Page 1 ! " " " # " " $ % % # & ' " ( ) (* + $ , + - % & + & & ( ! " # Page 2

! " " " #

"

"

$ %

%

#& ' "

( )(* +

$ ,

+ - %

& +& &(

!"#

! " ! # $% "

&$ &'

! "

# $ !% ! & $ ' ' ( $ ' # % % ) %* %' $

' + " %

& ' ! #

$, ( $ - $$$ . /"/"#012" )*3- +) * !4!& 5!6%55787& % 2 9:;567!%2 ?9:;567!; 0 @! * A & B ()*+, - ./* 0+,-1 . ! #

&$ 2(3' C 2 ! "C ( % 2

% % % % $ $ $C C

CC - $$$ /"/"#012" )*3- +) * !4!& 5!6%55787& %) C 9:;567!%D?9:;567!; 0 @! E & " E + CA

B ()*+, - ./* 0+,-1 . -C F8>78C /"/"#012" )*3- +)

"

& 8>78

Preface Gender discrimination has consistently been a subject of national and international concern over the years particularly from the 1970s. In the labour market, gender discrimination takes the form of unequal treatment of women or men relative to the other in the hiring decision and payment of wages by employers based largely on their subjective assessment of him or her.

There are a number of research works on this subject particularly in the developed world but the same cannot be said of developing countries including Ghana and this forms the basis of the production of this book. It provides a comprehensive analysis of various dimensions of gender differences in the Ghanaian labour market and show whether these gender differences indicate gender discrimination. The book draws on secondary dataset from four national population censuses and three nationally representative Ghana Living Standards Surveys (GLSS3,4&5), between 1960 and 2006 for the analysis and makes some recommendations to address the phenomenon of gender discrimination in the Ghanaian labour market.

The book is the first comprehensive piece on the subject of gender inequality related to employment and wages in Ghana and will be useful to economists and those interested in labour and gender issues. It will particularly provide a basis for gender activists, academics and policy makers to discuss gender issues related particularly to the labour market from a more informed perspective. It will also be useful for students studying gender and most especially those offering labour economics at the undergraduate and postgraduate level.

ŝ

This piece of work is an extract from my PK' 7KHVLV HQWLWOHG ³*HQGHU 3HUVSHFWLYH RI /DERXU 0DUNHW 'LVFULPLQDWLRQ LQ *KDQD´ DQG KDV LQGHHG benefited from useful comments from a team of supervisors comprising Professor Amoah Baah-Nuakoh, Professor Kwabia Boateng and Professor Ernest Aryeetey who deserve my heartfelt gratitude. I am also grateful to Professor Emmanuel Akyeampong and Dr. Margot Gill of Harvard University and Professor Kwadwo Asenso-Okyere whose initiative created an opportunity for me to pursue a coursework at Department of Economics, Harvard University in 2004-05 as part of my PhD programme. My sincere gratitude also goes to Professor Caroline Hoxby, formerly at the Department of Economics, Harvard University, Professor Lawrence Katz, and Professor Richard Freeman of the Department of Economics, Harvard University for granting me access to their labour economics class.

I am also grateful to my wife Afia Konadu and the children Kwaku Gyamfi and Akosua Akyiaa for their support, prayers and sacrifices. In all, the Almighty God who granted me the knowledge and strength to put these thoughts together deserves my highest honour and gratitude.

ŝŝ

Table of Contents Preface......................................................................................................... i Table of Contents ......................................................................................iii List of Tables...........................................................................................viii List of Figures ............................................................................................ x List of Boxes .............................................................................................xi List of Acronyms and Abbreviations.......................................................xii Chapter 1: Introduction ................................................................................. 1 1.1

Background....................................................................................... 1

1.2

The Concept of Gender and Sex....................................................... 5

1.3

Why Gender Discrimination an Economic Issue ............................. 6

1.4

Women as Discriminated Group ...................................................... 8

1.5

Objectives ....................................................................................... 12

1.6

Methodological Approach and Data Sources ................................. 13

1.7

Motivation....................................................................................... 14

1.8

Structure of the Book...................................................................... 15

Chapter 2: Overview of Gender Differences of the Ghanaian Labour Market ......................................................................................................... 17 2.0

Introduction..................................................................................... 17

2.1

A Brief Account of the Ghanaian Economy .................................. 17

2.2

Gender Dimension of the Labour Market in Ghana....................... 19

2.2.1

Labour Force Participation Rate............................................. 20

2.2.2

Employment, Unemployment & Underemployment................. 23

2.2.3

Working Children ..................................................................... 27

2.3

Gender Differences in Employment Classifications ...................... 29 ŝŝŝ

2.3.1

Occupation ............................................................................... 29

2.3.2

Employment Status ................................................................... 35

2.3.3

Employment Sector................................................................... 38

2.3.4

Industry or Economic Sectors .................................................. 40

2.4

Gender Differences in Employment Quality.................................. 42

2.5

Wage Differences by Gender ......................................................... 44

2.6

Gender Differences in Education and Training............................. 46

Chapter 3: Policy, Legal and Institutional Framework of Gender Issues ... 51 3.0

Introduction..................................................................................... 51

3.1

Policy Initiatives, and Gender Effects ............................................ 52

3.2

Legal Framework for Gender Equality........................................... 57

3.3

Institutional

Arrangements

and

Women

Economic

Empowerment..... ..................................................................................... 60 Chapter 4: Labour Market Discrimination by Gender: Issues and Types .. 63 4.0

Introduction..................................................................................... 63

4.1

Discrimination: Definition and Concept ........................................ 63

4.2

Pre-Market Discrimination ............................................................. 65

4.3

Market Discrimination.................................................................... 69

4.3.1

Occupational segregation by Sex............................................. 70

4.3.2

Wage Discrimination................................................................ 74

4.4

Post-Market Discrimination ........................................................... 77

Chapter 5: Theories of Segregation and Discrimination in the Labour Market ......................................................................................................... 78 5.0

Introduction..................................................................................... 78

5.1

Economic Models of Discrimination.............................................. 78 ŝǀ

5.1.1

Human Capital Model .............................................................. 78

5.1.2

Taste Hypothesis of Discrimination ......................................... 83

5.1.3

Monopsony Framework............................................................ 88

5.1.4

Statistical Theory of Discrimination ........................................ 90

5.2

Institutional Theories of Discrimination in the Labour Market ..... 93

5.2.1

Labour Market Segmentation Theory ...................................... 95

5.2.2

Crowding Hypothesis ............................................................... 97

5.3

Non-economic Models of Discrimination ...................................... 99

5.3.1

Patriarchy Argument.............................................................. 100

5.3.2

Feminist Model...................................................................... 101

5.4

Relevance of the Theories in Ghanaian Context .......................... 103

Chapter 6: Measuring Sex Segregation of Occupation in Ghana ............. 108 6.0

Introduction................................................................................... 108

6.1

Sex Segregation of Occupation .................................................... 108

6.2

Segregation Measures................................................................... 109

6.2.1

Duncan Index of Dissimilarity (ID) ...................................... 110

6.2.2

Marginal Matching Index (MM) ............................................ 112

6.2.3

Karmel Maclachlan Index (KM) ............................................ 114

6.2.4

Size-Standardized Index of Dissimilarity (Ds) ....................... 116

6.3

Computational Issues and Sources of Data .................................. 118

6.4

Analysis of Empirical Results of Segregation.............................. 120

6.4.1

Occupational Segregation of Total Employment ................... 121

6.4.2

Occupational Segregation of Paid-Employment and Self

Employment......................................................................................... 124 6.5

Accounting for Changes in Segregation....................................... 126 ǀ

6.5.1

Index of Decomposition Effects............................................. 127

6.5.2

Analysis of Empirical Results of Decomposition Effects ....... 128

6.5.3

Sex Composition and Occupational Mix effects of Changing

Occupational Distribution .................................................................. 130 Chapter 7: Wage Differentials between Men and Women in Ghana ....... 135 7.0

Introduction................................................................................... 135

7.1

Model Formulation ....................................................................... 135

7.2

Data Source and Descriptive Statistics......................................... 139

7.3

Estimation Procedure.................................................................... 143

7.4

Analysis of Empirical Results ...................................................... 144

7.4.1

Gender Wage Differentials overall and by Employment

Type..................................................................................................... 145 7.4.2

Determinants of Female and Male Wages ............................. 148

7.4.3

Gender Wage Differentials within Occupations .................... 151

7.4.4

Other Sources and Forms of Gender Wage Differentials...... 154

Chapter 8: Measuring Gender Wage Discrimination in the Labour Market in Ghana .................................................................................................... 160 8.0

Introduction................................................................................... 160

8.1

Wage Decomposition ................................................................... 160

8.2

Model Specification...................................................................... 161

8.2.1

Blinder-Oaxaca Decomposition Technique ........................... 162

8.2.2

Neumark and Oaxaca-Ransom Decomposition Approach .... 164

8.3

Estimation Procedure and Data Source ........................................ 167

8.4

Empirical Analysis of Gender Wage Decomposition .................. 169

Chapter 9: Summary, Conclusion and Policy Implications ...................... 178 ǀŝ

9.1

Summary of Empirical Findings .................................................. 178

9.2

Conclusion and Policy Recommendations ................................... 182

Bibliography........................................................................................... 189 Appendices............................................................................................. 210

ǀŝŝ

List of Tables Table 1.1: Sex Ratio of Population, Distribution of Labour Force and Employment and Labour Force Participation Rate, 1960±2000..............9 Table 2.1: Basic Indicators of Gender Differences in the Ghanaian Labour Market 1960±2006.................................................................................22 Table 2.2: Sex Distribution of Occupation, 1960±2006 (%).......................30 Table 2.3: Sex Distribution of Employment Status, 1970-2006 (%)..........36 Table 2.4: Distribution of Employment Sector by Sex, 1984 ± 2006.........39 Table 2.5: Sex Distribution of Industry of Employed Persons (15 yrs. and older), 1960±2006 (%)...........................................................................41 Table 2.6: Proportion of Paid Employees of 15+ years with Written Conditions of Service.............................................................................43 Table 2.7: Educational Attainment of Employed Persons (%)...................48 Table 2.8: Sex Distribution of Apprentices by Main Trade........................49 Table 6.1: Modified segregation table for calculating MM Index............114 Table 6.2: Sex Segregation of Occupation of Total Employment............122 Table 6.3: Sex Segregation of Occupation (1-digit classification) by Employment Status..............................................................................125 Table 6.4: Source of Over time Changes in Segregation Index, 19602006......................................................................................................129 Table 6.5: Changes in Female composition (Fi/Ti) of occupation and Occupational Distribution of Workforce (Ti/T) ..................................132 Table 7.1a: Means and Standard Deviations of Key Variables in the Wage Model...................................................................................................141 Table 7.1b: Maximum and Minimum Values of Key Variables in the Wage Model...................................................................................................142 Table 7.2: Results of Wage Regression Model (Equation 7.1) with Sample Selection: Heckman Selection Model (two-step estimates±Dependent Variable: log of monthly wage.............................................................147 ǀŝŝŝ

Table 7.3: Results of Wage Regression Model (Equation 7.1) within Occupation with Heckman Selection Model (two-step Estimates) ± Dependent Variable: log of monthly wages.........................................152 Table 7.4: Wage Regression Results with Heckman (two-step) Sample Selection: Dependent Variable: log monthly wages............................156 Table 8.1: Results of Blinder±Oaxaca Decomposition of Gender Wage Gap.......................................................................................................171 Table 8.2: Results of Nuemark-Oaxaca-Ransom Decomposition of Gender Wage

Gap

(based

on

non-discriminatory

wage

structure)...............................................................................................174

ŝǆ

List of Figures Figure 2.1: Average Annual Population Growth by Sex 1960-2000 (%).........................................................................................................19 Figure 2.2: Average Annual Growth Rate of Labour Force by Sex, 1960 ± 2000 (%).................................................................................................21 Figure 2.3: Annual Average growth of Employment, 1960-2000 (%)..........................................................................................................24 Figure 2.4: Annual Average Growth of Unemployment by Sex, 1960-2000 (%)..........................................................................................................27 Figure 2.5: Female-Male Representation by Occupations, 1960 ± 2000 (%)..........................................................................................................33 Figure 2.6: Female-Male Representation by Occupations, 1991±2006 (%)..........................................................................................................34 Figure 2.7: Female-Male Average Wage Ratio by Occupation, 19912006........................................................................................................46

ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ǆ

List of Boxes Box 1: Relative :DJHVLQ'LVFULPLQDWRU\/DERXU0DUNHW«««...«5 Box 2: Relative Profit of Discriminatory and Non-Discriminatory firms«««««««««««««««««««««««««««7 Box 3: Algebraic Analysis of Relative wages in the Discriminatory Monopsonistic Labour Market....................................................................89

ȱ ȱ ȱ ȱ

ǆŝ

List of Acronyms and Abbreviations CEDAW

Convention on the Elimination of all forms of Discrimination against Women

CSAE

Centre for the Study of African Economies

Ds

Size Standardised Index of Dissimilarity

DPs

Development Partners

DWM

31st 'HFHPEHU:RPHQ¶V0RYHPHQW

ENOWID

Enhancing Opportunities for Women in Development

GLSS

Ghana Living Standards Survey

GMES

Ghana Manufacturing Enterprise Survey

GPI

Gender Parity Index

GPRS I

Ghana Poverty Reduction Strategy

GPRS II

Growth and Poverty Reduction Strategy

GSS

Ghana Statistical Service

ID

Index of Dissimilarity

ILO

International Labour Organisation

ISCO

International Standard Classification of Occupation

JHS

Junior High School

KILM

Key Indicators of the Labour Market

KM

Kamel Maclachlan Index

LFPR

Labour Force Participation Rate

LIFO

Last In First Out

MDG

Millennium Development Goals

MLE

Maximum Likelihood Estimation

MM

Marginal Matching Index

NCWD

National Council on Women and Development

NDMW

National Daily Minimum Wage

NTC

National Tripartite Committee

OECD

Organisation

of

Development ǆŝŝ

Economic

Cooperation

and

OLS

Ordinary Least Squares

PAMSCAD

Programme of Action to Mitigate the Social Cost of Adjustment

PNDC

Provisional National Defence Council

RPED

Regional Programme on Enterprise Development

SAP

Structural Adjustment Programme

SSA

Sub-Saharan Africa

SSNIT

Social Security and National Insurance Trust

UN

United Nations

UNESCO

United Nations Educational Scientific and Cultural Organisation

UNFPA

United Nations Population Fund

US

United States of America

WID

Women in Development

ǆŝŝŝ

Chapter 1

Introduction 1.1

Background

There are a number of domestic and international laws and conventions that essentially frown on all forms of discrimination including wage and employment discrimination. For instance, Article 17(2) of the 1992 Constitution of the Republic of Ghana clearly prohibits discrimination of all forms on the grounds of race, gender, colour, ethnic origin, religion, creed or social or economic status. Similarly, in the *KDQD¶V/DERXU$FWRI 2003 (Act 651), discrimination against a person with respect to the employment or conditions of employment because the person is a member or an officer of a trade union is judged as unfair labour practice. The two ILO core conventions of equal remuneration for work of equal value (Convention No. 100, 1951) and employment and occupational discrimination (Convention No. 111, 1958)1 are a reflection of international concern on the issue of discrimination.

The perceived marginalisation of some groups of people all over the world has been found to be a contributory factor to religious, racial and ethnic tensions and conflicts among different groups. Generally, discrimination is said to exist when some superficial characteristics (e.g. skin pigmentation, religion, biological makeup etc) are used in an attempt to restrict LQGLYLGXDO¶V DFFHVV WR WKH DYDLODEOH HFRQRPLF SROLWLFDO DQG VRFLDO RSSRUWXQLWLHVIRUDGYDQFHPHQW'¶$PLFR 7KHFKDUDFWHULVWLFVEHLQJ 1 Convention 1958 (No. 111) regards as violation of rights discrimination on the basis of race, colour, sex, religion, political opinion, national extraction or social origin ϭ

uVHG IRU GLVFULPLQDWRU\ SXUSRVH DUH ODUJHO\ XQUHODWHG WR WKH LQGLYLGXDO¶V actual or potential talent, skills and drive.

On economic grounds, discrimination has generally been analysed in relation to employment

and wage differentials

among different

demographic groups (e.g. gender, race or colour, ethnicity, religion etc) on the basis of factors other than their productive characteristics. Essentially, discrimination in the labour market is perpetuated when society fails to recognise the consistent segmentation of occupations and distribution of tangible and intangible compensation on the basis of these superficial characteristics of individuals regardless of their productive characteristics.

Technically, there are two types of labour market discrimination, and these are occupational segregation and wage discrimination. Occupational segregation which measures the unequal distribution of occupation may constitute discrimination if it results from such individual characteristics as sex, race, ethnicity, religion and so on that are clearly unrelated to the LQGLYLGXDO¶V SURGXFWLYH FKDUDFWHULVWLFV 7KLV LPSOLHV WKDW RFFXSDWLRQDO segregation does not necessarily suggest discrimination if it is prompted by differences in skills or preferences of job seekers. Similarly, wage differentials between two groups of people indicate discrimination if the observed

differentials

are

caused

by

differences

in

superficial

characteristics rather than productive characteristics such as education, experience and effort among others.

One form of discrimination in the labour market that has caught the attention of the international community, civil society organisations, and many governments across the globe which is the main focus of this book Ϯ

has been gender discrimination2. Although, several communities worldwide experience one form of discrimination or another, the effects of gender discrimination which often takes the form of discrimination against women are profoundly observed in low-income economies. Indeed, gender differences in the Ghanaian labour market have been a subject of debate over the past two decades and beyond. Women advocates and gender activists have been at the forefront of the fight against the perceived gender inequality in decision making particularly against women in Ghana (see for example Tsikata, 2001, Allah-Mensah, 2005).

Unequal access to education and for that matter jobs have been cited as the major causes of gender wage disparity in Ghana (see Beaudry and Sowa, 1994; Verner, 1999; Schultz, 2003). Women representation in Parliament, Judiciary and the Executive has often been used as evidence of gender differences in Ghana. In the *KDQD¶V )RXUWK 5HSXEOLFDQ 3DUOLDPHQW WKH proportion of seats held by women dropped from 10 percent in the fourth Parliament (2005-2008) to 8.3 percent in the fifth Parliament (2009-2012). In the Judiciary, women accounted for 29 percent of Supreme Court judges and 25 percent of High Court Judges in 2008 (NDPC, 2010). Similarly, only 24 percent of Chief Directors of government ministries were women while 28 percent and 11 percent of District Assembly appointees and elected Assembly members respectively were women in 2008. The composition of women in the Executive arm of government and other public institutions follow similar pattern.

The growing recognition of the issue of gender discrimination globally and the need to reduce its incidence in recent times is reflected in a number of 2 6HHIRUH[DPSOH81)3$ ³6WDWHRI:RUOG3RSXODWLRQ´81:RUOG6XPPLWIRU6RFLDO Development 1995 ϯ

recommendations and conventions adopted at various international gatherings. For example, the 1995 World Summit on Social Development at Copenhagen committed countries to ensuring among others, balance and equity in decision making processes at all levels3. The Beijing Conference on Women in 1995 also recommended a number of actions including elimination of occupational segregation and all forms of employment GLVFULPLQDWLRQ HQVXUH ZRPHQ¶V HTXDO DFFHVV WR DQG IXOO SDUWLFLSDWLRQ LQ power structures and decision-making; and guarantee the rights of women and men to equal pay for work of equal value. In November 2008, the United

Nation

Economic

Commission

for

Africa

(UNECA)

in

collaboration with the African Union Commission (AUC) and the African Development Bank (AfDB) organised a pan-African conference on the WKHPH³(TXDOLW\IRU:RPHQLQ$IULFD´DQG came up with a Plan of Action to accelerate progress towards gender equality on the continent. The promotion of equality of opportunity and treatment for men and women workers has also been amplified by a number of recommendations such as Recommendation No. 165 of the ILO4.

7KHQHHGWRSURPRWHJHQGHUHTXDOLW\DQGZRPHQ¶VHPSRZHUPHQWWKURXJK elimination of gender disparity in primary schooling and improvement in the share of women in wage employment is clearly stated among the eight goals of the MDGs. In Ghana, the establishment of the National Council on Women and Development (NCWD) in the 1970s and the recent allocation RI RQH PLQLVWU\ UHVSRQVLEOH IRU ZRPHQ DQG FKLOGUHQ¶V DIIDLUV LV an acknowledgement of the problem of gender discrimination locally and the need to address it. Hence, there is a need for more systematic assessment of the extent of gender discrimination particularly from an economic 3 See Commitment 5B of United Nations Report of the Fourth World Conference on Women in Beijing, 4-15 September, 1995 4 www.ilo.org ϰ

perspective to provide an informed basis for development policy choices and enforcement.

1.2

The Concept of Gender and Sex

World Bank (2001) defines gender as socially constructed roles and the socially learned behaviours and expectations associated with females and males. On their part, Reeves and Baden (2000) refer to gender as how a SHUVRQ¶VELRORJ\ LV culturally valued and interpreted into locally accepted ideas of what it is to be a woman or man. Generally, the concept of gender may be thought of as the social attributes, roles and opportunities of male and females that are socially and culturally determined and acquired through the socialisation process. These social attributes and roles which are subject to change over time often result in unequal treatment and assignment of responsibilities between males and females with the implication of placing women and girls at a disadvantage compared to men and boys. 7KH FRQFHSWRI JHQGHU GLIIHUV IURP WKH FRQFHSW RI µVH[¶ RQ WKH EDVLV WKDW while the latter describes the differences in biological characteristics determined at conception, the former is a social construct which can be altered over time by education, advocacy or public policy among other measures. According to Watts (1998) there is a large overlap in the abilities of women and men based on biologically determined differences. For instance, while on average, men tend to have greater physical strength than women, some women are observed to be stronger than men. Similarly, while on average, women may seem to be more compassionate and loving towards children than men, there may indeed be some men who have more compassion and love for children than some women. This suggests the need to consider men and women as individuals rather than on the basis of ϱ

biologically determined average differences. In effect, there is a very limited basis to absolutely endorse the exclusion of women from occupations and jobs that require physical strength and also ascribe childbearing responsibility exclusively to women thereby laying the foundation for gender discrimination in the labour market.

1.3

Why Gender Discrimination an Economic Issue

The fundamental issue and concern about the existence and the extent of gender discrimination in the labour market is motivated by concerns for efficiency loss and for equity. Generally, group differences impede efficiency when resources are consistently allocated at least in part on the basis of sex without full regard for the productive characteristics of the LQGLYLGXDO&OHDUO\WKHUHLVWKHWHQGHQF\IRUVRFLHW\¶VDJJUHJDWHUHDORXWSXW to fall below the potential level if resources are distributed to members in a manner inconsistent with their productive capabilities. This is because there LVDWHQGHQF\IRUVRPHPRUHSURGXFWLYHPHPEHUVLQWKH³GLVFULPLQDWHG´RU less favoured group (females in this regard) to be under-allocated some scarce resources while some relatively unproductive members of the ³IDYRXUHG´JURXSZRXOGEHRYHU-allocated with scarce productive resources '¶$PLFR 7KH VKRUWIDOO LQ RXWSXW DFFRUGLQJ WR '¶$PLFR ZRXOG depend on the intensity of the discriminatory feelings, the ease with which members of the target group can be identified, and the size of the target population.

Anker (1998) corroborated the efficiency concern associated with labour market discrimination with the view that occupational segregation based on the sex of workers has a negative effect on labour market efficiency and functioning. This is because when most women are effectively excluded from most occupations, human resources are wasted with a potentially income declining effect. Thus, many of the best suited and most skilled ϲ

people are excluded from working in the occupation where they would be most productive.

Gender disparity is also bound to undermine social equity and increase inequality over time. As Loury (2003) sought to imply, the tendency to see racial disparities (and gender disparity in this instance) as communal rather than a societal problem contributes to inequality over time. This is because in the absence of any intervention, the low social conditions of women would persist and the negative social meanings ascribed to them are reinforced. Labour market discrimination that takes the form of segregation of occupation by sex has a negative effect on how men see women and even how women see themselves by reinforcing and perpetuating gender stereotypes (Anker, 1998). This tends to adversely affect the social status and economic empowerment of women. The effect is the perpetuation of many social problems including mortality and morbidity, poverty and inequality.

Essentially, the failure of society to address the issue of gender disparity becomes a recipe for deepened social inequality. An estimate of almost one-half of households headed by women in some parts of Africa by Buvinic (1995) give credence to the fact that any discriminatory tendencies that adversely affect their incomes could perpetuate poverty and inequality in these countries. The low pay and incomes for women workers that accompany occupational segregation are becoming increasingly important contributors to poverty and inequality in society as a whole (Anker, 1995).

Obviously, segregation of workers into different occupations on the basis of sex rather than relevant productive characteristics does not only have negative effect on education and training of future generations, but also contributes to gender wage differentials. Invariably, the decision on how ϳ

much education to provide girls and boys including the fields of study is often informed by labour market opportunities. This implies that the limited labour market opportunities for women and lower pay for female RFFXSDWLRQVFRQWULEXWHWRWKHSHUSHWXDWLRQRIZRPHQ¶VLQIHULRU SRVLWLRQLQ society.

One key concern about sex segregation of occupation is that it can also be a major source of labour market rigidity with its adverse effect on the labour PDUNHW¶V DELOLW\ WR UHVSRQG WR FKDQJHV LQ SROLFLHV $QNHU ,Q D highly sex segregated labour market, loss of jobs or retrenchment of mainly members of one demographic group (e.g. females) from one sector may find it difficult to fill job openings in sectors where members of another demographic group (males in this case) are preferred.

Additionally, occupational segregation that tends to keep many women out of wage employment altogether has an undesirable effect of raising fertility rates all things being equal (United Nations, 1985). Indeed, female wage employment, particularly in the formal sector, helps reduce fertility rates in developing countries. Anecdotal evidence shows that children of well educated women tend to do well in society compared with families where PRWKHUV¶ HGXFDWLRQ LV ORZ &OHDUO\ UHGXFLQJ JHQGHU LQHTXDOLW\ GRHV QRW only promise significant returns and economically empower women by improving their living conditions, but it also enables them to actively participate in the social and economic life of a country which may well be key for long-term sustainable development.

1.4

Women as Discriminated Group

Economic discrimination against women in the work environment generally takes the form of segregation of women into less prestigious and lowly rewarded jobs. This consequently tends to place them below their ϴ

male counterparts in terms of wages and social status. Gender segregation of occupation may result from unequal investment in education of children in favour of boys and cultural and religious practices that bar women from engaging in certain jobs.

In many countries across the world, the performance of women in the labour market is often perceived to be inferior to that of men. Generally, labour force participation rate of females remains lower than that of males. In Ghana, the labour force participation rate of females has often trended below that of men even though females constitute over half of the entire population (table 1.1). In addition, unemployment rate is estimated to be higher among women than men, whilst at the same time the share of females in wage employment is also lower than that of males (BaahBoateng and Turkson, 2005).

Table 1.1: Sex Ratio of Population, Distribution of Labour Force and Employment and Labour Force (LF) Participation Rate, 1960±2000 Year

Sex Ratio of Share of Females in LF Participation Rate Population* Lab. force Employment Crude Rate Activity Rate Male Female Male Female

1960 1970

1.04:1 0.98:1

38.41 38.74 49.3 32.1 44.19 45.17 43.8 34.2

89.0 83.5

56.7 63.6

1984 2000

0.97:1 0.98:1

51.12 51.37 49.70 49.53

83.5 76.7

81.7 72.6

44.9 44.6

45.9 43.1

*Sex Ratio is computed as Male±to±Female Population

Source: Computed by the Author from the Censuses In many countries including Ghana, women form a smaller proportion of the labour force, even though they account for a larger proportion of the population (Boateng, 2000). As evident in table 1, while the sex ratio of the ϵ

population in Ghana has trended below 99 percent (or 0.99) since 1970, the proportion of women in the labour force has largely been less than 50 percent. The only exception occurred in 1984 when the female share in the labour force reached 51 percent due partly to the economic recession that hit the country in the early 1980s which pushed many women into the labour market. This exception appears to confirm the outcome of the male chauvinist version of family utility model of labour supply, which indicates that participation of married women in the labour force to augment their huVEDQG¶VLQFRPHLQFUHDVHVGXULQJSHULRGVRIUHFHVVLRQ7KLVSKHQRPHQRQ LVUHIHUUHGWRDV³WKHDGGHGZRUNHUHIIHFW´VHH6SOHW]HU .

Women also receive lower pay than men throughout the world largely because women are generally found mostly in low paying and less prestigious jobs and are less well endowed with human capital compared with men. Differences in human capital endowment such as education and experience, between men and women which emanate from discrimination in the non-labour market largely account for gender pay differences within the same occupation (Gunderson, 1994). According to Bell-Lowther, (1985), Two-thirds of all the worked hours in the world are done by women but they receive less than one-WHQWK RI WKH ZRUOG¶V LQFRPH DQG RZQ OHVV thaQRQHKXQGUHGWKRIWKHZRUOG¶VSURSHUW\. This is a s a result of the nonUHZDUGLQJ QDWXUH RI PXFK RI ZRPHQ¶V ZRUN HIIRUW VXFK DV FKLOG EHDULQJ and nurturing and other domestic activities (Bell-Lowther, 1985).

Some empirical evidence in Ghana and elsewhere confirm the lower average wages of females than males. Anker (1998) estimated average female-male pay ratios in the world at roughly 75-80 percent based on an hourly reference period. In Ghana, Beaudry and Sowa (1994) found that ZRPHQ¶VPRQWKO\wages are 21 percent less than the average for men after controlling for individual productive and occupational characteristics. ϭϬ

Similarly, Verner (1999) revealed that female employees in Ghana are generally paid less than male employees and the negative wage premium received by females out-match their negative productivity. Schultz (2003) established a 23 percent gender pay gap in favour of men in Ghana with four-fifth of the gap attributable to differences in human capital endowment.

In terms of sex distribution of occupation across the world, women account for over half of clerical and sales jobs but less than one-quarter of managerial jobs which relatively pay better (see Psacharopoulos and Tzannatos 1989; and Schultz, 1990). Evidently, most workers in secretarial business, nursing and housekeeping are women while professional and technical, administrative and managerial, and production jobs which are highly remunerated are dominated by men. According to Boateng (2000), about 95 percent of secretaries in the civil service in Ghana are women compared with only 8.5 percent of administrative staff being women. In addition, women are more concentrated in unpaid agriculture than men in Ghana. Outside of agriculture, women are more concentrated in trading where one-quarter of all women work as against about 5 percent of men. Baah-Boateng (2007) found under representation of women in highly skilled and better rewarding occupations such as professional and technical, administrative and managerial, and clerical in 1998/99. In terms of employment type, women are more represented in self-employment than men but less represented in paid-employment.

