe, 2002; Ranchhod, 2010; Yu,
2013). The ‘gender contract’, as described in the literature review, would be especially
relevant to this variable. However, some have found that the presence of children may also
hinder men from entering the labour market (Cichello et al., 2014).
In the regressions, ‘young children’ in the household are defined as children who are six
years and younger, while ‘older children’ are children between the ages of 7 and 15. As older
children are more independent and do not require the intense supervision that young children
do, the effects of children are expected to be less severe when there are older children
present. Thus, from the evidence provided, one would expect the presence of young children
29
to have a negative impact on the LFP of women; however, there is also a possibility that this
could be true for men. Furthermore, this variable is especially important for young females
who are of peak childbearing age, and they are thus expected to be more likely to be NEA
when there are children in the home. Among labour force participants, children may also
negatively affect the chances of finding employment if small children hinder the job search,
particularly among women.
Number of pensioners and working-age adults in the household
When considering pensioners in the household, the income which pensioners contribute to
the household should be considered, as well as the potential which they provide for child
supervision. Having a grandparent in the household could allow women to enter the labour
market, as it provides funding for them to undertake job search and provides child
supervision, where grandparents are able to offer this support (Posel et al., 2006; Aassve et
al., 2012). In contrast, having a pensioner in the household could provide additional
household income, and mean that people have less of a need to enter the labour market.
Evidence to this effect is found by Ranchhod (2010) for the non-youth cohort of the
population, as well as by Dinkelman and Pirouz (2011) and Yu (2013). Having a pensioner in
the household may thus result in decreased LFP rates for individuals. In addition to this,
pensioners may need to be taken care of, also reducing the chances which those who live with
them have of entering the labour market, especially if they are sickly or disabled.
The same income effect as that expected when having a pensioner in the household, may be
present when there are working adults in the household. However, when there are
unemployed working-age adults in the home, this income effect could be diluted. Having
employed working-age adults in the home could mean that unemployed or NEA individuals
in the household will have greater access to information about available jobs leading to an
increase in the probability of them finding work. However, when the working-age adults in
the home are not employed, this benefit may not exist. The number of working-age adults in
the home could thus act as a hindrance to LFP, or could compel individuals in the home to go
out and search for work and potentially act as an information transmission mechanism.
Perceived socio-economic status at the age of 15
In the NIDS Adult questionnaire, participants are asked to “imagine a six step ladder where
the poorest people in South Africa stand on the bottom (the first step) and the richest people
in South Africa stand on the highest step (the sixth step)”. The perceived socio-economic
30
status is based on the first sub-question in this section which asks “On which step was your
household when you were 15?” Those who indicated that their households were on the first
or the second step were categorised as ‘lower income’, those on the third or fourth step,
‘middle income’ and those who chose the fifth or the sixth step were categorised as ‘higher
income’. One must keep in mind that this is a subjective measure, and that people would have
answered this question based on their perceptions, which are formed relative to the
households around them. Although there is no definitive way to determine this, one might
imagine that individuals who grew up in a township during the apartheid era, where they
might have viewed their families as relatively more wealthy than the other families around
them, would likely answer this question differently as might a ‘born-free’ youth growing up
in a different setting. Remaining cognisant of the subjective nature of the question, one would
nevertheless expect that those who grew up in households which were higher up the income
ladder to have had greater access to education, or working adults in the household and thus
have a greater probability of being employed, or of being a labour force participant.
Daily hardship in performing basic tasks
Daily hardship is included as a proxy for health status and reports the difficulty people
experience in performing daily tasks, as good health is conducive to entering the labour
market. This variable is a dummy variable and takes on a value of 1 if a respondent reported
having any difficulty, or was unable to dress, bathe, eat, or make use of ablution facilities due
to his or her physical state. This variable is 0 if the respondent had no difficulty in performing
any of these tasks. One would thus expect people who report any difficulty in performing
these tasks to be NEA, or if there is a serious need to obtain employment, it might be harder
to secure a job with any of the conditions mentioned. Those who have difficulty would thus
be expected to be NEA or unemployed.
Previous work experience
The number of years of previous work experience has also been found to be an important
variable in determining the labour market outcome of individuals, as individuals with little or
no work experience may not have as much knowledge about the labour market, or have the
necessary experience to secure certain types of jobs (Dinkelman & Pirouz, 2011). In the
NIDS survey, employed persons are not required to state whether or not they have any
previous work experience,
20
and for this reason, this variable is only included in those panel
20
It is assumed that an employed person naturally has work experience.
31
regressions which are based on a sample of individuals who were unemployed or NEA in
Wave 1. Due to the fact that having work experience counts in one’s favour when applying
for a job, those individuals who did report having work experience in a previous period
would be expected to be more likely to transition into employment in a subsequent period.
This variable is expected to be more important for females as they have a greater likelihood
of intermittent employment histories than men do, and would need certain skills to re-enter
the labour market, should they wish to do so.
3.5. Youth and non-youth sub-samples
As society, and the cultural and traditional norms within society, have rapidly been changing,
particularly gender norms, the analysis will be extended by disaggregating the samples into
youth and non-youth cohorts to determine whether the expected gender differences in the
variables are more or less pronounced amongst the youth. There is a possibility that gender
differences in labour market outcomes may be less pronounced amongst the youth, due to
greater opportunities for obtaining education for both men and women, and the affirmative
action policies adopted, which are likely to attract women to the labour market (as discussed
in the literature review). However, there is a possibility that gender differences may be even
more pronounced, as greater labour market opportunities cannot change the fact that women
are in the peak childbearing years of their youth. This has important implications for labour
market outcomes, as previously discussed.
3.6. Limitations
The investigation undertaken in this report has a number of limitations, which will be
discussed here.
Sample size in the sets of regressions in this study presents a great limitation. As the data is
disaggregated by gender and subsequently by age cohort throughout the report, sample sizes
do tend to become very small and in some instances (where indicated), results have been
omitted where observations were too few or not present. This is especially evident when the
data is disaggregated by age and in the panel where only continuing sample members are
used in the analysis. For example, in the panel regressions when the employed individuals in
Wave 1 are dropped from the sample, 81.75 percent of the Indian male sample is dropped and
86.68 percent of the white male sample is dropped (see Table 9).
32
Another important limitation of the study is the issue of endogeneity, which describes not
being able to determine whether causality runs from the explanatory variable to the
dependent variable, or the other way around. Examples of variables which are likely to be
endogenous include the marital status of an individual, and the location where an individual
resides. The marital status of an individual could determine whether someone is employed for
example, but being employed might increase someone’s chances of getting married.
Similarly, one would be unable to determine whether someone resides in a certain area,
because they are in a particular labour market state, or whether they are in a particular labour
market state as a result of residing in that area. An example would be someone residing in an
urban area having a greater likelihood of finding employment, due to the higher number of
employment opportunities available. However, an individual could be living in an urban area
if they had been offered an employment opportunity, and was obliged to relocate as a result
of this. Coefficients of variables with a likelihood of being endogenous should thus be
interpreted as correlations, rather than as indicating causality.
There is also the possible presence of unobserved heterogeneity, which is not accounted for.
While there are a number of variables accounted for in the study, there are certain
characteristics which could have a strong impact on an individual’s labour market outcomes,
but which are not accounted for in the model, as they are generally not observable or
measurable in the data, such as innate ability, or motivation. This is related to the issue of
sample selection, if individuals who are motivated or have higher innate ability are also more
likely to enter the labour force, then the sample is not random, and this might especially be
the case for women. The methods that account for these concerns are, however, beyond the
scope of this dissertation. Thus, although the results from this study provide useful insights,
one must be wary of potential bias introduced by unobserved heterogeneity and sample
selection (Hsiao, 2014).
An additional limitation, which one faces when working with panel data, is the possibility of
attrition bias. Attrition occurs when individuals who were interviewed in one period of a
study are not interviewed in a subsequent period of the study.
21
In the case of NIDS data, this
could occur when individuals in Wave 1 are not subsequently interviewed in Wave 2 and/or
Wave 3. Attrition bias, however, occurs when those individuals who have dropped out of the
sample “are behaviourally different from those who remain” in the sample (Baigrie & Eyal,
21
There are a number of reasons why individuals and households may cease to be part of a study, such as non-
contact, fieldwork errors, or refusal or death (Baigrie & Eyal, 2013: 3).
33
2013: 3). Panel weights are thus utilised in the set of regressions, which make use of Waves 2
and 3, as well to account for attrition bias as best as possible.
3.7. Conclusion
This chapter has described the data set, which was utilised to study the research questions
under investigation, as well as the analysis techniques which were adopted, the variables
which were included in the analysis, and the limitations of the study. It is clear that a number
of variables and circumstances need to be considered before drawing conclusions about the
labour market decisions men and women might make.
As the analysis is disaggregated predominantly by gender, the following chapter analyses
descriptively the variables used in the investigation for men and women. The observed
characteristics of both males and females in relation to the variables included in the analysis
will assist in the interpretation of the regression results which will follow in Chapter 5.
34
CHAPTER 4: DESCRIPTIVE RESULTS
4.1. Introduction
The descriptive statistics presented in this chapter disaggregate the variables used in the
analysis by gender, and in some cases by age. Age brackets used in these tables are 20 to 64
years for males, and 20 to 59 years for females, unless otherwise stated. The data are
weighted and all the tables and weights
22
are from the Wave 1 dataset unless otherwise stated.
For most of the statistical tables, the row percentages for each variable are displayed,
however, when viewed with column percentages, these are equally interesting and draw a
different picture from the data. All descriptive statistics with column percentages have thus
been included in the appendix.
4.2. Labour market outcomes by gender
South Africa’s labour market is unique in that it has been shaped by years of social injustice
and has had staggeringly high unemployment rates for decades, with these rates increasing
dramatically with the advent of democracy. While there has been a marked increase in the
participation of women in the labour force, the participation rates of men continue to be
higher (Casale & Posel, 2002; Floro & Komatsu, 2011). Table 1 suggests that for all ages,
women have lower LFP rates and higher unemployment rates than men, with the difference
in LFP being greatest between the older males (78.60%) and females (65.94%) and the
smallest difference being observed between young men (81.61%) and young women
(72.62%). Unemployment rates for women are higher than for men for all age groups with
young women having the highest unemployment rate (51.76%).
Table 1: Broad labour force participation and unemployment rates by gender and age cohort (%)
Male
Female
All
(20-64)
Youth
(20-35)
Non-Youth
(36-64)
All
(20-59)
Youth
(20-35)
Non-Youth
(36-59)
LFP rate
80.07
81.61
78.60
69.14
72.62
65.94
Unemployment rate
22.19
28.12
16.28
38.77
51.76
25.60
Notes:
1. Own calculations
2. LFP rate = sum of employed and unemployed (searching and non-searching) individuals as a share of the working-age
population.
3. Unemployment rate = unemployed (searching and non-searching) individuals as a share of the economically active population
(employed and unemployed individuals).
22
The weights referred to are Wave 1 post-stratified weights, discussed in Section 3.2 of this report.
35
It is clear that as women become older, they become less economically active, and thus
experience lower unemployment rates than do the youth (25.60 percent for older females
versus 51.76 percent for young females). The same pattern is noticeable amongst the males,
although the differences between the youth and non-youth cohorts of men are not as
pronounced as the differences amongst females.
Table 2: Labour market outcomes by gender and age cohort (%)
Male
Female
All
(20-64)
Youth
(20-35)
Non-Youth
(36-64)
All
(20-59)
Youth
(20-35)
Non-Youth
(36-59)
NEA
16.22
16.02
16.44
28.38
25.87
31.09
(0.89)
(1.26)
(1.17)
(1.09)
(1.49)
(1.43)
Unemployed
16.93
21.94
11.49
27.28
35.79
18.10
(1.09)
(1.58)
(1.20)
(1.18)
(1.65)
(1.14)
Employed
66.85
62.04
72.07
44.34
38.35
50.81
(1.50)
(2.00)
(1.77)
(1.26)
(1.54)
(1.64)
Total
100
100
100
100
100
100
N
4396
2153
2243
6439
3086
3353
Notes
1. The data are weighted, standard errors in parentheses.
Table 2 presents disaggregated data of the percentage of labour market states that each age
group occupies among working-age individuals. For both sexes, the overall group consisted
mainly of employed individuals, although the percentage of employed females (44.34%) was
significantly lower than the percentage of employed males (66.85%). Women had greater
percentages than men of NEA individuals (28.38 percent versus 16.22 percent for males) as
well as unemployed individuals (27.28 percent for females versus 16.93 percent for males).
Similar patterns were observable amongst the youth and non-youth cohorts, with men having
larger shares of individuals in employment than women and women having larger shares of
NEA and unemployed individuals than men. It is interesting that young women at peak
childbearing age have a large share of individuals who are unemployed (35.79%) and that
there is a smaller share of NEA individuals in this group compared to the older females
(25.87 percent for young women versus 31.09 percent for older women). While this could
point to the influx of youth into the labour market at a young age, it could also be a result of
increasing unemployment rates in the country as a whole, reducing the incomes that enter the
households where young women are living. Although the same argument could be made for
36
young males, they do experience greater employment rates than young females, while the
percentage of unemployed is not as high as that of the females.
4.3. Labour market outcomes by marital status
As expressed in the literature review, an individual’s marital status has interesting
implications for labour market outcomes, and significant differences amongst males and
females are often present. As the group of individuals who had never been married are the
largest group in the sample (47.40 percent of males and 41.50 percent of females), they make
up the majority of individuals in each labour market state, with the exception of employed
individuals, where married individuals make up the largest proportion, with 43.62 percent of
males and 39.18 percent of females (see appendix A 1).