Invariably, some cultural and religious practices, beliefs and childhood socialization tend to constrain women from accessing higher education which could facilitate their entry and access to paid employment in the labour market. Most often, women face various constraints related to social norms and values that govern the gender division of labour in production ϭϭ

and reproduction. It is not uncommon to observe young girls being persuaded to cut short their education midstream on the basis that money spent on girls would yield less return than that spent on boys. Ram (1982) reported that in West Africa, parents prefer to give their male children better education than female children with the belief that the money spent on girls would yield less return relative to that spent on boys.

Others also argue that higher education may cost them their marriage and child bearing beyond a certain age. These factors have largely contributed to the low educational attainment and low literacy rate of women relative to men in most developing countries. For example, Ghana recorded 49.8 percent and 42.5 percent adult literacy rate for women in 2000 and 2003 as against 66.4 percent and 66.2 percent for men (Ghana Statistical Service, 2002 and 2005b). Estimates from the fifth round of the Ghana Living Standards Survey (GLSSV) also indicates that only 8 percent of women in employment in Ghana as against 17 percent of men had at least secondary education in 2005/06. These developments tend to make it more difficult for women than men to escape poverty through paid work and higher incomes.

1.5

Objectives

The key objective of this piece of work is informed by the established link between human capital investment, which influences occupational choice and occupational segregation by gender, and gender wage differentials. Essentially, the broad objectives of the study are to empirically investigate the extent and sources of gender differences in occupations and wages in the Ghanaian labour market and ascertain whether these differences constitute gender discrimination.

ϭϮ

Specifically, I seek to assess the relevance of explaining gender discrimination in the Ghanaian labour market from economic perspective and measure the extent and changing trend of occupational segregation by gender in Ghana. An investigation into the extent and sources of gender wage disparity in the Ghanaian labour market and a measure of the share of gender wage gap emanating from sheer discrimination is also pursued. The book also attempts to ascertain the effect of female occupational composition on individual wage or earnings as means of testing the HPSLULFDOUHOHYDQFHRI³FURZGing hypothesis" of discrimination in Ghana. It also set out to establish whether female dominated occupations pay less than male dominated occupations; find out whether women engaged in male dominated occupations benefit from potentially higher wages assocLDWHGZLWK³PDOHRFFXSDWLRQV´; DQGZKHWKHUPHQZRUNLQJLQ³IHPDOH RFFXSDWLRQV´ DUH ³ZRUVH RII´ GXH WR ORZHU wages associated with female dominated occupations;

1.6

Methodological Approach and Data Sources

After a statistical overview of the Ghanaian labour market and theoretical and empirical review of the literature, the book adopts four major indices of occupational segregation to measure and assess the extent and changing pattern of gender segregation of occupation. The main data sources for this exercise are four population and housing censuses in 1960, 1970, 1984 and 2000 and the last three rounds of nationally representatives Ghana Living Standards Surveys (GLSS3, GLSS4 and GLSS5) of 1991/92, 1998/99 and 2005/06.

A quantitative estimation of wage differentials through the specification and estimation of different forms of Mincerian wage equations is also carried out. This is followed by the adoption of Blinder-Oaxaca and ϭϯ

Neumark-Oaxaca-Ransom gender decomposition techniques to estimate the proportion of gender wage gap that is attributable to discrimination.

1.7

Motivation

The absence of a comprehensive analysis of gender differences in the labour market and a measure of wage discrimination is the main source of motivation for the production of this book. Indeed, in spite of the local and international concerns about employment and wage discrimination along gender lines all over the world, a comprehensive study of the issue particularly from economic perspective in Ghana is yet to be carried out.

More often, studies on gender differences in the Ghanaian labour market tend to dwell on one aspect of the problem. For example, a study of wage discrimination by Barr and Oduro (2000) highlighted the relevance of premarket discrimination in the analysis of wage differentials between ethnic groups in the manufacturing sector. In an effort to establish a measure of occupational dissimilarity by gender, Boateng (2000) and Baah-Boateng (2007) showed the extent and changing trend of the degree of sex segregation in the labour in Ghana. However, they failed consider the extent to which occupational disaggregation affects the degree of occupational segregation.

The wage discrimination against women established by Boateng (1996) based on data from Social Security and National Insurance Trust (SSNIT) virtually focused on the formal sector while estimates of wage differentials by Beaudry and Sowa (1994) came out of a broader study of labour market in the era of adjustment (World Bank, 1994). The observed lower wages of females compared to males without regard to productivity by Verner (1999) were a partial outcome of an investigation into wage and productivity gap in Ghana. Wage differential for reproducible human ϭϰ

capital in Ghana and Ivory Coast was the focus of Schultz (2003) out of which a gender wage gap was established.

This piece of work differs substantially from other studies on gender differences in the Ghanaian labour market in many respects. It takes a more comprehensive view of the issue of gender discrimination in the labour market using a more current and previous nationally representative household survey dataset. It is supplemented by data from four different population censuses from 1960 to provide some empirical analysis of the issue from historical perspective. It also undertakes much more thorough review of the theoretical and empirical literature of labour market discrimination and assesses the relevance of economic and non-economic theories of discrimination to the Ghanaian labour market.

Besides, the establishment of female composition effect on wages, wage differentials between male dominated and female dominated occupations and gender wage differentials regardless of the type of occupation in Ghana is a significant addition to the existing literature. The findings in this book provide some basis for the formulation of policy strategies that could influence individual occupational choice decision in the direction that could reverse occupational segregation by gender in the country. The findings on the sources of the gender wage gap is also a source of information for the formulation of policy strategies to support the discriminatory gender group with the view to promoting equity and efficiency in income distribution and in resource allocation.

1.8

Structure of the Book

The book contains nine chapters beginning with the introductory chapter which raises the issues of gender discrimination in the labour market and outlines the objectives, methodology adopted and motivation for the work. ϭϱ

It is followed by a descriptive overview of gender differences in labour market covering labour force participation rate, employment and unemployment, education and wages are analysed in chapter two. Chapter three outlines gender policy, legal and institutional framework with chapter four highlighting the issues and types of labour market discrimination from gender perspective. A review of the theories of segregation and discrimination in the labour market is the main focus of chapter five followed by chapter six which applies four indices to assess the extent and changing pattern of sex segregation of occupation since 1960. In chapter seven, the book carries out empirical estimation and analysis of gender wage differentials followed by a measure of gender wage discrimination in chapter eight. Chapter nine provides a summary, conclusion and some policy implications.

ϭϲ

Chapter 2

Overview of Gender Differences of the Ghanaian Labour Market 2.0

Introduction

The focus of this chapter is to overview the gender dimension of the Ghanaian labour market in respect of gender differences in the labour force participation, employment, unemployment and underemployment as well as education and wages or earnings. It further looks at sex distribution of employment based on different classification of occupation, industry, type and economic sector.

2.1

A Brief Account of the Ghanaian Economy

Ghana is a democratic country located in the West Coast of Africa and gained independence from British rule in March 1957. The structure of the Ghanaian labour market largely mirrors the overall structure of the FRXQWU\¶VHFRQRP\Until 2006, when the national account was rebased, the agriculture was the major contributor to national output accounting for an average of about 40 percent of GDP between 1999 and 2006 followed by services sector with a third share of national output. Agriculture lost its dominance to service after the rebasing with a share of 28 percent to national output as against 50 percent by the services sector. The country began commercial production of oil in 2011 and this is expected to significantly boost the contribution of industry to national output which stood at 22 percent in 2010.

ϭϳ

The ORVW RI DJULFXOWXUH¶V SURPLQHQFH LQ QDWLRQDO RXWSXW KDV KRZHYHU QRW affected its status as the main source of livelihood for the estimated 24.2 million population5. Estimates from the fifth round of the Ghana Living Standards Survey (GLSSV) indicate that agriculture constitutes about 55 percent of total employment in 2006. The informal economy dominated by self-employment is very pervasive due to the sluggish growth of the public sector (Baah-Boateng and Turkson, 2005). Employment in the informal economy accounts for over 80 percent while the public sector has lost its dominance in the formal sector accounting for less than half of total formal VHFWRU HPSOR\PHQW *KDQD¶V HFRQRPLF UHIRUP ODXQFKHG LQ 83 which deemphasised the direct involvement of government in economic activity through privatisation and public sector retrenchment shed off public sector employment in favour of the informal and the private formal sector. The high level of informality implies high rate of vulnerable employment6 which is estimated at about three-quarters of total employment. This means that productive and decent employment with formal work arrangements in terms of adequate social security and effective social dialogue mechanism are limited in the Ghanaian labour market.

About a quarter of employed people in the Ghanaian labour market are estimated to be living in poor households indicating a 25 percent working poverty rate. The lower average public sector wages or earnings relative to that in private formal sector in the 1980s and early 1990s have been overturned in recent times on account of higher rate of unionisation in the public than the private sector. Over the past two decades and beyond, the Government of Ghana has carried out different forms of public sector pay reforms to improve salaries and working conditions of public sector 5 7KHSRSXODWLRQDQGKRXVLQJFHQVXVSXWV*KDQD¶VWRWDOSRSXODWLRQDW 6 Vulnerable employment rate measures the proportion of employed people engaged in own account work (i.e. self-employed without employees) and contribution family work (or unpaid family work). ϭϴ

As shown in figure 2.1, female population witnessed faster growth than male population over 1960-70 and 1970-84 with equal growth rate for the two sexes in 1984-2000. This resulted in a higher female population than male population in 2000. Indeed, in 1960, the total female population stood at 3.3 million representing 49.5 percent of the total population. This consistently increased to 9.6 million (or 50.5 percent of total population) in 2000. However, as the discussions in the subsequent sections indicate female representation in the labour market in general and in certain occupations in particular remains low.

2.2.1 Labour Force Participation Rate The faster rate of growth of the female population has largely translated into rapid growth of women in the labour force particularly between 1960 and 1984 culminating in their improved labour force participation rate. From figure 2.2, the average annual labour force growth rate of women increased remarkably from 3.5 percent between 1960 and 1970 to 4.9 percent over the 1970-84 period before slowing down to 2.3 percent during the period of 1984-2000. In contrast, the average annual growth of the PDOH¶V ODERXU IRUFH ZDV VORZHU ZLWK DQ LQFUHDVH IURP percent in the 1960s to 2.8 percent and 2.7 percent over the years 1970-84 and 1984-2000 respectively.

The observed faster growth in the size of the female labour force relative to that of males is largely a reflection of the significant improvement in the labour force participation rate (LFP) of women compared to men. The LFPR of women rose from 57 percent in 1960 to 73 percent in 2000 as DJDLQVWDGURSLQPHQ¶VODERXUIRUFHSDUWLFLSDWLRQUDWHIURPSHUFHQWWR 73 percent over the same period (see table 2.1). This has resulted in a significant decline in the gender participation gap from 32 percentage points in 1960 to 2 percentage points in 1984 before surging marginally to ϮϬ

1960

1970

67.5

71.7

Youth share in total unemp 59.9

2000

1991/92

1998/99

2005/06

--13.3

82.0 --76.5

--23.5

70.5 79.5 --69.6

35.8 --30.4

36.5

9.0 7.0 14.2 13.7 70.4 29.6 75.6 24.4

44.5 44.4 25.8 32.1

5.8 74.6

43.9

5.7 25.4

38.1

63.6 83.5 81.7 76.7 72.6 74.4 78.0 81.4 78.9 71.3 67.7 61.0 80.8 79.7 69.1 64.8 73.6 70.1 62.0 60.9 69.2 65.7 3.9 3.2 2.5 10.1 10.7 3.2 3.0 3.4 2.2 3.1 3.0 5.8 4.6 7.5 8.3 6.0 7.0 ----------5.0 5.2 5.4 4.6 6.9 6.1 6.8 8.3 12.5 18.6 13.2 13.9

Male Fem Male Fem Male Fem Male Fem 9.4 9.6 7.2 7.7 8.7 9.4 10.8 11.4 4.2 4.1 2.7 3.3 3.9 4.4 4.5 4.8

ϮϮ

Source: Population Censuses and Ghana Living Standards Surveys 3, 4, &5 (Ghana Statistical Service)

(2) Figures in italic are unemployment rates based on the broader definition (3) Adults refers to those aged 15 years and above (4) The GLSS figures were estimated by the Author

Activity Rate 7±14 yrs ----------- --15.8 15.0 7.7 8.0 25.5 30.3 16.4 13.9 Adult Illiteracy rate 64.0 82.0 48.0 65.5 35.0 51.8 33.1 44.5 39.2 61.5 35.8 62.6 41.0 63.2 Average basic hourly wage ----------------187 164 1,084 756 5,002 2,861 Note: (1) Labour Force Participation Rate is measured as Labour Force as a % of total working population of 15 yrs+

--86.7

---

---

83.5 77.0 7.6

Youth Unemployment rate ---

Underemployment (%) ----Sex Composition of Paid employment

1984

Male Fem Male Fem 4.25 4.31 6.1 6.2 1.9 1.5 2.7 2.9

56.7 53.7 5.2

Male Fem 3.4 3.3 1.7 1.1

89.0 Employment-to-pop ratio 83.2 Adult unemployment rate 6.5

L. F. Participation rate1

Estimated Population Estimated Labour Force

Relevant Indicators

Table 2.1: Basic Indicators of Gender Differences in the Ghanaian Labour Market 1960±2006 Population Census Ghana Living Standards Survey (GLSS)

Essentially, the performance of women in terms of labour force participation rate in Ghana appears to be better compared with the subSaharan African average. Available evidence shows that the 4 percentage and 7 percentage points gender difference in labour force participation rate in favour of men in Ghana in 2000 and 2006 respectively (see table 2.1) is better than the sub-Saharan African average of 24 percentage points in 2002 and 2006 (ILO, 2008). Indeed, the narrower gender difference in the participation rate in Ghana than in SSA stems from the rapid improvement in the participation rate of women vis a vis their male counterparts since 1960. While the average participation rate of women in SSA has consistently hovered around 63 percent compared with the average rate of 86 percent for men since 1996, the rate for women has averaged 78 percent compared with 80 percent for males in Ghana since 1991.

2.2.2 Employment, Unemployment & Underemployment The remarkable improvement in the participation rate of women is reflected in the increased employment-to-population ratio of women vis±àvis that of men. The ratio is a measure of quantity of employment only and does not relate to quality of employment. It provides information about the ability of the economy to provide employment (ILO, 2009). Quality of employment can be seen from the extent to which employment is reflected in improved wages, better conditions of work including social protection and decent living. An increase in the ratio accompanied by a rising share of wage employment indicates increased quantity and quality of employment.

Technically, there is no optimal employment-to-population ratio. However, evidence shows that developed economies tend to have lower ratios than developing countries (ILO, 2009). This is because higher productivity and incomes in developed economies suggest that fewer workers are required to meet the needs of the entire population. In contrast, very high ratio in Ϯϯ

3.5 percentage points in 2006. Relative to SSA, employment-to-population ratio is lower among males and higher among females in Ghana. This is evident in the 69.2 percent ratio for males in Ghana as against SSA average of 78 percent and 65.7 percent ratio for females in Ghana compared with SSA average of 57 percent in 2006.

Indeed, while the improved educational attainment of females and the sharp drop in their illiteracy rate since 1960 (see table 2.1) largely accounts for the improved employment-to-population ratio of women, the consistently increasing advocacy on women empowerment in Ghana and across the world may also be a major contributory factor. On the other hand, the desire of males to pursue higher education due probably to the difficulty in securing appropriate jobs may also explain the decline in the ratio for males. Even though, there has been remarkable increase in the proportion of both males and females with secondary education or better, males are found to have done better in the attainment of higher education than their female counterparts (see table 2.3 on page 36).

One relevant indicator used in assessing labour market performance around the world is the unemployment rate. It is a phenomenon of joblessness. Unemployment rate is conventionally measured by the proportion of labour force above a specific age (15 years and above in Ghana) who are jobless, available for work and actively looking for work. This conventional definition seems narrow when applied to the labour market in developing countries where many economically active jobless persons fail to seek work for various reasons, and t5hus understating the unemployment rates. It is therefore appropriate to also look at unemployment within a broader context that relaxes WKH ³DFWLYHO\ VHHNLQJ ZRUN´ FRQGLWLRQ. This helps in minimising WKH DGYHUVH ³GLVFRXUDJHG ZRUNHU´ HIIHFW RQ XQHPSOR\PHQW rates in many developing countries including Ghana. Unemployment rates Ϯϱ

based on the both the conventional (or narrow) and the broader definition are therefore used in the overview of the Ghanaian labour market.

Until 2000, unemployment in a narrow sense appeared to be a bigger challenge for men than women. As shown in figure 2.4, annual average growth of the number of adult unemployed women increased from 0.5 percent between 1960 and 1970 to 1.5 percent over 1970-1984 compared with 2.6 percent to 3.4 percent for men over the same period. The annual average growth of unemployed women also grew faster than that of men between 1984 and 2000. Based on census data, the narrow adult unemployment rate which was reported to be higher among men than women between 1960 and 1984 was reversed in 2000 with 0.6 percentagepoint higher rate for women than men. This may largely be due to the increased female participation in the labour market against the backdrop of limited available employment opportunities for women. Between 1984 and 2000, the narrow unemployment rate of adult women rose markedly from 2.5 percent to 10.7 percent compared with 3.2 percent rate among men to 10.1 percent over the same period (see table 2.1).

Among the youth, the narrow unemployment rate of women was observed to be higher than men in 1991/92 but the reverse was the case in 1998/99 and 2005/06. However, the proportion of young people aged 15-24 years in total unemployment is estimated to be generally higher for women than men (table 2.1). Broad unemployment rates of adult women relative to men were reported to be lower in 1991/92 but higher in 1998/99 and 2005/06. Among the youth however, the broad unemployment rates were higher among women than men between 1991 and 2006 indicating that more jobless young women than young men who are available for work often fail to actively seek work for various reasons. Ϯϲ

Ghana has ratified7. Any person below 15 years engaged in economic activity is considered as a working child. For children who are above 13 years, however, the law allows for engagement in light work. In this analysis, working children is defined as people aged between 7 and 14 years who are engaged in economic activity with no hazardous effect on him or her.

A statistical analysis of working children in Ghana suggests a higher proportion of working female children than working male children in the 1990s with the reverse occurring in 2000 and 2005/06. From table 2.1, an estimated 8 percent of girls and 7.7 percent of boys between the ages of 7 and 14 years were engaged in economic activities in 1991/92 and this surged to 30 percent for girls and 26 percent for boys in 1998/99. However in the year 2000, about 16 percent of male children as against 15 percent of female children were estimated to be engaged in one type of economic activity or the other. The activity rate for female children dropped to 14 percent in 2005/06 while that of male children remained at 16 percent indicating an increase in the male-female gap from 1 percentage point in 2000 to 2 percentage points in 2005/06.

This implies that while child work was prevalent among female children before 2000, the reverse is the case since then with the phenomenon being more pronounced among boys than girls. The reversed gender dimension of child work phenomenon may be partly due to the national campaign of encouraging girl child education. It is however essential to indicate the concern such that the boy child is not disadvantaged by the campaign to FUHDWH³UHYHUVHGLVFULPLQDWLRQ´In addition, the changing characteristics of occupation available to children, such as selling merchandise along the city main streets (a physically challenging job), which favoured the boy child 7 The minimum employment age of 15 represents the age for completing compulsory schooling. Ϯϴ

from the sedentary job of sitting by the family house and selling cold drink, which favoured the girl child explains increased incidence of working children among boys than girls after 2000.

2.3

Gender Differences in Employment Classifications

In this section, we discuss gender differences in employment from various dimensions. Technically and statistically, employment can be classified into different forms. These are occupation, type or status of employment, employment sector, and economic sector or industry. These major employment classifications form the basis of analysis in this section.

2.3.1 Occupation For the purpose of consistency in trend analysis of sex distribution of occupation, we use International Standard Classification of Occupation of 1968 (ISCO±68) to analyse differences of occupation under seven broad classifications between men and women. Table 2.2 reports on sex distribution of occupation between 1960 and 2006 based on population censuses and three rounds of GLSS and indicate that in non-agriculture occupations, women are mostly engaged in sales and service occupations. However, they are underrepresented in better paid occupations such as administrative and managerial, professional and technical, and clerical occupations. For instance, while at least 21 percent and 7 percent of the female workforce in 2000 and 2006 were estimated to be engaged in sales and service occupations, only 9 percent and 5 percent of male workforce were working in those occupations. On the other hand, at least about 8 percent and 7 percent of male workers in 2000 and 2005/06 as against 5 percent and 3 percent of female workers were engaged in professional and technical occupations.

Ϯϵ

--100 1.72

0.0

--100 1.42

0.9 25.7 1.5 54.5 15.9 --100 2.64

3.4 3.1 3.2 65.7 18.6

0.6

--100 2.79

1.4 24.0 1.6 55.8 14.3

0.1

1984 Male Fem 5.4 2.8 0.2 1.9 22.0 7.4 49.7 13.4

2.0 0.6 100 100 3.75 3.68

7.0 8.6 4.3 50.8 18.6

0.4

2000 Male Fem 8.3 4.8 0.4

0.0

0.4 3.7 7.8 5.5 59.0 17.9

0.1 1.2 27.3 4.0 50.3 14.4

1.4 5.8 4.4 58.6 22.2

0.8

ϯϬ

1.1 21.3 7.0 51.4 15.9

0.1

2005/06 Male Fem 6.8 3.2

------------100 100 100 100 100 100 3,561 4,337 3,854 4,633 6,863 7,512

3.2 1.5 4.3 23.7 3.1 2.2 66.6 59.1 17.1 10.4

1998/99 Male Fem 5.7 2.7

Ghana Living Standards Survey 1991/92 Male Fem 5.3 3.1

* N = Total Employment and figures for the GLSS represent sample and figures for census years are in millions Note: Some figures for 1991/92, 1998/99 and 2005/06 were the Authors own computation from the micro data source Source: Population Censuses and Ghana Living Standards Surveys 3, 4, &5 (Ghana Statistical Service)

--100 0.99

--100 1.56

Other Total N*

4.3 2.9 4.0 59.8 23.1

0.6

2.6 0.3 4.3 28.0 2.5 1.6 62.9 58.2 23.8 10.6

Clerical Sales Services Agriculture Production

0.1

Admin & Managerial

0.8

1970 Male Fem 5.3 2.0

Population Census

1960 Male Fem Professional & Technical 3.1 1.3

Occupation

Table 2.2: Sex Distribution of Occupation, 1960 ± 2006 (%)

Males are also better represented than their female counterparts in administrative and managerial occupations considered as a better and highly skilled occupation. In agriculture and production occupations, both sexes are almost equally represented with males having a slight edge over their female counterparts. In clerical occupations, males are found to be better represented than females which in sharp contrast with the general observation that clerical occupation is a female dominated occupation. Anker (1998), found overrepresentation of women in clerical occupations in most countries of the world reflecting the work done by women office workers who ensure that phones are answered, letters are typed, and data are entered. Only 6 out of about 57 study countries including Ghana and Nigeria saw underrepresentation of women in clerical occupation.

Generally, the sex distribution of occupation has undergone considerable changes since 1960. Estimates from the census data, show a shift in female representation in agriculture and sales occupations in favour of five other occupations between 1960 and 2000. Based on the GLSS data, between 1991 and 2006 agriculture, sales and clerical occupations shed female workers to the remaining four occupations. Similarly, there has been a shift in male workforce from agriculture, production and administrative and managerial occupations to the remaining four occupations between 1960 and 2000. Male representation in agriculture and clerical occupations experienced a decline in favour of the five remaining occupations between 1991 and 2006 based on the GLSS estimates.

As indicated in table 2.2, improvements in female representation occurred largely in high skilled occupations such as professional and technical and clerical as well as service occupations where the proportion of women, at least tripled between 1960 and 2000. Female representation in administrative and managerial occupations also doubled while that of ϯϭ

production occupation improved from 10.6 percent to 13.4 percent over the same period. Estimates from GLSS also show a substantial improvement in female representation in production occupation from 10 percent to 16 percent, and in service occupation from 2.2 percent to 7 percent between 1991 and 2006. Over the same period, a marginal improvement in the proportion of women in professional and technical and administrative and managerial occupations by 0.1 percentage point was recorded. A decline in female representation, however, occurred in agriculture and sales occupations between 1960 and 2000 by 8 and 6 percentage points respectively from 58 percent and 28 percent in 1960. The proportion of women in agriculture and sales occupations also shrunk by 8 and 3 percentage points between 1991 and 2006 based on the GLSS estimates. ȱ

There was a decline in male representation in agriculture and production occupations between 1960 and 2000 from 63 percent and 24 percent to 51 percent and 19 percent respectively while that of administrative and managerial occupation dropped from 0.8 percent to 0.4 percent. Estimates from GLSS also reported a dip in the proportion of male workforce in agriculture and clerical occupations from 67 percent and 3.2 percent in 1991/92 to 59 percent and 1.4 percent in 2005/06. The remaining five occupations experienced improvement in male representation by a minimum of 28 percent in professional and technical occupation to a maximum of 50 percent in administrative and managerial occupations between 1991 and 2006. The proportion of male workforce in sales and highly skilled occupations of professional and technical and clerical occupations at least doubled between 1960 and 2000 while that of service rose from 2.5 percent to 4.3 percent over the same period.

The relative improved or deteriorating trend in female and male representation is captured in the female-male representation ratio of various ϯϮ

2.3.2 Employment Status The improved educational attainment of women also has implications for participation of women in paid-employment. The proportion of women in paid employment has seen consistent improvement since 1970. It increased from 6.5 percent in 1970 to 9.7 percent in 2000 while that of men experienced a continuous decline from 35 percent to 22 percent over the same period (table 2.3). The trend was different between 1991 and 2006 with both men and women witnessing improvement in their respective proportion in paid-employment with men performing better than their female counterparts. The proportion of women of 10 percent in 2000 and 8 percent in 2006 in paid employment in Ghana is lower than the 14 percent average recorded in SSA in 2006. Similarly, the proportion of males in paid employment of 22 percent and 25 percent in 2000 and 2006 in Ghana falls short of the 29 percent average proportion of male workforce recorded in SSA.

In Ghana, women are better represented than men in self-employment with 79 percent and 57 percent of females compared with 68 percent and 54 percent of men engaged in that type of employment in 2000 and 2006. The gender difference of 11 and 3 percentage points in 2000 and 2006 is higher than the 1 percentage point gender difference recorded in 2006 in SSA based on 52 percent of men and 51 percent of women in self-employment. This implies that while both women and men are equally represented on average in self-employment in SSA, female representation in selfemployment is higher than male representation in self-employment in Ghana.

ϯϱ

100

1.42

35.0

53.5

7.9

3.6

100

Paid-employment

Self-employed

Unpaid fam. work

Other*

Total

Total Employment 1.72

2.64

100

2.6

9.4

62.6

25.4

2.79

100

0.7

15.5

76.4

7.4

Male Female

1984

3.75

100

4.4

5.7

68.0

21.8

3.68

100

3.4

7.8

79.0

9.7

Male Female

2000

3,561

100

19.2

1.2

58.9

20.7

4,337

100

19.4

1.7

71.4

7.5

Male Female

1991/92

3,863

100

0.3

10.7

65.9

23.0

ϯϲ

Source: Population Censuses and Ghana Living Standards Surveys 3, 4, &5 (Ghana Statistical Service

GLSS figures were computed by the Author

4,624

100

0.1

22.7

71.0

6.2

Male Female

1998/99

6,863

100

2.9

17.7

54.4

25.0

7,512

100

2.2

32.3

57.3

8.2

Male Female

2005/06

Ghana Living Standard Survey

Total employment is in millions for the census years while that of the GLSS represent sample size * Other includes domestic employees, apprentice and others

0.2

19.4

73.9

6.5

Male Female

1970

Status

Employment

Census Period

Table 2.3: Sex Distribution of Employment Status, 1970-2006 (%)

Data from the Census suggests an increase in the proportion of both females and males in self-employment between 1970 and 2000. However, sample data from GLSS indicate an increase in the incidence of selfemployment among males in the 1990s and a subsequent decline in 2006 while female representation in self-employment consistently declined between 1991 and 2006. Further classification of self-employment show that self-employment (non-agriculture) is dominated by women while men are more represented in self-employment (agriculture) than women. There is also a relatively high proportion of women in unpaid family work than men.

Relative to the SSA, the proportion of female and male workforce in unpaid family activity is quite low. For instance, about 8 percent and 27 percent of women in 2000 and 2006 were engaged in unpaid family work compared with 35 percent in SSA in 2006. Similarly, about 6 percent and 9 percent of men in 2000 and 2006 were represented in unpaid family work as against 19 percent in SSA in 2006. This shows a smaller gender representation differential in SSA than in Ghana with an estimated femalemale representation ratio of 16 percentage points in SSA against 18 percent in Ghana in 2006. Indeed, employment of men and women as unpaid family workers declined between 1970 and 2000 based on the census data with the proportion of women shrinking faster than that of men.

Clearly, the drop in the proportion of males in unpaid family activity coupled with the decline in paid-employment against a gain in selfemployment suggests a shift in male employment from family activity and paid-employment to self-employment. On the other hand, the decline in the proportion of females in unpaid family work compared with an increase in self- and paid-employment also indicate a shift in employment from the former to the latter two types of employment. Estimates from GLSS ϯϳ

however indicate appreciation of the proportion of both males and females in unpaid family activity with that of females rising faster than that of males. In effect, the improved educational attainment of both sexes appears to have positively impacted on the quality of employment between 1970 and 2000 with a shift away from unpaid family activity to either paid employment in case of women, and self-employment in both cases.

In contrast, estimates from GLSS suggest that the enhanced educational attainment may have triggered a shift in employment of men from other employment types (apprentice, domestic employees etc) to paid- and selfemployment as well as family activity. The impact of improved education of women is also seen to have caused a shift from self-employment and other forms of employment to paid employment and unpaid family activity which may not significantly have positive effect on female average wage.

2.3.3 Employment Sector The sex distribution of employment by sector indicates higher proportion of women than men in the informal sector and a lower representation of women than men in the formal sector. This is essentially a reflection of the relatively lower educational attainment of women than men. From table 2.4, the majority of the employed regardless of sex is employed in the informal sector. An increasing proportion of men and women were employed in the private formal sector between 1984 and 2000. In the public sector however, while the proportion of men declined between 1984 and 2000, the proportion of women employed in the public sector recorded a decline over the same period. An increasing proportion of men and a decreasing proportion women were engaged in the informal sector over 1984-2000.