Table 3: Labour market outcomes by marital status and gender (%)
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
Married
12.75
8.24
79.01
100
30.51
21.45
48.04
100
(1.18)
(0.99)
(1.63)
(1.91)
(1.63)
(2.07)
Cohabit
8.87
14.36
76.77
100
24.22
38.28
37.50
100
(1.60)
(2.32)
(2.74)
(2.63)
(2.87)
(3.15)
Widow
31.47
14.56
53.97
100
40.17
12.63
47.20
100
(7.10)
(6.06)
(7.50)
(3.31)
(2.22)
(3.26)
Divorced
12.56
7.89
79.56
100
15.45
10.34
74.21
100
(3.88)
(3.29)
(5.31)
(3.62)
(2.73)
(4.70)
Never-married
20.31
24.87
54.83
100
27.13
33.41
39.45
100
(1.30)
(1.75)
(2.08)
(1.44)
(1.67)
(1.54)
Total
16.24
16.90
66.86
100
28.36
27.27
44.36
100
(0.89)
(1.08)
(1.50)
(1.09)
(1.19)
(1.27)
N
4390
6426
Notes
1. The data are weighted, standard errors in parentheses.
Turning to the row percentages shown in Table 3, the majority of married males were
employed (79.01%), while 12.75 percent were NEA and 8.24 percent were unemployed. Of
the females who were married, 48.04 percent were employed, while a significantly higher
share was NEA (30.51%), compared to married males. The share of unemployed married
females and the share of all unemployed females are similar, with 21.45 percent of married
women being unemployed, and 27.27 percent of women being unemployed overall. The
group which had the largest share of NEA individuals were widows; amongst the widowed
37
females 40.17 percent were NEA and 31.47 percent of male widowers were NEA, although a
larger share of widowed males were in employment (53.97%) compared to widowed females
(47.20%). The group which had the largest proportion of employed individuals were divorced
individuals with 74.21 percent of divorced females employed, and 79.56 percent of divorced
males employed. There are a number of reasons why this could be the case, from divorced
people being more dedicated to their jobs on the one hand, to being financially vulnerable on
the other, and for that reason choosing to stay in employment. This, however, cannot be
determined with the given statistics.
While never-married individuals could have a lower reservation wage due to the lack of
income from a spouse or partner, they could also consist of young people still residing with
their families, with less of a need to work. This diverse group of people nevertheless have a
higher proportion of males in employment (54.83%) than females (39.45%), indicating that
even amongst those that have never been married, females are more likely to be NEA (27.13
percent for females and 20.31 percent for males).
Table 4: Labour market outcomes by marital status and gender (%) - Youth only
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
Married
3.74
7.01
89.25
100
26.52
28.43
45.05
100
(1.30)
(1.89)
(2.06)
(3.08)
(2.83)
(3.70)
Cohabit
6.73
11.80
81.47
100
20.30
44.59
35.11
100
(2.34)
(2.64)
(3.58)
(3.51)
(4.14)
(4.03)
Widow
-
22.80
77.20
100
22.80
29.08
48.12
100
-
(17.81)
(17.81)
(8.94)
(10.89)
(11.25)
Divorced
-
-
100
100
5.11
21.84
73.05
100
-
-
(0.00)
(3.70)
(13.10)
(13.14)
Never-married
20.26
26.83
52.91
100
27.37
37.03
35.60
100
(1.49)
(2.01)
(2.31)
(1.70)
(1.98)
(1.70)
Total
16.04
21.95
62.01
100
25.87
35.80
38.33
100
(1.26)
(1.58)
(2.00)
(1.49)
(1.66)
(1.55)
N
2150
3079
Notes:
1. Data are weighted, standard errors are in parenthesis.
2. The sample includes young males and females between the ages of 20 and 35.
3. The sample does not have any observations for widowed young males who are NEA and divorced young males who are NEA or
unemployed.
Table 4 displays labour market outcomes by marital status for the youth only. Among young
married males, a larger proportion are employed (89.25%) while the statistics for young
married females are even lower than the total sample of married females in Table 3 (45.05%).
38
It is interesting that not many cohabiting young females are employed (35.11%), but that a
larger proportion of them are unemployed (44.59%). For all marital states, young females
have higher shares of individuals in unemployment than do males, as well as higher shares of
individuals who are NEA.
4.4. Labour market outcomes by number of children in the household
Table 5 displays the mean values, disaggregated by gender, of the number of children in the
household by labour market state.The mean values for women are higher than they are for
men for all labour market outcomes, and for children of all ages, because children are more
likely to live with their mothers in South Africa. NEA and unemployed males and females
live with a larger number of children in the household when compared to employed males
and females. This suggests that children act as an impediment to employment, or that those
who are unemployed or NEA, and their children, are likely to move into households where a
social grant is received; indicating possible endogeneity of this variable. Furthermore, the
difference in the mean number of children between the employed and NEA or unemployed
individuals is larger for females than it is for males, but particularly so for young children.
Table 5: Mean values of the number of children in the household by gender
Male
Female
NEA
Unemployed
Employed
NEA
Unemployed
Employed
Number of young children
0.53
0.59
0.41
0.88
1.01
0.69
(0.05)
(0.07)
(0.03)
(0.04)
(0.04)
(0.04)
Number of older children
0.87
0.80
0.48
1.04
1.09
0.85
(0.06)
(0.10)
(0.03)
(0.05)
(0.05)
(0.04)
N
4396
6439
Notes
1. The data are weighted, standard errors in parentheses.
2. The number of young children in the household is the sum of all the children in a particular household who are 6 years and
younger, while the number of older children is the sum of all the children in a particular household who are between the ages of
7 and 15.
In Table 6, the mean values of young children in the home are greater than those displayed in
Table 5, while women once again have higher averages of children in the home than men.
Both young men and women who are unemployed live with larger numbers of young children
in the home, while those who are employed have fewer children in the home. These
differences are once again larger between those females who are employed and those females
who are in other labour market states than they are between the males.
39
Table 6: Mean values of the number of children in the household by gender – Youth only
Male
Female
NEA
Unemployed
Employed
NEA
Unemployed
Employed
Number of young children
0.58
0.63
0.45
1.09
1.15
0.82
(0.08)
(0.09)
(0.05)
(0.07)
(0.06)
(0.05)
Number of older children
0.89
0.81
0.39
1.07
1.07
0.72
(0.09)
(0.12)
(0.04)
(0.07)
(0.07)
(0.05)
N
2153
3086
Notes:
1. Data are weighted, standard errors are in parenthesis.
2. The sample includes young males and females between the ages of 20 and 35.
3. The number of young children in the household is the sum of all the children in a particular household who are 6 years and
younger, while the number of older children is the sum of all the children in a particular household who are between the ages of
7 and 15.
4.5. Labour market outcomes by education level
The majority of males and females in the dataset have Grade 8 to 11 schooling, these groups
also make up the largest proportion of NEA, unemployed, and employed individuals for both
males and females (see Appendix A 2). Table 7 below shows that, of those who had no
schooling amongst the men, 53.25 percent were employed, while this figure was only 35.04
percent for females; where most females with this level of education were NEA (51.3%).
This pattern is also noticeable amongst the individuals with Grade 1 to 7 schooling, where
18.88 percent of men were NEA compared to 38.84 percent of females. For both sexes, the
share of NEA individuals decreased as the level of education increased, although for all
levels of education, men still had the highest share of employed individuals; 61.55 percent for
Grade 8 to 11, 70.1 percent for Matric and 83.88 percent for a degree or diploma compared to
35.03 percent for Grade 8 to Grade 11, 49.64 percent for Matric and 71.99 percent for a
degree or diploma for women. The group of individuals who had the largest proportion of
unemployed individuals were those with Grade 8 to 11 for men as well as for women. The
importance of higher education for female employment is clearly noticeable with the share of
employment increasing from 49.64 percent for those with a Matric certificate to 71.99
percent for those with a degree or diploma. This difference is much less pronounced for men.
40
Table 7: Labour market outcomes by education level and gender (%)
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
No schooling
32.05
14.70
53.25
100
51.30
13.67
35.04
100
(3.69)
(2.95)
(4.03)
(2.68)
(1.53)
(2.70)
Gr1-7
18.88
17.31
63.81
100
38.84
22.95
38.21
100
(1.98)
(1.72)
(2.50)
(1.96)
(1.66)
(1.82)
Gr8-11
18.75
19.71
61.55
100
29.83
35.14
35.03
100
(1.50)
(1.85)
(2.44)
(1.60)
(1.68)
(1.42)
Matric
12.40
17.50
70.10
100
20.27
30.08
49.64
100
(1.97)
(1.81)
(2.84)
(1.95)
(2.06)
(2.20)
Certificate/Degree/Diploma
5.10
11.03
83.88
100
12.14
15.87
71.99
100
(1.16)
(2.04)
(2.38)
(2.24)
(2.47)
(2.87)
Total
16.21
16.97
66.81
100
28.39
27.26
44.35
100
(0.89)
(1.09)
(1.50)
(1.09)
(1.18)
(1.26)
N
4387
6429
Notes
1. The data are weighted, standard errors in parentheses.
For all the levels of education, the youth had a smaller share of employed individuals than did
the full sample displayed in Table 7; while they also had a greater share of individuals who
were unemployed for each level of education. When a young male is in possession of a
higher degree or diploma, the share of NEA individuals drops below five percent (4.93%);
while a mere 11.47 percent of young females in possession of a degree or diploma were
NEA. Young females had lower shares of individuals who were NEA, than for the full
sample of females in Table 7, with the exception of those with a Grade 8 to 11 schooling and
those with a Matric certificate.
Table 8: Labour market outcomes by education level and gender (%) - Youth only
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
No schooling
20.17
31.59
48.24
100
52.20
27.33
20.47
100
(6.60)
(9.49)
(9.04)
(8.18)
(6.16)
(5.58)
Gr1-7
18.07
25.66
56.27
100
32.74
36.35
30.91
100
(3.70)
(4.08)
(4.61)
(3.82)
(3.99)
(3.37)
Gr8-11
20.73
22.66
56.61
100
31.14
42.24
26.63
100
(1.86)
(2.27)
(2.82)
(2.13)
(2.10)
(1.69)
Matric
13.63
21.19
65.18
100
21.32
34.07
44.61
100
(2.50)
(2.27)
(3.36)
(2.25)
(2.51)
(2.86)
Certificate/Degree/Diploma
4.93
16.55
78.52
100
11.47
21.31
67.22
100
(1.61)
(3.62)
(3.85)
(2.03)
(3.61)
(3.78)
Total
15.99
21.98
62.03
100
25.88
35.79
38.34
100
(1.26)
(1.59)
(2.00)
(1.49)
(1.65)
(1.54)
N
2149
3083
Notes:
1. Data are weighted, standard errors are in parenthesis.
2. The sample includes young males and females between the ages of 20 and 35.
41
Comparing young men to young women, however, shows that women are still less likely to
be employed than men, regardless of the level of education. However, the differences in the
shares of employment between those with no schooling and those with a degree or diploma
are greater among women than men, indicating that higher education matters more for
women, even amongst the youth.
4.6. Labour market outcomes by race
Africans make up the majority of the South African population, thus, they will be the group
which makes up the majority of individuals in each labour market state, with African males
accounting for 79.74 percent of males in the dataset and African females 78.64 percent of
females (see Appendix A 3).
Table 9: Labour market outcomes by race and gender (%)
23
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
African
17.76
19.12
63.12
100
28.84
30.21
40.95
100
(1.05)
(1.23)
(1.74)
(1.13)
(1.20)
(1.19)
Coloured
10.08
15.15
74.77
100
26.70
20.46
52.84
100
(1.73)
(3.92)
(3.76)
(2.56)
(3.14)
(3.50)
Indian
9.03
9.21
81.75
100
29.45
10.66
59.89
100
(4.43)
(5.90)
(6.10)
(15.34)
(3.02)
(15.11)
White
10.53
2.78
86.68
100
25.99
14.41
59.60
100
(2.49)
(1.18)
(2.78)
(3.85)
(4.29)
(4.73)
Total
16.22
16.93
66.85
100
28.38
27.28
44.34
100
(0.89)
(1.09)
(1.50)
(1.09)
(1.18)
(1.26)
N
4396
6439
Notes
1. The data are weighted, standard errors in parentheses.
Indian and white males and females generally have better labour market outcomes than
coloured and African males and females. White males have a lower percentage of NEA and
unemployed individuals, while Indian males have a lower percentage of NEA individuals
than do coloured and African males. The same is true for the white women who have a lower
share of unemployed individuals than all the other race groups (14.41%), and only 25.99
percent of individuals who are NEA. Although the racial differences are clear, the gender
differences between males and females exist nonetheless; where males of all races have a
greater percentage of employed individuals than do the females, as well as lower percentages
23
Only the results which are significant in this table are reported on in the analysis.
42
of NEA and unemployed individuals. Thus, although labour market outcomes differ by racial
lines, gender differences still persist, even amongst those who are historically more racially
privileged.
4.7. Labour market outcomes by location
As expected, the largest proportion of employed individuals is located in urban areas; this is
true for both males (70.38%), and females (68.30%). The proportion of NEA individuals are
almost equally split between traditional and urban areas for males, 47.10 percent and 47.80
percent, respectively, and the difference for females is also small; 42.39 percent for
traditional areas and 48.31 percent in urban areas (see Appendix A 4). Turning to Table 10, in
traditional areas, men have higher shares of employment (45.63%) than do females (32.68%).
As women in traditional areas are predominantly NEA (38.85%), this could allude to the
possibility of more traditional roles which women might play in these areas. The majority of
men who reside in urban areas are employed (72.74%), with a small share of NEA
individuals (11.99%). Although the shares of NEA women in urban areas are smaller than
those in traditional areas (22.43%), the share of women residing in urban areas who are
employed is still less than half (49.54%).
The smallest portions of the population reside in commercial farm areas, with the majority of
men living on farms in employment (82.84%), while just under half of women in these areas
are employed (49.87%). As these areas are demarcated for commercial use, higher shares of
employment are not surprising.
Table 10: Labour market outcomes by location and gender (%)
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
Traditional
30.05
24.31
45.63
100
38.85
28.47
32.68
100
(2.22)
(1.78)
(2.12)
(1.91)
(1.62)
(1.55)
Urban
11.99
15.27
72.74
100
22.43
28.04
49.54
100
(0.97)
(1.42)
(2.00)
(1.27)
(1.69)
(1.83)
Farms
8.36
8.80
82.84
100
33.45
16.68
49.87
100
(2.52)
(2.53)
(3.88)
(4.70)
(2.13)
(4.81)
Total
16.22
16.93
66.85
100
28.38
27.28
44.34
100
(0.89)
(1.09)
(1.50)
(1.09)
(1.18)
(1.26)
N
4396
6439
Notes
1. The data are weighted, standard errors in parentheses.
43
4.8. Labour market outcomes by province
The largest share of NEA males resided in the Eastern Cape (24.10%), while the smallest
number of NEA males resided in the Northern Cape (2.50%). This is also true for females,
with 17.83 percent of NEA females residing in the Eastern Cape, and 2.39 percent of NEA
females residing in the Northern Cape. The largest share of employed men (34.38%) resided
in Gauteng, where this was also true for females (27.46%) (see Appendix A 5).