ϯϴ

Estimates from the GLSS dataset suggests that between 1991 and 2006, the proportion of men and women employed in the public sector consistently declined as against rising proportion of both men and women in the private formal sector over the same period. A declining proportion of women engaged in the informal sector were reported between 1991 and 2006. In contrast, the share of men in the sector dropped marginally in the 1990s and surged subsequently in 2006. Table 2.4: Distribution of Employment Sector by Sex, 1984 ± 2006 Employment Sector

Census Period 1984 2000

Ghana Living Standard Survey3

1991/92 1998/991 2005/06 Male Fem Male Fem Male Fem Male Fem Male Fem Public 15.7 4.9 11.8 6.3 10.5 3.6 9.8 3.3 6.8 2.9 Priv. Formal 9.7 2.5 11.1 6.4 8.0 2.1 13.0 2.9 12.9 5.3 Informal

74.6 92.6 77.2 87.3

81.5 94.3 77.2 93.8

80.3 91.8

Total Total

100

100

100

100

100

100

100

100

100

100

Employment2 2.64 2.79 3.75 3.68 3561 4337 3863 4624 6,863 7512 Notes: 1. Employed people aged 15-64 while others are aged 15+ years 2. Figures for GLSS represent sample size and census figures represents population in millions 3. GLSS figures were computed by the Author Source: Pop. Census Data and Ghana Living Standard Survey 3, 4, & 5 (GSS)

The shift in employment from the public to the informal sector, particularly in agriculture in the second half of the 1980s and early part of the 1990s, underscores the effect of the implementation of some programmes that bother on employment and incomes under the Structural Adjustment Programme (SAP). Beaudry and Sowa (1994) confirmed the move of retrenched public sector workers in Ghana to agriculture between midϯϵ

1980s and early 1990s. The decline in the proportion of both male and female in public sector employment between 1984 and 1991 was largely attributed to the retrenchment exercise which saw the loss of public sector jobs particularly among those in the lower echelon of the job ladder. The low level of education and limited employable skills of both men and women who lost their jobs in the public sector made it difficult for them to move into the private formal sector. Indeed, the informal sector provided a refuge for both men and women who lost their jobs in the public sector because of the flexible nature of activities in the sector while others particularly women became unemployed as reflected in the higher unemployment rate of women than men since 1991.

2.3.4 Industry or Economic Sectors Based on International Standard Classification of Industries (ISCI), agriculture continues to serve as the key source of employment for both sexes with 61 percent of males and 53 percent of females in employment earning their livelihood in that sector in 2005/06. The second major source of employment for both men and women is the service sector employing at least 25 percent of men and 33 percent of women in employment in 2005/06 (table 2.5). Within the service sector, trade and hospitality constitute the major source of employment for women accounting for about 25 percent of working women compared with 9 percent of men in the same sub-sector. In the industrial sector, a larger proportion of employed women than employed men are in manufacturing while a higher proportion of working men than working women are found in the construction, mining and quarrying and energy sub-sectors. Over the years, the representation of women and men in agriculture has been declining in favour of other sectors. The proportion of women in agriculture dropped from 58 percent to 52 percent between 1960 and 2000. ϰϬ

2.9

8.6

0.9

5.5

18.0

6.0

12.2

100

1.57

Mining/Quarrying

Manufacturing

Electricity/Water/Gas

Construction

Service

Trade

Other Services

Total

N* (in millions)

0.2

0.0

15.1

0.2

15.5

54.5

100

20.8

3.9

1.42

100

3.9

26.1

24.7 30.0

4.1

0.7

9.7

1.7

16.2

0.99 1.72

100

3.3

27.7

31.0

0.3

0.0

10.0

0.3

10.6

59.1

56.0

0.1

0.1

14.0

0.1

2000

100

17.2

11.0

28.2

5.0

0.5

10.1

1.9

17.5

54.3

3.68

100

10.9

23.8

34.7

1.0

0.2

11.2

0.9

13.3

52.0

Male Fem

2.64 2.79 3.75

100 100

16.0 5.2

4.2 24.4

20.2 29.6

2.3

0.5

7.5

0.9

11.2 14.3

66.4

Male Fem

1984

ϰϭ

Note: GLSS figures were computed by the Author Source: Population Censuses and GLSS Survey Data (Ghana Statistical Service);

17.9

Industry

58.4

Male Fem

Male Fem

63.9

1970

1960

Agriculture

Industry

100

6.4

25.0

31.4

0.1

0.1

9.4

0.1

9.7

58.9

100

19.3

7.4

26.7

2.8

0.4

8.9

1.4

13.5

59.8

100

8.0

27.4

35.4

0.2

0.1

13.9

0.1

14.3

51.1

Male Fem

1998/99

53.2

0.1

0.1

13.3

0.3

100

16.2

9.0

100

8.3

24.7

25.2 33.0

3.5

0.3

8.5

1.1

13.4 13.8

61.4

Male Fem

2005/06

3,561 4,337 3,863 4,624 6,863 7,512

100

18.7

4.7

23.4

2.5

0.2

6.7

1.0

10.4

66.2

Male Fem

1991/92

Table 2.5: Sex Distribution of Industry of Employed Persons (15 yrs. and older), 1960±2006 (%) Census Period Ghana Living Standard Survey

In contrast, the proportion of women in industry rose from 11 percent to 13 percent and in service from 31 percent to 35 percent over the same period (table 2.5). Similar trends were observed between 1991 and 2006 based on the GLSS estimates with a drop of 6 percentage points of female representation in agriculture in favour of 4 percentage points and 2 percentage point gain in their representation in industry and service sectors respectively. A decline in the proportion of men occurred in both agriculture and industry (specifically mining, construction and utilities) between 1960 and 2000 in favour of service. The GLSS estimates however show a decline in the proportion of men engaged in agriculture by 5 percentage points in favour of industry and service between 1991 and 2006.

The general shift in employment from agriculture to industry and service is due partly to the apparent loss of interest in agriculture as a result of the limited support, if any for the sector. The food crop sector which absorbs the chunk of agricultural labour continues to suffer from heavy reliance on nature, limited access to farmland and affordable credit, absence of market and guaranteed prices for farm produce among others culminating in lower incomes in the sector. The observed improvement in education also explains the shift in employment from agriculture to other sectors since the skill requirement in agriculture (based on perception and reality) is low relative to other sectors.

2.4

Gender Differences in Employment Quality

While increased labour force participation and employment rate as well as declining unemployment rate provide a clear indication of improved labour market performance, it does not necessarily indicate enhanced quality of employment. The improvement in the quality of employment needs to be reflected in the form of an enhancement in the occupation and employment status of the people and more importantly improved condition of service. ϰϮ

Essentially, apart from better remuneration, employment quality should be measured by the provision of employment contract, medical benefits, paid holidays and leave, pension and social security benefits and freedom to join XQLRQV RI RQH¶V FKRLFH among others. Clearly, these provisions are generally unavailable in the informal sector and amongst the self-employed where women dominate. They are largely available in paid and/or the formal sector employment where females are underrepresented suggesting a better quality of employment of men than women.

Table 2.6: Proportion of Paid Employees of 15+ years with Written Conditions of Service Service Condition 1998/99 2005/06 Male Female Male Female Signed Contract 52.8 52.0 44.8 47.0 Trade Union 47.9 44.3 39.0 37.1 Paid Holidays

61.8

63.4

48.6

53.1

Paid Sick Leave Maternity Leave Both Sick & Maternity Leave

66.7 -----

62.3 -----

46.9 2.6 4.3

22.1 4.4 30.9

Pension Subsidised Medical Care Other Soc. Security benefits Training since last six months

36.2 53.8 51.9 27.4

31.5 40.4 51.9 25.2

39.4 38.1 31.0 9.4

39.8 33.9 32.9 11.2

Source: Computed by Author from GLSS4&5, Ghana Statistical Service

The improved representation of women in paid and/or formal sector employment where conditions of service are better compared to informal sector and/or self-employment, and increased proportion of women in highly skilled RFFXSDWLRQV PD\ EH VHHQ DV LPSURYHG TXDOLW\ RI ZRPHQ¶V employment since 1960. Differences in service condition as a measure of quality of employment between the two sexes have been mixed. As ϰϯ

reported in table 2.6, a higher proportion of men than women, had contracts with their employers, were entitled to receive any retirement pension and had received training six months prior to the survey in 1998/99. The reverse was the case in 2005/06 in favour of women with a greater proportion of women than men benefiting from these employment conditions. An equal percentage of women and men in employment were entitled to social security benefits in 1998/99. In 2005/06 however, a greater proportion of women than men was reported to be entitled to social security benefits.

A higher proportion of women than men were also observed to have benefited from paid holidays, maternity leave and both sick and maternity leave in 1998/99 and 2005/06 with greater proportion of women than men in paid employment benefiting from these provisions. A higher proportion of male workers than female workers in paid employment enjoyed better conditions of service than their female counterparts in terms of subsidised medical care, sick leave and belonging to trade union. Generally, quality of employment in terms of service conditions for both men and women worsened considering the fact that the proportion of women and men benefiting from all but one of the listed conditions dropped between 1998/99 and 2005/06 (see table 2.6).

2.5

Wage Differences by Gender

The increased participation rate of women coupled with their improved employment-to-population ratio and their rising proportion in paid employment do not seem to have had any significant enhancing effect on their wages or earnings. In spite of the improved performance of women in the labour market since 1960 as per most of the employment indicators, average hourly wage differentials between the two sexes have been widening. Estimates from GLSS provide evidence to suggest a consistent ϰϰ

declining female-male wage ratio from about 88 percent in 1991/92 to 70 percent in 1998/99 and 57 percent in 2005/06. This may largely be explained by the increased proportion of women in unpaid family activity and high prevalence of subsistence economic activity of women relative to men. In addition, although the composition and representation of women in wage employment may be on the ascendency, it is not quite clear whether they are equally placed with equal responsibility, position and status as their male counterparts.

Analysis of gender differences in wages by occupation shows that all but one major occupation reported relatively lower average female wages than that of males. Apart from clerical occupation that reported female-male average monthly wages or sex ratio of earnings of 1.19, all the six other major occupations recorded female-male average sex ratio of wages or earnings of less than unity. This ranges from the highest of 0.73 in production occupation to the lowest of 0.23 in service occupations (figure 2.7). The declining female-male average wage or earnings ratio was observed in most occupations over the period. The ratio declined in professional and technical, administrative and managerial, service, agriculture and production occupations (figure.2.7). Average wages of women in clerical and sales occupations however appreciated relative to their male counterparts in 2005/06 after dropping in 1998/99 from the 1991/92 level resulting in an increase in female-male wage or earnings ratio in these occupations.

The relatively lower average wage of women against the backdrop of improved share of women in wage employment indicates that women are probably engaged at the lower echelon of the job ladder. For instance, in agriculture, most women are engaged in subsistence agriculture and/or unpaid agricultural activity. In professional and technical occupations, ϰϱ

differences in the distribution of occupation by sex and gender wage differentials in favour of men have been linked to the limited access of girls to education. Adult literacy rate and educational attainment of women have often lagged behind their male counterparts. In 2005/06, over 63 percent of adult women compared with 41 percent of adult men was estimated to be illiterate (table 2.1). In terms of educational attainment, about 62 percent of working females as against 41 percent of working males (table 2.7) were found to have either completed primary education or never been to school. On the other hand, over 17 percent of working men compared with 8 percent of working women were observed to have secondary education or better. Furthermore, 42 percent of working men as against 30 percent of working women has either tasted or completed middle or Junior High School (JHS).

The trend analysis of educational attainment of the working people shows improvement in education of both men and women. Table 2.7 reports that the proportion of working women with secondary education or better increased continuously from less than 1 percent in 1960 to at least 12 percent in 2000. At the same time, the proportion of women in employment with no education dropped substantially from 91 percent to 57 percent over the same period. A similar trend was also observed between 1991 and 2006 using estimates from the GLSS with improvement in the proportion of employed females with secondary education or better by 4.2 percentage points. This was accompanied by a drop in the proportion with no education by about 10 percentage points.

Education of men has also witnessed considerable improvement since 1960 and this largely accounts for the improvement in male representation in the highly skilled occupations. The 2.4 percent of working men with secondary education or better in 1960 surged to 20.3 percent in 2000 while those with ϰϳ

no education declined from 76 percent to 42 percent over the period. Similarly, the proportion of the working men with secondary education or better also rose from 10.0 percent to 17.4 percent between 1991 and 2006. Table 2.7: Educational Attainment of Employed Persons (%) Education Level Sex None Male

Population Census Ghana Living Standard Survey 1960 1970 1984 2000 1991/92 1998/99 2005/06 75.9 63.5 45.5 42.1 37.1 21.5 27.6

Female Male Female Middle/JHS Male

90.6 82.7 64.3 7.1 8.2 7.9 4.7 7.4 9.1 14.3 20.4 37.3

Primary

Female 4.0 Secondary+ Male 2.4 Female 0.6

8.3 7.8 1.6

56.6 5.1 6.1 32.6

56.3 20.2 22.3 32.7

43.1 20.1 24.9 41.5

44.7 13.1 17.0 41.9

23.3 25.0 9.2 20.3 2.7 12.2

17.6 10.0 3.9

26.9 16.9 5.2

30.2 17.4 8.1

Note: figures for 2005/06 were computed by the Author Source: Pop. Census and Ghana Living Standards Surveys 3, 4&5 (GSS)

The improved educational attainment of both men and women is largely linked to government commitment to promoting the quality of human capital through education with establishment of many schools at all levels since independence. With only one public university at the time of independence in 1957, data from Ministry of Education show that the country as at 2010 had about 8 public universities and over 50 private universities, 10 polytechnics, 296 Technical and Vocational Institutions and 38 Teacher training Colleges. There are over 17,881 primary schools, 10,213 Junior High Schools (JHS) and 670 Senior High Schools (SHS) as of 2009 compared with very few schools at the time of independence.

ϰϴ

A critical examination of the relative improvement in educational attainment of the two sexes shows a better performance of women than men. From table 2.7, it is estimated that the ratio of female-to-male proportion of those with secondary education or better improved markedly from 0.25 to 0.6 between 1960 and 2000. The ratio also increased from 0.28 to 0.77 for those with middle/JHS education over the same period. Similarly, the ratio increased from 0.39 to 0.47 for those with secondary education or better between 1991 and 2006 and from 0.54 to 0.72 for those with middle/JHS education over the same period. Enrolment of girls relative to boys in school has also improved considerably at all levels of education over the past decade. Table 2.8: Sex Distribution of Apprentices by Main Trade Main trade

1998/99

Main trade

2005/06

Learnt Carpentry Masonry Tailoring

Male Female 16.0 0.4 8.5 --13.2 64.4

Learnt Food processing Building Textile/furnishing

Male Female 0.5 7.9 27.8 0.2 16.8 58.9

Blacksmithing Mechanical Electronics/Electricals Painting/spraying Other

4.0 17.2 8.9 4.8 27.5

Automotives Mechanical Electrical Personal services Transportation etc

16.3 8.6 6.8 1.2 17.3

0.4 0.1 0.3 29.9 0.1

2.7

2.2

100

100

--0.3 --1.1 33.7

Other Total 100 100 Total Source: Computed by the Author from GLSS4&5, GSS

The Gender Parity Index (GPI) in primary and JHS increased from 0.92 and 0.88 respectively in 2001/02 to 0.96 and 0.92 in 2009/10 while that of SHS improved from 0.78 to 0.85 between 2005/06 to 2009/10 (see appendix table 2a). The GPI at tertiary level also rose by about 1.3 ϰϵ

percentage points from 0.40 between 2001 and 2005 with female share in total enrolment at the University of Ghana (the university with the largest population) increasing from 24 percent in 1991/92 to 41 percent in 7KH VXEVWDQWLDO LPSURYHPHQW LQ IHPDOHV¶ HGXFDWLRQ FOHDUO\ KDV implications for relative representation of men and women in various occupations.

In apprenticeship training females are commonly found in vocation or trade that involves activities similar to domestic chores and require less physical strength. As reported in table 2.8, females are better represented in textile/apparel/furnishing,

personal/ground

services

and

food

preparation/processing in 2005/06 while building, transportation/material moving trade and automotives as well as electrical are common among males than females. Similarly, in 1998/99, a higher proportion of females (64 percent) were found in tailoring with mechanical, electronics/electrical, carpentry, masonry, painting/spraying and blacksmithing recording better representation of males than females.

ϱϬ

Chapter 3

Policy, Legal and Institutional Framework of Gender Issues 3.0

Introduction

The changing trend and pattern of gender differences in employment, education and wages in Ghana are the outcomes of economic and social policies, institutional arrangements and the legal framework established over the years. The apparent general marginalisation of women in Ghana and across many developing countries has largely influenced the design of policies with gender considerations in favour of women. Generally, the intensification of campaigns for economic empowerment of women across the globe was essentially triggered by the 1995 Beijing Conference on women which recommended among other things, elimination of occupational segregation and all forms of employment discrimination and guaranteeing the rights of women and men to equal pay for work of equal value.

Over the years, gender activists have consistently advocated for policy, legal and institutional measures to promote gender equality in the pursuit of the economic, political and social development agenda in Ghana. Consequently, a number of legal and institutional arrangements have been established over the years to propagate the agenda of economic empowerment of women and promote active involvement of women in all spheres of national life. Indeed, many development policy initiatives have strived to ensure the inclusion of measures aimed at ensuring economic empowerment of women and the realisation of gender equality as enshrined ϱϭ

in the MDGs. The focus of this section is to overview the policy initiatives and establishment of legal and institutional framework for the promotion of gender equality and women involvement in national development effort.

3.1

Policy Initiatives, and Gender Effects

Ghana is yet to develop a comprehensive gender policy to provide some degree of coherency in addressing gender issues in the country. Generally, *KDQD¶V SROLFLHV UHODWHG WR JHQGHU KDYH RIWHQ IRFXVHG RQ UDLVLQJ WKH economic and social status of women with the intent of promoting gender equality. Over the years policy that concern women has been largely influHQFHG E\ WKH GHYHORSPHQW SDUDGLJP RI ³:RPHQ LQ 'HYHORSPHQW´ (WID). In the 1980s when the Structural Adjustment Programme (SAP) took off, the WID activities shifted towards increasing the productivity of women through the provision of micro credit to women in the informal sector.

The implications of the adoption of measures outlined in the SAP on economic activities of women and men could be traced from varying sources. The implementation of fiscal and monetary measures towards the realisation of macroeconomic stability and sectoral reforms impacted on men and women disproportionately. The removal of direct support for food crop agriculture as expenditure reduction measures coupled with the relatively high concentration of women in food production implies that female farmers were more adversely affected than male farmers by SAP. On the other hand, the export oriented trade measures that saw improved incentives to the cocoa sector including an enhanced producer price of cocoa benefited male farmers more than female farmers in terms of increased farm incomes because of limited representation of women in the non-food cash crop sector. ϱϮ

The adoption of monetary and financial policies that tightened the availability of credit did not only constrain access to credit of both sexes, but also exacerbated the plight of women in the credit market. In a study of female traders in Accra and Kumasi during the adjustment, Clark and Manuh (1991) found women complain, about capital squeeze to be partly due to the lack of access to formal credit and the crowding out of the informal credit market by new entrants particularly men, some working in traditionally female areas. The women, however, found the end of extensive regulation and serious harassment associated with market regulations appropriate.

One major aspect of the SAP that had employment implications was the SXEOLF VHFWRU UHWUHQFKPHQW SURJUDPPH RU WKH ³UHGHSOR\PHQW´ H[HUFLVH which started in 1987. The main objective of the exercise was to reduce the growth of public expenditure among others. The employment effect of the redeployment exercise was the loss of many public sector jobs. Boateng (2001) estimates that the number of public sector workers redeployed between 1987 and 1991 were 49,873 representing 15 percent of public sector employment. According to him, since in all cases, the principle of last-in-first-out (LIFO) was applied, redeployees were mostly young workers, labourers, cleaners, sweepers, messengers, drivers and other grades in the lowest echelons in the public sectors.

Essentially, it was expected that the loss of jobs as a consequence of the redeployment exercise would affect female workers much more than male workers considering the potentially higher concentration of women than men at the lowest echelon of the public sector. Indeed, the decline in the proportion of the female workforce in public sector from 4.9 percent in

ϱϯ

1984 to 3.6 percent in 19918 (see table 2.4) may be explained by the apparent huge job loss of women in the public sector as a result of the redeployment exercise carried out within the period. In addition, the decompression of civil service pay scales under public sector reform, that gave higher rewards to senior grades, may have widened gender wage differentials, given the concentration of women in lower grades (Gregory et al, 1992).

In spite of the seemingly adverse effect of SAP on the agenda of promoting gender equality, some of the interventions carried out to minimise the impact of the adjustment had a limited component targeted at ensuring gender equality in employment and incomes. One of such interventions was the Programme of Action to Mitigate the Social Cost of Adjustment (PAMSCAD). One component of the programme was the Enhancing Opportunities for Women in Development (ENOWID) that was targeted at promoting economic development of women. The ENOWID project which was the only component, mainly involved the disbursement of credit to SURPRWH ZRPHQ¶V VPDOO VFDOH HQWHUSULVHV WKURXJK WKH IRUPation of credit groups and revolving loan funds. Evaluation of this component of the PAMSCAD identified as a drawback the limited attention devoted to the production and marketing issues in the project, considering the fact that activities that were supported by the project were the usual traditional economic activities of women including food and fish processing, handicraft and trading.

In the mid to late 1990s, Ghana like many other developing countries put together strategies under the Poverty Reduction Strategy Programme (PRSP) supported by the World Bank and other Development Partners 8 The comparison of 1984 census estimates and the 1991/93 GLSS3 estimates is done advisedly and with caution. ϱϰ

(DPs) to address the poverty challenges in the country. The design and implementation of *KDQD¶VYHUVLRQRIWKLVSURJUDPPHWKH Ghana Poverty Reduction Strategy (GPRS) appear to have given little policy attention to gender issues. In the GPRS I (2003-2005), gender concerns were captured under the thematic area of vulnerability and exclusion. As argued by some analysts, such characterisation fails to adequately portray the important role played by women in society (Agboli, 2007).

The second phase of the GPRS dubbed Growth and Poverty Reduction Strategy (GPRS II) appeared to have better incorporated gender issues in the overall strategies. However, policy measures outlined towards gender equality have tended to emphasize participation of women in governance or political decision making. Indeed, policy strategy outlined in the GPRS II related to economic empowerment of women was the policy of bridging gender gap in access to education. This involved strategies towards attaining gender parity in access to education and meeting the objectives of Millennium Development Goals (MDG), Goal 3. This includes provision of LQFHQWLYHV RU VFKRODUVKLS VFKHPHV WR LQFUHDVH JLUOV¶ HQUROPHQW retention and completion of schooling, particularly in deprived areas; and sensitising SDUHQWVDQGFRPPXQLWLHVRQWKHLPSRUWDQFHRIJLUOV¶HGXFDWLRQ

In the MDG, the third goal seeks to promote gender equality and empower women in developing countries by the year 2015. The main target of the MDG3 is to eliminate gender disparity in primary and secondary education preferably by 2005, and in all levels of education not later than 2015. An official assessment (IMF EBS/06/72 p36 T9) and cited in Agboli (2007) indicates that as of 2005, the Gender Parity Index (GPI) had reached 93 females per 100 males at the primary level and 88 females per 100 males at secondary level. This indicates that the goal of totally eliminating gender disparity by 2005 could not be realised in Ghana. As reported in appendix ϱϱ

table 2a, the GPI at the primary level was 96 females per 100 males and 92 females per 100 males at the JHS level in 2010. At the SHS level, it rose by about 7 percentage points from 78 females per 100 males to 85 females per 100 males between 2005 and 2010. The GPI in primary and JHS in 2010 represents considerable improvement over the 1991 figures of 0.85 (or 85 females per 100 males) and 0.65 (or 65 females per 100 males) respectively (see appendix table 2a).

The GPI at the tertiary level reached 0.53 (or 53 females per 100 males) in 2005 from 0.3 (or 30 females per 100 males) in 1991. This suggests that the task of reaching GPI of 1.0 (equal number of females and males at the tertiary level by 2015 is a daunting one. The improvement in GPI at the tertiary level is largely on account of the increased numbers of private tertiary institutions and expansion of existing public universities in the country. In addition, WKHLPSOHPHQWDWLRQRIWKH³DIILUPDWLYH´DFWLRQpolicy LQWKHXQLYHUVLWLHV¶DGPLVVLRQSURFHVV in recent times has also contributed considerably to the improved GPI at the tertiary level.. Over the past decade and beyond, the University of Ghana, for instance, has consistently lowered the cut-off points by an aggregate score of 1 for females9. This has resulted in a consistent appreciation of the composition of females in total enrolment from 23.6 percent in 1991/92 academic year to 41.2 percent in 2008/09 academic year (see appendix table 2a).

As already pointed out, in 2010, there were over 50 private and 8 public universities accredited to run various degree programmes in Ghana besides the 10 polytechnics. In addition, all Teacher Training colleges have been upgraded to diploma awarding institutions. The combination of the above has implication for increased access of both men and women to tertiary 9 In 2008/2009 admission, the cut-off point for male applicants with General Arts from Senior High School into Bachelor of Arts degree programme was 15 against 16 for girls. ϱϲ

education in the country. One major concern however is the unequal distribution across programmes by sex. While the Sciences, Engineering, and to a lesser extent Business Administration programmes are dominated by males, females are highly represented in Fine Arts, Home Science and to some extent General Arts. From appendix table 2b, at least 58 percent and 82 percent of undergraduate students who obtained Bachelor of Fine Arts and Bachelor of Science in Home Science were females. In contrast, about 62 percent, 76 percent and 84 percent of students who graduated in Bachelor of Science, Bachelor of Science in Agriculture and Bachelor of Science in Engineering respectively in 2008 were males. Some notable positive outcomes of WKHLPSURYHPHQWLQZRPHQ¶VHGXFDWLRQ are the reduced adult illiteracy rate of women and increased women representation in wage employment. However, the improved education and employment performance of women over the years is yet to translate into increased wages or earnings given the widening female-male earning gap since 1991.

3.2

Legal Framework for Gender Equality

The growing international concerns about gender inequality in education, employment and incomes over the years is largely reflected in the signing of various international conventions, treaties and enactment of legislation to minimise the situation. The formulation and implementation of national policies essentially draw its strength from relevant legislations and conventions. Generally, policy measures aimed at ensuring gender equality are often designed and implemented within the relevant legal framework.

Over the years Ghana has ratified a number of international conventions and treaties designed for the elimination of all forms of discrimination against women especially in employment, remuneration and education. ϱϳ

One of such conventions is the ILO Convention 111 of 1958 ratified by Ghana in April 1961. This Convention enjoins member countries of the ILO to design and pursue national policies to promote equality of opportunity and treatment in respect of employment and occupation with the view to eliminating discrimination. The ILO Convention 100 of 1966 which was ratified by Ghana in March 1968 also seeks to ensure the application to all workers, the principle of equal remuneration for men and women workers for work of equal value. Ghana has also ratified the UN Convention on the Elimination of All Forms of Discrimination against Women (CEDAW) of 1979. This Convention which was ratified by Ghana LQ-XO\HQFRPSDVVHVVWHSVWRHQVXUHZRPHQ¶VHTXDOLW\RIHFRQRPLFLQ addition to civil, political, social and cultural life of the country.

There are a number of domestic laws and decrees enacted over the years that seek to promote economic empowerment of women and tackle the problem of widening gender inequality. The 1992 Constitution of Ghana outlines a number of provisions to deepen the advocacy for economic empowerment of women and promote gender equality. In article 36(6) of the Constitution, the state is required to afford equality of economic opportunities to all citizens; especially to take steps to ensure full integration of women into the mainstream of economic development of the country. Related to article 36(6) is article 17(4) that frowns on discrimination of all forms including gender.

A number of provisions that prohibit discrimination and seek to facilitate the economic empowerment of women and promote gender equality have also been outlined in the *KDQD¶V Labour Act 2003, Act 651. In section 57(1) of the Labour Act, women are entitled to three-month maternity leave. This is in line with Article 27(1) of the constitution that accords paid leave and provides special care to mothers during a reasonable period ϱϴ

before and after childbirth. The Labour Act also outlines a number of provisions to protect women from night work or overtime. Section 55(1) of the Act prohibits an employer from assigning or employing a pregnant woman to do work between 10:00pm and 7:00 am and forbids an employer from engaging for overtime, a pregnant woman worker or a mother of a child of less than eight months. The Act provides that women give their consent before they are assigned night and overtime work.

The open discrimination against women workers in the public sector during the colonial era partly informed the enactment of the 1967 Labour Decree which sought to grant equal rights of employment to women and men. Until 1963, women were forced to resign on pregnancy and were not eligible for entry to the administrative class in the civil service regardless of their qualification. The Labour Decree of 1967 granted women six weeks maternity leave with pay and that it was illegal to dismiss a woman who was absent on maternity leave. In 1971, women were granted three months fully paid maternity leave in the public service. This was expected to provide women the opportunity to have time off for nursing their infants and still maintain their jobs.

These legal measures are intended to promote employment of women, protect them from hazardous employment conditions and enhance their economic status in society. However, these measures may suffer some challenges in the quest to promote the quantity and quality of employment of women and their wellbeing in the labour market. For instance, some of these labour legislations, conventions and treaties are often difficult to enforce in the informal sector where women workers constitute the majority. In the formal sector where these legal measures are largely complied with, it tends to discourage employment of women particularly ϱϵ

by employers in the private formal sector. Invariably, the hiring decisions of private employers are often driven by profit maximisation motive.

Undoubtedly, the offer of paid maternity leave to women has cost implications for employers. This tends to compel employers in their recruitment decision to choose men over women regardless of their qualification. In a situation where women are even offered employment, they are more likely to be engaged at lower echelon of the job ladder with lower responsibility and pay relative to their male counterparts. In addition, women workers may suffer from slower career progression compared to male workers. For instance, a study carried out in Tema in the 1980s revealed that maternity leave provisions were seen by some employers as reason for not recruiting women (Date-Bah, 1986). In addition, women were more likely than men to be in junior positions, controlling for education level and that men were more likely to gain promotion. This may partly account for the declining average wage of women in most of the broad category of occupations since 1991 despite the rising proportion of women in paid-employment.

3.3

Institutional Arrangements and Women Economic Empowerment

Over the years, a number of organisations and institutions have been HVWDEOLVKHGWRSURPRWHZRPHQ¶VHPSRZHUPHQWDQGJHQGHUHTXDOLW\7KHVH LQVWLWXWLRQV DQG RUJDQLVDWLRQV DUH ZRPHQ¶V RUJDQLVDWLRQV ZLWK PRUH IRFXV on political participation and empowerment of women. Issues related to ZRPHQ¶V HFRQRPLF HPSRZHUPHQW DUH RIWHQ VXERUGLQDWHG WR RWKHU objectives. In 1975, the National Council on Women and Development (NCWD) was established with the mandate to play an advisory role in planning policies to facilitate the full integration of women in development. Some of the major activities of the Council that relate to economic ϲϬ

empowerment of women were income generating activities for women, dissemination of labour saving technologies for rural women and job counselling in schools for girls to shape their future career choice.

The operations and activities of NCWD were later overshadowed by the emergence of 31st 'HFHPEHU :RPHQ¶V 0RYHPHQW ':0 LQ ,Q 1986, the entire Council was dissolved by government and replaced by an Interim Management Committee largely made up of prominent members of DWM. Some of the fundamental objectives of the DWM were to improve WKH RYHUDOO VLWXDWLRQ RI ZRPHQ DQG WR IDFLOLWDWH ZRPHQ¶V DFWLYH participation in the political process. The Movement engaged in a number of economic activities and mobilisation efforts that gave some women some respite from economic challenges by creating the enabling environment for them to engage in certain economic activities however minimal (Allah-Mensah, 2005). 6RPHZRPHQ¶VJURXSVDIILOLDWHGWRWKH0RYHPHQWZHUHSURYLGHGZLWKJDUL processing machines and other supports to enhance their economic status. The Movement also established a number of day care centres in some cities, towns and villages and set up numerous small scale income generating activities and agricultural projects nationwide (Botei-Doku, 1990). The affiliation of the movement to the then ruling government of the Provisional National Defence Council (PNDC) created the general perception that it was a political organisation that sought to promote the political aspiration of a section of Ghanaians rather than seeking to promote economic empowerment of all women. ,QWKHJRYHUQPHQWHVWDEOLVKHGWKH:RPHQDQG&KLOGUHQ¶V0LQLVWU\WR formulate and facilitate the promotion of gender mainstreaming across all sectors of the economy towards the realisation of gender equality and ϲϭ

empowerment of women. The Ministry is also charged to design and initiate the implementation of policies towards the survival, growth and development of children and ensure the protection of the rights of women and children. As cited in Agboli (2007) the work of the Ministry includes advocacy for enactment of gender responsive laws and adoption of gender sensitive policies; consultations and partnerships with stakeholders for advocacy, gender training and sensitisation; collaboration with other public sector entities for gender-sensitive policy development, mainstreaming and SURJUDPPH LPSOHPHQWDWLRQ DQG VXSSRUW WR ZRPHQ¶V HFRQRPLF DQG political empowerment. Since its establishment, the Ministry is yet to make any significant impact in its mandate to ensuring economic empowerment of women through LQLWLDWLRQ RI SROLFLHV WR VXSSRUW ZRPHQ¶V HFRQRPLF DFWLYLWLHV 1RQHWKHOHVV RQH QRWDEOH SURJUDPPH IRFXVHG RQ ZRPHQ¶V economic empowerment has been targeted support to women in the business sector with access to finance and credit under a Japanese Grant ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ

ϲϮ

Chapter 4

Labour Market Discrimination by Gender: Issues and Types 4.0

Introduction

Gender differences in terms of employment, wages and education discussed in chapter 2 are driven by a combination of factors, prominent among them is gender discrimination in the labour market. In this section, we delve into the definitional issues and types of labour market discrimination dwelling on local and global examples.

4.1

Discrimination: Definition and Concept

The widely accepted definition of discrimination has been linked to Becker (1957) who found the phenomenon to exist when two equally qualified individuals are treated differently solely on the basis of characteristics that are irrelevant in a particular context. The phenomenon generally exists in the labour market if individual workers with identical productive characteristics are treated differently due to external characteristics such as sex, colour, race, ethnicity, beauty, religion, age, etc. Following Altonji and %ODQN ODERXU PDUNHW GLVFULPLQDWLRQ FRXOG EH UHIHUUHG WR DV µD situation in which persons who provide labour market services and who are equally productive in a physical or material sense are treated unequally in a ZD\WKDWLVUHODWHGWR«VH[¶10%\µXQHTXDO¶LWLVLPSOLHGWKDWWKHVHSHUVRQV receive different wages or face different demands for their services at a given wage. 10 See page 3168 of Handbook of Labour Economics, Vol. 3C ϲϯ

Early research work on labour market discrimination generally focused on the issue of race, but a discussion of the phenomenon in the 1960s and 1970s was extended to cover other areas including gender. Earlier researchers on gender issues focused generally on gender differences in employment. In the 19th century when few married women were engaged in market work, commentators such as Gillman (1898) and Engels (1884) concentrated on the gender division of labour itself and espoused the HPDQFLSDWLQJ HIIHFWV RI ZRPHQ¶V SDUWLFLSDWLRQ LQ WKH ODERXU IRUFH 2YHr time, interest of economists on gender issues including growth in womHQ¶V participation in the labour force, sex segregation of occupation and gender wage differences has grown.

The literature classifies economic discrimination into two main types: discrimination that takes place before the individual enters the market (premarket discrimination) and that which occurs within the context of the labour market (called current market discrimination). There is also another type of discrimination ± post market discrimination which has not been widely highlighted in the literature, in view of its limited significance in terms of economic efficiency loss or in the perpetuation of discrimination.

There is a two-way link between pre-market discrimination and current market discrimination. Essentially, pre--market discrimination tends to affect current labour market outcomes to the extent that it influences how workers from different groups (male and female) prepare to enter the current labour market. There is also a possible reverse effect such that the perpetuation of current labour market discrimination can reinforce premarket discrimination in the form of unequal access to physical and human resource endowment that prepares the individual for the current labour market. The sections that follow discuss in more detail, the three types of ϲϰ

labour market discrimination: pre-market discrimination, current market discrimination and post market discrimination.

4.2

Pre-Market Discrimination

Pre-market discrimination refers to the differential treatment accorded members of less favoured group (e.g. females) prior to their entry into the labour force. According to Boateng (2000), pre-market discrimination occurs when certain members of society, for example, parents and teachers, consciously or unconsciously influence the skill and interests of the other usually younger members of society in a way that later affects their human capital decisions. In the case of girls and boys, such decisions in turn generate differences in the future career commitments between the two sexes.

Pre-market discrimination can take different forms. The first form involves SDVW ODERXU PDUNHW GLVFULPLQDWLRQ WKDW WHQGV WR DIIHFW LQGLYLGXDO¶V FKRVHQ level of investment in human capital (see Benabou, 1994; Darlauf, 1992; Loury, 1977). Indeed, pre-market discrimination of this form tends to manifest as a confluence of school, home, neighbourhood deficiencies which cause females on average to be less well prepared and thus less marketable than their male counterparts.

Within the Ghanaian context, it is not uncommon to observe young girls being persuaded to end their education at the primary or at most secondary level on the grounds that higher education may cost them good husbands, or they may not be able to have children after a certain age. This attitude is reinforced by early childhood socialization that encourages female children to play with dolls while male children play with toy cars, plane toy tools etc. By implication, female children may grow up only looking forward to ϲϱ

becoming mothers, while male children aspire to become engineers and doctors.

Invariably, occupational choice in the labour market is also influenced by division of labour and notions of work in the household (Boateng, 2000). Roles that take children outside the homes are assigned to boys while girls are always encouraged to stay in around the kitchen. In similar vein, boys are often encouraged to acquire marketable skills with girls persuaded to engage in petty trading that requires limited skills and education thereby contributing to differences in labour market skills between boys and girls. In a typical third world household, significant differences may exist in SDUHQW¶VWUHDWPHQWRIVRQVUHODWLYH to daughters, while gender biases may be nonexistent in relatively wealthier European and North American households (Ray, 1998). It is argued that some traditional societies require sons to take care of parents in their old age, while daughters are married off early into a new marital home (Strauss and Thomas, 1995).

In Ghana, given the cultural and traditional practices that emphasisHJLUOV¶ commitment to housekeeping activities resulting in a lower school attendance rate and low educational attainment, their representation in highly skilled and well paid jobs would be low and this may cause widening wage disparity between men and women in Ghana (BaahBoateng, 2007). He observes that apart from biological factors which may prevent women from engaging in certain activities, the cultural and religious beliefs that have established some negative perceptions and discourage women from acquiring higher education to become more productive still persist. He concludes that until these negative beliefs and perceptions that are fuelled by cultural and religious entrenchment are removed, the gender gaps in terms of income, employment and education would be difficult to address in the country. ϲϲ

Osafo-Kwaako, (2001) however observed some positive changes in SDUHQWV¶ DWWLWXGH WRZDUGV JLUOV¶ HGXFDWLRQ $FFRUGLQJ WR KLP LQ UHFHQW WLPHV SDUHQWV¶ SUR-VRQ ELDVHV LQ WKH LQYHVWPHQW RI WKHLU FKLOGUHQ¶V education is changing in Ghana such that most Ashanti parents realise the best educational investment for their last penny should be put to their ³VPDUWHVWFKLOG´DQGQRWQHFHVVDULO\³DVRQ´+HKRZHYHUFRQFHGHVWKDWLQD hypothetical case where the academic performance of both boys and girls were at par, 98 percent of parents interviewed choose to invest in the education of their sons rather than their daughters. The reason is that cultural emphasis on marriage and motherhood often distracted young girls from completing secondary school.

The second form of pre-market discrimination may occur in other markets such as market for social services. For instance, the access to quality of schooling varies (Maxwell, 1994)11, and even the influence of parents in the labour market also varies. Indeed, discrimination in the acquisition of skills or human capital such that male students compared with their female counterparts are encouraged to pursue science and mathematics courses may largely account for the underrepresentation of women in such professions as medicine and engineering among others. In high school, boys are more likely than girls to take advanced mathematics and science courses while girls are more likely to take foreign languages (Marini, 1989; Fennema and Sherman, 1979). In the university, women enrol in very different majors than do men and while some of these sex differences have narrowed over time, large differences still exists (see, for example Boateng and Ofori-Sarpong 2001). Essentially, past discrimination, can also affect current behaviour relating to human capital investments and occupational 11 See also Neal and JohnsoQ DQG2¶1HLO ϲϳ

choice by denying the discriminated group (e.g. women) the necessary role models (see e.g. King and Knapp, 1978). Other sources of pre-PDUNHW GLVFULPLQDWLRQ VXFK DV SDUHQWV¶ HGXFDWLRQDO EDFNJURXQG RU DUHD RI UHVLGHQFH PD\ DIIHFW WKH LQGLYLGXDO¶V SHUVRQDO characteristics which eventually affect his or her access to the current labour market. Some empirical studies have provided evidence to link premarket discrimination to the current labour market outcomes. The effect of pre-market discrimination on the current labour market outcomes essentially occurs through prejudice or discriminatory approach to human capital investment which influences educational attainment of individuals, their occupational choice and eventually, their wages or earnings. As noted by Gunderson (1994) differences in human capital endowment such as education and experience which are caused by non-labour market factors and pure discrimination as well as dual labour market are critical factors that precipitate gender pay differences within the same occupation.

Carneiro et al (2005) demonstrated that factors determined outside of the market play a major role in accounting for wage differentials in modern labour market. They sought to re-examine the Neal-Johnson analysis (see Neal and Johnson, 1996) that endowments acquired before people enter the market explain most of the minority-majority wage gap. They show that cognitive test scores taken prior to entering the labour market are influenced by schooling, and adjusting the scores for racial or ethnic differences in education at the time the test is taken reduces their role in accounting for the wage gaps. Thus, for all members of the discriminated groups, but black males, adjusting for abilities that the discriminated group brings to the market eliminates the wage gap.

ϲϴ

Brown and Corcoran (1996) observed that differences in field of highest degree or college majors contributed to a significant part of male-female wage gap among college graduates in the United States. This is based on the fact that college majors influenced the kind of occupations and industries college graduates engaged in. They however, found little effect of high school courses on the wage gap due to the fact that the courses typically pursued by girls are just as wage-enhancing as those typically elected by boys. In addition, differences in test scores in high schools were observed to account for little of the wage gap. In effect, gender wage and productivity differentials would be difficult to address if unequal access of women to education which reflects pre-market discrimination is not resolved.

The role of pre-market discrimination in the determination of labour market outcomes in Ghana and other developing countries through educational effect on male-female wage gap has been established by a number of studies (see Verner 1999; Schultz, 2003; Kabubo-Mariara, 2003). Barr and Oduro (2000) observed large earnings differentials between ethnic groups in the Ghanaian manufacturing sector largely as a result of varying educational attainments and family background across ethnic groups. Using data drawn from the fifth wave of the Ghana Manufacturing Enterprise Survey (GMES) they concluded that differences in educational attainments and family background among ethnic groups account for a significant proportion of the wage differentials underscoring the importance of premarket discrimination in the Ghanaian context.

4.3

Market Discrimination

Given predetermined productive characteristics of the worker an attempt to treat some group of workers (e.g. women) differently from others (e.g. men) in terms of wages, responsibility etc. implies current market ϲϵ

discrimination. One fundamental concern about market discrimination is that, it has feedback effects on occupational and educational choices that underlie the acquisition of productive characteristics (Ehrenberg and Smith, 1997).

Market discrimination takes two main forms namely occupational and wage or earnings discrimination. Generally, occupational discrimination which takes the form of sex segregation of occupation exists if discriminated group (notably women) is confined to lower paying occupations or levels of responsibility without regards to their training and productive characteristics. In wage discrimination, employers are sometimes suspected of paying women less than men with the same level of education, training, and experience and working under the same conditions in the same jobs.

The two forms of market discrimination have common links such that the perpetuation of one tends to sustain the existence of the other. On one hand, segregation of women into low paying and less prestigious occupations and jobs may partly account for gender wage or earnings differentials. On the other hand, low wages or earnings of women emanating from discriminatory practices would determine their returns on education and FOHDUO\LQIOXHQFHLQYHVWPHQWGHFLVLRQRISDUHQWVRQZRPHQ¶VHGXFDWLRQDQG their occupational choice and hence contribute to the perpetuation of gender segregation of occupation.

4.3.1 Occupational segregation by Sex Occupational segregation by sex is the tendency for men and women to be employed or engaged in different occupations from each other and the phenomenon exists in virtually all countries (see, Blau, Ferber and Winkler, 1998). As already noted, occupational segregation does not necessarily ϳϬ

imply discrimination if it results from actual or perceived differences in VNLOOV RU HGXFDWLRQ DWWDLQPHQW DQG MRE VHHNHUV¶ SUHIHUHQFHV +RZHYHU WKH phenomenon may indicate discrimination if LW LV EDVHG RQ LQGLYLGXDO¶V characteristics such as sex, ethnicity, colour or religion which are largely unrelated to the productive characteristics of the individual.

After a period of declining segregation from 1870 to 1900 in the United States, the degree of sex segregation of occupation remained stable throughout the first half of the 20th century (see Gross, 1968, Jacobs, 1989, and Bertaux, 1991). It began to trend downwards from the 1960s and accelerated markedly downwards between 1970 and 1980 (see Beller, 1985; Bianchi and Rytina, 1986). The declining trend continued in the 1980s only at a slower pace (Blau, Simpson and Anderson, 1998) with the underlying shifts in the sex composition of occupations and in the occupation mix accounting for the reduction in segregation. Changes in sex composition within occupations largely on account of entry of women into traditionally male jobs constituted at least two-thirds of the decline in segregation.

A cross-national review of empirical studies of sex segregation of occupation by Anker (1995) shows different degrees of segregation across the globe. Blau and Ferber (1992) found the degree of sex segregation of occupation highest in Latin America, the Caribbean and Middle East using 'XQFDQ¶V ,QGH[ RI 'LVVLPLODULW\ with 1-digit occupational classification. The lowest degree of occupational segregation by sex was recorded in subSaharan Africa and East Asia. The highest occupational segregation was also reported in Latin America and lowest in Africa and Asia by Boulding (1976). A similar degree of segregation (lower than Latin America but higher than Africa and Asia) was observed in Europe and North Africa/Middle East. ϳϭ

In a study of 8 OECD countries, the lowest occupational segregation was also reported in Japan and highest in Australia and Norway (OECD, 1985) while among 23 OECD countries, Italy, Japan, Greece and Portugal saw lowest degree of occupational segregation by sex in the 1980s (OECD, 1988). Reubens and Harrison (1983) confirmed the lowest value of segregation index in Japan using 3-digit occupational classification with Sweden recording the highest segregation among 9 industrialised countries. A study of 45 countries across the globe by World Bank (1994) using 1 digit occupational classification reported highest degree of occupational segregation in North Africa and lowest in West Africa.

Apart from being a major source of labour market rigidity and related adversHFRQVHTXHQFHRQODERXUPDUNHW¶VDELOLW\WRUHVSRQGWRFKDQJHVVH[ segregation of occupation also has potential effect on male-female wage differentials. Gunderson (1994) identified a number of causes of malefemale pay differentials that are directly linked with occupational VHJUHJDWLRQE\VH[7KHVHDUHGLIIHUHQFHVLQSD\IRUZRUNRI³HTXDOYDOXH´ (resulting from the relationship between pay level in an occupation and the degree to which it is feminized), differences in job desired, and differences in job availability all of which are related to sex segregation of occupation. The persistence of occupational segregation has been identified as a major IDFWRU LQ ZRPHQ¶V FRQWLQXHG LQIHULRU HDUQLQJV DQG VWDWXV LQ HPSOR\PHQW (Blackburn, Jarman and Siltanen, 1993). This clearly contradicts the assertion by Jacobs and Lim (1992) that occupational segregation is mainly of interest as a subject because of its effect on male-female pay gap.

Empirical evidence suggests that between 12 percent and 37 percent of the gender wage gap in the United States can be explained by occupational segregation (Ehrenberg and Smith, 1997; Sorensen, 1990). Similarly, about ϳϮ

10 percent to 40 percent of gender wage differences in many countries are accounted for by occupational segregation (see Rimmer, 1991; Pike, 1982; Treiman and Hartmann, 1981). Groshen (1991) observed a much larger effect of occupational segregation on the gender wage gap using highly detailed occupational categories while Macpherson and Hirsch (1995) demonstrated that occupational segregation explains only 5 percent of the gender wage gap based on a fixed effect model. OECD (1985) claimed that the most important explanation for the continuing gender differentials in ZDJHV LV RFFXSDWLRQDO VHJUHJDWLRQ DQG ZRPHQ¶V FRQFHQWration in low paying occupations.

Boateng (1996) found an increase in the index of dissimilarity between men and women distribution of employment by sector in Ghana from 37 percent to 49 percent between 1975 and 1993 using Social Security and National Insurance Trust (SSNIT) data. Changes in gender composition and sectoral composition of employment accounted for 20 percent and 37 percent respectively of the increase in occupational segregation while the remaining 43 percent was due to the interaction between changes in sector composition and gender composition of employment. A low degree of segregation by employment type, occupation, industry, and employment sector was estimated in Ghana based on five different indices 12 using 2000 Population and housing census data (Baah-Boateng, 2007). The highest degree of segregation of 28.5 percent was recorded in the employment distribution by industry using size-standardized dissimilarity index. He however concedes that the generally low segregation indexes obtained was due to the limited disaggregation of the employment distribution.

12 7KH ILYH LQGH[HV XVHG LQ WKH PHDVXUHPHQW RI VHJUHJDWLRQ ZHUH 'XQFDQ¶V ,QGH[ RI dissimilarity, size-standardized index of dissimilarity, index of concentration, Karmel Maclachlan index and index of segregation. ϳϯ

4.3.2 Wage Discrimination Wage Discrimination is a major form of labour market discrimination and has become an important research focus in labour economics research. This is based on the fact that among the various labour market outcomes, wage is of fundamental importance as a key determinant of economic welfare for the employed and of potential gain to market employment for those who are out of work.

The residual wage gap as evidence of gender

discrimination in wages has been based on the premise that if after FRQWUROOLQJ IRU ZRUNHUV¶ SURGXFWLYLW\ D JDS SHUVLVWV WKHQ LW PXVW UHIOHFW discrimination. For example, Blau and Ferber (1987) found that even when refined measures of productivity-related characteristics are held constant, blacks and women in US earn less than whites and men.

Evidence of wage discrimination across countries is well documented in the literature. Blau and Kahn (2000) observed a gender pay gap tied to the gender division of labour indirectly through its impact on the strength of statistical discrimination against women in the United States. They KRZHYHU UHFRJQLVHG WKH GHFOLQLQJ SDWWHUQ RI ZRPHQ¶V SULPDU\ responsibility of house work and child care as families respond to rising labour market opportunities for women that increases the opportunity cost of home work. In recent times, there has been a decline in the size of the gender wage gap and thus a reduction in discrimination all over the world (see for example Goldin, 2002, Blau and Khan, 1997). This appears to confirm Becker (1957) claims that increased competition in the product market would reduce discrimination against women and minorities. This tends to suggest a positive correlation between market power and economic discrimination.

Nonetheless, the persistence of gender wage or earnings differentials with increasing globalization has led some researchers to conclude that it must ϳϰ

reflect unobserved productivity differences rather than discrimination )XFKV 2¶1HLO )RU LQVWDQFH *ROGLQ DWWULEXWHG WKH decrease in wage discrimination over a long run to the emergence of ³FUHGHQWLDOLVHG¶RFFXSDWLRQVWKDWFRXOGQRWEHSROOXWHGE\WKHSUHVHnce of women. Goldin further argued that some of the decline from the 1960s to the 1980s may have been due to anti-discrimination legislation and to an environment in which discrimination was less tolerated. Blau and Khan DOVR REVHUYHG WKDW ZRPHQ¶V gain in work experience and RFFXSDWLRQDOVWDWXVH[SODLQPXFKRIWKHLPSURYHPHQWLQZRPHQ¶VUHODWLYH wages. The improvement in observable characteristics and the reduced labour market discrimination against women have also contributed to the narrowing male-female wage gap.

A study of 15 Latin American countries revealed that on average discrimination accounts for about 88 percent of male advantage in pay after correcting for selectivity biases (Psacharopoulos and Tzannatos, 1992). In the East Asian labour market, Horton (1996) observed that generally, differences in returns to male and female characteristics account for not less than half the gap between male and female earnings, although this differential appears to be narrowing over time.

A link between gender wage differentials and discrimination has also been established in some African economies including Tanzania, Guinea&RQDNU\8JDQGD(WKLRSLD&RWHG¶,YRLUH.HQ\DDQG*KDQD,Q7DQ]DQLDQ manufacturing firms, Knight and Sabot (1982) found that 17 percent of gender wage differentials were attributable to factors other than observed characteristics. Glick and Sahn (1997) also observed that in GuineaConakry, differences in gender characteristics accounted for 45 percent of the male-female gap in earnings from self-employment and 25 percent of the differences in earnings from public sector employment. ϳϱ

According to Appleton et al (1999), differences between actual and pooled returns account for much of the gender wage gap in Ethiopia and Uganda, DQG UDWKHU OHVV LQ &RWH G¶,YRLUH EDVHG RQ 1XHPDUN GHFRPSRVLWLRQ technique. They concluded that in all three countries, the wage gap is narrower than it might otherwise be because women are overrepresented in the better-paid public sector. Sebaggala (2007) also concluded that discrimination against women in Uganda is more relevant than nepotism towards men in explaining gender wage gap. Kabubo-Marara (2003) found favouritism towards men to be more pronounced in private and public sectors with seemingly no evidence of discrimination against women in all sectors in Kenya although she found marked difference in the gender wage gaps.

In Ghana, lower wages or earnings of women relative to men have empirically been established by Beaudry and Sowa (1994). Using single wage equation approach and controlling for education, experience, occupation, household status, region of residence, industry and nationality, WKH\ IRXQG WKDW ZRPHQ¶V PRQWKO\ wages are 21 percent less than the DYHUDJH IRU PHQ LQGLFDWLQJ VRPH OHYHO RI GLVFULPLQDWLRQ 7KH ³ZDJH VKRUWIDOO´PHDVXUHGE\WKHHVWLPDWHGFRHIILFLHQWVRIWKHIHPDOHGXPP\LQ the wage equation for the various sectors were highest in the farming sector and lowest in the formal sector. They attributed the high wage shortfall for women in the farming sector in Ghana to the fact that majority of women are usually in subsistence production and argued further that the better performance of women in the formal sector in contrast is due largely to the influence of unions and wage policy in the public sector.

Using the Regional programme on Enterprise Development (RPED) dataset produced by the Oxford University, Verner (1999) revealed that on ϳϲ

average, females are paid 17 percent less than male workers in Ghana, controlling for differences in qualification and individual and other characteristics. In addition, the wage gap between men and women is notable at a general level in the manufacturing sector.

4.4

Post-Market Discrimination

Post-market gender discrimination of labour generally relates to unequal treatment of individuals in terms of government tax and transfer policy outside the labour market on the basis of sex. For instance, differences in tax credits and unemployment and pension benefits available to men and women outside the labour market particularly after leaving the labour market temporarily or permanently constitute post-market discrimination. In certain traditional areas of Ghana, customary practices that exclude women from inheriting men and from owning land may fall under this type of sex discrimination (Boateng, 2000).

Post market discrimination in the form of differences in benefits coverage could largely result from occupational segregation and associated wage differentials. In many countries the computation of pension benefits as a IXQFWLRQRIRQH¶V wage or earnings in the labour market may suggest that the segregation of women into low-paid jobs would cause gaps in pension benefits in favour of men. Investigation by Currie and Chaykowski (1992) revealed some gender gaps in benefits coverage due to occupational segregation. They disclosed that while workers in female jobs are more likely to have leave provisions, they are less likely to have pension coverage than workers in male jobs, even when entry-level wages and other characteristics of the jobs are controlled for. Related to this is the link of high poverty rates among elderly women due to lack of pension coverage (Galarneau, 1991; and Beller, 1981). ϳϳ

Chapter 5

Theories of Segregation and Discrimination in the Labour Market 5.0

Introduction

The question of why discrimination exists in the labour market, the sources and how it is sustained has been explained by several models or theories in the literature. The sources of labour market discrimination can conveniently be classified broadly into economic, institutional and noneconomic theories of discrimination.

5.1

Economic Models of Discrimination

The neoclassical framework of all forms of economic discrimination (including gender discrimination) upholds the rationality (i.e. profitmaximising) assumption on the part of both employers and workers. Thus, it is based on the efficient functioning of the labour markets. There are both demand and supply dimensions to the discussion of neoclassical theories of labour market discrimination. The book uses four theories to explain labour market discrimination from economic perspective. These are the human FDSLWDO PRGHO %HFNHU¶V WDVWH RU SUHMXGLFH K\SRWKHVLV PRQRSVRQ\ framework, and statistical theory.

5.1.1 Human Capital Model The human capital theory of labour market discrimination has supply and demand dimensions. The supply side of the model from a gender SHUVSHFWLYH HPSKDVLVHV ZRPHQ¶V ORZHU OHYHOV RI HGXFDWLRQ DQG ODERXU ϳϴ

market experience that essentially affect their occupational choice and earnings. Indeed, besides tending to have less education and less relevant fields of study women are also observed to often accumulate limited experience in the labour market. This is largely explained by the intermittent withdrawal from the labour market due to marriage and childbearing. By implication, women would rationally choose occupations that are flexible in terms of entry and working hours with relatively weak sanctions for temporary withdrawal. For example, considering the fact that occupations such as sales, production, agriculture and some services occupations are mostly are dominated by informality with high degree of flexibility, women are found to be highly represented.

On the demand side of human capital theory of discrimination, it is argued that the hiring decision of employers is often motivated by the concern to minimize employment costs, including training costs. The type of occupations or jobs offered to women by employers is often influenced negatively by the perception of women as higher cost workers relative to men due to their high absenteeism and labour turnover rates in the market. In addition, with relatively lower level of education and less relevant field of study for women, they are less likely to be offered occupations that require a relatively high level of education. Most often, education and experience have always been considered to be central in the signalling and selection of suitable applicants in the job markets. Thus the theory argues that women are not seen in many high status positions because they tend to leave jobs intermittently which prevents them from obtaining skills and relevant experience required to qualify for high-status and high-income position and occupations.

Essentially, the productivity-related variables of education and labour market experience influence both the occupational choice of women and ϳϵ

HPSOR\HUV¶ UHFUXLWPHQW GHFLVLRQ 7KLV LQ WXUQ FRQWULEXWHV WR WKHLU confinement to low status and less rewarding occupations and the creation or exacerbation of occupational segregation. In recent times however, there has been an increasing labour market experience of women (Anker, 1998). This has been attributed to the decline in household based work and increasing importance of female headed households that has the tendency to push more women to work continuously. Nonetheless, occupational segregation remains very high in many countries across the world (Anker, 1998) and this appears to contradict Baah-Boateng (2007) who found sex segregation of occupation to be low in Ghana.

A number of empirical studies have corroborated the role of human capital on female-male occupational choice decision and wage differentials. The role of human capital in the occupational choice decision of workers culminating in occupational segregation in the literature is established through the formulation and estimation of a multinomial logistic regression model that links occupational choice with education and labour market experience. The theory of human capital is largely the starting point in the analysis of occupational choice of households (Becker, 1991)13.

Indeed, differences in human capital endowments between men and women are found to a significant extent to account for occupational differences between the two sexes. Based on cross sectional data of farm households, Mwabu and Evenson (2008) found education as one of the key factors in the transformation of occupational structure in rural Kenya using maximum likelihood estimation of a logit model. Atieno and Teal (2006) observed that in Kenya, women are more likely than men to have a public sector job at very high levels of education while the probability gap of a private sector job between women and men widens as educational levels increase. Other 13 See also Schultz (1961) ϴϬ

studies elsewhere established that variables capturing human capital ± education and experience are important determinants of occupational choice and worker status in occupation (see Schmidt and Strauss, 1975; Nickel, 1982; Greenhalgh and Stewart, 1985). Empirical analysis is documented in Baah-Boateng (2009) and it is expected to be developed further in other forms.

Empirically, the gains made by women in education and experience worldwide in recent times have been observed to account for the narrowing male-female wage gap. For instance, Blau and Khan (1997) regressed wage on education and experience and estimated that gains in these variables reduced the wage gap in the United States by 7.6 percent. On their part, 2¶1HLOO DQG 3RODFKHN 3) found that one-third to one-half of the narrowing gender wage gap between the mid-1970s and the late 1980s in the United States was due to relative changes in schooling and work experience (see also Ashraf, 1996). Changes in experience have however been observed to be more important than changes in education in closing the male-female wage gap (Altonji and Blank, 1999). The increased accumulation of labour force experience by women on account of their increased labour force participation implies that the male-female wage gap is becoming increasingly narrow. According to Blau and Khan (1997) changes in accumulation of labour market experience of women have been far larger and explain a much larger share of the increase in female-male wage ratio than do changes in education.

In Kenya, besides other demographic factors, education is found to have a significant influence on the choice of sector of employment and earnings (Kabubo-Mariara, 2003). This led her to recommend that investment in instruments to reduce gender inequalities in access to education must be pursued. On their part, Appleton et al (1990) ascribed low participation of ϴϭ

ZRPHQLQWKHODERXUIRUFHLQ&RWHG¶,YRLUHRQWKHOLPLWHGDFFHVVRIZRPHQ to education. In addition, Schultz (2003) estimated that three-fifth of wage JDSLQ&RWHG¶,YRLUHDQGIRXU-fifth of the gap in Ghana was accounted for by differences in human capital endowment. Using the OLS estimation method, Schultz found a substantial wage differential associated with HGXFDWLRQ LQ &RWH G¶,YRLUH EXW D PRGHUDWe wage differential in Ghana. Schultz attributed this observation to both the relatively large supply of educated workers and the slow growth of national economy between 1960 DQGLQ*KDQDFRPSDUHGZLWK&RWHG¶,YRLUH+HREVHUYHGDUHODWLYHO\ larger educational effect on wages using Instrumental Variable estimation method treating education as the exogenous variable with height and bodymass index as endogenous variables.