Table 11: Labour market outcomes by province and gender (%)
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
Western Cape
12.40
7.12
80.48
100
25.54
20.87
53.59
100
(1.90)
(2.07)
(3.20)
(2.54)
(2.94)
(3.24)
Eastern Cape
32.87
22.36
44.77
100
41.57
25.85
32.58
100
(3.43)
(4.13)
(3.99)
(3.90)
(2.79)
(3.02)
Northern Cape
16.30
14.12
69.58
100
30.51
28.21
41.28
100
(2.84)
(2.31)
(4.07)
(2.71)
(2.61)
(2.91)
Free State
12.59
23.99
63.42
100
24.25
32.16
43.59
100
(1.80)
(3.23)
(4.14)
(5.71)
(3.61)
(4.65)
KwaZulu-Natal
17.34
21.27
61.39
100
25.44
26.47
48.09
100
(2.45)
(2.51)
(3.83)
(2.59)
(3.17)
(3.21)
North West
13.25
20.84
65.91
100
27.27
35.38
37.35
100
(2.89)
(3.50)
(3.65)
(5.87)
(4.34)
(3.17)
Gauteng
8.49
13.05
78.46
100
20.65
29.96
49.38
100
(1.34)
(2.05)
(2.92)
(2.11)
(2.92)
(3.25)
Mpumalanga
15.51
17.14
67.34
100
23.14
29.98
46.88
100
(2.32)
(3.28)
(4.37)
(2.53)
(3.33)
(4.07)
Limpopo
27.41
17.38
55.21
100
46.05
20.64
33.31
100
(5.36)
(3.48)
(7.07)
(3.13)
(2.52)
(3.49)
Total
16.22
16.93
66.85
100
28.38
27.28
44.34
100
(0.89)
(1.09)
(1.50)
(1.09)
(1.18)
(1.26)
N
4396
6439
Notes
1. The data are weighted, standard errors in parentheses.
The majority of men residing in each province were employed, with the highest proportion of
employed men found in the Western Cape (80.48%), and the Free State having the largest
proportion of unemployed men (23.99%), with the Eastern Cape having the largest share of
NEA men (32.87%). The Western Cape consisted of the largest share of employed females
(53.59%), the North West Province had the largest share of unemployed females (35.38%),
while Limpopo had the largest share of NEA females (46.05%). While it is difficult to draw
inferences from the statistics obtained by province, an investigation into the types of jobs
44
available to women and men in each province may shed light on the gender differences.
However, this is beyond the scope of this report.
4.9. Transition matrices
The previous set of descriptive statistics was based on the cross-sectional data from Wave 1.
Transition matrices are presented here to determine what percentage of individuals
transitioned into different labour market states and what percentage stayed in the same labour
market state over the two periods studied in the panel regression. These are presented
separately for males and females.
From Table 12, it can be confirmed that women are more likely than men to move between
labour market states, which is consistent with evidence found by Cichello et al. (2014);
possible reasons could include women moving in between states due to marriage or
childbearing responsibilities. Of those who were unemployed in Wave 1, men were more
likely to transition into employment in Wave 2, whereas women were more likely to
transition into being NEA in Wave 2. Of the males who were NEA in Wave 1, 23.34 percent
transitioned into employment, 16.62 percent into unemployment, and 60.04 percent remained
NEA. For NEA women, 18.34 percent moved into employment and a larger percentage of
females moved into unemployment than men (20.81%), with 60.84 percent remaining NEA.
During the two Waves, more males remained in employment, with 75.28 percent of those
who were employed in Wave 1 remaining employed in Wave 2, and 65.7 percent of women
who were employed in Wave 1 remaining employed in Wave 2. For both genders, most
unemployed individuals transitioned into a different labour market state, with 39.83 percent
of men moving into employment and 42.93 percent of females becoming NEA.
Table 12: Transition matrices of labour market states between Wave 1 and Wave 2 (%)
Male
Female
Wave 1
status
Wave 2 status
Wave 1
status
Wave 2 status
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
NEA
60.04
16.62
23.34
100
NEA
60.84
20.81
18.34
100
Unemployed
35.47
24.69
39.83
100
Unemployed
42.93
29.69
27.37
100
Employed
14.71
10.01
75.28
100
Employed
24.69
9.611
65.70
100
Total
27.99
13.96
58.05
100
Total
40.68
18.55
40.78
100
Notes:
1. The data are weighted using Wave 2 panel weights.
45
The Wave 1 to 3 period paints a similar picture, although a much lower percentage of men
and women remained NEA (38.49 percent of men and 50.73 percent of women) between the
two Waves, compared to the period between Wave 1 and Wave 2 (60.04 percent of men and
60.84 percent of women). A similar number of males and females also remained in
employment (75.35 percent of men and 64.9 percent of women) when compared to the
previous table. A likely explanation for the variation in results between the two matrices
could be that the transition period between Wave 1 and Wave 3 is longer than the transition
period between Wave 1 and 2, allowing for more time to move between employment states.
Table 13: Transition matrices of labour market states between Wave 1 and Wave 3 (%)
Male
Female
Wave 1
status
Wave 3 status
Wave 1
status
Wave 3 status
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
NEA
38.49
27.2
34.31
100
NEA
50.73
24.59
24.67
100
Unemployed
17.3
33.26
49.43
100
Unemployed
32.71
31.89
35.4
100
Employed
13.87
10.78
75.35
100
Employed
22.78
12.32
64.90
100
Total
19.54
18.11
62.36
100
Total
34.25
21.68
44.06
100
Notes:
1. The data are weighted using Wave 3 panel weights.
4.10. Conclusion
The descriptive statistics indicate that being NEA is a labour market state, which seems to be
more attractive to females, although the regression analysis in the next chapter will likely
shed light on the possible reasons for this. There are also clear differences between the youth
statistics and the statistics for all ages, especially for the females who display greater LFP
rates. Nonetheless, the LFP rates among young females are still lower than they are among
young males, while unemployment rates are higher. Using the variables studied in the
descriptive statistics, the regression analysis will likely shed more light on the ways in which
the variables mentioned affect the labour market outcomes, employment probabilities and
LFP rates of men and women differently, overall as well as among the youth and the non-
youth cohorts.
46
CHAPTER 5: REGRESSION ANALYSIS
5.1. Introduction
Results for all the regressions described in the methodology chapter are displayed and
analysed in this section. The first set are probit regressions, which analyse LFP, followed by
the multinomial logistic regressions, which determine the probabilities of being in a particular
labour market state. These regressions are split by gender and age cohort. Lastly, the logistic
panel results display the probability of someone moving from being NEA or unemployed in
one period to becoming employed in the next. These regressions are split by gender only, as
the sample size would fall by too much with a further disaggregation by age cohort.
24
As the purpose of this report is to investigate gender differences, the analysis in this chapter
will focus only on those variables in which gender differences are expected to be present or
are actually found in the results, such as the geographical location, educational attainment,
marital status, the number of household members, the individual’s perceived socio-economic
status, and daily hardship. Variables such as age, race and the province in which an individual
resides are not expected to produce significant or consistent gender differences, and they are
merely included as controls. The discussion which follows on the regression results will thus
not include an analysis of these variables.
5.2. Cross-sectional probit regression analysis on labour force participation
probabilities
As explained in the introduction, having as many people as possible in the labour force is a
desirable outcome for any economy; and it is thus useful for policy-makers to know how
certain factors assist in driving men and women, young and old into the labour market. The
binary probit regressions describe those individual factors that correlate with men and women
being labour force participants, as well as how these factors differ between the youth and
non-youth cohorts. The regression output is displayed in Table 14.
24
All the regressions were re-run for a restricted African sample to verify that results were not driven by racial
differences, however, the results obtained from this restricted sample did not differ much to the results included
in this report. These regression results are thus not included in this analysis.
47
Men and women of all ages are more likely to be labour force participants if they reside in an
urban area, compared to those who live in traditional areas. The coefficients for individuals
living in urban areas are all significant, while the magnitudes are greater for the women than
men, with the exception of the youth cohort. Men residing on farms are more likely to be
labour force participants, while the magnitude for the older cohort of males is greater than
that for the overall sample and the young males. The coefficients for females of all ages are
also positive, although none of the coefficients are significant. Women and men of all ages
thus have a greater probability of being in employment when they live outside of a traditional
area, with women having a greater likelihood of being a labour force participant in an urban
area, and men having a greater likelihood of being a labour force participant on a farm.
It is clear that education is one of the most important variables to consider when making a
decision to participate in the labour force or not, as education raises the returns which an
individual gains in the labour market. For women in the whole sample and the older cohort of
women, the magnitudes of the coefficients for all levels of education are greater than those of
their male counterparts, with the exception of Grade 1 to Grade 7 schooling, where the
magnitudes for the men are greater. The results for the youth differ slightly, with young
females having coefficients which are positive, significant and greater in magnitude for all
levels of education compared to the coefficients for young males; while only having a higher
degree or diploma was significant for young males. For women, the likelihood of being a
labour force participant becomes greater with every level of education, with the effects being
especially strong for young females, while for men, this in only evident among the older
cohort (although the coefficient for Grade 1 to Grade 7 schooling is greater than the
coefficient for Grade 8 to Grade 11 schooling). From the results, one can thus conclude that
the higher the level of education which an individual possesses, the higher the likelihood of
that individual participating in the labour force, however, this effect is especially important
for young women. Furthermore, only amongst the women, those who reported being fluent in
English (a proxy for quality) were more likely to be active labour force participants. The
coefficient for this variable was not significant for any of the male regressions.
Marriage is an especially important variable, with the coefficients for males all being
significant and positive and being negative for females, although none of the female
coefficients are significant. Compared to those who had never been married, married men are
thus more likely to be labour force participants, and women are less likely to be labour force
48
participants. These results are expected, as numerous studies have found that being married
has a positive effect on the LFP of men and a negative effect on the LFP of women (Naudé &
Serumaga-Zake, 2001; Ntuli, 2007; Ntuli & Wittenberg, 2013). The coefficients for divorced
females are also highly significant and positive, indicating the greater need that women have
to work if they do not have financial support from a spouse, or to re-enter the labour force if
they had chosen not to participate while they were married. The coefficient for divorced
young females has the greatest magnitude, while the coefficients for divorced men are all
positive, but none are significant.
The number of children in the household is surprisingly not significant in any of the
regressions, while the signs of the coefficients also do not produce any consistent results. The
number of working-age adults in the household do have gender differences, with the signs of
the coefficients being negative for males and positive for females, indicating that for each
additional working-age adult in the household, a male is less likely to be a labour force
participant and a female is more likely to be a labour force participant. However, only the
coefficient for older males is significant with the magnitude of this coefficient also being
greater for this group of males than the youth cohort of males. Having working-age adults in
the household could enable women to enter the labour market if the working-age adults are
unemployed individuals who are able to take care of children, but could provide the financial
resources to undertake job search, if there are employed working-age adults in the home.
The number of pensioners in the household has the same effect for males, with all the male
coefficients being negative, however only the older male coefficient is significant when
disaggregated by age. The coefficients for females are also negative with the exception of
young females who have a positive coefficient. While the coefficient for young females is not
significant, it could allude to the possibility of the income from the pension facilitating job
search for these young women, or the presence of a grandparent providing childcare, which
enables young women to enter the labour market (Aassve et al., 2012; Posel et al., 2006). The
significant negative coefficient for the older cohort of women could indicate to the possibility
of a pensioner needing care, resulting in these women being unable to enter the labour
market.
The perceived socio-economic status of an individual’s family when they were young is
particularly important for older females, with those who considered themselves to be in the
49
middle and higher income groups at age 15 being more likely to be in the labour force than
those in the lower income groups. Only the coefficients for males of all ages and older males
are significant for the daily hardship variable, the coefficients for all the groups are negative,
indicating that those who have difficulty performing basic tasks on a daily basis are less
likely to enter the labour force, as expected.
50
Table 14: Probability of labour force participation, by gender and age cohort
Binary probit model
Labour Force Participation
All
Youth
Non-Youth
Male
Female
Male
Female
Male
Female
Age
0.238***
0.187***
0.797***
0.449***
0.167**
0.068
(0.02)
(0.02)
(0.15)
(0.11)
(0.08)
(0.08)
Age Squared
-0.313***
-0.247***
-1.331***
-0.724***
-0.231***
-0.118
(0.03)
(0.02)
(0.27)
(0.21)
(0.08)
(0.09)
Race (Ref: African)
Coloured
0.303*
-0.089
0.879***
0.234
0.027
-0.351**
(0.16)
(0.11)
(0.23)
(0.15)
(0.19)
(0.15)
Indian
-0.241
-0.618*
0.276
-0.420
-0.518**
-0.836**
(0.26)
(0.32)
(0.54)
(0.34)
(0.22)
(0.37)
White
0.105
-0.491***
0.672*
-0.101
-0.214
-0.793***
(0.18)
(0.16)
(0.38)
(0.28)
(0.20)
(0.20)
Province (Ref: Gauteng)
Western Cape
-0.417**
-0.065
-0.211
-0.184
-0.529**
0.048
(0.19)
(0.12)
(0.28)
(0.19)
(0.22)
(0.17)
Eastern Cape
-0.730***
-0.406***
-0.771***
-0.360**
-0.758***
-0.455***
(0.15)
(0.11)
(0.19)
(0.16)
(0.20)
(0.14)
Northern Cape
-0.401**
-0.104
-0.366
-0.228
-0.451**
0.006
(0.16)
(0.12)
(0.23)
(0.19)
(0.22)
(0.17)
Free State
-0.194
-0.089
0.097
-0.037
-0.411*
-0.092
(0.15)
(0.21)
(0.19)
(0.23)
(0.23)
(0.27)
KwaZulu-Natal
-0.086
0.190*
-0.022
0.280*
-0.238
0.120
(0.14)
(0.11)
(0.20)
(0.17)
(0.19)
(0.15)
North West
-0.105
0.067
0.151
0.034
-0.389*
0.108
(0.16)
(0.18)
(0.19)
(0.25)
(0.23)
(0.17)
Mpumalanga
-0.316*
0.065
-0.286
0.090
-0.297
0.041
(0.17)
(0.11)
(0.22)
(0.18)
(0.26)
(0.14)
Limpopo
-0.655***
-0.484***
-0.816***
-0.766***
-0.472**
-0.060
(0.19)
(0.13)
(0.24)
(0.18)
(0.24)
(0.17)
Geographical Location (Ref:
Traditional)
Urban
0.263**
0.280***
0.301**
0.301**
0.264**
0.283***
(0.10)
(0.08)
(0.14)
(0.12)
(0.12)
(0.10)
Farms
0.722***
0.114
0.525*
0.022
0.929***
0.259*
(0.20)
(0.12)
(0.29)
(0.16)
(0.19)
(0.14)
Educational Attainment (Ref: No
Schooling)
Grade 1 to Grade 7
0.343***
0.178**
0.169
0.557***
0.372***
0.088
(0.13)
(0.09)
(0.29)
(0.20)
(0.13)
(0.10)
Grade 8 to Grade 11
0.322**
0.393***
0.150
0.698***
0.308*
0.373***
(0.14)
(0.09)
(0.28)
(0.19)
(0.17)
(0.12)
Matric
0.506***
0.689***
0.344
1.031***
0.570**
0.695***
(0.17)
(0.12)
(0.30)
(0.20)
(0.25)
(0.19)
Diploma/Degree
0.858***
0.971***
0.660**
1.259***
0.859***
0.969***
(0.18)
(0.15)
(0.32)
(0.24)
(0.25)
(0.20)
Fluent English
0.118
0.173**
0.060
0.167*
0.234
0.220*
(0.09)
(0.07)
(0.11)
(0.09)
(0.15)
(0.12)
Marital Status (Ref: Never-married)
Married
0.395***
-0.068
0.532***
-0.119
0.397***
-0.133
(0.09)
(0.07)
(0.18)
(0.12)
(0.13)
(0.09)
Cohabiting
0.189
0.069
0.104
0.168
0.280*
-0.151
(0.13)
(0.11)
(0.20)
(0.15)
(0.17)
(0.14)
Widow
0.051
-0.059
-
-0.123
-0.040
-0.108
(0.28)
(0.11)
-
(0.32)
(0.27)
(0.13)
Divorced
0.313
0.480***
-
0.879**
0.287
0.454**
(0.20)
(0.17)
-
(0.40)
(0.21)
(0.19)
Young Children
0.067
0.006
0.069
-0.010
0.050
-0.021
(0.05)
(0.03)
(0.07)
(0.04)
(0.06)
(0.04)
Older Children
-0.063
-0.019
-0.010
-0.024
-0.078
0.003
(0.04)
(0.03)
(0.06)
(0.03)
(0.05)
(0.03)
Working-age adults
-0.032
0.014
-0.034
0.008
-0.069**
0.030
(0.03)
(0.02)
(0.04)
(0.03)
(0.03)
(0.02)
Pensioners
-0.184**
-0.031
-0.058
0.103
-0.442***
-0.202**
(0.08)
(0.06)
(0.09)
(0.07)
(0.11)
(0.08)
51
Perceived socio-economic status
(Ref: Lower Income)
Middle Income
-0.035
0.037
-0.169
-0.084
0.110
0.164*
(0.09)
(0.07)
(0.12)
(0.11)
(0.13)
(0.09)
Higher Income
-0.051
0.361
-0.300
0.147
0.158
0.647**
(0.27)
(0.23)
(0.36)
(0.36)
(0.35)
(0.31)
Daily Hardship
-0.491**
-0.295
-0.163
-0.214
-0.496**
-0.260
(0.22)
(0.18)
(0.55)
(0.40)
(0.22)
(0.20)
Constant
-3.369***
-3.213***
-10.685***
-6.994***
-1.831
-0.493
(0.49)
(0.41)
(1.94)
(1.57)
(1.86)
(1.96)
N
4345
6375
2113
3060
2217
3315
Notes
1. Source: NIDS, 2008
2. The data are weighted, standard errors in parentheses.
3. *** p<0.01 ** p<0.05 * p<0.10.
4. Age groups are restricted as follows: All male (19-64); All female (19-59); Youth male and female (19-35); Older male (36-64);
Older female (36-59).
5. No observations for divorced and widowed young males, as sample sizes are too small.
52
5.3. Cross-sectional multinomial logistic regression analysis on
probabilities of different labour market outcomes
The previous set of regressions investigated what the factors were which affected the LFP of
men and women differently. However, making a decision to be a labour force participant
does not guarantee employment. For this reason, a multinomial logistic model was utilised to
determine which of the factors, utilised in the previous set of regressions are likely to result in
an individual being employed, unemployed or NEA. The base outcome used in this model is
unemployed, with the first half of the regressions determining what the likelihood of someone
being NEA is, relative to being unemployed, and the second displaying the likelihood of
someone being employed relative to being unemployed.
Amongst those individuals who reside in farm areas, there are clear gender differences, with
males residing on farms being less likely to be NEA, while females are more likely to be
NEA. This could be due to the types of jobs that are available on commercial farms, and
could allude to the possibility that men could be working on farms while their families live
with them, possibly in remote areas, where other employment is not easily obtainable. All
groups have a greater probability of being employed compared to unemployed if they reside
on a farm, with older males having the greatest probability of being employed. Surprisingly,
none of the coefficients for the urban variable are significant for the employed outcome,
although both men and women – of all ages – are less likely to be NEA when residing in an
urban area, with the results for females being significant and the magnitudes of the
coefficients being greater than those of the males.
Men and women are less likely to be NEA than unemployed (i.e. in the labour force) for
every level of education, compared to those who have no schooling, although the magnitude
and significance of the coefficients are much stronger for women. Both men and women have
a greater likelihood of being employed rather than unemployed when they are in possession
of a degree or diploma. Once again, these coefficients, are stronger for women than they are
for men (while the youth male coefficient is not significant), emphasising the important role
which education plays in ensuring that women are not only labour force participants, but that
they are also able to obtain employment. It is also interesting to note that only a post-
secondary education seems to matter for obtaining employment.
53
Those who reported being fluent in English amongst the females, were less likely to be NEA
and more likely to be employed compared to unemployed, while the men who are fluent in
English were also less likely to be NEA, but less likely to be employed compared to
unemployed. Although most of the coefficients for the employed and NEA outcomes for this
variable are not significant, it does reflect the possibility that language proficiency is more
important for the kinds of jobs women do.
Marital status is an important variable in determining whether someone is in employment or
not. Married and cohabiting males are more likely to be employed relative to being
unemployed, whereas married and cohabiting females are more likely to be unemployed than
employed, with the exception of older females for whom there is no significant effect. This is
possibly indicating the ease of engaging in employment for men, when there is a partner
present to take care of reproductive work in the home. Divorced individuals are also more
likely to be employed relative to being unemployed, possibly reflecting the need for divorced
individuals to be financially independent, as the supporting income of a partner is not present.
The coefficient for older divorced females is especially strong.
Women, apart from older females living in a home with young children, are interestingly all
less likely to be NEA when there are children in the home, however, all women are less likely
to be employed as well when there are children in the household. The result for this was
negative and significant for females in the whole sample, as well as for young females for
whom the coefficients are greater. This could indicate that there is a need for women to work,
where they report being part of the labour force, but where they might have difficulty
securing employment as their children require supervision. Men are also less likely to be
employed, compared to unemployed, when there are older children in the household, with the
result being negative and significant for men in the whole sample.
Working-age adults in the home resulted in men being less likely to be employed, with the
result being negative and significant for older men in particular, while this result was
negative – but not significant – for women in the whole sample and for young women, but
positive for older women.
A number of studies have found that living with a pensioner reduces the employment
probabilities for both males and females (Dinkelman & Pirouz, 2011; Ranchhod, 2010).
Having a pensioner in the household has a negative effect on the probability of employment
for all groups, with the coefficients for men being stronger than those of the women. Once
54
again, there is a possibility that some of these individuals may be amongst the non-searching
unemployed, which would explain the convenience of remaining unemployed when there are
intra-household transfers of income, or it could allude to the fact that unemployed individuals
tend to relocate to households where there is a pensioner present (Dinkelman & Pirouz,
2011).
While the perceived socio-economic statuses of individuals do not necessarily display
consistent results for either one of the outcomes, women are more likely to be NEA when
they have difficulty performing basic tasks, particularly older women. This might indicate
that difficulty in performing daily tasks, combined with being an ageing individual, could
have a greater than anticipated impact on labour market outcomes. Interestingly, women who
have difficulties in performing basic tasks are more likely to be employed than unemployed,
again particularly women in the non-youth cohort. This may be as a result of women being
able to obtain jobs which may not require a lot of physical effort, such as office jobs, whereas
the same result is not found for men.
55
Table 15: Probability of labour market outcomes, by gender and age cohort
Multinomial logistic model
Reference outcome: Unemployed
Not Economically Active
All
Youth
Non-Youth
Males
Females
Males
Females
Males
Females
Age
-0.382***
-0.281***
-1.286***
-0.855***
-0.558***
-0.373*
(0.05)
(0.04)
(0.30)
(0.21)
(0.21)
(0.20)
Age Squared
0.516***
0.415***
2.182***
1.477***
0.695***
0.514**
(0.06)
(0.05)
(0.56)
(0.39)
(0.22)
(0.21)
Race (Ref: African)
Coloured
-0.800*
0.397*
-2.262***
-0.072
0.549
0.755***
(0.41)
(0.24)
(0.58)
(0.34)
(0.46)
(0.28)
Indian
0.467
1.503***
-1.150
1.745***
3.817***
1.138
(0.95)
(0.49)
(1.60)
(0.36)
(0.79)
(0.90)
White
0.993*
1.086**
0.147
0.729
1.643***
1.297***
(0.52)
(0.42)
(1.11)
(0.64)
(0.63)
(0.48)
Province (Ref: Gauteng)
Western Cape
1.187***
0.279
1.013
0.473
0.815*
-0.015
(0.43)
(0.30)
(0.68)
(0.41)
(0.49)
(0.41)
Eastern Cape
0.775**
0.588**
0.808**
0.547*
0.960**
0.611**
(0.31)
(0.23)
(0.36)
(0.32)
(0.45)
(0.29)
Northern Cape
0.731**
0.037
1.051**
0.227
0.304
-0.162
(0.34)
(0.25)
(0.51)
(0.37)
(0.54)
(0.32)
Free State
-0.206
-0.020
-0.830**
-0.021
0.549
-0.138
(0.29)
(0.36)
(0.38)
(0.43)
(0.50)
(0.47)
KwaZulu-Natal
-0.044
-0.170
-0.213
-0.437
0.258
0.246
(0.29)
(0.26)
(0.43)
(0.35)
(0.42)
(0.32)
North West
-0.158
-0.328
-0.669
-0.218
0.576
-0.519
(0.35)
(0.38)
(0.46)
(0.46)
(0.43)
(0.42)
Mpumalanga
0.283
-0.087
0.097
-0.057
0.379
-0.239
(0.30)
(0.24)
(0.38)
(0.36)
(0.55)
(0.31)
Limpopo
0.738*
0.842***
0.948*
1.074***
0.480
0.293
(0.40)
(0.27)
(0.51)
(0.35)
(0.50)
(0.37)
Geographical Location (Ref:
Traditional)
Urban
-0.337
-0.477***
-0.369
-0.544**
-0.312
-0.422*
(0.23)
(0.18)
(0.30)
(0.22)
(0.27)
(0.22)
Farms
-0.372
0.345*
-0.204
0.392*
-0.251
0.252
(0.43)
(0.19)
(0.64)
(0.23)
(0.44)
(0.28)
Educational Attainment (Ref: No
Schooling)
Grade 1 to Grade 7
-0.521*
-0.491***
-0.093
-0.864**
-0.578*
-0.325
(0.27)
(0.18)
(0.58)
(0.36)
(0.32)
(0.22)
Grade 8 to Grade 11
-0.411
-0.915***
0.097
-1.175***
-0.615
-0.870***
(0.32)
(0.18)
(0.56)
(0.35)
(0.40)
(0.26)
Matric
-0.610
-1.111***
-0.220
-1.454***
-0.629
-1.058**
(0.38)
(0.25)
(0.60)
(0.36)
(0.62)
(0.46)
Diploma/Degree
-0.989**
-1.026***
-0.655
-1.385***
-0.797
-0.819*
(0.45)
(0.31)
(0.70)
(0.44)
(0.64)
(0.48)
Fluent English
-0.235
-0.223
-0.089
-0.243
-0.674*
-0.263
(0.19)
(0.14)
(0.23)
(0.18)
(0.37)
(0.25)
Marital Status (Ref: Never-married)
Married
0.114
0.096
0.152
0.111
-0.138
0.232
(0.24)
(0.13)
(0.55)
(0.21)
(0.34)
(0.18)
Cohabiting
-0.033
-0.327
0.352
-0.508*
-0.621
0.010
(0.33)
(0.21)
(0.49)
(0.28)
(0.39)
(0.27)
Widow
-0.017
0.456
-
20.905***
0.266
-0.233
0.661**
(0.62)
(0.28)
(0.96)
(0.67)
(0.66)
(0.30)
Divorced
0.113
-0.093
0.658
-1.432
-0.250
0.199
(0.58)
(0.38)
(0.54)
(1.03)
(0.60)
(0.41)
Young Children
-0.095
-0.054
-0.109
-0.035
-0.053
0.004
(0.10)
(0.05)
(0.14)
(0.06)
(0.14)
(0.09)
Older Children
0.001
-0.027
-0.081
-0.031
0.091
-0.034
(0.10)
(0.05)
(0.13)
(0.06)
(0.12)
(0.07)
Working-age Adults
-0.011
-0.028
0.026
-0.027
-0.075
-0.012
(0.05)
(0.03)
(0.07)
(0.04)
(0.09)
(0.06)
56
Pensioners
0.107
-0.082
-0.041
-0.200
0.322
-0.000
(0.16)
(0.11)
(0.17)
(0.13)
(0.24)
(0.19)
Perceived socio-economic status
(Ref: Lower Income)
Middle Income
0.174
-0.099
0.481*
0.093
-0.288
-0.366
(0.20)
(0.15)
(0.27)
(0.23)
(0.30)
(0.23)
Higher Income
-0.277
-0.941
0.132
-0.946
-0.012
-0.477
(0.49)
(0.60)
(0.53)
(0.84)
(1.02)
(0.78)
Daily Hardship
0.723
0.869**
0.138
0.423
0.727
1.300**
(0.55)
(0.43)
(1.36)
(0.77)
(0.64)
(0.53)
Constant
6.564***
5.337***
17.926***
13.220***
11.113**
7.126
(0.98)
(0.74)
(3.96)
(2.87)
(5.20)
(4.56)
Employed
All
Youth
Non-Youth
Male
Female
Male
Female
Male
Female
Age
0.064
0.105***
0.201
-0.116
-0.356*
-0.315
(0.04)
(0.04)
(0.27)
(0.23)
(0.19)
(0.20)
Age Squared
-0.066
-0.066
-0.280
0.375
0.385*
0.389*
(0.05)
(0.05)
(0.50)
(0.42)
(0.20)
(0.22)
Race (Ref: African)
Coloured
-0.334
0.407
-0.777
0.674**
0.612
0.210
(0.37)
(0.25)
(0.52)
(0.28)
(0.38)
(0.28)
Indian
-0.047
0.483
-0.645
1.402
2.911***
-0.389
(0.72)
(0.47)
(0.73)
(0.89)
(0.71)
(0.77)
White
1.203**
0.276
1.629*
0.902**
1.256**
-0.096
(0.47)
(0.34)
(0.87)
(0.41)
(0.54)
(0.42)
Province (Ref: Gauteng)
Western Cape
0.534
0.210
0.912
0.192
-0.115
0.075
(0.42)
(0.25)
(0.64)
(0.30)
(0.45)
(0.32)
Eastern Cape
-0.743**
-0.218
-0.812*
-0.173
-0.535
-0.278
(0.33)
(0.21)
(0.42)
(0.29)
(0.41)
(0.28)
Northern Cape
0.084
-0.274
0.611
-0.419
-0.497
-0.205
(0.28)
(0.23)
(0.42)
(0.30)
(0.45)
(0.29)
Free State
-0.781***
-0.300
-1.004***
-0.210
-0.278
-0.410
(0.24)
(0.22)
(0.30)
(0.29)
(0.48)
(0.26)
KwaZulu-Natal
-0.291
0.214
-0.293
0.051
-0.267
0.550*
(0.24)
(0.24)
(0.33)
(0.26)
(0.38)
(0.33)
North West
-0.428
-0.418*
-0.467
-0.401
-0.152
-0.530
(0.27)
(0.24)
(0.42)
(0.26)
(0.40)
(0.40)
Mpumalanga
-0.391
0.031