Verner (1999) used the RPED dataset produced by Oxford University from the 1994 survey in Ghana to demonstrate that the more the years of education a worker has completed, the higher the wage the worker received conditional on a variety of individual and enterprise covariates, including occupation. His findings revealed that the size of the wage premium on education increases rapidly with completed level of education. These research outcomes are in line with those of Canagarajah and Thomas (1997) who observed that workers in Ghana who had completed tertiary education earned 2.7 times more than illiterate workers in 1991. In addition, Verner (1999) further indicated that education enhances firm level productivity in Ghana. The forgoing empirical evidence underscores the importance of human capital theory on gender differences in occupational choice which accounts for sex segregation of occupation and gender wage differentials across the world.

ϴϮ

5.1.2 Taste Hypothesis of Discrimination One famous model of discrimination that sought to explain the underlying reasons for the existence of discrimination within the market framework where firms operate as profit maximisers is the one developed by Becker ,QKLV ³7KHRU\ RI 7DVWH IRU 'LVFULPLQDWLRQ´KH DVVHUWV WKDW ZKHUH employers feel a disutility in hiring a less favoured group of workers solely EHFDXVHRIWKHZRUNHU¶VGHPRJUDSKLFFKDUDFWHULVWLFVZKLFKDUHLUUHOHYDQWWR WKH ZRUNHU¶V SK\VLFDO SURGXFWLYLW\ WKHQ HPSOR\HUV PD\ EH VDLG WR Ee prejudiced. Consequently, members of the less favoured group would be confined to less prestigious and low paid jobs and receive lower remuneration compared with the favoured counterparts on the basis of the HPSOR\HUV¶SUHMXGLFHDJDLQVWWKHP

According to Becker, discrimination can also originate from co-workers and customers particularly if the employer shares the same taste of prejudice or cannot do without the services of the prejudiced workers or customers. Clearly, customers or co-ZRUNHUV¶SUHMXGLFe also has the effect of causing wage differentials between members of the favoured and the less favoured group. For instance, employers who share the prejudice tendencies of co-workers and customers may be compelled to pay a wage premium to them to get them accept to deal with (work or buy from) the less favoured group. Applying the theory to racial discrimination, Borjas and Bronars (1989) established that minority self-employed workers earn lower incomes than white self-employed workers because white consumers dislike

purchasing

goods

from

self-employed

minority

workers.

Nonetheless, consumer-based discrimination generally plays a minor role in the differences in average wages received by race and sex groups. 7KHUH DUH LPSOLFDWLRQV RI %HFNHU¶V WDVWH K\SRWKHVLV RI GLVFULPLQDWLRQ IRU the profitability and efficiency of the discriminatory employer. Regardless ϴϯ

of the motivating factor of discrimination, the prejudiced employers incur additional cost for their taste of discrimination through measures taken to VHJUHJDWH WKH ZRUNHUV WR VDWLVI\ WKH FXVWRPHUV RU ZRUNHUV¶ WDVWH RI discrimination. The additional cost could also take the form of paying extra wages to the favoured workers to entice them to work with their less favoured counterpart. The discriminatory customer may also be compelled to pay a premium for his or her desire to indulge in discrimination. Becker observes that employers behave as if the price associated with hiring a worker from the less favoured group (females in this case) is their wage SOXVDQDGGLWLRQDODPRXQWZKLFKKHFDOOV³FRHIILFLHQWRIGLVFULPLQDWLRQ´

Furthermore, the existence of a wage gap between favoured and less favoured workers could be explained from three major sources: (i) the number of non-discriminatory employers; (ii) the extent to which employers discriminate; and (iii) the size of the population of the less favoured group. While being economically segregated, workers in the less favoured group would not suffer from economic discrimination if there were sufficient unprejudiced employers. Cain (1986) explained that if the unprejudiced employers hire all the minority workers (or workers in the less favoured group) and some of the non-minority workers (or workers in the favoured group) the wages of the two groups will be equal (Box 1).

The extent of wage gap between the discriminated or less favoured and favoured group also depends on the extent to which employers become discriminatory or exhibits their discriminatory tendencies. This is measured by the slope of the labour demand curve (Box 1) such that highly discriminatory employer depicted by a greater slope of labour demand curve results in lower wage of the less favoured group relative to wages of the favoured group. ϴϰ

Box 1: Relative Wages in Discriminatory Labour Market

ȱ

The horizontal portion of the labour demand curve represents demand by non discriminatory firm. Given the size of labour of less favoured group as S1 the non-discriminatory firms pay the same wage (i.e. 1.0) for all workers regardless of the differences in their non-market characteristics and employ E1 of the less favoured group. This is made possible because the size of labour demand of non-discriminatory firms is large enough to absorb all the workers in the less favoured group. If the size of labour supply of the less favoured group increases beyond the demand capacity of the non-discriminatory firms to S2, then only E0 of the less favoured group will be hired by nondiscriminatory firms leaving E2-E0 being hired by discriminatory firms at the relative wage (Wd/Wf) of 0.75. An increase in labour demand by non-discriminatory firms shown by an extension of the horizontal section of the aggregate labour demand FXUYHIURPSRLQW$WRSRLQW$¶UHGXFHVWKHH[WHQWRIDYHUDJHZDJHJDS (i.e. equilibrium relative wage greater than 0.75 and less than 1.0). This is because a reduced proportion of the less favoured group will receive the relative wage of less than unity. Similarly, the relative wage increases and the average wage gap declines if discriminatory firms become less discriminating with the labour demand curve rotating outwards about point A from AD to say AD1. ϴϱ

The extent of the relative wage between the less favoured and the favoured group can also be explained by number or the size of the population of the less favoured group (Box 1). Particularly, when the less favoured group constitutes a small fraction of the labour market, only a few firms without distaste for workers in the less favoured group are necessary to ensure that the market does not discriminate against the less favoured group.

Essentially, the cost of discrimination incurred by the prejudiced employer tends to have a negative consequence on the profit which could make its survival in a competitive market difficult in the long run (Box 2). Thus, the NH\ SUHGLFWLRQ RI %HFNHU¶V WDVWH K\SRWKHVLV RI GLVFULPLQDWLRQ LV WKDW WKH phenomenon would be eliminated in the long run due to the cost associated with discrimination. Clearly, with free entry and exit, discriminatory employers are expected to be driven out of the market in the long run by non discriminatory employers who are not affected by any additional cost for indulging in discriminatory practices. This suggests that discrimination is unlikely to persist in a competitive environment.

The recent narrowing of gender wage gap in an era of increased competition through international trade and deregulation might seem to RIIHUVXSSRUWLYHHYLGHQFHRI%HFNHU¶VDUJXPHQW$VREVHUYHGE\%ODFNDQG Brainerd (2004), the residual gender wage gap narrowed more rapidly in concentrated industries than in competitive industries. This was largely in response to trade shock indicating the implication of international trade and regulation on gender wage gap.

However, the existence of wage differentials between men and women in many countries and the survival of discriminatory employers in competitive economies such as the United States and other market oriented countries ϴϲ

Box 2: Relative Profit of Discriminatory and Non-Discriminatory firms

A discriminatory firm faced with a wage rate of w1 will hire Ed of the less favoured group rather than Ef. This is an inefficient choice which KDVDGYHUVHHIIHFWRQWKHGLVFULPLQDWRU\ILUPV¶SURILW In the absence of discrimination, the actual productivity of the worker will be represented by the MRPLA, the total revenue for the firm will be equal to the area OAGEf , and given the wage rate w1 the total wage bill will be equal to OW1GEf. The total profit will be equal to the area AW1G if Ef is employed at the wage rate w1. If as a result of discrimination the marginal productivity of the less favoured group is deliberately devalued and set at MRPL d, a discriminatory firm will employ only Ed. The discriminatory firm earns rent equal to the area AFCB for paying the less favoured group less than their marginal productivity. However, it still makes a loss equal to the area FCG and this is the price discriminatory firm pays to satisfy their taste for discrimination. ϴϳ

SRLQWV WR DQ REYLRXV ZHDNQHVV RI %HFNHU¶V K\SRWKHVLV 7KLV PD\ EH explained by the relatively lower share of women in the labour force which stands below 50 percent in many countries across the world. The persistence of discriminatory practices of some employers in an increasingly competitive and globalised world economy indicates that these employers probably are able to find means of covering the cost of discrimination (Boateng, 2000).

A study by Neumark (1996) and cited in Altonji and Blank (1999) appears to suggest that hiring discrimination continues to persist in the more FRPSHWLWLYH HFRQRP\ RI WKH 8QLWHG 6WDWHV WKHUHE\ XQGHUPLQLQJ %HFNHU¶V prediction of the elimination of discrimination through competition in the long run. In an audit study of sex discrimination in the restaurant industry in United States Neumark found men to be more likely to receive interviews and job offers in high priced restaurants with women more likely to secure jobs in low priced restaurants. Neumark, however, provided limited evidence to indicate that wages are higher in high priced restaurants and also that the relative probability that a male is hired in a high priced restaurant is positively related the percentage of men among the clientele.

5.1.3 Monopsony Framework The monopsonist model of discrimination is based on the proposition that workers are captive in a market where there is only one employer, or where a group of employers collude and act as one buyer (Robinson, 1934). This provides a consistent model for discrimination by postulating a more inelastic supply curve of labour for less favoured workers relative to the favoured group resulting in a greater gap of average and marginal labour costs between the less favoured and the favoured group. Consequently, ϴϴ

Box 3: Algebraic Analysis of Relative wages in the Discriminatory Monopsonistic Labour Market 1. Relationship between Marginal Labour Cost and Labour supply elasticity dTLC wL wL ww ww MLC w. L w L (5.1) dL L wL wL wL Where TLC=total labour cost; L = labour supply MLC=marginal labour cost; w=wage wL w es . (5.2) ww L Where es=elasticity of labour supply Taking an inverse of (5.2) and rearranging, we obtain ww 1 w 1 ww L . . (5.3) wL es L es wL w Substituting (5.3) into (5.1), we obtain

§ 1 w· 1 MLC w L¨¨ . ¸¸ w w e L e s © s ¹

§ 1· w¨¨1 ¸¸ © es ¹

(5.4)

2. Determining Relative Wages between favoured and less favoured groups We can write (5.4) for favoured group as

§ 1 ·¸ w f ¨1 ¨ e ¸ sf ¹ © and less favoured or discriminated group as MLC f

§ 1 · ¸¸ wd ¨¨1 e sd ¹ ©

MLCd

(5.5a)

(5.5b)

and equilibrium condition in the monopsonistic labour market as MRP

MLC f

MLCd

where subscripts f and d represents favoured and less favoured or discriminated group respectively § § 1 · 1 · ¸¸ Therefore MLC f w f ¨¨1 ¸¸ wd ¨¨1 © esd ¹ © esf ¹

wf wd

§ 1 ¨¨1 © ed § ¨1 1 ¨ e f ©

· ¸¸ ¹ · ¸ ¸ ¹

(5.6)

A greater elasticity of labour supply of the favoured group than the less favoured group implies higher wage for the favoured than the less favoured group i.e. esf ! esd w f ! wd . ϴϵ

since both groups are paid differently from their common marginal revenue product, a discrepancy arises between the less favoured and the favoured group. The differential wage between the two groups is the consequence of differing supply elasticities (see Box 3).

By implication, the existence of labour market discrimination may be sustained by some form of monopsonistic power for employers. As argued by Black (1995), when workers find it costly to search for a job and employers have a degree of monopsonistic power, it results in economic discrimination against less favoured group. He explained that while firms without prejudice would employ the discriminated workers, these firms tend to exploit the less attractive labour market alternative of these discriminated workers and thus offer them lower wages relative to members of the favoured group.

A major weakness of the monopsonist model is the limited empirical support for the prevalence of monopsony and lower-than-competitive wages (Bunting, 1962). An expanded geographic boundary of the labour market on account of increasing access to information technology and increasing degree of labour mobility do not make it easy to expect workers to accept wages lower than competitive wages (see Cain, 1986). Cain concedes though, that there are some workers who are trapped by a combination of industry-specific skills and a decline in the number of firms competing for their skills, and who suffer long-lasting exploitation. Nonetheless, these are not conditions that could be generalised to the entire labour market.

Additionally and more importantly, the assumption of a discriminated group having relatively less elastic labour supply has been difficult to back by empirical evidence. Regarding gender discrimination, there is a good ϵϬ

deal of empirical evidence and theoretical support for the findings of a greater elasticity for the labour supply of women than that of men and that this larger elasticity refers to the market rather than individual firms (Cain, 1986). He however argued that as a firm (or group of firms) becomes monopsonistic, then the distinction between the supply of a factor to the labour market and to the monopsonist firm tends to vanish. This implies that the larger labour supply elasticity of women in the labour market indicates larger elasticity to the monopsonist which is the reverse of the requisite condition for the existence of wage discrimination against women based on taste.

5.1.4 Statistical Theory of Discrimination The statistical model of discrimination based on imperfect information provides another reason for the persistence of labour market discrimination across the world. This model of labour market discrimination was pioneered by Phelps (1972) and Arrow (1973) on the basic premise that firms have limited information about skills and reliability of job applicants. Consequently, employers tend to use correlated and observable FKDUDFWHULVWLFV VXFK DV VH[ WR ³VWDWLVWLFDOO\ GLVFULPLQDWH´ DPRQJ ZRUNHUV Thus, in the traditional statistical models of discrimination, less favoured workers and their favoured counterparts have the same distribution of abilities, but employers are more accurate judges of the talents of favoured workers than less favoured workers. To the extent that women for example are observed to be less meritorious than men in the market place, statistical discrimination is essentially a shortcut mechanism to minimize cost and/or risk while screening for better qualified workers (Tuma, 1995).

From the optimal statistical rule, less favoured workers (e.g. women) with above-average ability are likely to earn an expected wage below that of favoured workers (e.g. men) of similar ability. Such information ϵϭ

asymmetries may result in less favoured or discriminated workers investing less in human capital which would generate wage differentials between favoured and less favoured workers with similar abilities (Lundberg and Startz, 1983). Arrow (1973) and Coate and Loury (1993) found that due to feedback via, for example, investment in human capital, biased stereotypes might be self confirming. Statistical discrimination of this form in the Ghanaian context could be causing both current labour market discrimination and feedback effects and promoting or worsening gender wage differentials.

Another form of statistical discrimination concerns the effect of group differences in the precision of the information available to employers about individual productivity on the basis of the quality of the match between worker skills and job requirement (Aigner and Cain, 1977). Those groups for which more precise information is available earn a premium. However, the premium is eroded by worker experience if no underlying productivity difference is found as initially disadvantaged workers gain more exposure to the market. As noticed in competitive markets, labour market discrimination has also been described as unstable in markets with asymmetric information. Cain (1986) argued that wage differentials due to information as\PPHWULHV ZRXOG QRW ODVW EHFDXVH HPSOR\HUV¶ XQFHUWDLQW\ DERXW ZRUNHU¶V SURGXFWLYLW\ PD\ EH LQH[SHQVLYHO\ UHGXFHG E\ REVHUYLQJ WKHZRUNHU¶VSHUIRUPDQFHRQWKHMRE+RZHYHUDVDUJXHGE\0LOJURPDQG Oester (1987) where firms have no incentive to reveal that information, such wage differentials may persist.

One fundamental concern about statistical discrimination is the possible efficiency loss that may arise. As Lundberg (1991) pointed out, the failure of firms to use group specific equations to estimate the productivity of an individual will reduce accuracy of their estimated productivity. Hence, if ϵϮ

output is significantly influenced by the quality of the match between the job and the worker, then the reduced accuracy may result in efficiency loss. Invariably, the accuracy of skill assessment may be undermined by cultural and language differences and more importantly sex differences as well as social networks that tend to run along gender and racial lines. Obviously, groups that are poorly represented in higher level positions may be at an information disadvantage (Montgomery, 1991). In effect, labour market discrimination against women arising from wrong assessment of skills and competence of male and female job seekers by employers on account of wrong judgment and beliefs may undermine efficiency and create social inequality.

The theory of statistical discrimination appears to ignore the role of occupational

segregation

by

sex

in

perpetuating

labour

market

discrimination into the next generation. As cited in Anker (1998), because women are discriminated against, they are likely to obtain less education than men and to pursue work careers that reinforce this current situation. In addition, the theory is less relevant in explaining discrimination in promotion as compared to recruitment, since information costs for the former may be less than for the latter in many enterprises.

5.2

Institutional Theories of Discrimination in the Labour Market

The underlying proposition of the institutional models of labour market discrimination is that labour market opportunities and outcomes are not fixed by market forces alone since the market itself is not competitive as assumed in neoclassical models. Rather, they are fixed by institutional forces such as gender stereotyping and differences in child experiences DPRQJRWKHUV)RUH[DPSOHFKLOGKRRGH[SHULHQFHWHQGVWRVKDSHZRPHQ¶V ϵϯ

beliefs that make their life in paid employment is potentially short and unstable. Hence, being rational, they tend to choose training and occupations that have high transferability of skills and experiences to the home environment. According to the institutional model, the structural condition for discrimination is the socialisation process. The institutional theories of labour market discrimination also rely on well established economic thought and neoclassical logic and recognise the role of institutions (e.g. unions and large enterprises) in determining hiring, firing, promotion and payment of wages.

The institutional approach sometimes cuts across several disciplines including psychology and theories of adaptive behaviour (Cain (1986). The initial placement of disadvantaged workers into low-wage, low-status jobs creates attitudes and habits that perpetuate their status (Piore, 1970). Arrow (1973) provided a related model in which the psychological theory of cognitive dissonance rationalises market exchanges that result in a suboptimal equilibrium. Essentially, expectations are formed by employers about the inferiority of the discriminated group which cause an internalisation of these expectations resulting in underinvestment in human capital as a confirmation of those expectations.

An objection to this argument stems from the desire of members of the discriminated group to reverse the expectation and the interest of employers to prefer to augment the supply of labour by encouraging more investment in human capital and positive attitudes towards investment and work (Cain, 1986). Incidentally, there are many models within the institutional framework that explain the persistence of labour market discrimination in many countries. In this book, we review two related institutional models namely segmentation theory and crowding hypothesis that underpin labour market discrimination. ϵϰ

5.2.1 Labour Market Segmentation Theory The dual labour market theory, one of the best known segmented market theories by Doeringer and Piore (1971) distinguishes between the ³SULPDU\´ VHFWRU DQG WKH ³VHFRQGDU\´ VHFWRU 7KH MREV LQ HDFK VHFWRU DUH supposed to have their own unique characteristics. In the secondary sector, jobs tend to be relatively poor quality, with low pay, poor chances for promotion, poor working conditions and little protection or quality. Jobs in the primary sector on the other hand, are found to be relatively good jobs in terms of pay, job security, opportunity for advancement and working conditions. The primary-secondary sector segmentation is synonymous to other labour market segmentation theories that divide the labour market into formal and informal economy (ILO, 1972) and progressive and static segment (Standing, 1999).

Adapting the segmentation theory to occupational segregation by sex with RQH VHJPHQW FRPSULVLQJ ³IHPDOH´ RFFXSDWLRQV DQG DQRWKHU ³PDOH´ occupations could provide an explanation for the persistence of sex segregation of occupation and gender wage differentials in many countries. As an illustration, more men have historically been engaged in market work than women. For example, in the teaching profession, women dominate the lower levels of education while men are highly represented at the tertiary level and assume higher managerial responsibility in the sector. 7KLV PD\ ODUJHO\ DFFRXQW IRU PHQ¶V GRPLQDQFH LQ WKH LQWHUQDO ODERXU markets and ports of entry into careers with good promotional systems.

Clearly, since jobs in the primary sector are described to be more secure, with firm-specific experience and low labour turnover, male employers would tend to be favoured by employers in that sector. In addition, the higher wage paid by employers in the primary sector would enable them to attract best qualified workers which may put women at the disadvantage ϵϱ

due to their relatively low productive characteristics in terms of education and labour market experience. This suggests that women would be underrepresented in the primary sector and over represented in the secondary sector with the potential of perpetuating discrimination against them. According to Boateng (2000) women are discriminated against in this respect because they tend to have unstable work histories, but these histories are themselves as a result of being unable to break into the primary labour market.

Related to labour market segmentation is statistical discrimination theory described as one of the economic theories of discrimination. Essentially, due to differences in productivity, skills, experiences etc between men and women vis à vis high search and information cost of identifying whom to hire, rational employers discriminate on the basis of information available to them thereby promoting discrimination. Thus, statistical discrimination theory provides an explanation for how an entire occupation can consist of mostly male even though many women have more ability, education and experience than many men (Anker, 1998).

Really, the dual labour market model of the segmentation theory offers useful explanation to the existence and persistence of discrimination in the labour market worldwide. However, it does not explain what initially caused the discriminated group to be confined into secondary occupations. As Anker (1998) pointed out, the labour market segmentation theory is rather better at explaining vertical occupational segregation by sex (i.e. why men are more likely than women to have better quality jobs in the same occupation, which is a major source of female-male wage differentials) than horizontal occupational segregation.

ϵϲ

5.2.2 Crowding Hypothesis The crowding hypothesis is one of the oldest theories of gendered pay differentials. The crowding hypothesis of labour market discrimination could be deemed as drawing its strength on segmentation of occupations such that it becomes difficult for females to gain access to male occupations. This model postulates that discrimination against women occurs through the exclusion of women from occupations considered ³PHQ¶VZRUN´,QGHHGVLQFHWKHVHMREVDUHUHVHUYHGIRUPHQUHODWLYHO\IHZ women are hired into these positions.

The theory was first propounded by Millicent Fawcett in 1918, and restated in neoclassical form by Edgeworth (1922). Social pressures that emphasize WKH WUDGLWLRQDO UROH RI ZRPHQ PD\ DIIHFW WKH FKRLFHV RI ZRPHQ¶V occupation. In this case, many women may choose traditional female jobs since information on them is easily available. By implication therefore, the VXSSO\ RI ZRPHQ LQFUHDVHV ³ZRPHQ¶V ZRUN´ ZKLFK LQ WXUQ UHGXFHV WKHLU wages. That is, crowding of women into certain occupations tends to compress the wages to a level below that of similar qualified workers in other occupations. Thus, the key prediction of the crowding model is that because of discrimination women and men are segregated into different occupations and that those engaged LQ³ZRPHQ¶VZRUN´HDUQOHVVWKDQWKRVH GRLQJ³PHQ¶VZRUN´HYHQWKRXJKDOOZRUNHUVDUHHTXDOO\ZHOOTXDOLILHGIRU both jobs.

The model implicitly assumes equal abilities of women and men that could make them receive equal pay in the absence of discrimination. The wage differential created by the crowding is worsened by the fact that the interoccupational mobility is not enough to equalize wages. Clearly, the feminization of certain occupations and lower relative wage for women DOVR UHIOHFWV DQ RFFXSDWLRQDO ³FURZGLQJ´ VXFK WKDW ZRPHQ FDQ EH ϵϳ

concentrated in particular occupations based on their preferences or on past and current barriers to alternative occupations (Bergmann, 1974). There is HYLGHQFHWRVXJJHVWWKDWZRPHQ¶VRZQSUHIHUHQFHVIRUVSHFLILFMREVRUWDVNV promote sex segregation in occupation. Gunderson (1989) admitted that there is considerable evidence to support the belief that gender differences in preferences play some role in gender differences in occupations and this tends to cause further widening of male-female wage gap in the labour market. 7KH K\SRWKHVLV GXEEHG ³TXDOLW\ VRUWLQJ´ UHODWHG WR WKH H[SODQDWLRQ RI the wage-JHQGHU FRPSRVLWLRQ VXJJHVWV WR VRPH H[WHQW WKDW ZRPHQ¶V occupational crowding may be derived from quality of labour. The quality sorting model asserts that if women rather than men are concentrated in low rewarded occupations as a result of discriminatory barriers, the gender composition of an occupation turns to be a lower quality index for men and to lesser extent for women. Women may have different perspectives over the life cycle since they tend to be more concerned with home production than men.

Becker (1983) developed a model to suggest that women seek jobs with attributes consistent with household production where they have a comparative advantage and also household chores where they bear primary responsibility. These perspectives affect ZRPHQ¶V HGXFDWLRQDO DWWDLQPHQW and training on the job. Since the expected short duration of attachment to the firm lowers the return to investments, women may invest less. In effect, low-rewarded occupations with high concentration of women over time would attract low-skilled men and would lose high skilled women. In this case, the occupational crowding could be explained from the angle of labour quality, resulting in lower wages. This explanation is more DSSOLFDEOH WR PHQ¶V ZDJHV VLQFH WKH\ ZRXOG KDYH PRre choices in the ϵϴ

selection of occupations. As a result all workers in highly female concentrated occupations would have lower average wages and productivity.

The basic limitation of this model is that it does not have a complete formulation of the discrimination process (Sorensen, 1990). For instance, how occupations become female or male dominated is not addressed by the model. According to Macpherson and Hirsch (1995), although, the crowding model is useful in explaining the negativity of female composition effect on female wages or earnings, it cannot equally explain why wages of men in predominantly female jobs are relatively low. This is because on the basis of the argument that men do not face similar barriers14 facing women there is no justification for men to accept lower wages in predominantly female occupations when higher wages are available in predominantly male occupations. That is, given that women face barriers to better rewarded occupations, low-remunerated occupations would attract a disproportionately large number of women against a small proportion of men to generate a negative correlation between

female occupational

composition and male and female wages. The crowding hypothesis cannot also explain why competitive pressures do not erode discrimination in the long run. Nonetheless, Sorensen (1990) agrees that the crowding hypothesis does provide a link between occupational segregation and lower wage in female dominated jobs.

5.3

Non-economic Models of Discrimination

There are two fundamental theories of gender discrimination of employment and wages that are equally relevant and fall outside the economic and institutional framework of discrimination. These are the noneconomic theories of labour market discrimination which are often taken as 14 These include cultural, biological, social and religious barriers. ϵϵ

given in the economic and institutional analysis of gender discrimination in the labour market. In this section, we review two non-economic models of labour market discrimination: WKH ³SDWULDUFK\ K\SRWKHVLV´ DQG IHPLQLVW model.

5.3.1 Patriarchy Argument 7KH³SDWULDUFK\K\SRWKHVLV´RIODERXUPDUNHWGLVFULPLQDWLRQSRVLWVWKDWRQ average men are more highly motivated than women to achieve and maintain positions of high status (Heath and Ciscel, 1988). The basic patriarchy model argues that decision-making vis à vis the assignment of production responsibilities and the distribution of consumption goods within the household are dominated by roles in the households. Most often, jobs held by women are low in the hierarchy of jobs. In addition, industries dominated by women tend to be those at the end of the hierarchy of industries in terms of wages, social importance and power (Bergmann, 1989). These may stem from the desire of men to assert their authority in VRFLHW\ RYHU ZRPHQ DQG PD\ LPSHGH ZRPHQ¶V HQtry into occupations which are seemingly dominated by men.

In explaining possible reasons why men may oppose women entry into their occupations, Cohn (2000) argued that men view work as an expression of masculinity. Consequently, the introduction of women may be perceived by men as a threat to their identity as men by invading their province that defines their gender (Hartmann, 1976; Rose 1992). Goldin (2002) also developed a model that appears to blame men for discrimination against women in support of Cohn (2000). The model treats discrimination as the consequence of a desire by men to maintain their occupational status or prestige, distinct from the desire to maintain their HDUQLQJV $FFRUGLQJ WR *ROGLQ ³PHQ¶V ZRUN´ LV SHUFHLYHG DV EHWWHU WKDQ ³ZRPHQ¶VZRUN´DQGREVHUYLQJDZRPDQGRLQJDPDQ¶VMREVLJQDOOHGWKDW ϭϬϬ

WKHPDQ¶VMREKDVEHHQGRZQJUDGHG7KLVLQDVWDWLFFRQWH[WFRXOGEHDQ instrument through which occupational segregation between men and women is institutionalized.

Cockburn (1988) also argued that men would have no objection to women working but would object feminization of manly skills that defines their special masculine competence. These assertions according to Williams (1989) are however flawed on the grounds that introduction of women into armed forces in many countries over the years has had almost no effect on the masculine image of the modern soldier.

5.3.2 Feminist Model Related to the patriarchy argument of gender discrimination is the feminist model that emphasises the perpetuation of discrimination against women in the labour market on account of the oppression of women and male domination in society. A basic premise of the feminist theory is that ZRPHQ¶V GLVDGYDQWDJHG SRVLWLRQ LQ WKH ODERXU PDUNHW LV D UHIOHFWLRQ RI ZRPHQ¶V subordinate position in society and the family. In all societies, home production and child care are perceived as the main responsibility of women. On the other hand, men are seen as mainly the breadwinners. This division

of responsibilities

influences the

determination

of less

accumulation of human capital of women compared to men in the prelabour market. Thus, women are seen as having less need for the labour market and that explains why female children tend to receive less education and are less likely to pursue fields of study that are considered relevant for the labour market (e.g. sciences and technology than male children).

Goldin (2002) asserted that the institutionalization of occupational barriers, as was the case for firms hiring office workers in the 1930s contributed to the existence of lags in the labour market. Reskin and Hartman (1986) also ϭϬϭ

attributed gender discrimination in society to the institutional barriers that have historically excluded women from particular pursuits or impeded their XSZDUG SURJUHVVLRQ 7KH H[LVWHQFH RI ³PDUULDJH EDUV´ KDV WHQGHG WR prevent married women from holding certain jobs and some employer policies in the past have excluded women from certain jobs (Goldin, 1990).

Undoubtedly, feminist or gender theory provides valuable explanation to occupational segregation by sex and gender wage differentials by pointing out how closely the characteULVWLFVRI ³IHPDOH´RFFXSDWLRQVFRUUHVSRQGWR typical stereotypes of their women and their supposed abilities (Anker, 1998). According to Anker, the positive stereotype of women (e.g. caring nature, skill and experience at household-related work, greater honesty, physical appearance and manual dexterity) makes them better qualified for occupations such as nursing, doctor, teacher, social worker, maid, housekeeper, cleaner, cook, waiter, launderer, hairdresser, spinner, weaver, knitter, dressmaker, midwife, salesperson, accountant, receptionist, shop assistant.

In contrast, five negative stereotypes (e.g. disinclination to supervise others, less physical strength, less ability to do science and mathematics, less willingness to travel, and less willingness to face physical danger and XVH SK\VLFDO IRUFH DGYHUVHO\ DIIHFW ZRPHQ¶V DFFHSWDQFH RI YDULRXV occupations which consequently makes these occupations become male occupations. These stereotypes make it difficult for women to qualify for such types of occupations as managers, supervisors, construction workers, miners, well drillers, engineers, architects, pilots, ship officers, firefighters, police and security officers etc. Essentially, as argued by the Anthropo-biological school, the process of social reproduction, the breederfeeder role of women, which prevents women from specializing in production roles or engaging in physically strenuous, more demanding and ϭϬϮ

risky but high-paying occupations underscores the poor performance of women in the labour market. Limited policy attention to reverse the ideological differentiation between women and men will continue to undermine the effort to addressing gender discrimination in the labour market across the globe particularly in Ghana and Africa.