-0.561
0.175
-0.235
-0.238
(0.32)
(0.25)
(0.48)
(0.30)
(0.41)
(0.30)
Limpopo
-0.652*
0.015
-0.767*
-0.489
-0.509
0.232
(0.33)
(0.25)
(0.43)
(0.33)
(0.54)
(0.37)
Geographical Location (Ref:
Traditional)
Urban
0.201
0.002
0.239
-0.056
0.246
0.070
(0.21)
(0.15)
(0.28)
(0.17)
(0.24)
(0.21)
Farms
1.329***
0.834***
1.168***
0.815***
1.760***
0.891***
(0.32)
(0.21)
(0.39)
(0.27)
(0.38)
(0.29)
Educational Attainment (Ref: No
Schooling)
Grade 1 to Grade 7
0.131
-0.305
0.491
0.108
0.047
-0.258
(0.24)
(0.20)
(0.46)
(0.42)
(0.30)
(0.24)
Grade 8 to Grade 11
0.208
-0.474**
0.698
-0.090
-0.124
-0.369
(0.30)
(0.20)
(0.50)
(0.42)
(0.36)
(0.24)
Matric
0.397
0.151
0.767
0.559
0.385
0.205
(0.34)
(0.25)
(0.53)
(0.45)
(0.49)
(0.43)
Diploma/Degree
0.706*
0.855***
1.038*
1.245***
0.778
0.967**
(0.39)
(0.28)
(0.61)
(0.47)
(0.54)
(0.42)
Fluent English
-0.051
0.121
-0.058
0.108
-0.248
0.175
(0.17)
(0.14)
(0.19)
(0.18)
(0.30)
(0.22)
Marital Status (Ref: Never-married)
Married
1.131***
-0.025
1.517***
-0.170
0.784***
0.017
(0.20)
(0.14)
(0.42)
(0.18)
(0.26)
(0.21)
Cohabiting
0.559**
-0.405**
0.861***
-0.469**
0.005
-0.399
(0.24)
(0.17)
(0.30)
(0.20)
(0.33)
(0.25)
Widow
0.239
0.506**
0.289
0.092
-0.224
0.643**
(0.59)
(0.25)
(1.21)
(0.62)
(0.62)
(0.26)
57
Divorced
0.843*
0.939***
21.982***
0.574
0.377
1.173***
(0.50)
(0.32)
(0.61)
(0.83)
(0.52)
(0.39)
Young Children
0.048
-0.090*
-0.017
-0.144**
0.110
-0.048
(0.07)
(0.05)
(0.10)
(0.07)
(0.12)
(0.07)
Older Children
-0.167*
-0.113**
-0.161
-0.189***
-0.073
-0.043
(0.10)
(0.05)
(0.12)
(0.06)
(0.11)
(0.06)
Working-age adults
-0.112**
-0.001
-0.051
-0.009
-0.259***
0.048
(0.06)
(0.04)
(0.06)
(0.05)
(0.08)
(0.06)
Pensioners
-0.370**
-0.247**
-0.279*
-0.041
-0.641***
-0.474***
(0.15)
(0.11)
(0.17)
(0.15)
(0.21)
(0.18)
Perceived socio-economic status
(Ref: Lower Income)
Middle Income
0.154
-0.051
0.325*
-0.108
-0.141
-0.134
(0.12)
(0.13)
(0.18)
(0.20)
(0.22)
(0.20)
Higher Income
-0.462
-0.414
-0.584
-1.366**
0.225
0.723
(0.52)
(0.42)
(0.63)
(0.54)
(0.88)
(0.73)
Daily Hardship
-0.176
0.593*
-0.339
0.280
-0.217
1.070**
(0.53)
(0.35)
(0.77)
(0.49)
(0.70)
(0.53)
Constant
-0.302
-2.177***
-2.976
0.323
9.874**
6.991
(0.86)
(0.82)
(3.66)
(3.32)
(4.44)
(4.72)
N
4345
6375
2128
3060
2217
3315
Notes:
1. Source: NIDS, 2008.
2. The data are weighted, standard errors in parentheses.
3. *** p<0.01 ** p<0.05 * p<0.10.
4. Age groups are restricted as follows: All male (19-64); All female (19-59); Youth male and female (19-35); Older male (36-64);
Older female (36-59).
5. Widowed youth males for the “not economically active” outcome, produced a large and significant coefficient. The size of this
coefficient is most likely due to the small sample size of widowed youth males in the dataset. This coefficient should thus be
interpreted with caution.
58
5.4. Panel logistic regression analysis on employment probabilities
The panel logistic regression results, which determine the likelihood of transitioning into
employment over a period, are displayed in Table 16. While the binary probit model and the
multinomial logistic model presented in the last two sections gave insight into the conditions
which may result in LFP and the likelihood of obtaining employment at a particular time, the
panel regressions allow for an investigation into the employment probabilities of an
individual across time, based on the factors used in the previous regressions.
As mentioned in the methodology chapter, the employed individuals are dropped from the
sample, to determine which factors are likely to result in an unemployed or NEA individual
in Wave 1 obtaining employment in Wave 2 or Wave 3.
25
As each period presents its own
unique challenges and circumstances relevant to that time, many coefficients in the regression
output did not render consistent results. However, the most relevant findings are discussed in
this section.
The periods under consideration are Period I (2008-2010/11) and Period II (2008-2012).
These regressions are not disaggregated by age, as disaggregating by youth and non-youth
cohorts significantly reduce the sample sizes. The sample sizes for males of all ages were
significantly reduced when the employed individuals were dropped from the sample with
66.85 percent of males and 44.34 percent of females dropped from the sample (see Table 2).
26
An additional variable included in these regressions is a dummy variable stating whether the
individual was searching unemployed,
27
or not, in Wave 1. This variable is positive and
significant for males across both periods, and implies that being searching unemployed in
Wave 1 was very likely to result in employment in a subsequent Wave for men. For women,
the variable was only significant and positive for the Period I regression.
In Period I, women were more likely to obtain employment if they resided in urban areas or
farms, compared to those who resided in traditional areas, although these coefficients are
25
These regressions were also run for an additional period, Wave 2 to Wave 3, using Wave 2 characteristics as
the base, this is included as an appendix (see Appendix A 7). Appendix A 8 provides the shares of labour market
outcomes by gender for Wave 2, to accompany the regression analysis.
26
The regressions were also run with a sample of unemployed individuals only (excluding NEA individuals)
and these results are displayed in Appendix A 6. In these regressions, the sample sizes of the females especially
become significantly reduced, and therefore, these results should be interpreted with caution.
27
The searching unemployed refers to those individuals who are without work, reported wanting to work, have
actively searched for work in the last 4 weeks, and are able to accept a job within the next week.
59
negative in Period II. The same result was obtained when the NEA individuals were removed
from the sample (see Appendix A 6).
Only women with a Matric certificate in Period I or degree/diploma in Periods I and II were
more likely to transition into employment, with none of the coefficients being significant for
men. This is consistent with stronger coefficients obtained for females on educational
variables in the cross-section regressions.
The marital status variables did not produce any noteworthy results, although cohabiting
males were significantly more likely to transition into employment in Period I, while
divorced males were significantly more likely to transition into employment in Period II. It is
worth noting that the sample size for NEA and unemployed married males, which is the
sample included for these regressions are small, likely resulting in insignificant results.The
coefficients for the number of children in the household were not significant across the two
periods, while the effect of having a working-age adult in the home mattered for men, who
had negative and significant coefficients in both periods. Having a pensioner present in the
household had the same effect, with the coefficients being significantly negative for men, and
the coefficients being negative but not significant for women across both periods.
Reporting difficulty in performing basic daily tasks is correlated significantly with a
reduction in the employment probabilities of males in Period II, while the coefficients for
females were positive but not significant. This is consistent with results from the cross-
sectional regressions, once again alluding to the possibility of women being able to obtain
jobs where a lot of physical effort is not necessary.
The last effect of interest which the panel produced was that having previous work
experience only seemed to matter for women, who were more likely to transition into
employment if they reported having previous work experience in Wave 1 for both sets of
regressions, periods I and II.
60
Table 16: Probability of employment across periods, by gender.
Binary logistic model
Period
I
II
Wave 1 to Wave 2
Wave 1 to Wave 3
Dependent Variable
Wave 2 outcome
Wave 3 outcome
Male
Female
Male
Female
Searching unemployed in Wave 1
0.353*
0.281*
0.714***
0.171
(0.21)
(0.15)
(0.18)
(0.14)
Age
0.137**
0.152***
0.100**
0.157***
(0.05)
(0.05)
(0.05)
(0.04)
Age Squared
-0.211***
-0.212***
-0.185***
-0.223***
(0.07)
(0.07)
(0.07)
(0.06)
Race (Ref: African)
Coloured
0.245
0.100
0.890**
0.554**
(0.46)
(0.31)
(0.44)
(0.26)
Indian
0.728
-1.304
-2.723**
-1.670*
(1.06)
(0.82)
(1.37)
(0.87)
White
-0.691
-0.301
1.128
0.030
(0.96)
(0.46)
(0.69)
(0.40)
Province (Ref: Gauteng)
Western Cape
-0.697
-0.028
-0.517
-0.094
(0.61)
(0.30)
(0.47)
(0.28)
Eastern Cape
0.013
0.201
0.238
-0.389*
(0.36)
(0.24)
(0.33)
(0.22)
Northern Cape
0.099
-0.781**
0.136
-0.712***
(0.40)
(0.30)
(0.40)
(0.26)
Free State
0.080
-0.105
0.706**
-0.110
(0.37)
(0.27)
(0.33)
(0.26)
KwaZulu-Natal
-0.289
-0.247
0.201
-0.298
(0.35)
(0.25)
(0.32)
(0.22)
North West
0.108
-0.636**
0.156
-0.931***
(0.37)
(0.30)
(0.36)
(0.27)
Mpumalanga
-0.163
-0.446*
0.411
-0.383
(0.36)
(0.27)
(0.34)
(0.25)
Limpopo
-0.388
-0.107
0.657*
-0.310
(0.38)
(0.28)
(0.34)
(0.25)
Geographical Location (Ref: Traditional)
Urban
-0.332
0.290*
0.076
-0.172
(0.23)
(0.17)
(0.20)
(0.15)
Farms
0.155
0.724***
0.453
-0.051
(0.36)
(0.25)
(0.35)
(0.24)
Educational Attainment (Ref: No Schooling)
Grade 1 to Grade 7
-0.139
0.340
-0.920***
-0.251
(0.36)
(0.31)
(0.35)
(0.25)
Grade 8 to Grade 11
-0.469
0.520
-0.558
-0.113
(0.37)
(0.32)
(0.34)
(0.25)
Matric
-0.175
0.851**
-0.187
0.436
(0.42)
(0.36)
(0.39)
(0.29)
Diploma/Degree
0.058
1.516***
0.616
0.963***
(0.56)
(0.41)
(0.48)
(0.35)
Fluent English
0.180
-0.058
-0.071
0.115
(0.21)
(0.17)
(0.18)
(0.15)
Marital Status (Ref: Never-married)
Married
-0.407
-0.056
0.268
-0.231
(0.33)
(0.19)
(0.30)
(0.18)
Cohabiting
0.761**
0.008
0.263
0.132
(0.37)
(0.24)
(0.40)
(0.21)
Widow
0.352
0.190
-0.802
-0.018
(0.76)
(0.34)
(0.86)
(0.33)
Divorced
0.145
-0.544
1.425**
0.191
(0.69)
(0.53)
(0.64)
(0.48)
Young Children
-0.104
-0.056
0.148
0.064
(0.10)
(0.07)
(0.09)
(0.06)
Older Children
0.037
-0.049
-0.005
-0.056
(0.09)
(0.06)
(0.08)
(0.05)
Working-age adults
-0.191**
0.061
-0.174***
-0.026
61
(0.08)
(0.05)
(0.06)
(0.04)
Pensioners
-0.448**
-0.044
-0.574***
-0.164
(0.18)
(0.15)
(0.17)
(0.13)
Perceived socio-economic status (Ref: Lower Income)
Middle Income
-0.043
-0.213
-0.057
0.066
(0.22)
(0.17)
(0.19)
(0.14)
Higher Income
0.030
0.744
-0.739
0.064
(0.68)
(0.49)
(0.62)
(0.50)
Daily Hardship
-0.513
0.273
-1.399**
0.217
(0.75)
(0.48)
(0.68)
(0.39)
Previous work experience
0.115
0.283*
-0.313
0.387***
(0.23)
(0.16)
(0.22)
(0.15)
Constant
-1.790*
-4.531***
-1.047
-3.189***
(1.05)
(0.89)
(0.94)
(0.78)
N
1383
3001
1464
3100
Notes
1. Source: NIDS, 2008, 2010/11, 2012.
2. The data are weighted, standard errors in parentheses.
3. *** p<0.01 ** p<0.05 * p<0.10.
4. Wave 2 panel weights were utilised to analyse period I and Wave 3 panel weights were utilised to analyse period II.
62
5.5. Conclusion
The three sets of regressions thus all served a different purpose, where the binary probit
investigates how demographic and socio-economic characteristics of men and women affect
LFP differently, while the multinomial logistic regressions indicate whether these
characteristics are likely to result in being employed, NEA or unemployed. Lastly, the
logistic panel regressions considered the way in which these characteristics may be likely to
affect the employment probabilities of a NEA or unemployed individual across time. The
regressions were all run with the purpose of investigating gender differences, as well as
gender differences across different age groups.