5.4

Relevance of the Theories in Ghanaian Context

The nature and characteristics of the labour market in developing countries bring to the fore the applicability of economic models of labour market discrimination in the Ghanaian context. The economic models of discrimination largely consider the employer as the principal source of discrimination against women and other groups in the labour market. The prejudice hypothesis of labour discrimination originates primarily from the employer and in some instances co-workers and customers especially if the prejudicial tendencies of the two are in line with that of the employer (Becker, 1957). The monopoly power of an employer or group of employers in collusive cartel tends to form the basis of discrimination against women or black race in terms of wages and occupational status under the monopsony model of discrimination. Similarly, discrimination ZLWKLQWKHVWDWLVWLFDOIUDPHZRUNUHVXOWVIURPHPSOR\HUV¶GHFLVLRQWRDVVHVV the productivity and to offer jobs to women and other discriminated groups on the basis of the characteristics of the group that they belong to.

From the arguments outlined above, it is quite clear that economic models of labour market discrimination are largely applicable to paid-employment which is common in developed economies and where the employer and the worker are two separate agents. In many developing countries however, this type of employment accounts for a smaller proportion of total employment. In Ghana for instance, paid employment constituted about 16 percent of total employment in 2005/06 implying that economic models of ϭϬϯ

discrimination can explain less than one-fifth of discrimination in the entire labour market.

Obviously, the dominance of self-employment in many developing countries including Ghana makes it difficult to explain labour market discrimination within the framework of economic models that blame discrimination largely on the hiring decision of employers. In selfemployment which does not differentiate the employer from the employee, there is limited basis for employer discrimination to persist. Thus, unlike the paid-employed worker, the self-employed worker is not subject to employer discrimination in hiring and promotion that culminates in the perpetuation of sex segregation of occupation and wage gap between two demographic groups (male and females). Borjas and Bronas (1989), note that the larger self-employment rates and incomes of whites than blacks is difficult to interpret in the traditional framework of an employer discrimination model since self-employed persons have no reason to discriminate against themselves.

One notable economic model of discrimination applied to gender differences in self-HPSOR\PHQW DQG QRW UHODWHG WR WKH HPSOR\HUV¶ discriminatory behaviour though, is customer discrimination proposed by Becker (1971). In customer discrimination, differences in earnings between self-employed men and women may result from the tendency of customers to dislike purchasing goods and services from women or other less favoured group. As noted by Moore (1983) FRQVXPHU¶V GLVFULPLQDWLRQ may influence female-male wage ratio in self-employment as in paid HPSOR\PHQW VXFK WKDW D FRQVXPHU¶V HYDOXDWLRQ RI SHUVRQDO VHUYLFHV purchased may depend on price, reliability and race or sex of the person providing such service. ϭϬϰ

The model of customer discrimination is however based on assumption of perfect information where consumers without any cost are aware of the sex or race of the seller and the price of the good being sold. The existence of imperfect information and the cost associated with obtaining information may undermine the discriminatory behaviour of the consumer thereby reinforcing

the

challenge

of

explaining

occupational

and

wage

discrimination among self-employed workers within the framework of economic theories of discrimination.

Gender differences in occupational distribution and earnings in selfemployment may also be explained by gender differences in human capital endowment as discussed under economic framework of discrimination but unrelated to employer discrimination. Relatively limited human capital endowment of women that tends to obstruct their access to some lucrative offers in wage employment raises the incentive to engage in some occupations as self-employed. This is augmented by the disadvantaged worker argument that claims that workers without an attractive mix of human capital become self-employed where they are unable to obtain a wage job. This may explain the relatively large number of women in selfemployment in Ghana and other developing countries with its negative wage effect.

Furthermore, the relatively low educational attainment and limited accumulation of labour market experience of women force them into low rewarding occupations thereby contributing to lower wages of selfemployed women relative to men. Some research works suggest that highly skilled women receive greater returns to human capital in self-employment than wage employment (Devine, 1994). This then underscores the need to bridge gender wage gap through the promotion of female education and training. In addition, there is a stronger positive effect of experience on ϭϬϱ

earnings among self-employed women (Williams, 2000). However, the generally lower labour market experience and education of self-employed women deprive them of this opportunity to close the wage gap between them and self-employed men.

Some reasons assigned to gender differences in self-employment in terms of occupational status and earnings can best be situated under institutional and/or non economic theories of discrimination. For instance, institutional ERWWOHQHFNV WKDW OLPLW ZRPHQ¶V DELOLW\ WR RZQ RU LQKHULW DVVHWV LQFOXGLQJ land or obtain credit from financial institutions may impede their access to capital intensive occupations and industries culminating in occupational segregation. In effect, the resulting concentration of women in less profitable occupations in self-employment may reduce their earnings and worsen the gender wage gap.

The overconcentration of women in self-employment for various reasons seems to be in line with the crowding hypothesis that attributes gender wage differentials to crowding of women into certain occupations that cause their earnings to decline. Women may prefer to engage in occupations mainly organised on self-employment basis because it offers them a flexible work strategy to combine competing responsibilities from employment and families. Indeed, the perception of women about selfemployment as a work and family balancing strategy implies that they may be more willing to engage in less lucrative self-employment occupations than men in exchange for benefits derived from undertaking family responsibilities. Thus, men may enter more lucrative jobs and sectors at a higher rate than women as a result of the combination of homework making women earn less than men. In addition, gender wage differentials may also result from the more discretion associated with self-employment ϭϬϲ

that makes self-employed women work fewer hours and put in less effort than self-employed men to accommodate family demands.

Clearly, the discussion of the gender perspective of labour market discrimination in Ghana and Africa using economic models that attribute discrimination to the employer may ignore an important and substantial segment of the market where gender discrimination is explained largely by institutional and other non-economic factors. To better understand gender discrimination in the Ghanaian labour market, it is important that the analysis of gender discrimination does not only consider the labour market as one unit but also discuss the issue from the perspectives of paid employment and self-employment separately. This argument which is one of the prime focus of this book is strongly supported by research conclusion reached by Glick and Sahn (1997) and echoed by Atieno and Teal (2006) that the divide between self-employment and wage activities is where the role of gender is most important in terms of occupational choice. ȱ

ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ȱ ϭϬϳ

Chapter 6

Measuring Sex Segregation of Occupation in Ghana 6.0

Introduction

In chapter two, the study discussed the distribution of occupation by sex in the Ghanaian labour market. This chapter attempts to measure sex segregation of occupation as a means of assessing the extent and changing pattern of unequal distribution of occupation between the two sexes in the labour market. The study adopts four widely used segregation indices in the literature and draws on four population censuses over 1960-2000 and datasets of three nationally representative household surveys between 1991 and 2006 for the computation of the indices.

6.1

Sex Segregation of Occupation

Occupational segregation refers to the unbalanced distribution of the sexes across occupations in a manner inconsistent with their overall shares of employment, irrespective of the nature of job allocation (see Watts and Rich, 1992; and Jonung, 1982). The term segregation has sometimes been used interchangeably with concentration but the two terms are not the same in the strict sense. Segregation concerns the tendency for men and women to be employed in different occupations from each other across the entire spectrum of occupations. On the other hand, concentration is concerned with the sex composition of the workforce in an occupation or set of occupations. The two concepts differ from the concept of dissimilarity which describes the difference in the proportions of female and male workforce in a particular occupation. ϭϬϴ

Several

authors

have

argued

that

occupational

segregation

and

concentration of women in low-paying occupations largely account for the continuing gender discrepancies in wages15. Empirical studies on gender wage differences in some countries such as Australia and the United Kingdom have found that occupational segregation accounts for between 10 to 40 percent of gender wage differences (Pike, 1982 and Rimmer, 1991).

Occupational segregation can be classified into two: vertical occupational segregation and horizontal occupational segregation (Hakim, 1979). Vertical segregation exists when men are working in higher level jobs and women engaged at the lower level jobs within an occupation or vice versa. In contrast, horizontal segregation which is the primary focus of this chapter arises when men and women are disproportionately working in different occupational groups. Data on employment distribution across levels within occupation for the measurement of vertical segregation is difficult to obtain. Consequently, the book focuses on the measurement of horizontal segregation of occupation by sex.

6.2

Segregation Measures

The extent and magnitude of gender differences in occupational distribution is often measured by an index of segregation. The outcome of different measures of segregation is largely influenced by the choice of index. Various studies on segregation have used different indices to measure the extent of segregation in the labour market. There is, however, QR DJUHHPHQW DERXWWKH FRUUHFW LQGH[ DQG DV D UHVXOW LQGH[ ³ZDUV´ EUHDN out from time to time (Karmel and Maclachlan, 1988)16. Indeed, none of 15 See Gunderson, 1985; OECD, 1985; Reskin and Padavic, 1994) 16 See also Blackburn, Jarman and Siltanen (1993), Grusky and Charles (1998), Watts (1994, 1998) etc. ϭϬϵ

the segregation indices have been proven to be absolutely perfect or without any flaws.

The search for a single summary index has become counter-productive (Hakim, 1992) and no single index has been proven to be sufficient. The results of an analysis often depend less on the choice of index than on other methodological choices. Watts (1994) agUHHV ZLWK +DNLP¶V DVVHUWLRQ WKDW no index can capture both the horizontal and vertical dimensions of occupational sex segregation. Regardless of which segregation index used, it does not provide information about the direction of the imbalance and this remains a fundamental weakness of measures of segregation.

In this book, we use four of the widely used indices in the literature to capture and assess the extent and the changing trend of horizontal segregation of occupation by sex in the Ghanaian labour market between 1960 and 2000. These are the Duncan index of dissimilarity (ID), the Marginal Matching index (MM), the Karmel Maclachlan index (KM), and the size-standardised index of dissimilarity (Ds).

6.2.1 Duncan Index of Dissimilarity (ID) Duncan and Duncan (1955) proposed this dissimilarity index, which has undoubtedly become the most widely used index in the research literature for the study of sex segregation of occupation. It indicates the proportion of males (or females) that would have to change occupations in order to maintain gender ratio of each occupation equal to the gender ratio of workers as a whole. The index measures the absolute sum of the difference between the proportion of the female workers and the proportion of the male workers in each occupation. The index is expressed as:

ϭϭϬ

ID

1

n

F M 2¦ F i

i 1

i

(6.1)

M

where i represents occupation; force in occupation i ;

Mi

M

Fi

F

is the proportion of female labour

is the proportion of male labour force in

occupation i ; and i 1,2,........., n .

The index is interpreted as the proportion of the workers that has to be shifted in order to make the two distributions equal though such transfers may not be feasible. Intuitively, when the gender composition of all occupations is the same, it would necessary equal the gender composition of the workforce and there would be no occupational segregation. The index is assumed to take values ranging from 0 as minimum (when there is no difference between male and female occupational distribution) to a maximum of 1 (when the difference between male and female occupational distribution is unequally high). Thus, a higher index implies greater degree of occupational segregation by sex and vice versa.

One major advantage associated with the use of the ID index is its simplicity and widespread usage. Another potential advantage of the index is its intuitively understandable meaning (Anker, 1998). However, a great deal of confusion and misinterpretation of the index is witnessed in the literature. Anker (1998) provides evidence to show the correct interpretation of the index actually as the proportion of women workers who have to change occupation in addition to the proportion of male workers who would have to change occupation in order to bring about a situation where all occupations have the same percentage female. Anker was quick to admit that the correct interpretation is not intuitively very meaningful. ϭϭϭ

There are some fundamental limitations inherent in the use of the index as a measure of sex segregation of occupation. First, the index fails to provide a threshold above which one could conclude that segregation is getting out of hand. Secondly, and perhaps more importantly, it is argued that the value of segregation index measured by ID index is sensitive to changes in the occupational structure and sex composition of the workforce. As Blackburn et al (1993) put it, the index lacks sex composition invariance and gendered composition invariance making a comparison of sex segregation overtime fundamentally difficult.

6.2.2 Marginal Matching Index (MM) The MM index was developed to measure changes over time in occupational segregation by sex resulting exclusively from changes in the sex composition of occupations. The index was suggested by Blackburn et al (1993) and later named the index of segregation. It is designed to among other things address the inherent sensitivity of the value of the ID index to changes in occupational structure and gender composition of overall employment or workforce. The MM index is expressed as:

MM

W

f

u M m Wm u M f

(6.2)

FuM

where W f and Wm represent women in female occupation and women in male occupation respectively; M f and M m represent men in female occupation and men in male occupation respectively; F and M refer to the total number of women and men respectively in employment.

The MM index may be interpreted as the measure of the extent to which sex composition of occupation and structure of occupation vary together, i.e. how male occupations are staffed by men and female occupations ϭϭϮ

staffed by women. The difference between the ID index and the MM index LV EDVHG RQ WKH ZD\ ³PDOH´ DQG ³IHPDOH´ RFFXSDWLRQV DUH GHILQHG 7KH LQGH[ UHIHUV WR ³IHPDOH´ RU ³PDOH´ RFFXSDWLRQV DV WKRVH IRU ZKLFK WKH proportion of women (or men) in that occupation is greater than female (or male) share in the labour force. One key advantage of the MM index over the ID index is that the former provides segregation measures that are comparable across situations while the obvious simplicity in the computation of the latter makes it more appealing.

The computation of the MM index is done as follows: (i) ordering occupations according to their degree of female concentrations in the occupations in ascending order from lowest to highest female share in occupation; (ii) calculating the cumulative distribution of the employed along this ordering up to a point where the sum of employees (male and female) in the ordered occupations just equals the number of male employees in the total workforce. At this point, the number of employees in the remaining occupations will be just equal to the number of women in employment. In this case, marginal totals N m and N f are respectively matched to M and F . This can be represented in a modified basic table (table 6 ZKHUHURZVDQGFROXPQVDUHIRUFHGWREHHTXDORU³PDWFKHG´

There are clear methodological issues regarding the relationship between the MM and the ID indices. According to Watts (1994), there is a basic weakness in the computation of the MM index. He argues that the reallocation of employees to ensure that symmetry of 2×2 is achieved, required by the MM index is generally incompatible with the sex composition of the marginal occupations. This appears to suggest that the MM index is not immune to changes in sex composition that undermines trend comparison of segregation index. ϭϭϯ

Table 6.1: Modified segregation table for calculating MM Index Men

Women

³0DOH´RFFXSDWLRQV

Mm

Wm

Nm

³)HPDOH´RFFXSDWLRQV

Mf

Wf

Nf

Marginal Totals

M

F

N

Note: M m Wm Nm N f

Nm

M f Wf

N

and M F

Mm M f

Nf

Marginal Total

M

Wm W f

F

N

Source: Constructed from Anker (1998)

Some methodological relationships have been identified to exist between the MM index and the ID index. Algebraically, the two indices are basically the same when total employment is equally divided between the two sexes (see Anker, 1998 pp 92-93). Anker also used a graphical approach by plotting the difference between the values of ID and MM indices against percentage of females in the non-agricultural workforce based on data from 41 countries and observed a very strong relationship between ID-MM differences and percentage of the female non-agricultural workforce. In a regression analysis, he found a highly significant and nonlinear relationship between the two indices. This is confirmed by the regression results shown in appendix table 6b that the two indices have a significantly non-linear relationship.

6.2.3 Karmel Maclachlan Index (KM) The Karmel Maclachlan index (KM) also measures the disparity of gender shares of employment across occupation. It was first proposed by Duncan (1965) and developed and advocated by Karmel and Maclachlan in 1988 (Jones, 1992). Although, the ID index is related to KM index according to ϭϭϰ

the overall gender shares, the former cannot be readily decomposed to highlight the contribution of different occupational groups to the overall level of segregation (Watts & MacPhail, 2004). The KM index is expressed as:

1T ¦ F aM n

KM

i

i

Fi

(6.3)

i 1

where T denotes total employment of the two sexes, M and F ; a is female share of total employment; n is the number of occupations; Fi is the number of females in the ith occupation; and M i is the number of males in the ith occupation; and i 1,2,........., n .

The KM index is interpreted as the fraction of total employment that would have to relocate with replacement to achieve zero gender segregation, but maintains the occupational structure and the total overall gender shares of employment (Watts, 1994). The index assumes values ranging from a minimum of 0 to a maximum of 1. The closer the index to zero, the smaller the disparity of the gender shares of employment across occupation while a value closer to 1 gives an indication of larger disparity of gender shares of employment across occupation.

One key weakness of the index like other indices is that it fails to provide a threshold by which one can readily infer the seriousness or otherwise of occupational dissimilarity by gender. In addition, like many other indices, the KM index appears to increase with the number of disaggregated groups such that the higher the number of disaggregated occupations, the larger the index. Consequently, it becomes difficult to use the index to undertake a cross-national comparison. This may be due to different occupational classifications adopted in different countries and over time for one country ϭϭϱ

as well as new occupational classification introduced because of innovations and development.

Arguably, given the change in female share of total employment over time and its accompanying change in occupational structure and female shares of employment by occupation, gross changes in the value of KM index becomes an inappropriate measure of the rate of integration of the sexes in occupations (Watts, 1993). Rather, the composition effect which measures the percentage change in the overall gender shares and changes in the occupational structure becomes a better option as noted by Watts and Rich, (1991). Nevertheless, Karmel and Maclachlan (1988) use linear transformation of occupational distributions of employment by sex to show that changes in their index can readily be decomposed into composition and mixed effects.

6.2.4 Size-Standardized Index of Dissimilarity (Ds) The size-standardized dissimilarity index which was proposed by Gibbs (1965) controls for the effect of occupational structure, using all occupations as if they were of the same size, computed over a fixed number of comparable occupational categories (Williams 1979)17. The index is expressed as:

Ds

1

n

2¦ i 1

§Mi · § Fi · ¨ T ¸ ¨ Ti ¸¹ i ¹ © n© n §M · § Fi · ¨ T ¸ ¦¨ i T ¸ ¦ i i ¹ © ¹ i 1 i 1 ©

(6.4)

where Ti denotes total number of males and females in occupation i ; the numerators

Fi

Mi index the female and male proportions in Ti and Ti

17 See also Senyonov & Scott (1983); Charles and Grusky (1995). ϭϭϲ

occupation i while the denominators adjust such proportions on the proportions in other occupations n 1 .

The index takes values ranging from 0 to 1. The closer the index to 1 the greater the degree of dissimilarity and if the index is close to 0, the extent of dissimilarity is deemed to be low. Since the index standardizes each of the ith occupation to the same size, it is not affected by the shape of the occupational distribution. By not allowing changes in the size of the occupations in time to affect the value of the index makes it immune to occupational effects.

Hakim (1992) argues in favour of using the size standardised index on the grounds that the observation of male or female dominance is quite independent of the numbers in the ocFXSDWLRQ,QDGGLWLRQ\RXQJSHRSOHV¶ perceptions of the accessibility of occupations according to male or female dominance are unlikely to be affected. However, this argument completely ignores the vagaries of occupational classification, whereby occupations may and do differ markedly in size and hence significance in the analysis of segregation (Watts, 1993).

The Ds index also suffers from similar weaknesses inherent in other indices. The index fails to provide a benchmark or critical value with which one can determine when segregation occurs or the extent of segregation. In addition, it increases with the number of disaggregated groups thereby posing some degree of difficulty in making comparison among different classifications when the number of disaggregated group differs. Moreover, the fact that the index potentially solves the problem of size is an indication that the weighting procedure used generates a biased estimate. As noted by

ϭϭϳ

Jacobs (1989) and Semyonov (1980) among others18, while this standardization eliminates a kind of marginal dependence, it has the perverse effect of introducing a new dependence on the rate of female labour force participation. Nonetheless, the use of this index in addition to three other indices would provide a good assessment of the extent and pattern of sex segregation of occupation in the Ghanaian labour market overtime.

6.3

Computational Issues and Sources of Data

The main data sources for the study are the 1960, 1970, 1984 and 2000 population censuses as well as the last three rounds of GLSS dataset of 1991/92, 1998/99 and 2005/06. The three GLSS datasets are nationally representative datasets that are largely consistent with each other. The study draws on employment and time use module of the dataset with LQGLYLGXDOV¶ PDLQ MRE EHLQJ WKH PDLQ IRFXV RI HPSOR\PHQW 7KH FHQVXV data (except the 1970 census) reports one digit classification whilst the GLSS dataset gives disaggregated three-digit classification which can be grouped into two or one-digit classification. The occupational classification adopted is the International Standard Classification of Occupation, 1968 (ISCO±68). The use of the ISCO±68 instead of a more recent one in 2008 is meant largely to ensure consistency with occupational classification based on historical data to facilitate trend analysis while taking cognisance of the emergence of new occupations.

Another nationally representative household datasets available that could have been used in the analysis are the Core Welfare Indicators Questionnaire (CWIQ) conducted in 1997 and 2003 and the first two rounds of GLSS in 1987/88 and 1988/89. These dataset could not be used in this analysis because the occupational classifications are not consistent 18 For example, Charles and Grusky (1995), Jacobsen (1994) ϭϭϴ

with the census and the last three rounds of the GLSS. In addition, the sampling frame and instruments particularly related to employment were not consistent with the sampling frame and survey instruments of the last three household living standards surveys.

Generally, the value of the segregation index depends to a greater extent on the degree of disaggregation of occupations. Segregation indices are often sensitive to the level of disaggregation of occupational data. The greater the level of disaggregation, the greater the value of the segregation index all things being equal. Occupation as per ISCO can be classified into one-digit, or more detailed two-digit or three-digit classification. Anker (1998) provides a very important implication with regard to the implications of the degree of disaggregation of occupational data for segregation indices. Much more widely available one-digit occupational data tends to conceal a great deal of occupational segregation by sex. This indicates that moving from one digit to a two digit and three-digit classification scheme should have an increasing effect on the values of occupational segregation. This may suggest that occupational data at the 1-digit level are likely to provide a relatively weak measure of the extent of occupational segregation.

In this book, we compute measures of occupational segregation by sex using one-digit and three digit classification schemes. We compute segregation indices based on available one-digit or broad occupational classification schemes using four census data, i.e. 1960, 1970, 1984 and 2000 as well as GLSS data of 1991/92, 1998/99 and 2005/06. This will provide a means of assessing the changing trend of occupational segregation by sex in broad terms since 1960. The study also computes segregation measure using an available three-digit detailed disaggregated occupational data. In addition, we compute index of segregation for paid and self-employment separately and for total employment or full sample to ϭϭϵ

capture the potential differences in occupational segregation based on types of employment.

6.4

Analysis of Empirical Results of Segregation

We present and discuss results of the segregation measure under two broad headings. First, the results of occupational segregation by sex for all employed persons using the four sex segregation indices based on 1-digit and 3-digit classifications of occupations are discussed. This is followed by analysis of differences in outcome of segregation measure using the 1-digit occupational classification separately for those in self-employment and paid-employment. This will provide the basis of assessing the differences in the degree of occupational segregation in self-employment and paidemployment.

The overall results of the segregation indices confirm the methodological issues about the relationship between the ID and the MM indices. Indeed, the outcome of the segregation measure as reported in table 6.2 and 6.3 show minimal differences in the value of ID and MM. Following the observation by Anker (1998), the female share in the workforce for almost all the years was close to 50 percent. According to Anker, as the share of female in employment approaches 50 percent, the value of MM and ID become equal. This observation is supported by the regression results reported in appendix table 6b suggesting a strong and statistically significant negative relationship between differences in values of ID and MM indices and the percentage of females in employment. The results also points to a non-linear relationship between the two variables considering the statistically significant coefficient of the square of percent of females in employment.

ϭϮϬ

6.4.1 Occupational Segregation of Total Employment The outcome of the segregation measure based on 1-digit and 3-digit classifications of occupation for the period 1960 to 2006 are reported in table 6.2. The results of the segregation measure based on 3-digit occupational classification are represented in the parenthesis. To ensure consistency and harmony in the trend analysis of the outcome of segregation measures, we discuss the results of the indices based on the census data separately from the results obtained from the GLSS dataset.

The results based on 1-digit occupational classification suggest a generally low or at worse, moderate degree of sex segregation of occupation signifying a lower degree of unequal occupational distribution by sex. This is based on the range and benchmark suggested by Jahn, et al (1947) to judge segregation index as low, moderate or high. According to their criteria, index ranging from 0 to 0.3 is low, 0.3 to 0.6, moderate and above 0.6 deemed high. Considering the fact that three segregation measures (ID, MM and KM) all produced figures lower than 0.3 is an indication of a lower degree of sex segregation of occupation in Ghana. The sizestandardised index produced a low to moderate degree of segregation ranging from a low of 0.28 in 2000 to 0.45 in 1970.

Indeed, the generally low coefficients confirms the low degree of segregation using ID index, KM index, index of concentration, index of segregation and size standardised index obtained by Baah-Boateng (2007) based on the

2000 population census. He observed a low degree of

segregation in terms of the sex distribution of occupation, employment type and employment sector. In contrast, Boateng (1996) found a 49 percent (or 0.49) index of dissimilarity in 1993 among workers who are Social Security and National Insurance Trust (SSNIT) contributors and who predominantly operate in the formal sector of the labour market. ϭϮϭ

Table 6.2: Sex Segregation of Occupation of Total Employment Population Census

Ghana Living Standards Survey 2000 1991/92 1998/99 2005/06 0.165 0.195 0.189 0.176 (0.343) (0.305) (0.316) (0.311) 0.165 0.109 0.144 0.161 (0.344) (0.305) (0.316) (0.302)

Index ID

1960 0.237

1970 0.228

1984 0.208

MM

0.239

0.228

0.209

KM

0.112

0.113

0.104

0.082 (0.171)

0.097 (0.152)

0.094 0.088 (0.157) (0.155)

Ds

0.415

0.445

0.334

0.283 (0.523)

0.314 (0.723)

0.324 0.310 (0.732) (0.581)

Figures in parenthesis represent segregation indices based on 3-digit classification or disaggregated occupations with number of disaggregated occupations as 79 in 1984; 76 in 1991/92; 82 in 1998/99; 121 in 2005/2006 Source: Computed by the Author from Population Census & GLSS (GSS)

The outcome of the segregation measure based on the 3-digit classification of occupation also indicates a generally low and moderate degree of segregation. With the exception of the size-standardised index that produced a higher degree of segregation of over 0.70 in the 1990s, the three other measures showed indices of less than 0.6 (based on benchmark suggested by Jahn et al, 1947) implying a low to moderate degree of segregation since 1984. The higher segregation index based on 3-digit occupational classification relative to 1-digit classification confirms the general observation that the more disaggregated the occupation, the greater the degree of segregation.

The fundamental difficulty associated with trend analysis of segregation indices based on the 3-digit classification is the variation in the number of occupations over time. As reported in table 6.2, the number of occupations ϭϮϮ

in the 3-digit classification varied over time. For instance, a total of 79 disaggregated occupations were reported in 1984 compared with 76 in 1991/92, 82 in 1998/99 and 121 in 2005/06. The increasing number of occupations in the 3-digit classification has implications for the value of the segregation index, making it inappropriate for trend analysis to be carried out.

The empirical results over time based on the 1-digit classification of occupation (containing seven broad occupations in all the years) show that the degree of segregation has generally trended downwards since 1960. This indicates a gradual and consistent move towards achieving equal distribution of occupation by sex in the Ghanaian labour market. The results in table 6.2 suggest that the index of dissimilarity and marginal matching index which produced similar results witnessed a larger decline between 1984 and 2000 by at least 0.04 point compared with about 0.01 and 0.02 point drop in 1960s and 1970s respectively.

The general decline in the occupational dissimilarity index between males and females observed over a four decade period is in sharp contrast with the increase in the index of dissimilarity of sectoral distribution of employment by sex from 0.37 to 0.49 between 1975 and 1993 based on the SSNIT dataset (Boateng, 1996). With SSNIT data covering mostly the formal sector of the labour market implies that while the entire labour market experienced a declining trend in occupational segregation by sex, the same cannot be said of employment distribution by sex in the formal segment of the labour market particularly between the late 1970s and early 1990s.

The KM and Ds indices however reported a consistent fall in segregation from 1970 after a marginal increase in the indices in the 1960s. The results ϭϮϯ

show a 0.03 point rise in Ds index in the 1960s as against 0.11 and 0.51 points decline in the 1970s and 1984-2000. Similarly, the KM index recorded a 0.001 point rise in the index in the 1970s followed by a 0.009 and 0.022 points decline in the 1970s and 1984-2000 respectively. The declining trend in the degree of segregation in Ghana is also observed in the 1990s and beyond using the GLSS dataset particularly based on the ID and KM measures. The results show a consistent fall in the value of ID index from 0.195 to 0.176 between 1991/92 and 2005/06 with the values of KM index following the same trend from 0.097 to 0.088 over the same period. The MM index however reported a gradual increase in the degree of segregation from 0.109 to 0.144 in the 1990s and further up to 0.161 in 2005/06. The value of the Ds index increased marginally in the 1990s by about 0.01 point but subsequently dropped from 0.324 in 1998/99 to 0.31 in 2005/06.

Overall, Ghanaian labour market has enjoyed a general decline in the degree of occupational segregation by sex since 1960. In effect, this general declining pattern of sex segregation of occupation in Ghana over the past four and half decades is a positive development that needs to be sustained towards achieving equal distribution of occupations between the two sexes and to overcome gender imbalances in employment.

6.4.2 Occupational Segregation of Paid-Employment and Self Employment The extent of occupational segregation and the changing pattern is observed to differ among workers in paid employment and self employment. The results presented in table 6.3 based on the 1-digit occupational classification indicate a higher degree of occupational segregation by sex among self-employed workers than among paidemployees based on all the four indices. This is in line with some research ϭϮϰ

works that demonstrate greater occupational segregation in selfemployment than in wage employment (see e.g. Wharton, 1989).

The relatively lower degree of occupational segregation by sex among paid employees could be explained by the lower differential in human capital endowment between men and women. There are more self-employed women with no or limited education than self-employed men culminating LQ ZRPHQ¶V XQGHUUHSUHVHQWDWLRQ LQ SURIHVVLRQDO

and

technical,

administrative and managerial and clerical occupations and overrepresented in sales, service and to a limited extent production occupations. Table 6.3: Sex Segregation of Occupation (1-digit classification) by Employment Status 1991/1992

1998/1999

2005/2006

Index

Paid

Paid Self

Paid

IID MM KM

Employment 0.203 0.220 0.191 0.228 0.114 0.117

Employment 0.207 0.310 0.170 0.231 0.108 0.153

Employment 0.278 0.365 0.179 0.301 0.107 0.183

Ds 0.299 0.462 % of Female in 27.50 58.40 Employment * include unpaid family workers

0.283 0.410 25.20 55.60

0.264 0.548 26.20 54.80

Self*

Self

Source: Computed by the Author

The nature of the changing pattern of segregation depends on the type of index used in the measurement of segregation. While the ID index reported continuous increase in segregation over time among paid employees, the D s and KM indices suggested a continuous decline with MM index recording a decline in the 1990s before increasing thereafter. In self-employment, the results show a consistent rising trend of occupational segregation based on ϭϮϱ

ID, MM and KM indices with Ds reporting a rise in occupational segregation by sex between 1998/99 and 2005/06 after initial drop from the 1991/92 level. The general picture however suggests an increasing pattern of occupational segregation overtime among self-employed against a general declining pattern among workers in paid employment since 1991. This appears to suggest increasing access of both men and women into any occupation in paid employment due largely to little differences in the education of men and women while the reverse is the case in selfemployment.