Women and men outside of traditional areas had a greater probability of being labour force
participants and being employed, while women were more likely to transition into
employment if they were on a farm or in an urban area. This is likely pointing to the limited
economic opportunities in traditional areas.
The most noteworthy results were that education was important for males and females, but
played an especially important role for women, who were more likely to be labour force
participants, more likely to be employed and more likely to transition out of unemployment
or inactivity, the higher the level of education. For women, the level of post-secondary
education was especially important, and in addition to this, previous work experience also
mattered for women in transitioning into employment.
Married men were not only more likely to be labour force participants, but they were also
more likely to be employed. The opposite was true for married women, who were more likely
to be NEA and less likely to be employed. Although the number of children in the home did
not have the expected negative effect on women’s LFP, this reduced the chances of women
being in employment in Wave 1, especially young women who are of peak childbearing age.
The composition of the household was especially important for men, who were less likely to
be labour force participants, less likely to be employed, and less likely to transition into
employment if there was a working-age adult or a pensioner in the household. While having a
pensioner in the home enabled young women to enter the labour market, it hindered the
participation of older women. Lastly, the daily hardship variable provided interesting results,
as men were less likely to be labour force participants and less likely to transition into
63
employment, while women were less likely to be NEA, but interestingly, more likely to be
employed, if they reported having difficulty in performing certain basic daily tasks.
64
CHAPTER 6: CONCLUSION
This study attempted to explore whether there are gender differences in the demographic and
socio-economic characteristics, which determine an individual’s labour market outcomes. A
number of regressions were run to determine how these characteristics influence women’s
LFP decisions differently to men’s, and how these characteristics impacted employment
probabilities differently for men and women.
Data from Wave 1, 2, and 3 of the NIDS surveys were used in this report. The methods
utilised to investigate the research questions posed in the introduction, included estimating a
binary probit model to determine how characteristics of individuals impacted the likelihood
of being a labour force participant or not. This was followed by a multinomial logistic model
to determine how these characteristics impacted the probability of an individual being NEA,
employed or unemployed. Lastly, a binary logistic model was used to analyse the way in
which individual characteristics affect the likelihood of an unemployed or NEA individual
obtaining employment in a subsequent period.
In the literature review, a number of theories were investigated, which hypothesised on the
determinants of LFP of individuals, and particularly women. Empirical evidence was then
used to support or refute the hypotheses forwarded by the theoretical views presented. The
human capital theory, an individual’s reservation wage, the issue of reproductive labour, and
the added worker effect were analysed in the context of the study, and how these will
possibly affect men and women’s LFP differently.
The main findings of the report related to the geographical location of an individual, the
effects of education, the marital status of an individual, and the household composition of the
individual. These variables produced interesting gender differences in the ways in which they
affect labour market outcomes.
It was found that there may be restrictive gender relations in traditional areas and
opportunities which may not be as widespread for females as they are for males. In contrast
to this, farms were highly conducive for the employment probabilities of men and women;
while women were more likely to be labour force participants and employed if they resided in
an urban area. This suggests that there may be job segmentation in the types of jobs which are
available to men and women in urban areas, traditional areas and on farms. In addition, this
65
could point to the fact that more opportunities are available in urban areas than in traditional
areas, and that those who reside on farms are more likely to be employed as these areas have
been demarcated for productive activity, i.e. commercial farming.
Although education was an important factor for both men and women, it had a stronger effect
in inducing women to enter the labour market and in obtaining employment. This was evident
from all three sets of regressions, with higher levels of education, such as the possession of a
martic certificate, a degree, or a diploma having significantly positive effects on the labour
force participation and employment probabilities of women. This is consistent with findings
from Dinkelman and Pirouz (2011), Mlatsheni and Rospabé (2002) and Van Der Westhuizen
et al. (2007). In addition to this, the youth also gained greatly from increased levels of
education, with young women especially, choosing to enter the labour market as opposed to
remaining inactive, the higher the level of education. Having previous work experience
affected women’s employment probabilities positively as well, with women being more
likely to transition into employment, having been unemployed or inactive in a previous
period, if they have previous work experience. This may be as a result of women obtaining
jobs, where a certain level of education and previous work experience is valued by
employers, such as clerical and office jobs, whereas men may more easily obtain jobs which
involve manual labour or where on the job training may be more appropriate.
Women were not likely to be NEA when there were children present in the home, although
their employment probabilities were reduced. Given the fact that many women are heading
households, this could allude to a lower reservation wage when there are children present in
the home, but an inability to obtain employment as children and in particular young children,
may be in need of supervision. The presence of older children in the home also reduced the
employment probabilities of men. Young women and men were less likely to be employed
and less likely to be NEA, compared to unemployed, than their older counterparts, when
there were children present in the household. In addition to this, they also had higher mean
values of children, both young and older in the household, thus making it difficult for young
women in particular to obtain employment. This indicates that young people with children in
the household may have a need to work, thus choosing to be active in the labour force, but
that they may find it difficult to obtain employment. Once again, this could allude to the
reservation wage of young people with children in the home being lower than for the older
cohort who may not have the burden of many young children in the household, but who
interestingly have larger shares of NEA individuals.
66
The presence of pensioners and working-age adults in the household had an adverse effect on
the employment probabilities and LFP of males. Having a pensioner who receives a pension
in the home, could mean that men are benefitting from intra-household transfers and choosing
not to work as a result. Furthermore, young women were more likely to be labour force
participants when there is a pensioner in the home, possibly suggesting that they are
facilitating job search and providing childcare responsibilities. This is consistent with
findings from Aassve et al. (2012), Posel et al. (2006) and Ranchhod (2010).
The study illustrates that, gender differences do exist in the ways in which certain
characteristics affect the labour market outcomes of men and women. These differences also
exist among the youth cohort in many cases, suggesting that traditional roles around the
reproductive age play a larger role than changing norms in society around gender roles.
Recommendations
Assuming that many women residing on farm areas are NEA due to them accompanying their
husbands to their places of work, work opportunities, which are appropriate for women, could
be explored on farms to make use of the potential labour these women provide; that is, if they
wish to enter into employment while living in these areas.
Education is also an area where inroads can be made, as the report has shown the strong
effects which education has for both sexes, and especially for women and the youth in
entering the labour force, as well as obtaining employment. Greater access to quality
education would prove beneficial to the South African labour market, as this would increase
the opportunity cost of choosing to stay NEA, and would also increase the likelihood of
individuals obtaining employment.
Since women with children are more likely to be unemployed, relative to being NEA, and
more likely to be unemployed relative to being employed, the possibility of women wanting
or needing to enter the labour market, but not being able to find work, due to childcare
responsibilities is demonstrated. It would thus be in the interest of these women to have work
environments where children could be taken care of, perhaps in a day-care facility. Policy-
makers would do well in encouraging firms to accommodate young women in managing their
childcare responsibilities.
Areas for future research
While the results presented some interesting gender differences even among the youth cohort,
the study faced a number of limitations. The limitations, as discussed in Chapter 3, were
67
mainly related to causality, which was difficult to determine in certain instances; as well as
sample size, particularly in the panel regressions. While these challenges are beyond the
scope of this report, this suggests that there is still room for further investigation should better
data become available.
68
APPENDICES
A 1: Labour market outcomes by gender and marital status (column
percentages)
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
Married
28.97
18.00
43.62
36.91
38.91
28.45
39.18
36.17
(2.34)
(1.89)
(2.20)
(1.83)
(2.10)
(2.10)
(1.75)
(1.33)
Cohabit
6.03
9.38
12.68
11.04
9.76
16.03
9.66
11.43
(1.08)
(1.50)
(1.10)
(0.78)
(1.11)
(1.47)
(1.19)
(0.83)
Widow
3.56
1.58
1.48
1.84
9.25
3.02
6.95
6.53
(0.97)
(0.74)
(0.27)
(0.27)
(0.99)
(0.56)
(0.70)
(0.46)
Divorced
2.17
1.31
3.34
2.81
2.38
1.66
7.31
4.37
(0.64)
(0.55)
(0.64)
(0.46)
(0.56)
(0.42)
(1.09)
(0.55)
Never-married
59.26
69.72
38.87
47.40
39.70
50.84
36.91
41.50
(2.62)
(2.83)
(1.83)
(1.63)
(1.88)
(2.25)
(1.69)
(1.34)
Total
100
100
100
100
100
100
100
100
N
4390
6426
Notes
1. The data are weighted, standard errors in parentheses.
A 2: Labour market outcomes by gender and education level (column
percentages)
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
No schooling
15.00
6.57
6.05
7.59
13.13
3.64
5.74
7.26
(1.83)
(1.39)
(0.70)
(0.64)
(1.20)
(0.46)
(0.56)
(0.52)
Gr1-7
23.74
20.79
19.48
20.39
25.65
15.78
16.15
18.75
(2.28)
(1.79)
(1.49)
(1.17)
(1.68)
(1.25)
(1.18)
(0.88)
Gr8-11
39.42
39.59
31.41
34.10
40.18
49.29
30.20
38.24
(2.54)
(2.48)
(1.67)
(1.25)
(1.82)
(1.71)
(1.64)
(1.21)
Matric
16.83
22.71
23.10
22.02
14.35
22.18
22.50
20.10
(2.65)
(2.17)
(1.55)
(1.23)
(1.40)
(1.45)
(1.45)
(0.97)
Certificate/Degree/Diploma
5.00
10.33
19.96
15.90
6.69
9.11
25.41
15.65
(1.13)
(2.00)
(1.80)
(1.36)
(1.47)
(1.09)
(2.32)
(1.44)
Total
100
100
100
100
100
100
100
100
N
4387
6429
Notes
1. The data are weighted, standard errors in parentheses.
69
A 3: Labour market outcomes by gender and race (column percentages)
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
African
87.29
90.07
75.29
79.74
79.92
87.09
72.63
78.64
(2.07)
(2.60)
(2.75)
(2.26)
(2.93)
(2.35)
(3.08)
(2.38)
Coloured
4.82
6.94
8.67
7.75
8.34
6.65
10.56
8.86
(0.90)
(2.12)
(1.31)
(1.11)
(1.47)
(1.76)
(1.66)
(1.45)
Indian
1.37
1.34
3.01
2.46
2.56
0.96
3.33
2.47
(0.85)
(1.34)
(1.19)
(1.08)
(2.09)
(0.48)
(1.14)
(1.03)
White
6.52
1.65
13.03
10.05
9.18
5.30
13.48
10.03
(1.87)
(0.70)
(2.25)
(1.71)
(2.09)
(1.67)
(2.46)
(1.72)
Total
100
100
100
100
100
100
100
100
N
4396
6439
Notes
1. The data are weighted, standard errors in parentheses.
A 4: Labour market outcomes by gender and location (column percentages)
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
Traditional
47.10
36.51
17.35
25.42
42.39
32.33
22.82
30.97
(3.93)
(4.29)
(1.93)
(2.34)
(3.24)
(3.15)
(2.35)
(2.43)
Urban
47.80
58.34
70.38
64.68
48.31
62.84
68.30
61.14
(3.77)
(4.39)
(2.72)
(2.62)
(3.10)
(3.23)
(2.65)
(2.52)
Farms
5.10
5.15
12.27
9.90
9.30
4.83
8.88
7.89
(1.71)
(2.01)
(2.40)
(1.97)
(2.76)
(1.11)
(1.86)
(1.73)
Total
100
100
100
100
100
100
100
100
N
4396
6439
Notes
1. The data are weighted, standard errors in parentheses.
70
A 5: Labour market outcomes by gender and province (column percentages)
Male
Female
NEA
Unemployed
Employed
Total
NEA
Unemployed
Employed
Total
Western Cape
7.28
4.00
11.46
9.52
9.69
8.24
13.02
10.77
(1.16)
(0.99)
(1.43)
(0.98)
(1.21)
(1.41)
(0.97)
(0.75)
Eastern Cape
24.10
15.71
7.97
11.90
17.83
11.54
8.94
12.17
(2.24)
(3.40)
(0.86)
(0.98)
(1.51)
(1.90)
(1.26)
(1.07)
Northern Cape
2.50
2.07
2.59
2.49
2.39
2.30
2.07
2.22
(0.41)
(0.35)
(0.33)
(0.22)
(0.33)
(0.28)
(0.20)
(0.17)
Free State
4.82
8.80
5.89
6.21
4.81
6.63
5.53
5.62
(0.77)
(1.27)
(0.60)
(0.42)
(1.16)
(0.92)
(0.71)
(0.44)
KwaZulu-Natal
16.53
19.43
14.20
15.46
17.31
18.75
20.95
19.32
(2.22)
(2.69)
(1.32)
(1.10)
(2.07)
(1.99)
(1.66)
(1.07)
North West
6.86
10.34
8.28
8.40
6.40
8.64
5.61
6.66
(1.19)
(2.20)
(0.94)
(0.79)
(1.68)
(0.94)
(0.44)
(0.51)
Gauteng
15.33
22.57
34.38
29.29
17.94
27.08
27.46
24.66
(2.61)
(3.43)
(2.38)
(1.97)
(2.09)
(2.92)
(2.06)
(1.60)
Mpumalanga
7.43
7.87
7.83
7.77
6.57
8.86
8.52
8.06
(1.18)
(1.64)
(0.84)
(0.65)
(0.89)
(1.36)
(1.03)
(0.75)
Limpopo
15.14
9.20
7.40
8.96
17.05
7.95
7.89
10.51
(2.39)
(1.68)
(1.47)
(0.96)
(1.62)
(0.96)
(1.18)
(0.84)