6.5

Accounting for Changes in Segregation

One major limitation with the use of segregation index such as ID and MM in examining changes in segregation over time is its sensitivity to changes in occupational structure and sex composition of the workforce. The common practice to remedy this challenge is to disaggregate ID values into three component parts: (i) a sex composition effect; (ii) an occupational composition effect or occupational mix effect; and (iii) an interaction effect of both sex and occupational influence termed as the residual.

The sex composition effect represents the changes in the index over time due to changes in sex composition within occupations with the size of each occupation remaining fixed at its initial level. The occupational composition or occupational mix effect accounts for the changes in the index as a result of changes in the size of occupation holding sex composition within occupations constant. The residual interaction effect however has been observed to have no statistical meaning (Watts, 1998; Anker, 1998). It only shows the effect due to interaction between the changes in sex composition and occupational structure.

ϭϮϲ

6.5.1 Index of Decomposition Effects The indices to capture these sex and occupational composition effects are measured and specified as:

ª º n mit Ti 0 f i 0Ti 0 mi 0Ti 0 » 1 « n f it Ti 0 'SEX «¦ ¦ 2 i 1 ¦ f it Ti 0 ¦ mit Ti 0 i 1 ¦ f i 0Ti 0 ¦ mi 0Ti 0 » «¬ i »¼ i i i º ª n mi 0Tit f i 0Ti 0 mi 0Ti 0 » 1 « n f i 0Tit 'OCC «¦ ¦ 2 « i 1 ¦ f i 0Tit ¦ mi 0Tit i 1 ¦ f i 0Ti 0 ¦ mi 0Ti 0 »» i ¼ ¬ i

(6.5a)

(6.5b)

where f i and mi represent percent of female and male respectively in occupation i , Ti denotes total number of workers in occupation i M i Fi ; subscripts 0 and t are the initial and subsequent periods respectively. 'SEX represents changes in sex composition of occupation holding

occupational structure or composition constant; and 'OCC represents changes in occupational structure or composition holding sex composition of employment constant.

Generally, one fundamental problem with this standardisation methodology as echoed by Anker (1998) is that the weights (i.e. distributions) chosen for this standardisation (i.e. the initial year distribution) do not provide unique results. This is because the results will be different if other weights (e.g. the ODWWHU\HDU¶VGLVWULEXWLRQ DUHXVHG. Moreover, the changes calculated as due to sex composition of occupation and changes in the occupational structure of the workforce do not necessarily sum up to the actual observed change. The difference is what is known as the residual, interpreted as interaction effect resulting from changes in sex composition and employment mix over the period (Blau and Hendricks, 1979). ϭϮϳ

6.5.2 Analysis of Empirical Results of Decomposition Effects Table 6.4 reports the outcome of decomposition of changes in values of ID index to underscore the sources of changes in the values of dissimilarity index over time. The decline in sex segregation of occupation is reported to be statistically significant over the period 1970-84 and 1984-00. This is reflected in the negative values of the upper and lower limits of 95 percent confidence interval of decline in segregation index as reported in table 6.4. Decline in segregation in the 1960s and 1990s as well as 1999-2006 were however not statistically significant based on the negativity of lower limit and positivity of upper limit of the confidence interval.

The results of the changing pattern of sex segregation of occupation generally indicate stronger residual effect across the years except in 19842000 when the effect fell below 10 percent. Boateng (1996) also observed relatively stronger residual effect of changes in dissimilarity index using SSNIT data. He estimated that about 43 percent of the 12.7 percentage point increase in the index of dissimilarity between the distribution of employment by sector of men and women over 1975-93 period was accounted for by the residual effect. About 37 percent of the increase in the index emanated from changes in sectoral composition while the remaining 20 percent was accounted for by sex composition within the employment sector.

The decline in occupational segregation in the 1960s, although not statistically significant was largely due to changes in occupational mix effect of 28 percent compared with a 2 percent sex composition effect. A stronger occupational mix effect relative to sex composition effect was observed between 1991/92-1998/99 and 1998/99-2005/06. The reverse was the case during the period 1970-84 with 36 percent sex composition effect as against 4 percent occupational mix effect. Sex composition effect was ϭϮϴ

marginally stronger than occupational mix effect by 1 percentage point between 1984 and 2000 (table 6.4). Palaz and Bandirma (1999) found similar results in Turkey using Karmel Maclachlan decomposition approach to the effect that an increase in the gender composition of LQGLYLGXDO¶VRFFXSDWLRQVZDVWKHPDLQUHDVRQIRUWKHLQFUHDVHLQWKHOHYHO of segregation. Table 6.4: Source of Over time Changes in Segregation Index, 1960-2006

Index ǻLQ,'LQGH[ Standard error

National Population Census 1960±70 1970±84 1984±00 ±0.9 0.871

±1.97** 0.897

±4.33** 1.746

GLSS 1991±99 1999±2006 ±0.59 0.832

±1.29 1.790

95% Conf. Interval (-2.6, 0.8) (-3.7, -0.2) (-7.8, -0.9) (-2.2, 1.0) (-4.8, 2.2) Due to: Sex Composition ±0.015 ±0.708 ±1.971 ±0.036 ±0.252 (1.7%) Occupational Mix Residual

(35.8%)

±0.248 ±0.074 (27.6%) (3.8%) ±0.637 ±1.193 (70.7%) (60.4%)

(45.6%)

(6.1%)

(19.5%)

±1.939 (44.8%) ±0.414 (9.6%)

±0.143 (24.4%) ±0.407 (69.5%)

±0.301 (23.4%) ±0.736 (57.1%)

Figures in parenthesis are percentage contribution to overtime changes in occupational segregation by sex composition, occupational mix and residual effect; ** significant at 5% * significant at 1% Source: Computed by the Author from Population Census & GLSS (GSS)

The relatively weak occupational mix effect reported in 1970-1984 relative to other period confirms the dismal economic performance of Ghana over the period which saw the country slipping into recession, particularly in the late 1970s and early 1980s. A strong economic recovery from the mid 1980s with the introduction of Structural Adjustment measures that saw a revamp and expansion of the ailing industrial and services sectors largely ϭϮϵ

contributed to the strong occupational mix effect in 1984-2000. In addition, the increased demand for highly skilled workforce in response to increasing globalization from the 1980s which resulted in a shift from training of lower skill to higher skill manpower contributed to a shift from low skilled to high skilled occupation. As Baah-Boateng and Turkson (2005) noted, employment of skilled workforce in many developing countries including Ghana has been increasing in response to globalisation.

A weak sex composition effect in the 1960s does not come as a surprise since gender advocacy in favour of women empowerment was quite low or virtually non-existent at the time. Gender consciousness in Ghana began in the early 1980s and was largely championed by the 31st December :RPHQ¶V 0RYHPHQW19. The Beijing conference on women empowerment in 1995 and inclusion of gender equality and women empowerment in the Millennium Development Goals (MDG3) also amplified gender advocacy for women. The outcome of these initiatives is the strong sex composition effect of 36 percent and 46 percent recorded in 1970-84 and 1984-2000 respectively. A stronger sex composition effect reported in the 1990s and beyond reflects the intensification of the struggle against gender inequality and marginalisation of women in all spheres of the national life and work towards achieving the third goal of the MDG that seeks to promote gender equality and empowerment of women.

6.5.3 Sex Composition and Occupational Mix effects of Changing Occupational Distribution For better understanding of the relative contribution of major occupations to changes in occupational structure and sex composition that accounted for the declining trend in dissimilarity index, we present in table 6.5 relative changes in sex composition of occupation and distribution of total 19 A Women Non-Governmental Organisation (NGO) led by the wife of the then Head of State ϭϯϬ

employment by occupation. Essentially, as equations 6.5a and 6.5b implicitly indicate, the role of a specific occupation to the changes in sex composition and occupational structure depend largely on the initial size of occupation. For instance, with agriculture accounting for over 60 percent of total employment in 1960, any change in the share of agriculture in total employment would have a greater impact on the occupational structure effect.

The results in table 6.5 indicate that the 45 percent declining occupational structure effect reported over 1984-2000 was largely due to the declining relative importance of agriculture which initially accounted for 61 percent of total employment. Indeed, agriculture shed over 10 percentage point share in total employment to non-agriculture occupations except administrative and managerial occupation over 1984-2000. Similarly, the 3.7 percentage point decline in the proportion of employment in agriculture and to a limited extent sales and administrative and managerial occupations over 1960-1970 contributed to occupational mix effect of 28 percent. This eventually translated partly into a drop in the dissimilarity index during the period. Between 1970 and 1984 however, the reduction in proportion of the employed engaged in production and service occupations (which jointly accounted for about 25 percent of total employment in 1970) could only overturn the gains in the share of agriculture, sales and professional and technical occupations to account for occupational mix effect of the less than 4 percent.

The contribution of the occupational structure effect of 24 percent to changes in segregation index in the 1990s was largely as a result of the 7 SHUFHQWDJH SRLQW GURS LQ DJULFXOWXUH¶V VKDUH LQ WRWDO HPSOR\PHQW EHWZHHQ 1991/92 and 1998/99. A decline of 6 and 3 percentage points in the share of sales and professional and technical occupations and a 2 percentage point ϭϯϭ

GURS LQ DJULFXOWXUH¶V VKDUH LQ WRWDO HPSOR\PHQW RXWZHLJKHG WKH DQG percentage point gain in services and production occupation to produce a 23 percent negative occupational structure effect over the period of 1998/99-2005/06. In effect, the mere size of agriculture makes it a key occupation in the determination of the extent of changes in occupational structure and eventually changes in index of dissimilarity as a measure of changes in occupational segregation by sex in Ghana. Table 6.5: Changes in Female composition (Fi/Ti) of occupation and Occupational Distribution of Workforce (Ti/T) Occupation Prof/Tech

1960±70 1970±84 1984±00 1991-99 Fi/Ti Ti/T Fi/Ti Ti/T Fi/Ti Ti/T Fi/Ti Ti/T 4.0 1.4 12.0 0.3 0.5 2.5 ±4.9 3.7

Adm/Mgerial ±7.3 ±0.2 8.8 0.0 Clerical 7.9 1.1 15.1 ±0.4 Sales Service

7.5 ±0.3 ±5.2 0.7

Agriculture Production

6.0 ±3.7 4.4 13.5 0.9 9.4

Total

6.4

---

1.0 11.1

6.2

1999 2006 Fi/Ti Ti/T ±1.7 ±2.9

24.1 0.0 14.3 0.0 ±1.4 ±8.8 2.1 ±11.4 ±0.5 21.4

0.3 ±0.6

0.6 ±17.5 1.4 ±0.5 28.1 3.4

±3.7 1.7 ±4.0 ±6.3 ±0.5 1.2 24.4 6.3 3.2 1.7 ±10.4 0.8 ±6.6 ±2.4 ±1.6 ±3.3 ±4.8 1.0 8.7 0.6 ±6.2 4.9 --- ±1.8 --0.7 --- ±2.2 ---

Source: Computed by the Author from Population Census and GLSS (GSS)

7KHH[WHQWWRZKLFKFKDQJHVLQWKHIHPDOH¶VVKDUHLQRFFXSDWLRQFRQWULEXWHV to sex composition effect depends on a number of factors. Clearly, a gain in the percent of females in predominantly male jobs or a decline in the percent of females in traditionally female jobs would generally cause a negative sex composition effect. This has the implication of feeding into a decline in dissimilarity index. In addition, though, the sex composition effect holds changes in the size of occupations constant, a given increase in the composition of women in an occupation would generally have a greater ϭϯϮ

effect if the initial size of the occupation is large. It is important to observe WKDWDQLQLWLDOO\ ³PDOH dominated´MRELQZKLFKIHPDOHFRPSRVLWLRQULVHV by more than the increase in the share of women in total employment overtime, causes a negative sex composition effect. An initially ³female GRPLQDWHG´job in which the proportion of female in a particular occupation UHPDLQVXQFKDQJHGRUULVHVE\OHVVWKDQWKHLQFUHDVHLQZRPHQ¶VVKDUHLQ total employment also contributes to a negative sex composition effect.

The empirical results in table 6.5 show that the negative sex composition effect in the 1960s reflects the appreciable increase in the female share in agriculture and a rise in the percent of females in initially male dominated clerical and production occupations by more than the 6.4 percentage point rise in female representation in employment. This was counteracted by the positive sex composition effect associated with the 7.5 percentage point increase in female composition in female dominated sales occupation and a considerable percentage point decline of 5.2 in female composition in male dominated service occupation. The net effect is the marginal negative sex composition effect of 2 percent between 1960 and 1970 as shown in table 6.4.

Between 1970 and 1984, changes in sex composition of the seven occupations were associated with a negative sex composition effect. As shown in table 6.5, the percent of females in male dominated occupations of professional and technical, administrative and managerial, clerical, service and production occupation increased by more than the increase in women representation in total employment of 6.2 percentage points. In addition, the only female dominated occupation of sales rose by less than WKH LQFUHDVH LQ ZRPHQ¶V UHSUHVHQWDWLRQ LQ WRWDO HPSOR\PHQW DQG FRupled with the appreciable increase in female composition in agriculture resulted ϭϯϯ

in a 36 percent sex composition effect over the 1970-1984 period reported in table 6.4.

The empirical results in table 6.5 reports a drop in female representation in total employment accompanied by a considerable percentage point increase in composition of women in male dominated occupations of service, administrative and managerial and professional and technical occupations with a negative sex composition effect between 1984 and 2000. This is reinforced by a substantial percentage point decline in the composition of women in female dominated sales occupation resulting in 46 percent negative sex composition effect during the period.

In the 1990s, the 6 percent negative sex composition effect was generally accounted for by the 9 percent increase in composition of women in initially male dominated production occupation and 4 percent drop in female dominated sales occupation. The 20 percent sex composition effect between 1998/99 and 2005/06 was also accounted for by substantial increase in female composition in initially male dominated occupation of service and clerical occupations and 4 percentage point drop in the composition of female in female dominated sales occupation. In effect, the combination of changes in composition of women in initially male dominated occupations vis à vis initially female dominated occupations explains to a large extent the sex composition effect that accounts for changes in the value of dissimilarity index overtime.

ϭϯϰ

Chapter 7

Wage Differentials between Men and Women in Ghana 7.0

Introduction

In chapter six, the book examined the extent and changing pattern of occupational segregation by sex which is a reflection of one form of gender differences in the Ghanaian labour market. This chapter focuses on another dimension of gender differences in the labour market that have pecuniary considerations.

The chapter primarily estimates and analyses wage

differential between men and women from various angles. It examines gender wage differentials among workers in self-and paid-employment, determines the female composition effect and establishes whether women or men in male dominated occupations earn better than their counterparts in female dominated occupations. The data used in the estimation of different wage equations is drawn from the fifth round of Ghana Living Standards Survey (GLSS5) conducted in 2005/200620.

7.1

Model Formulation

Wages or Earnings differentials between two demographic groups have often been linked to differences in human capital accumulation and labour market experience as well as occupational segregation. Indeed, differences in personal and demographic characteristics such as age, sex and marital status have also been found to contribute to wage differentials in the labour

20 The main source of data for model estimation has been discussed under section 5.4 in chapter five ϭϯϱ

market. In this chapter, we examine different types of wage models that explain wage differentials between men and women in Ghana.

We specify and estimate different forms of wage equations to establish the differences in wages between men and women after controlling for all the relevant personal, productive and job characteristics. First, we specify a Mincerian human capital wage function which assumes that the proportional change in wages or earnings is a function of the personal and productive characteristics of the individual worker including age, sex, marital status, education, experience, occupation and residence of the individual. The model which is a complex relationship between human and social capital embedded in the regional and ecological framework conditioned on occupation is formulated as:

ln Wi

X i ' E Df P i

(7.1)

where lnWi is the log of individual monthly wage received from work including bonuses in main occupation; X is vector of characteristics, relevant in determining wages including age (in years), married dummy (married=1: single=0), education (measured by number of years in school), experience (measured by number of years in main occupation), effort (proxied by number of hours spent in main occupation per week), residence of individuals in terms of rural dummy (rural=1: urban=0), and ecological dummy (coastal, forest, and savannah zones), and occupation engaged in by the individual21. The parameter ȕ represents a matrix of coefficients of X control variables. The variable f is female dummy (measured as female=1: male=0) and the 21 Some important variables such as trade union membership could not be controlled for due to limited observations for the variables in the GLSS5 dataset ϭϯϲ

coefficient Į is the coefficient of female dummy to capture the wages or earnings differences between male and females. The µ is a random error term assumed to be normally distributed with zero mean and constant variance, i.e. P ࡱ N (0, 1).

The chapter also attempts to test the prediction of the crowding hypothesis in the Ghanaian labour market. As discussed in chapter five, one major SUHGLFWLRQRIWKH³FURZGLQJ´K\SRWKHVLVRIODERXUPDUNHWGLVFULPLQDWLRQLV that the overcrowding of women into some designated occupations ascribed DV³IHPDOH´RFFXSDWLRQVFRmpresses wages below that of equally qualified workers in other occupations. In this regard, we specify and estimate wage equation of the form that captures the female composition effect as:

ln Wi

X i ' E Dfemcompi Pi

(7.2)

where femcomp refers to female composition of occupation measured as percentage of workers in occupation i who are women. The parameter Į captures the female composition effect on wages that underpins the ³FURZGLQJ´ K\SRWKHVLV $ VLJQLILFDQWO\ QHJDWLYH Į is an indication of declining effect of increasing female composition on wages or earnings and WKXVFRQILUPLQJWKHSUHGLFWLRQRIWKH³FURZGLQJ´K\SRWKHVLV

From equation (7 LIWKHSUHGLFWLRQRIWKH³FURZGLQJ´K\SRWKHVLVKROGV it follows then that female dominated occupations would pay less than other occupations with lower female composition. Consequently, we further seeks to test the hypothesis that female dominated occupations potentially pay less relative to male dominated occupations which may largely account for the lower wages of women than men. The Mincerian

ϭϯϳ

wage regression model is therefore modified to capture female occupation dummy as:

ln Wi

X i ' E Dfemoccup P i

(7.3)

where femoccup refers to the female occupation dummy. In this analysis, we define female occupation as the occupation in which the female share is higher than the share of females in total employment. By implication, ³PDOH´RFFXSDWLRQis defined as the occupation in which the share of males in that occupation is higher than maleV¶ VKDUH LQ total employment. From GLSS 5 dataset used in the estimation, occupations that qualify to be considered as female occupations based on this definition are represented in appendix table 7a.

We further investigate wage differentials between men in male dominated and men in female dominated occupations and between women in male dominated and women in female dominated occupations. Thus, considering that female occupations are potentially low rewarding, does it follow then that men in female dominated occupations are worse off in terms of wages compared to their counterparts in male dominated occupations? Similarly, if male dominated occupations are found to pay better than female dominated occupations, does it then suggest that women in male occupations would potentially be better off in terms of wages compared with women in female occupations? To answer these questions, a wage equation is formulated with the introduction of interactive dummies of working men and women in female and male occupations as:

ln Wi

X i ' E D1W f D 2 M f D 3 M m D 4Wm Pi

ϭϯϴ

(7.4)

where Wf is the dummy of women working in female occupations generated as an interactive dummy of female with female occupation dummy; Mf is the dummy of men working in female occupations generated as an interactive dummy of male with female occupation dummy; Mm is the dummy of men working in male occupations generated as an interactive dummy of male with male occupation dummy; and Wm is the dummy of women working in male occupations generated as an interactive dummy of female with male occupation dummy.

7.2

Data Source and Descriptive Statistics

The dataset used in the empirical analysis of gender wage differentials is drawn from the GLSS5 conducted in 2005/06 by the Ghana Statistical Service. It is a nationwide representative survey covering a sample of 8,687 households in 580 enumeration areas, containing 37,128 household members. An alternative source of data for the empirical analysis would have been the Regional Programme on Enterprise Development (RPED) data collected by the Centre for the Study of African Economies (CSAE) based in Oxford. The dataset contains comprehensive panel data set on a sample of firms within the manufacturing sector in Ghana located in four major cities of Accra, Kumasi, Takoradi and Cape Coast. However, the motivation for the choice of GLSS5 dataset is that it is the most recent nationwide survey that covers all occupations and sectors of the economy.

We make use of information collected on the demographic characteristics, educational background and labour market experience as well as employment status of about 14,360 adults aged 15 year and above who reported to have done some work 7 days prior to the survey. Individual wages are measured by payment received from the main occupation including any bonus. We also generated variables for age, female dummy, ϭϯϵ

married dummy, rural dummy and ecological dummy (i.e. coastal, forest, and savannah belts).

In addition, variables such as education measured by number of years spent in school, experience and experience squared measured by the number of years engaged in the main occupation, and effort in terms of the number of hours spent in the main occupation per week are also generated in the estimation of gender wage models. Occupational dummy for seven broad occupations based on International Standard Classification of Occupations, 1968 (ISCO±68) are also created in the estimation process (see appendix table 6a for classification of occupations).

The means and standard deviations of the variables are reported in table 7.1a whilst the minimum and maximum values of the variables are represented in table 7.1b. Out of about 14,360 reported to be working, 59 percent was engaged in self-employment compared with 16 percent in paid-employment with the remaining 25 percent as unpaid family workers and other workers. The self-employed were dominated by women who accounted for 52 percent while paid-employment was divided into 75 percent males and 25 percent females. This underscores the dominance of women in self-employment and men in paid-employment and the implication for wage differentials between the two sexes.

Gender disparities were observed to be more pronounced in education, experience and effort in self-employment than in paid-employment in favour of men (table 7.1a). A marginally smaller log of monthly wage or earnings is observed in self-employment than in paid employment. Male workers are also observed to be averagely older than their female counterparts while married men were estimated to be more than married women in both paid and self-employment. The difference in the minimum ϭϰϬ

0.41

ϭϰϭ

0.30 0.49

0.22

Agriculture 0.58 0.49 0.51 0.50 0.10 Production 0.22 0.42 0.16 0.37 0.40 Source: Computed from the GLSS 5, Ghana Statistical Service

0.18 0.03 0.10 0.36 0.34

0.16 0.21 0.24 0.36

0.03 0.00 0.01 0.15 0.13

0.03 0.05 0.06 0.15

0.25

0.09 0.12 0.21 0.25

0.49 0.29

9.5 21.5 1.0 0.50 0.47 0.50

0.10 0.14

0.01 0.12 0.10 0.19

0.34

0.38 0.11

7.3 43.1 13.2 0.40 0.20 0.51

0.29 0.35

0.10 0.33 0.30 0.39

0.47

0.49 0.32

8.9 19.9 1.0 0.49 0.40 0.50

0.77 0.14

0.00 0.00 0.04 0.03

0.01

0.46 0.33

17.7 39.8 12.6 0.69 0.74 0.21

0.42 0.34

0.04 0.04 021 0.18

0.11

0.50 0.47

13.6 18.3 1.3 0.46 0.44 0.41

0.41 0.21

0.00 0.00 0.21 0.16

0.00

0.45 0.24

12.9 37.1 12.4 0.56 0.62 0.31

0.49 0.40

0.02 0.00 0.41 0.37

0.06

0.50 0.42

12.8 20.7 1.3 0.50 0.49 0.46

0.07

0.38 0.09

9.6 50.5 13.5 0.56 0.32 0.52

0.01 0.01 0.04 0.07

0.50 0.45

12.2 19.9 1.3 0.50 0.48 0.45

Admin/Managerial Clerical Sales Service

0.44 0.29

12.3 36.7 12.4 0.57 0.65 0.28

Prof/Technical

0.50 0.45

12.8 19.9 1.3 0.49 0.48 0.45

0.43 0.28

Female

Forest Savannah

Male

Self-employment

14.1 42.8 12.9 0.58 0.63 0.29

Female

Experience (years) Effort (in hours) Log of monthly wage Married Rural Coastal

Male

Age (in years) Education (years)

Female

Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. 39.3 14.7 38.4 14.2 38.2 11.9 34.6 11.6 43.1 14.6 41.6 14.0 7.0 5.2 4.6 4.9 10.5 4.6 10.5 5.2 5.8 4.9 4.5 4.6

Male

Table 7.1a: Means and Standard Deviations of Key Variables in the Wage Models Variable Full Sample (Total employment) Paid Employment

0 0 5.3 0 0 0

0 0

0

0 0 0 0

Age (in years) Education (years)

Experience (in years) Effort (in hours) Log monthly wage Married Rural Coastal

Forest Savannah

Prof./Technical

Admin/Managerial Clerical Sales Service

1 1 1 1

1

1 1

98 126 19.8 1 1 1

MAX 99 24

0 0 0 0

0

0 0

0 0 5.1 0 0 0

MIN 15 0

1 1 1 1

1

1 1

98 126 18.2 1 1 1

MAX 95 24

Female

0 0 0 0

0

0 0

0 0 7.1 0 0 0

MIN 15 0

Male

Agriculture 0 1 0 1 0 Production 0 1 0 1 0 Source: Computed from the GLSS 5, Ghana Statistical Service

MIN 15 0

Male

ϭϰϮ

1 1

1 1 1 1

1

1 1

98 119 19.4 1 1 1

MAX 99 24

0 0

0 0 0 0

0

0 0

0 0 9.6 0 0 0

MIN 15 0

1 1

1 1 1 1

1

1 1

44 105 17.2 1 1 1

MAX 81 24

Female

Table 7.1b: Maximum and Minimum Values of Key Variables in the Wage Models Variable Full Sample (Total Employment) Paid Employment

0 0

0 0 0 0

0

0 0

0 0 5.3 0 0 0

MIN 15 0

1 1

1 1 1 1

1

1 1

98 126 19.8 1 1 1

MAX 98 20

0 0

0 0 0 0

0

0 0

0 0 5.1 0 0 0

MIN 15 0

Female

Self-employment Male

1 1

1 1 1 1

1

1 1

70 126 18.2 1 1 1

MAX 95 18

and maximum values of the continuous variables between male and female were not too large except age, years of experience and effort in paid employment where maximum values for females were substantially lower than their male counterparts (table 7.1b). The minimum and maximum values of discrete variables were 0 and 1 indicating that those variables are dummy variables.

7.3

Estimation Procedure

The main technique adopted in estimating wage equations is the Heckman selectivity approach to correct for underreporting of wages or earnings by respondents. Out of the total of workforce of over 14,000, about 9,500 people reported their wages leaving as many as 4,500 who did not provide information on their wages. In addition, zero wages was recorded for unpaid family workers which tend to create missing values when natural logarithm of wages is generated.

Clearly, the use of OLS estimation technique without correcting for underreporting could cause the estimated parameters to be biased. The general practice is to correct for sample selectivity bias using Heckman (1979) twostep procedure. The procedure involves the estimation of a probit model of reporting in the first stage to obtain an inverse Mills ratio. The ratio derived from stage one is used in the second stage OLS estimation of the wage function as a regressor to correct for possible biases22.

In estimating equation (7.1), we recognise the heterogeneity of the labour market to the extent that the outcome of estimated wage differentials between 22 The study uses STATA software package to estimate the earnings equations. ϭϰϯ

men and women in paid employment could be significantly different from self-employment. We therefore estimate equation (7.1) for the full sample (total number of people in employment), workers in self-employment and for workers in paid-employment. In addition, equation (7.1) is estimated for males and females separately to assess differences in returns between the two sexes. Equation (7.1) is also estimated for each of the seven broad categories of occupations to find out the wage differentials between male and female within each of the occupations.

7.4

Analysis of Empirical Results

Table 7.2 reports the results of equation (7.1) estimated using a two-step procedure of Heckman selection model for five different sets of samples covering total employment, the self-employed, the paid-employed, male workers and female workers in the labour market. The statistical significance of the inverse Mills ratio based on the sample covering full sample (i.e. total employment) and reported in column 1 of table 7.2 indicates a potentially sample selection bias of parameter estimates of wage equation (7.1) if OLS technique had been applied. Thus, the statistical significance at 1 percent level of the inverse Mills ratio suggests the appropriateness of using the Heckman two-step procedure to arrest the problem of selectivity bias from underreporting that would have resulted from the adoption of OLS procedures This is also true for the estimated results of equations (7.2), (7.3) and (7.4) for the full sample, covering the total employed people in the labour market as ZHOO DV VDPSOH RI ³PDOH´Rccupations reported in table 7.4. In each of these estimated results, the inverse Mills ratio was found to be statistically significant.

ϭϰϰ

In contrast, the sample involving males and that involving females as well as the samples of workers in self-employment and in paid-employment in table 7.2 are found to be free from selectivity bias from underreporting of individual monthly wages. This is based on the evidence that the inverse Mills ratios of these samples are not statistically significant and may be as a result of the reduced number of observations (and for that matter degrees of freedom) of results based on these samples. By implication, the adoption of OLS technique in the estimation of equation (7.1) based on these samples would have not suffered from selectivity bias. This also applies to the estimated results of wage equation 7.1 within occupation reported in table 7.3 and female occupation reported in table 7.4 where inverse Mills ratios are not statistically significant.

However, to ensure consistency in the analysis, we proceeded to use Heckman two-stage procedure in estimating the wage equation for self- and paid employment as well as male and female samples separately. This is because the results would not significantly differ from that obtained based on OLS estimation technique. The results also show that the statistical significance of the Wald statistic at 1 percent level signifies the rejection of the null hypothesis of all coefficients in the five different samples being equal to zero. Thus, in all cases, the variation in log of monthly wages is significantly explained jointly by the explanatory variables.

7.4.1 Gender Wage Differentials overall and by Employment Type The estimated results of equation (7.1) reported in tables 7.2 and 7.3 point to wage differentials in favour of men based on the full sample (i.e. total employment) and in different employment types and occupations in line with general expectation. The extent of the wage differentials between the two ϭϰϱ

sexes has been found to exist based on the type of employment and occupation. Evidently, a wider gender wage differentials is reported in selfemployment than in paid-employment. The results shown in the first column of table 7.2 indicate that controlling for all the relevant variables that influence LQGLYLGXDOV¶ PRQWKO\ wages, women earn 35 percent less than men in the entire labour market.