Total
100
100
100
100
100
100
100
100
N
4396
6439
Notes
1. The data are weighted, standard errors in parentheses.
71
A 6: Probability of employment across periods, by gender - excluding NEA
individuals
Binary Logistic Model
Period
I
II
Wave 1 to 2
Wave 1 to 3
Dependent Variable
Wave 2 outcome
Wave 3 outcome
Male
Female
Male
Female
Searching unemployed in previous Wave
0.067
0.032
0.414*
0.071
(0.26)
(0.19)
(0.22)
(0.16)
Age
0.009
-0.047
0.049
0.066
(0.07)
(0.06)
(0.06)
(0.05)
Age Squared
-0.010
0.098
-0.051
-0.039
(0.10)
(0.09)
(0.09)
(0.08)
Race (Ref: African)
Coloured
-0.217
-0.224
1.098*
0.616
(0.51)
(0.43)
(0.63)
(0.38)
Indian
0.677
-2.052*
-2.832**
-1.670
(1.31)
(1.05)
(1.28)
(1.27)
White
-
0.386
1.480
0.743
-
(0.72)
(1.14)
(0.60)
Province (Ref: Gauteng)
Western Cape
1.852**
0.522
-0.391
0.188
(0.81)
(0.49)
(0.54)
(0.39)
Eastern Cape
0.279
0.262
0.061
-0.521**
(0.44)
(0.30)
(0.38)
(0.26)
Northern Cape
0.089
-0.410
0.577
-0.641*
(0.47)
(0.37)
(0.49)
(0.34)
Free State
-0.080
-0.013
0.715*
-0.182
(0.47)
(0.33)
(0.37)
(0.31)
KwaZulu-Natal
-0.184
0.047
0.122
-0.339
(0.43)
(0.31)
(0.36)
(0.26)
North West
0.248
-0.404
0.227
-0.735**
(0.48)
(0.34)
(0.42)
(0.31)
Mpumalanga
0.302
0.064
0.349
-0.774***
(0.49)
(0.33)
(0.38)
(0.28)
Limpopo
0.071
0.190
0.558
-0.461
(0.51)
(0.35)
(0.39)
(0.30)
Geographical Location (Ref: Traditional)
Urban
0.006
0.444**
-0.055
-0.407**
(0.30)
(0.21)
(0.22)
(0.18)
Farms
0.481
0.974**
0.419
-0.124
(0.56)
(0.43)
(0.40)
(0.32)
Educational Attainment (Ref: No Schooling)
Grade 1 to Grade 7
-0.081
0.481
-1.800***
-0.233
(0.52)
(0.45)
(0.41)
(0.32)
Grade 8 to Grade 11
-0.474
0.484
-1.191***
-0.200
(0.53)
(0.45)
(0.41)
(0.32)
Matric
-0.063
0.708
-0.824*
0.224
(0.58)
(0.50)
(0.46)
(0.36)
Diploma/Degree
-0.791
1.452***
0.113
0.520
(0.69)
(0.56)
(0.61)
(0.44)
Fluent English
0.143
-0.076
-0.035
0.135
(0.27)
(0.20)
(0.21)
(0.17)
Marital Status (Ref: Never-married)
Married
0.045
0.364
0.145
-0.213
(0.44)
(0.26)
(0.35)
(0.22)
Cohabiting
1.117**
-0.008
-0.112
0.111
(0.48)
(0.30)
(0.46)
(0.25)
Widow
1.175
0.771
-1.925*
-0.202
(1.34)
(0.47)
(1.11)
(0.43)
Divorced
-0.146
-0.635
3.553***
0.153
(1.21)
(0.71)
(1.21)
(0.68)
Young Children
-0.140
-0.006
0.136
0.092
(0.13)
(0.08)
(0.11)
(0.07)
Older Children
0.014
-0.045
0.105
-0.113*
72
(0.12)
(0.08)
(0.09)
(0.06)
Working-age adults
-0.089
0.053
-0.176**
-0.014
(0.09)
(0.06)
(0.07)
(0.05)
Pensioners
-0.061
0.019
-0.475**
-0.107
(0.24)
(0.18)
(0.20)
(0.16)
Perceived socio-economic status (Ref: Lower Income)
Middle Income
-0.038
-0.267
0.261
0.119
(0.28)
(0.20)
(0.22)
(0.16)
Higher Income
0.011
0.568
-0.951
-0.023
(0.83)
(0.62)
(0.68)
(0.66)
Daily Hardship
0.000
1.425**
-0.882
0.512
(.)
(0.72)
(0.80)
(0.53)
Previous work experience
0.131
0.332
-0.361
0.274
(0.30)
(0.21)
(0.26)
(0.18)
Constant
0.532
-0.903
0.495
-1.111
(1.32)
(1.12)
(1.11)
(0.93)
N
662
1286
1022
1758
Notes
1. Source: NIDS, 2008, 2010/11, 2012.
2. The data are weighted, standard errors in parentheses.
3. *** p<0.01 ** p<0.05 * p<0.10.
4. Wave 2 panel weights were utilised to analyse period I and Wave 3 panel weights were utilised to analyse period II.
5. The dataset does not contain any unemployed white males in period I, thus the coefficients are omitted.
73
A 7: Probability of employment across periods, by gender - Wave 2 to Wave 3
Binary Logistic Model
Dependent Variable: Wave 3 outcome
Excludes NEA individuals
Male
Female
Male
Female
Searching unemployed in Wave 2
0.738***
0.101
0.415**
-0.034
(0.17)
(0.16)
(0.20)
(0.17)
Age
0.291***
0.245***
0.170***
0.113**
(0.04)
(0.04)
(0.05)
(0.05)
Age Squared
-0.419***
-0.308***
-0.223***
-0.089
(0.06)
(0.06)
(0.08)
(0.07)
Race (Ref: African)
Coloured
-0.430
0.639**
-0.365
0.624*
(0.37)
(0.27)
(0.44)
(0.33)
Indian
0.740
-0.424
1.556
-0.103
(0.66)
(0.68)
(1.09)
(0.97)
White
0.404
0.005
0.709
1.302
(0.59)
(0.76)
(0.87)
(1.07)
Province (Ref: Gauteng)
Western Cape
-0.170
0.237
-0.196
0.387
(0.42)
(0.31)
(0.46)
(0.38)
Eastern Cape
-0.724**
-0.173
-0.792**
-0.194
(0.33)
(0.27)
(0.38)
(0.30)
Northern Cape
0.276
0.002
0.599
0.026
(0.34)
(0.29)
(0.46)
(0.34)
Free State
0.107
0.360
-0.029
0.520
(0.32)
(0.27)
(0.37)
(0.33)
KwaZulu-Natal
-0.384
0.330
-0.207
0.494*
(0.27)
(0.24)
(0.32)
(0.29)
North West
-0.251
-0.332
-0.377
-0.220
(0.32)
(0.29)
(0.39)
(0.35)
Mpumalanga
-0.105
0.221
-0.139
-0.036
(0.30)
(0.26)
(0.34)
(0.30)
Limpopo
-0.204
0.389
0.005
0.398
(0.33)
(0.28)
(0.40)
(0.33)
Geographical Location (Ref: Traditional)
Urban
-0.110
0.070
-0.072
-0.064
(0.20)
(0.16)
(0.23)
(0.20)
Farms
0.257
-0.153
0.147
0.059
(0.27)
(0.25)
(0.30)
(0.28)
Educational Attainment (Ref: No Schooling)
Grade 1 to Grade 7
-0.659**
0.324
-1.177***
0.120
(0.32)
(0.26)
(0.41)
(0.37)
Grade 8 to Grade 11
-0.723**
0.244
-1.115***
0.025
(0.32)
(0.26)
(0.41)
(0.38)
Matric
-0.444
0.582*
-0.844*
0.405
(0.37)
(0.31)
(0.45)
(0.42)
Diploma/Degree
-0.299
1.063***
-0.706
0.516
(0.43)
(0.37)
(0.53)
(0.49)
Fluent English
0.183
0.330**
0.061
0.302*
(0.19)
(0.16)
(0.21)
(0.18)
Marital Status (Ref: No Schooling)
Married
0.560*
-0.548***
0.543
-0.490**
(0.30)
(0.19)
(0.34)
(0.22)
Cohabiting
0.333
0.184
0.169
0.363
(0.34)
(0.24)
(0.40)
(0.29)
Widow
1.260**
-0.078
1.979*
-0.094
(0.52)
(0.29)
(1.15)
(0.38)
Divorced
-0.793
-0.613
-0.932
-0.910*
(1.22)
(0.48)
(1.19)
(0.54)
Young Children
0.055
-0.007
0.025
0.031
(0.09)
(0.06)
(0.10)
(0.07)
Older Children
-0.013
0.032
0.098
0.028
(0.07)
(0.05)
(0.09)
(0.07)
Working-age adults
-0.094*
-0.034
-0.103*
-0.060
(0.05)
(0.04)
(0.06)
(0.05)
Pensioners
-0.384**
-0.276**
-0.188
-0.248
74
(0.15)
(0.14)
(0.18)
(0.16)
Perceived socio-economic status (Ref: Lower Income)
Middle Income
0.081
-0.151
0.280
-0.118
(0.17)
(0.15)
(0.20)
(0.17)
Higher Income
-0.685
0.775*
-0.931
1.160*
(0.73)
(0.45)
(0.76)
(0.65)
Daily Hardship
-0.319
-0.269
0.031
-0.097
(0.39)
(0.34)
(0.54)
(0.40)
Previous work experience
0.176
-0.329
0.323
-0.187
(0.35)
(0.36)
(0.47)
(0.45)
Constant
-4.192***
-6.026***
-1.512
-3.115***
(0.82)
(0.79)
(1.02)
(0.92)
N
1884
3200
1165
1685
Notes
1. Source: NIDS, 2010/11, 2012.
2. The data are weighted, standard errors in parentheses.
3. *** p<0.01 ** p<0.05 * p<0.10.
4. Wave 3 panel weights were utilised in this panel analysis.
A 8: Wave 2 labour market outcomes by gender (%)
Male
Female
NEA
25.79
39.73
(1.11)
(1.09)
Unemployed
14.36
18.86
(0.90)
(0.89)
Employed
59.86
41.41
(1.34)
(1.15)
Total
100
100
N
3552
5118
Notes:
1. Source: NIDS, 2010/11
2. Data are weighted with Wave 2 weights
75
A 9: Mean characteristics of variables by gender
Variable
Total
Male
Female
Variable
Total
Male
Female
NEA
0.372
0.163
0.283
Farms
0.083
0.100
0.079
(0.01)
(0.01)
(0.01)
(0.02)
(0.02)
(0.02)
Unemployed
0.191
0.171
0.274
Never-married
0.493
0.474
0.415
(0.01)
(0.01)
(0.01)
(0.01)
(0.02)
(0.01)
Employed
0.437
0.666
0.443
Married
0.318
0.369
0.361
(0.01)
(0.02)
(0.01)
(0.01)
(0.02)
(0.01)
Age
36.473
36.650
36.197
Cohabiting
0.088
0.110
0.114
(0.30)
(0.34)
(0.22)
(0.01)
(0.01)
(0.01)
African
0.792
0.799
0.787
Widow
0.069
0.019
0.066
(0.02)
(0.02)
(0.02)
(0.00)
(0.00)
(0.00)
Coloured
0.079
0.076
0.088
Divorced
0.032
0.028
0.044
(0.01)
(0.01)
(0.01)
(0.00)
(0.00)
(0.01)
Indian
0.024
0.025
0.025
No Schooling
0.091
0.075
0.073
(0.01)
(0.01)
(0.01)
(0.00)
(0.01)
(0.01)
White
0.104
0.100
0.100
Grade 1 to 7
0.197
0.205
0.188
(0.02)
(0.02)
(0.02)
(0.01)
(0.01)
(0.01)
Gauteng
0.245
0.289
0.245
Grade 8 to 11
0.396
0.340
0.381
(0.01)
(0.02)
(0.02)
(0.01)
(0.01)
(0.01)
Western Cape
0.097
0.094
0.108
Matric
0.187
0.219
0.201
(0.01)
(0.01)
(0.01)
(0.01)
(0.01)
(0.01)
Eastern Cape
0.132
0.120
0.122
Diploma/Degree
0.128
0.160
0.157
(0.01)
(0.01)
(0.01)
(0.01)
(0.01)
(0.01)
Northern Cape
0.022
0.025
0.022
Fluent English
0.462
0.463
0.469
(0.00)
(0.00)
(0.00)
(0.01)
(0.02)
(0.02)
Free State
0.057
0.062
0.056
Number of young children
0.688
0.464
0.834
(0.00)
(0.00)
(0.00)
(0.03)
(0.03)
(0.03)
KwaZulu-Natal
0.187
0.157
0.193
Number of older children
0.905
0.600
0.969
(0.01)
(0.01)
(0.01)
(0.03)
(0.04)
(0.03)
North West
0.073
0.084
0.067
Number of working-age adults
2.750
2.559
2.895
(0.00)
(0.01)
(0.01)
(0.07)
(0.08)
(0.08)
Mpumalanga
0.079
0.079
0.081
Number of pensioners
0.284
0.141
0.164
(0.01)
(0.01)
(0.01)
(0.01)
(0.01)
(0.01)
Limpopo
0.108
0.090
0.105
Lower Income
0.680
0.708
0.675
(0.01)
(0.01)
(0.01)
(0.01)
(0.02)
(0.01)
Traditional
0.321
0.256
0.310
Middle Income
0.295
0.265
0.301
(0.02)
(0.02)
(0.02)
(0.01)
(0.01)
(0.01)
Urban
0.596
0.644
0.611
Higher Income
0.024
0.027
0.024
(0.02)
(0.03)
(0.03)
(0.00)
(0.00)
(0.00)
Daily Hardship
0.026
0.017
0.019
(0.00)
(0.00)
(0.00)
Notes:
1. Source: NIDS, 2008
2. Data are weighted, standard errors in parenthesis.
76
REFERENCES
Aassve, A., Arpino, B. & Goisis, A. (2012). Grandparenting and mothers’ labour force
participation: A comparative analysis using the generations and gender survey.
Demographic Research, 27(3), 53-83.
Addabbo, T., Favaro, D. & Magrini, S. (2012). Gender differences in productivity rewards:
The role of human capital. International Review of Economics, 59(1), 81-110.
Albelda. R. (1999). Marxist political economics. In J. Petersen & M. Lewis (Eds.), The Elgar
companion to Feminist Economics (pp. 536-543). Chelthenham: Edward Elgar.
Alexander, N. (2006). Affirmative action and the perpetuation of racial identities in post-
apartheid South Africa. Transformation, 63, 92-126.
Baigrie, N. & Eyal, K. (2014). An evaluation of the determinants and implications of panel
attrition in the National Income Dynamics Survey (2008-2010). South African
Journal of Economics, 82(1), 39-65.
Bakker, I. (1999). Development policies. In J. Petersen & M. Lewis (Eds.), The Elgar
companion to Feminist Economics (pp. 83-95). Chelthenham: Edward Elgar.
Barker, D. K. (1999a). Gender. In J. Petersen & M. Lewis (Eds.), The Elgar companion to
Feminist Economics (pp. 390-396). Chelthenham: Edward Elgar.
Barker, D. K. (1999b). Neoclassical economics. In J. Petersen & M. Lewis (Eds.), The Elgar
companion to Feminist Economics (pp. 570-576). Chelthenham: Edward Elgar.
Bbaale, E. & Mpuga, P. (2011). Female education, labour force participation and choice of
the employment type: Evidence from Uganda. International Journal of Economics
and Business Modeling, 2(1), 29-41.
Becker, G. S., Murphy, K. M. & Tamura, R. (1994). Human capital, fertility and economic
growth. In G. S. Becker (Ed.), Human capital: A theoretical and empirical analysis
with special reference to education (3rd ed.) (pp. 323-350). Chicago, IL: University of
Chicago Press.
77
Bridges, S., & Lawson, D. (2008). A Gender-based Investigation into the Determinants of
Labour Market Outcomes: Evidence from Uganda. Journal of African Economies,
18(3), 461-495.
Bridges, S., Lawson, D., & Begum, S. (2011). Labour market outcomes in Bangladesh: The
role of poverty and gender norms. European Journal of Development Research, 23(3),
459-487.
Brown, M., Daniels, R. C., De Villiers, L., Leibbrandt, M. & Woolard, I., eds. (2012).
National Income Dynamics Study Wave 2 User Manual. Cape Town: Southern Africa
Labour and Development Research Unit.
Burger, R. & Von Fintel, D. (2009). Determining the causes of the rising South African
unemployment rate: An age, period and generational analysis (Working Paper No.
24). Retrieved from University of Stellenbosch, Department of Economics website:
www.ekon.sun.ac.za/wpapers/2009/wp242009/wp-24-2009.pdf.
Casale, D. & Posel, D. (2002). The continued feminisation of the labour force in South
Africa: An analysis of recent data and trends. South African Journal of Economics,
70(1), 156-184.
Casale, D. & Posel, D. (2010a). English language proficiency and earnings in a developing
country: The case of South Africa. The Journal of Socio-Economics, 40, 385-393.
Casale. D. & Posel, D. (2010b). The male marital earnings premium in the context of bride
wealth payments: Evidence from South Africa. Economic Development and Cultural
Change, 58(2), 211-230.
Cichello, P., Leibbrandt, M. & Woolard, I. (2014). Winners and losers: South African labour-
market dynamics between 2008 and 2010. Development Southern Africa, 31(1), 65-
84.
Coleman, M. S. (1999). Labour force participation. In J. Petersen & M. Lewis (Eds.), The
Elgar companion to Feminist Economics (pp. 500-504). Chelthenham: Edward Elgar.
Colman, R. (1998). The economic value of unpaid housework and child-care in Nova Scotia.
GPI Atlantic Technical Briefing. Retrieved from
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&
78
uact=8&ved=0ahUKEwigm_GL2d3KAhUKWRoKHQYKB98QFggkMAE&url=http
%3A%2F%2Fwww.gpiatlantic.org%2Fpublications%2Fsummaries%2Fhouseworksu
mm.pdf&usg=AFQjCNG0bcrebHkV00Dlk-
sGyyqfpNLIBQ&sig2=m3FeXDDHtXVUMjrl8sw8aA.
Cramer, J. S. (1991). The Logit Model. New York: Routledge.
De Villiers, L., Brown, M., Woolard, I., Daniels, R. & Leibbrandt, M., eds. (2014). National
Income Dynamics Study Wave 2 user manual. Cape Town: Southern Africa Labour
and Development Research Unit.
De Villiers, L., Brown, M., Woolard, I., Daniels, R. & Leibbrandt, M., eds. (2013). National
Income Dynamics Study Wave 3 User Manual. Cape Town: Southern Africa Labour
and Development Research Unit.
Delaunay, C. (2010). Gender differentiation and new trends concerning the division of
household labour within couples: The case of emergency physicians. Journal of
Comparative Research in Anthropology and Sociology, 1(1), 33-56.
Dinkelman, T. & Pirouz, F. (2011). Unemployment and labour force participation in South
Africa: A focus on supply-side (Working Paper No. 28). Retrieved from Economic
Research Southern Africa website:
http://www.econrsa.org/system/files/publications/working_papers_interest/wp28_inte
rest.pdf.
Duflo, E. (2012). Women empowerment and economic development. Journal of Economic
Literature, 50(4), 1051-1079.
Escudero, V. & Mourelo, E. L. (2013). Understanding the drivers of the youth labour market
in Kenya (ILO Research Paper No. 8). Retrieved from
http://www.ilo.org/wcmsp5/groups/public/---dgreports/---
inst/documents/publication/wcms_222527.pdf.
Fernandes, R. & de Felício, F. (2005). The entry of the wife into the labor force in response
to the husband’s unemployment. A study of the added worker effect in Brazilian
metropolitan areas. Economic Development and Cultural Change, 56(4), 887-911.
79
Floro, M. S. & Komatsu, H. (2011). Labour force participation, gender and work in South
Africa: What can time use data reveal? Feminist Economics, 17(4), 33-66.
Folbre, N. (1994). Who pays for the kids? Gender and the structures of constraint. London:
Routledge.
Hsaio, C. (2014). Analysis of Panel Data (3rd Ed.). [Online]. Retrieved from http://0-
dx.doi.org.oasis.unisa.ac.za/10.1017/CBO9781139839327
ILC. (2011). Retirement Age. An ILC Global Alliance Circular. Retrieved from
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahU
KEwj-
4vP6stbKAhWH6RQKHRn2ACIQFgggMAA&url=http%3A%2F%2Fwww.ilc-
alliance.org%2Fimages%2Fuploads%2Fpublication-
pdfs%2FCircular_state_pension_age_2_.pdf&usg=AFQjCNGup01-
z0xkjwOV8ZB9ery4oBx9UA&sig2=zDvM_dYWlQHoQSncN-tCHw&cad=rja.
Jacobsen, J. P. (1999). Human capital theory. In J. Petersen & M. Lewis (Eds.), The Elgar
companion to Feminist Economics (pp. 443-448). Chelthenham: Edward Elgar.
Kabeer, N. (2012). Women’s empowerment and inclusive growth: Labour markets and
enterprise development (Working Paper No. 1). Retrieved from International
Development Research Centre website:
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&
uact=8&ved=0ahUKEwit0YGxtvzKAhVGpnIKHaGTBasQFggdMAA&url=http%3A
%2F%2Fwww.idrc.ca%2FEN%2FDocuments%2FNK-WEE-Concept-
Paper.pdf&usg=AFQjCNEmK-
90ECJMEwMCqMBA5_1AC5Q8XA&sig2=0Wfb61THUvpnWRzUrpEG7A.
King. M. C. (1999). Labour market segmentation. In J. Petersen & M. Lewis (Eds.), The
Elgar companion to Feminist Economics (pp. 505-511). Chelthenham: Edward Elgar.
Klasen, S. & Lamanna, F. (2009). The impact of gender inequality in education and
employment on economic growth: New evidence for a panel of countries. Feminist
Economics, 15(3), 91-132.
Knight, G. & Kingdon, J. (2000). The incidence of unemployment in South Africa. TIPS
conference 2000 Annual Forum. Retrieved from http://www.tips.org.za/files/384.pdf.
80
Lam, D., Leibbrandt, M. & Mlatsheni, C. (2008). Education and youth unemployment in
South Africa (Working Paper No. 22). Retrieved from SALDRU webiste
http://hdl.handle.net/11090/33.
Leibbrandt, M., Woolard, I., McEwan, H. & Koep, C. (2010). Employment and inequality
outcomes in South Africa: What role for labour market and social policies. Southern
African Labour and Development Research Unit. Retrieved from
http://www.oecd.org/employment/emp/45282868.pdf.
Lewis, G. (2001). Black women’s employment and the British economy. In K. Bhavnani
(Ed.), Feminism & ‘Race’ (pp. 297-318). New York, NY: Oxford University Press.
Lim, L. L. (2002). Female labour force participation. Background paper for the United
Nations Population Division, Expert Group Meeting on completing the fertility
transition, March. Retrieved from
http://www.un.org/esa/population/publications/completingfertility/RevisedLIMpaper.
PDF.
Lips, H. (2013). The gender pay gap: Challenging the rationalizations. Perceived equity,
discrimination, and the limits of human capital models. Sex Roles, 68, 469-185.
Maja, B & Nakanyane, S. (2006). Women in the South African labour market 1995-2005
(Department of Labour). Retrieved from
http://www.labour.gov.za/DOL/downloads/documents/useful-documents/labour-
market-research-and-statistics/Labour%20Market%20Research%20-
%20Women%20in%20the%20South%20African%20Labour%20Market%201995%2
0-%202005.pdf.
Matthaei, J. (1999). Race. In J. Petersen & M. Lewis (Eds.), The Elgar Companion to
Feminist Economics (pp. 653-657). Chelthenham: Edward Elgar.
Mehtabul. A, Aimee, C. & Nishith, P. (2013). The returns to English-language skills in India.
Economic Development and Cultural Change, 61, 335-367.
Mjoli-Mncube, N. (1998). Gender neutrality or gender equality: Access to worker housing in
South Africa. In A. Larsson, M. Mapetla & A. Schlyter (Eds.), Changing gender
relations in Southern Africa: Issues of urban life (pp. 206-225). Lesotho: The Institute
for Southern African Studies
81
Mlatsheni, C. & Rospabé, S. (2002). Why is youth unemployment so high and unequally
spread in South Africa? (Working Paper No. 65). Retrieved from Development Policy
Research Unit website:
https://open.uct.ac.za/bitstream/handle/11427/7202/DPRU_WP02-
065.pdf?sequence=1.
Naidoo, V. & Kongolo, M. (2004). Has affrimative action reached South African women?
Journal of International Women’s Studies, 6(1), 124-136.
Naudé, W. & Serumaga-Zake, P. (2001). An analysis of the determinants of labour force
participation and unemployment in South Africa’s North-West province.
Development Southern Africa, 18(3), 261-278.
NIDS. (2014). Inclusion of Census 2011 geographic variables in NIDS (household level).
Retrieved from http://www.nids.uct.ac.za/documents/docs-and-files/211-inclusion-of-
census-2011-geographic-variables-in-nids.
Ntuli, M. & Wittenberg, M. (2013). Determinants of Black women’s labour force
participation in post-Apartheid South Africa. Journal of African Economies, 22(3),
347-374.
Ntuli, M. (2007). Determinants of South African women’s labour force participation, 1995-
2004. Institute of Labour Studies. Retrieved from http://ftp.iza.org/dp3119.pdf.
O’Higgins, N. (2001). Youth unemployment and employment policy: A global perspective.
Geneva: International Labour Organization.
Posel, D., Casale, D. & Vermaak, C. (2014). Job search and the measurement of
unemployment in South Africa. South African Journal of Economics, 82(1), 66-80.
Posel, D., Fairburn, J. A. & Lund, F. (2006). Labour migration and households: A
reconsideration of the effects of the social pension on labour supply in South Africa.
Econometric Modelling, 23, 836-853.
Ranchhod, V. (2010). Labour force participation and employment in South Africa: Evidence
from Wave 1 of the National Income Dynamics Study. Journal for Studies in
Economics and Econometrics, 34(3), 111-127.
82
Seguino, S. (1997). Gender wage inequality and export-led growth in South Korea. The
Journal of Development Studies, 34(2), 102-132.
Seguino, S. (2000). Gender inequality and economic growth: A cross-country analysis. World
Development, 28(7), 1211-1230.
Serumaga-Zake, P. A. E. & Kotze, D. (2004). Determinants of labour force participation of
married women in South Africa. Journal for Studies in Economics and Econometrics,
28(3), 99-111.
Siphambe, H. & Motswapong, M. (2010). Female participation in the labour market of
Botswana: Results from the 2005/06 labour force survey data. Botswana Journal of
Economics, 7, 65-78.
Smith, J. P. & Ward, M. P. (1985). Time-series growth in the female labour force. Journal of
Labor Economics, 3(1), S59-S90.
So, Y. & Kuhfeld, W. F. (1995). Multinomial logit models. In SUGI 20 Conference
Proceedings, Cary, NC: SAS Institute Inc. Retrieved from
http://www.sascommunity.org/sugi/SUGI95/Sugi-95-206%20So%20Kuhfeld.pdf.
ILO (http://www.ilo.org).
Thomas, A. (2002). Employment equity in South Africa: Lessons from the global school.
International Journal of Manpower, 23(3), 237-255.
Tsani, S., Paroussos, L., Fragiadakis, C., Charalambidis, I. & Capros, P. (2012). Female
labour force participation and economic development in Southern Mediterranean
countries: What scenarios for 2030? MEDPRO Technical Report No. 19. Retrieved
from
http://aei.pitt.edu/59153/1/No_19_Tsani_et_al_Female_Labour_Force_Participation.p
df.
Van Der Merwe, A. (2010). Does human capital theory explain the value of higher
education? A South African case study. American Journal of Business Education,
3(1), 107-118.
Van Der Westhuizen, C., Goga, S. & Oosthuizen, M. (2007). Women in the South African
labour market 1995-2005 (Working Paper No. 06). Retrieved from Development
83
Policy Research Unit website:
http://www.dpru.uct.ac.za/sites/default/files/image_tool/images/36/DPRU%20WP07-
118.pdf.
Walker, R. (2003). Reservation wages – Measurement and determinants: Evidence from the
Khayelitsha/Mitchell’s Plain (KMP) survey (Working Paper No. 38). Retrieved from
CSSR website: http://opensaldru.uct.ac.za/bitstream/handle/11090/632/csssr-saldru-
wp38.pdf?sequence=1.
Wittenberg, M. (2002). Job search in South Africa: A nonparametric analysis. South African
Journal of Economics, 70(8), 1163-1196.
Woolridge, J. M. (2010). Econometric Analysis of cross section and panel data (2nd ed.).
London: MIT press.
Yu, D. (2013). Youth unemployment in South Africa revisited. Development Southern Africa,
3(4-5), 545-563.
Data
Statistics South Africa, Quarterly Labour Force Trends, 2008-2015, Pretoria.
Wave 1 Data: Southern Africa Labour and Development Research Unit. National Income
Dynamics Study 2008, Wave 1 [dataset]. Version 5.3. Cape Town: Southern Africa Labour
and Development Research Unit [producer], 2015. Cape Town: DataFirst [distributor], 2015.
Wave 2 Data: Southern Africa Labour and Development Research Unit. National Income
Dynamics Study 2010-2011, Wave 2 [dataset]. Version 2.3. Cape Town: Southern Africa
Labour and Development Research Unit [producer], 2015. Cape Town: DataFirst
[distributor], 2015.
Wave 3 Data: Southern Africa Labour and Development Research Unit. National Income
Dynamics Study 2012, Wave 2 [dataset]. Version 1.3. Cape Town: Southern Africa Labour
and Development Research Unit [producer], 2015. Cape Town: DataFirst [distributor], 2015.