These observations are consistent with Beaudry and Sowa (1994) who established that the monthly wages of women in Ghana are 21 percent less than the average for men. The outcome of male-female wage differentials is also confirms Verner (1999) observation that female workers are paid 17 percent less than male workers in Ghana. The results further demonstrate in column 2 and 3, of table 7.2 that women in self-employment earn 51 percent less than self-employed men as against 25 percent in paid employment. The wage differentials recorded after controlling for the relevant factors of wage differentials could be interpreted to mean the existence of wage discrimination against women which is higher among workers in self-employment than in paid employment.

The estimated results of equation 7.1 lend empirical support to the role of human capital in the determination of wage differentials between men and women. The increasing wage effect of labour market experience also has implication for gender wage differentials. From the empirical results reported LQWDEOHDQDGGLWLRQDO\HDUVSHQWLQRQH¶VPDLQRFFXSDWLRQVKDVWKHHIIHFW of raising his/her monthly wage by 2.6 percent in total employment compared with 2.8 percent and 2.4 percent in self-employment and paid-employment respectively. It is estimated that the mean years of labour market experience of women is 12 years as against 14 years for men implying that potentially, ϭϰϲ

Table 7.2: Results of Wage Regression Model (Equation 7.1) with Sample Selection: Heckman Selection Model (two-step estimates) Dependent Variable: log of monthly wages Explanatory Variables

All 1

Self-Employed Paid-Employed Males 2 3 4

Females 5

Female Age Married Experience

±0.3495** ±0.0048** 0.1820** 0.0261**

±0.5097** ±0.0037** 0.1900** 0.0283**

±0.2490** 0.0017 0.2285** 0.0235**

----±0.0023 ±0.0030 0.2969** 0.1156** 0.0191** 0.0421**

Experience2 Education

±0.0004** 0.0400**

±0.0004** 0.0297**

±0.0004** 0.0719**

±0.0002** ±0.0008** 0.0400** 0.0445**

Effort Rural

0.0053** ±0.1135**

0.0069** ±0.1043**

0.0044** ±0.0480

0.0067** 0.0041** ±0.1204** ±0.1064**

Savannah (reference ecological zone) Coastal 0.1405** 0.3555** Forest 0.1468** 0.4233** Agriculture (reference occupation) Prof./Tech 1.0390** 0.7605** Adm/M¶Dgerial 1.5907** 2.3878** Clerical 0.9177** 0.7353 Sales 0.6294** 1.0034** Service 0.6998** 1.0997**

0.1996** 0.0534

0.2061** 0.2133**

0.2971** 0.2792**

0.3382** 0.6677** 0.2421* ±0.0027 0.0664

1.1195** 1.6907** 0.8775** 0.8139** 0.8117**

1.3243** 1.5716** 1.2248** 0.7385** 0.8174**

Production Constant Mills±lambda Wald Chi2 No of obs.

0.7178** 11.8892**

1.0397** 11.3046**

0.1576 11.9197**

0.7743** 0.8788** 11.4723** 11.2063**

±0.3277** 4775.30** 14,368

0.3849 1835.96** 8,487

±0.5889 933.35** 2,274

0.1022 ±0.2129 2211.21** 2605.54** 6,858 7,507

Censored obs.

4,719

1,402

26

1,735

2,981

Uncensored obs.

9,649

7,085

2,248

5,123

4,526

* Statistically significant at 5%

** Statistically significant at 1%

ϭϰϳ

women would generally receive lower wages that accrue from labour market experience than men. Clearly, the commitment of women to home production in response to societal norms and tradition as well as biological consideration that limits their accumulation of labour market experience can also be blamed for lower wages or earnings of women in Ghana.

Similarly, the significantly positive effect of education on wages coupled with limited labour market endowment of women has implications for wage differentials between men and women. The results show that additional year of education has an increasing effect on wages by 4 percent in the entire labour market as well as 3 percent and 7 percent in self-employment and paidemployment respectively. In effect, with the meaQ \HDUV RI ZRPHQ¶V education of 5 years compared with 7 years for men, women are bound to reap lower benefits in terms of higher wages associated with education than men. This observation is in line with the human capital theory of discrimination and earlier findings that support the importance of education on wage differentials in Ghana (see e.g. Canagarajah and Thomas, 1997; Verner, 1999; Schultz, 2003). The findings also support the argument that premarket discrimination in the form of limited access of girls to education has negative effect on ZRPHQ¶VODERXUPDUNHWSHUIRUPDQFHLQUHODWLRQWRZDJHV 7.4.2 Determinants of Female and Male Wages Columns 4 and 5 of table 7.2 present empirical results of equation (7.1) for male and female samples respectively. The results point to stronger returns to education and experience for women than men suggesting that women earn higher than men for additional year of schooling. Specifically, an additional \HDURIZRPHQ¶VHGXFation has a 4.5 percent increasing effect on their wages compared with 4 percent for men. Similarly, an additional year of labour ϭϰϴ

PDUNHW H[SHULHQFH LQFUHDVHV ZRPHQ¶V wages or earnings by 4.2 percent compared 1.9 percent for men. Apparently, this observation does not support observation by Ram (1982) that in West Africa, parents prefer to give their male children better education than female children on the conviction that the money spent on girls would yield less return relative to that spent on boys.

The observed high return to education and experience of women may be partly linked to the smaller number of educated and experienced women than men in employment and in the labour force relative to demand for their skills. In 2005/06, about 8 percent of women in employment had secondary education or better compared with 17 percent of men in employment with similar level of educational attainment. In contrast, about 45 percent of women in employment as against 28 percent of men had no education (table 2.7 in chapter 2). Women are also reported to have an average labour market experience of 12 years as against 14 years by men in 2005/06 (table 7.1a). Clearly, few highly educated and experienced women relative to the demand for their skills tend to raise their returns to education and labour market experience. This observation is reinforced by the inherent segmentation of the labour market into male and female dominated occupations such that few women who are able to break the barrier through education and enter the VXSSRVHGO\³PDOH´GRPLQDWHGRFFXSDWLRQVDWWUDFWSRWHQWLDOO\KLJKHUSULFHIRU their services. (VVHQWLDOO\KLJKHUUHWXUQVWRZRPHQ¶VHGXFDWLRQUHODWLYHWRPHQVXJJHVWWKDW policies geared towards increasing women access to education have a stronger effect of bridging the wage gap between the two sexes. In addition, the removal of cultural and institutional bottlenecks that contribute to intermittent EUHDNV LQ ODERXU PDUNHW DFWLYLWLHV FRXOG HQKDQFH ZRPHQ¶V DFFXPXODWLRQ RI ϭϰϵ

labour market experience and facilitate the bridging of gender wage gap in Ghana. This is consistent with the conclusion by Blau and Khan (1997) to the effect that the narrowing male-female gap in recent times can largely be traced to gains made by women in education and labour market experience.

The effort of men is better rewarded than that of women such that an additional hour per week spent on a job by a male worker increases their monthly wage by 0.7 percent compared with 0.4 percent for women. Both married men and married women are also reported to earn 30 percent and 12 percent more than their unmarried counterparts. The empirical results also show differences in wages between men and women based on location. Rural women are reported to earn 11 percent less than their urban colleagues compared with 12 percent lower wages of rural men relative urban men. Coastal women are also found to earn 30 percent more than women in savannah belt while wages of coastal men are estimated at 21 percent more than savannah men. In addition, women in forest zone are also estimated to earn 28 percent more relative to savannah women compared with 21 percent higher wages of men in forest zone relative to men in the savannah belt.

Differences in wages between the two sexes on the basis of broad occupational categorization are also reported in columns 4 and 5 of table 7.2. The results show that both men and women earn highest in administrative and managerial occupation followed by professional and technical, and clerical occupations in that order. Service occupation is observed to be the fifth highest paid occupation for both men and women. Sales occupation reported to be the fourth highest paid occupation for men becomes the sixth highest paid occupation for women while the sixth highest paid occupation (i.e. ϭϱϬ

production) for men is observed to be the fourth highest paid occupation for women.

Specifically, relative to agriculture, women in professional and technical, and clerical occupations earn 132 percent and 122 percent more while men in the same occupations earn 112 percent and 88 percent more respectively. In addition, wages of women in service and production occupations are estimated at 82 percent and 88 percent higher relative to women in agriculture with 81 percent and 77 percent higher wages recorded for men engaged in these occupations relative to agriculture. Women engaged in administrative and managerial, and sales occupations are observed to earn 157 percent and 74 percent relative to women in agriculture. Similarly, men working in administrative and managerial, and sales occupations earn 169 percent and 81 percent relative to male workers in agriculture. It is worth noting that these observations do not provide strong conclusion of clear wage differences between men and women by occupation. This makes it imperative to discuss wage differentials between the two sexes within each of the seven broad occupational categorisations.

7.4.3 Gender Wage Differentials within Occupations Table 7.3 reports the outcome of equation (7.1) estimated for each of the seven broad occupational categories in assessing the determinants of wage within occupations. In all, five equations reported gender wage differentials in favour of men based on the statistically significant negative coefficient of female dummy. Agriculture reported a wider gender wage differentials followed by sales, service, production, and professional and technical occupations in that order. Women in agriculture are said to earn 56 percent less than men, which is consistent with general expectation. ϭϱϭ

ϭϱϮ

Table 7.3: Results of Wage Regression Model (Equation 7.1) within Occupation with Heckman Selection Model (two-step Estimates) ± Dependent Variable: log of monthly wage Explanatory Prof./Tech Adm./M¶gerial Clerical Sales Service Agriculture Production Variable 1 2 3 4 5 6 7 Female ±0.2384** ±0.5764 0.1239 ±0.4880** ±0.3955** ±0.5612** ±0.2997** Age 0.0064 ±0.0100 0.0130 ±0.0041 ±0.0034 ±0.0114 0.0029 Married 0.1761* ±0.0717 0.0954 0.1501* 0.1005 0.2069** 0.1293** Experience 0.0258** 0.0424 0.0334 0.0478** 0.0700** 0.0199** 0.0306** Experience2 ±0.0003* ±0.0014 ±0.0007 ±0.0010** ±0.0021** ±0.0002* ±0.0006** Yrs of education 0.1184** 0.1511* 0.0625* 0.0409** 0.0591** ±0.0050 0.0358** Effort 0.0077** 0.0019 0.0029 0.0035* 0.0020 0.0057** 0.0055** Rural ±0.1649* ±0.0581 ±0.1174 ±0.3338** ±0.1000 0.0723 ±0.0949* Coastal 0.2482* 0.4709 0.3383 0.1517 0.1449 0.0022 0.1390 Forest 0.0859 0.0605 0.3793* 0.1958 0.1301 0.3293** 0.0962 Paid employed 1.2583 dropped 0.1597 0.3879 ±0.0732 0.6660** 0.1428 Self-employed 0.9000 0.9559 0.0575 0.2303 0.0101 0.0043 0.1804 Unpaid family work (reference dummy) Constant 9.9641** 12.1581** 11.4408** 12.8741** 12.5162** 12.4928** 12.4071** Mills ± Lambda 0.3598 ±1.7213 ±0.1791 ±0.3097 ±0.4885 ±0.7380 0.5145* Wald Chi2 416.81** 6.97 38.69** 511.6** 457.38** 1525.44** 872.25** Censored observations 26 1 4 139 117 4,021 408 Uncensored observ. 653 58 162 1,216 1,217 4,226 2,117 Savannah zone is reference dummy for ecological zone * Statistically significant at 5% ** Statistically significant at 1%

7KHKXJH³ZDJHVKRUWIDOO´RIZRPHQPHDVXUHGE\WKHHVWLPDWHGFRHIILFLHQWRI female dummy is in support of Beaudry and Sowa (1994) who also found wider gender wage differentials in the farming sector. They attributed the high wage shortfall for women to their concentration in subsistence farming. Clearly, the current lower wages of women relative to men could be largely linked to the substantial proportion of women working in agriculture as unpaid workers. Estimates from the GLSS5 indicate that about 55 percent of women in agriculture are working as unpaid labourers compared with only 18 percent of men in unpaid agriculture occupation. Essentially, this substantial proportion of women in agriculture who earn virtually nothing largely accounts for the wider male-female wage gap in that occupation.

In sales and service occupations which are reported to be dominated by women (see table 2.2 in chapter 2), men earn 49 percent and 40 percent respectively more than women. Similarly, women are also observed to earn 30 percent and 24 percent less than their male counterparts in production and professional and technical occupations respectively. Clearly, in production, sales and service occupations men are mostly engaged as wage employees compared with women who are mostly working as self-employed with relatively lower wages.

Indeed, estimates from GLSS5 show that over 63 percent of men against 13 percent of women in production occupation are working for pay while 32 percent of men and 76 percent of women work as self-employed. Similarly, about 47 percent of men compared with 7 percent of women in production and 36 percent of men against 6 percent of women in sales are engaged as paid employees. In contrast, about 39 percent of men compared with 80 percent of women in production and 60 percent of men against 85 percent of women in ϭϱϯ

sales occupation operate as self-employed workers in these occupations. Undoubtedly, the high representation of men as paid-employees and women as self-employed workers in these occupations could also explain gender wage differentials after controlling for the relevant determinants of wages.

In professional and technical occupations, men are estimated to earn 24 percent higher than their female counterparts. Besides administrative and managerial occupation, professional and technical occupation also recorded higher returns to education than the other five occupations. Specifically, an additional year of education increases individual wages or earnings by 12 percent in professional and technical and 15 percent in administrative and managerial occupations compared with less than 9 percent in other occupations. Estimates from GLSS5 indicate that 43 percent of men compared with 25 percent of women in professional and technical occupation have higher than secondary education. The higher educational attainment of men relative to women implies that men would be potentially engaged in higherranked jobs, and would receive higher wage than women in the occupation. The lower educational attainment of women relative to men does not make them derive much benefit from higher returns to education in professional and technical occupations.

7.4.4 Other Sources and Forms of Gender Wage Differentials One potential source of male-female wage gap could be linked to the declining effect of increasing female occupational composition on individual wages or HDUQLQJV2QHPDMRUSUHGLFWLRQRIWKH³FURZGLQJ´K\SRWKHVLVLVWKHFURZGLQJ or concentration of women into some specific occupations resulting in reduced wage of workers in such occupations. The outcome of the estimated wage equation (7.2) provides empirical evidence for this argument. Columns 1, 2 ϭϱϰ

and 3 of table 7.4 present the outcome of empirical investigation into the female occupational composition effect on wages or earnings. The empirical results show that wage differentials in favour of men can be traced to the negative female composition effect on wages. In the labour market as a whole (full sample or total employment), a one percentage point increase in the female share in an occupation exerts a 0.4 percent declining effect on monthly wages of workers. The effect is higher on female wages than male wages. Specifically, a one-percentage point increase in female occupational composition results in a decline in wages by 0.41 percent among females compared with a decline in wages of 0.29 percent among males. Thus, a declining female occupational composition effect on wages is observed to be stronger among females than males.

The significantly negative female composition effect on wages or earnings FOHDUO\ OHQGV HPSLULFDO VXSSRUW WR WKH ³FURZGLQJ´ K\SRWKHVLV RI LQVWLWXWLRQDO theory of discrimination that the concentration of females into specific occupations tends to compress wages in that occupation. Essentially, the negative female composition effect on wages or earnings has implication for wage differentials between male and female workers in male occupations and female occupations. This is in line with the key prediction of the crowding hypothesis of discrimination that suggest that individuals confined into women dominated occupation earn less than their counterparts in men dominated occupations.

The estimated result of equation (7.3) represented in column 6 of table 7.4 gives empirical evidence to support this hypothesis that male dominated occupation pay better than female dominated occupations. From the results in ϭϱϱ

---------

±0.0040**

1 -----------

±0.0029**

2 ---

Equation 7.2 Full Sample Males

---------

±0.0041**

3 ---

Females

0.2256** 0.2383**

Coastal Forest

0.0058

0.1816** 0.2048**

0.2837** 0.0175** ±0.0003** 0.0575** 0.0088** ±0.0900*

Savannah belt (reference or omitted dummy)

0.2284** 0.0219** ±0.0004** 0.0685** 0.0090** ±0.0887**

0.0067*

Married Experience Experience2 Yrs of education Effort Rural

Age

ϭϱϲ

0.3721** 0.3455**

0.1180** 0.0383** ±0.0008** 0.0510** 0.0050** ±0.1032*

0.0031

Women in male occupations (reference or omitted dummy)

Female occupation Women in Fem Occupation Men in Fem Occupation Men in Male Occupation

Fem composition

Female

Explanatory Variables

0.2308** 0.2360**

0.2265** 0.0217** ±0.0004** 0.0712** 0.0095** ±0.1028**

0.0071

±0.1941** -------

---

4 ---

Equation 7.3

0.28095** 0.2674**

0.2333** 0.0161** ±0.0003** 0.0833** 0.0113** ±0.0695**

0.0135**

---------

---

5 ±0.4792**

Male Occupation

0.2687** 0.2817**

0.1173** 0.0472** ±0.0009** 0.0385** 0.0025* ±0.1518**

±0.0089*

---------

---

6 ±0.3353**

Female Occupation

0.2727** 0.2648**

0.1951** 0.0212** ±0.0004** 0.0628** 0.0082** ±0.1130**

0.0074*

--0.1464** 0.4848** 0.4863**

---

7 ---

Equation 7.4

Table 7.4: Wage Regression Results±Heckman (two-step) Sample Selection: Dependent Variable: log monthly wages

0.7520**

0.8154** 0.8077**

Sales

Service Production

0.8032** 0.7001**

0.8879**

1.0720** 1.5614** 0.8617**

Uncensored observations No. of observation

5,123 6,858

2168.46** 1,735

0.6496**

10.9083**

4,526 7,507

2431.81** 2,981

0.2050

10.7714**

1.1009** 1.1515**

1.0281**

1.3563** 1.6469** 1.4715**

ϭϱϳ

** Statistically significant at 1%

9,649 14,368

Wald Chi2 Censored observations

* Statistically significant at 5%

0.5227**

4440.92** 4,719

Mills ± lambda

10.6722**

Constant

Agriculture (reference or omitted occupation dummy)

1.0775** 1.5343** 1.0116**

Professional/Technical Admin/Managerial Clerical

9,649 14,368

4385.09** 4,719

0.5545**

10.4324**

0.847** 0.8986**

0.8025**

1.1210** 1.6369** 1.0550**

6,317 10,533

3442.68** 4,216

0.9747**

9.9217**

0.8129** 0.8012**

2.5555**

1.1610** 1.5838** 0.9574** 0.7970** 0.8043**

±0.3760** ±0.3602*

3,332 3,832

697.11** 500

±0.6917

9,649 14,368

4699.06** 4,719

0.5939**

10.1952**

0.7202**

±0.4363**

13.1790**

1.1610** 1.6252** 1.0444**

±0.0770 dropped dropped

table 7.4, ³IHPDOH´ RFFXSDWLRQV DUH UHSRUWHG WR SD\ SHUFHQW OHVV WKDQ ³PDOH´RFFXSDWLRQV&OHDUO\WKLVPD\EHRQHRIWKHXQGHUO\LQJUHDVRQVIRU the lower wage of women than men in the labour market.

The larger declining effect of female composition on wages of women than men have potential effect on wage differentials between men and women ZLWKLQ ³IHPDOH´ DQG ³PDOH´ RFFXSDWLRQV &ROXPQV DQG RI WDEOH which report the estimated results of equation (7.1) for male and female occupations provide evidence to indicate lower wages of women than men in both occupations. Women are reported to earn 34 percent less than men in female occupations compared with 48 percent less in male occupation after controlling for all the relevant personal, demographic, and productive characteristics as well as occupation. This observation is consistent with earlier findings that show stronger negative female composition effect on female wages than male wages. This appears to suggest a higher wage discrimination against women in male occupation than female occupation.

Considering the fact that male occupations are found to pay better than female occupations, it is expected that women engaged in male occupations would be paid better than women working in female occupations in spite of the seemingly wage discrimination facing them in both occupations. On the other hand, the evidence of lower wages associated with female occupations may imply that men engaged in these occupations may potentially suffer in terms of lower wages relative to men working in male dominated occupations. This is captured by equation (7.4) and the estimated results represented in column 7 of table 7.4.

The empirical results show that contrary to expectation that women in male occupations would benefit from the high wages associated with that occupation relative to women in female occupations, the opposite is the ϭϱϴ

case. Indeed, women in female occupations are reported to rather earn 15 percent higher relative to their colleagues in male occupation based on coefficient 0.1464 of women in female occupation variable. This clearly shows greater wage differentials against women in male occupation compared to that which occurs in female occupations. On the other hand, the lower wages associated with female occupation did not have substantial adverse effect on male wages in that occupation relative their colleagues in male occupations. Specifically, relative to women in male occupations (reference dummy), men in male occupations are observed to earn 48.63 percent higher compared with 48.48 percent higher wages of men in female occupations.

Undoubtedly, the substantial wage differentials against women in male occupation make them worse off for engaging in that occupation relative to women in female occupations. In contrast, men in female occupations relative to their colleagues in male occupations only suffer adverse marginal effect of lower wages for engaging in female occupations. These observations reinforce the seemingly wage discrimination that women face in the Ghanaian labour market regardless of which occupation they find themselves.

ϭϱϵ

Chapter 8

Measuring Gender Wage Discrimination in the Labour Market in Ghana 8.0

Introduction

In chapter 7, we investigated the existence of wage discrimination by estimating wage function with a focus on female dummy as the variable of interest and controlling for relevant determinants of wages. The primary focus of this chapter is to measure the extent of wage differentials emanating from discrimination. We adopt the Oaxaca-Blinder and Nuemark-Oaxaca-Ransom wage decomposition approaches to show the proportion of gender wage differentials that can be explained and the proportion that cannot be explained as a measure of the extent of discrimination. The main source of data for this exercise is the most recent nationally representative household survey, the GLSS5.

8.1

Wage Decomposition

The coefficient of female dummy in equation (7.1) in chapter 7 facilitates the assessment of lower or greater female wages relative to their male counterpart. Having controlled for relevant variables that influence individual wages or earnings, the greater the coefficient of the female dummy the larger the monetary impact of discrimination. However, this approach only indicates how much lower (if coefficient of female dummy is negative) female wages are compared with male wage holding all relevant variables constant. It could be used to explain the existence of wage discrimination but does not necessarily show what proportion of this ϭϲϬ

wage difference is attributable to differences in productive and other relevant characteristics and what proportion is due to discrimination.

A wage decomposition approach widely used in the literature is therefore adopted to determine the proportion of gender wage gap that could be explained by relevant observable variables and the proportion that would reflect unobservable gender differences which is termed in the literature as wage discrimination.

8.2

Model Specification

The wage decomposition approach classifies the sources of gender wage gap into the part explained by differences in personal and productivity enhancing characteristics, as well as location and occupation and the part ³XQH[SODLQHG´ E\ VXFK GLIIHUHQFHV 7KH XQH[SODLQHG wage gap reflecting unobserved differences between men and women that influence wage is termed in the literature as wage remnant, and is considered to be a measure of the extent of wage or earnings discrimination.

The original Oaxaca-Blinder decomposition approach (Oaxaca, 1973; Blinder, 1973) is used to measure the proportion of wage differences that could

be

explained

by differences

in

observable

characteristics

(endowment) and the part that could be attributed to discrimination. There is another standard decomposition framework by Neumark (1988) and Oaxaca and Ransom (1994) which is an improvement over Oaxaca-Blinder decomposition that breaks male-female wage gap into part explained by differences in endowments and other parts attributable to male advantage and female disadvantage.

ϭϲϭ

8.2.1 Blinder-Oaxaca Decomposition Technique The Blinder-Oaxaca Decomposition technique separates the sources of wage gap LQWR WKH SDUW ³H[SODLQHG´ E\ GLIIHUHQFHV LQ observable FKDUDFWHULVWLFV DQG WKH SDUW ³XQH[SODLQHG´ E\ VXFK GLIIHUHQFHV 7KH approach begins with the specification of wage decomposition function in a Mincerian form similar to equation (7.1) for male and female respectively as:

ln Wi m X im ' E jm P i ln Wi

f

(8.1a)

X i ' E j Pi f

f

(8.1b)

Pi ~ N (0,1) where X is a vector of characteristics relevant in determining wages including age, marital status, experience, education, effort, residence of an individual in terms of rural-urban and ecological zone, and occupation; ȕj denotes a vector of coefficients of (or returns to) these characteristics; and µ i represents a random error term assumed to be normally distributed with zero mean and constant variance. The superscripts m and f denote males and females respectively and subscript i represent observations of individual workers.

Based on the properties of OLS estimation technique, we write the wage differential equation in terms of average between male and female as:

m

f

ln W i ln W i

m

f

X i ' E jm X i ' E jf

(8.1c)

where ln W i m and ln W i f are mean log monthly wages of male and female and

Xi

m

and

X

m i

are vectors containing the means of independent variables ϭϲϮ

³HQGRZPHQWV´ IRU PDOHV DQG IHPDOHV UHVSHFWLYHO\ 7KHVH DUH WKH characteristics or means of explanatory variables that influence wages of males and females. The estimated ȕm and ȕf are the estimated parameters obtained from the Mincerian-type wage functions of male and female f

respectively. By subtracting and adding X i ' E jm from equation (8.1c), we obtain

m

f

m

f

f

f

X i ' E jm X i ' E jm X i E jm X i ' E jf ,

ln W i ln W i

which eventually yields

m

f

ln W i ln W i

X

m' i

f

X i ' E jm X i f ' E jm E jf

(8.1d)

Equation (8.1d) is the Oaxaca-Blinder wage decomposition. The first term of the equation reflects the differences in mean wage due to differences in observable or explained characteristics between male and female. It is based on estimates of what a female would receive if she faced the male wage structure. The second term represents the differences in the average wage due to the shift coefficient ȕ i.e. the differential returns to the explained variables or observable characteristics between males and females7KLVLVZKDWLVUHIHUUHGWRLQWKHOLWHUDWXUHDV³ZDJHUHPQDQW´DQG is used to measure the extent of wage discrimination.

The key problem with the Oaxaca-Blinder decomposition however is an LQKHUHQW ³LQGH[ QXPEHU SUREOHP´ (TXDWLRQ 8.1d) could have been expressed in a manner which would show the first term being expressed in terms of how much a male would earn if paid according to the female wage structure, i.e. ϭϲϯ

m

f

ln W i ln W i

X

m' i

f

X i ' E jf X im ' E jm E jf

(8.1G¶¶

The wage decomposition can be quite sensitive to which wage structure is used. As noted by Cotton (1988) and later by Oaxaca and Ransom (1994), the choice of non-discriminatory wage structure (either male or female) will yield different results. For instance, a study by Appleton et al (1999) indicate a 38 percent wage differential due to differences in returns based on female wage structure but 212 percent differential using male wage VWUXFWXUH LQ &RWH G¶,YRLUH 7KH\ DOVR UHSRUWHG D percent wage differential due to differences in returns using female wage structure compared with 74 percent differentials based on male wage structure in 8JDQGD 7R DYRLG WKH SRVVLELOLW\ RI EHLQJ FDXJKW LQ WKH ZHE RI ³LQGH[ QXPEHU SUREOHP´ ZH DGRSW DQRWKHU wage decomposition approach, the Neumark-Oaxaca-Ransom technique, considered to be an improvement over Blinder-Oaxaca decomposition techniques in the gender wage decomposition analysis.

8.2.2 Neumark and Oaxaca-Ransom Decomposition Approach The Neumark and Oaxaca-Ransom approach attempts to provide a remedy IRUWKH³LQGH[QXPEHU´SUREOHPLQKHUHQWLQ%OLQGHU-Oaxaca decomposition technique by using a weighted average of the female and male wage structures in a pooled model for both male and female. According to Neumark (1988), the appropriate decomposition depends on the type of discrimination hypothesised. Thus, employers may adopt behaviours that favour men or discriminate against women. Extending the argument to cover self-employment impliHVWKDWWKH³IDYRXULWLVP´LQIDYRXURIPHQDQG discrimination against women in this regard may emanate from customers or the market system shaped by cultural norms and practices among others.

ϭϲϰ

In this gender wage decomposition technique, in a situation where males are favoured (i.e. pure nepotism), and females are paid the competitive wage and males overpaid, implies that females¶ wage function would provide an estimate of the non-discriminatory structure. On the other hand, in situations where females face discrimination (i.e. pure discrimination approach) and are underpaid with their male counterpart receiving competitive wage suggests that coefficients of male wage function are taken as the non-discriminatory wage structure. Based on the restriction thaW HPSOR\HUV¶ SUHIHUHQFHV DUH KRPRJHQRXV RI degree zero within each type of labour (male or female), Neumark shows that the non-discriminatory wage structure can be estimated from a wage function over the pooled sample (both males and females). As demonstrated by Appleton et al (1999), the non-GLVFULPLQDWRU\RU³SRROHG´ wage or wage structure ȕp is a weighted average of the male and female wage structures:

E :E m (1 :)E f

(8i)

7KHZHLJKWLQJPDWUL[DFFRUGLQJWR2D[DFDDQG5DQVRP LV

: ( X ' X ) 1 ( X m' X m )

(8ii)

where X is the observation matrix for the pooled sample and X m, the corresponding matrix for males only. This weighting is a generalisation of the one proposed by Cotton (1988) which used the proportion of men and women in employment.

ϭϲϱ

Generally, the Neumark decomposition approach adjusts the BlinderOaxaca decomposition model to show the difference in mean wages between male and female is decomposed as: ln W i ln W i m

f

X

m i

>

X i E j E j E j X i E0 E0 f

p

m

p

m

m

p

@ >E

p j

E jf X i f E 0p E 0f

@

(8.2) where the estimated ȕp refers to the Oaxaca-Ransom and Neumark nondiscriminatory wage structure that is obtained by estimating the parameters based on the pooled sample of both the male and female demographic groups; and the other terms in the equation remained as already defined.

The first term of equation (8.2) reflects the component of average wage difference between males and females that is attributable to differences in means of the explanatory variables, weighted by the estimated coefficients of the non-discriminating wage structure. The second and the third terms reflect the contribution of differences between actual and pooled returns for men and women referred to as deviation of male and female returns respectively. Indeed, the two parts of the equation measure the different manner with which the labour market rewards the characteristics of male and female workers relative to the benchmark pooled wage structure. The second term is interpreted as the advantage of males and the third term as disadvantage of females.

In Ghana, advantage of men that feeds into male-female wage gap could take the form of easy access to credit and ownership of assets by men relative to women particularly in the informal sector. Clearly, one of the PDLQIDFWRUVWKDWKDYHVKDSHGWKHGLUHFWLRQRIZRPHQ¶VHFRQRPLFDFWLYLWLHV in Ghana has been the male-biased allocation of traditional entitlements and modern assets, a relic of neo-patriarchal ideologies of gender relations ϭϲϲ

found in both patrilineal and matrilineal kinship systems in all parts of the country (Aryeetey, 2000). On the other hand, disadvantage of women could stem from biological characteristics of women that affect their role in the home (e.g. pregnancy, childbearing etc) and in effect prevent them from being assigned higher responsibility associated with higher wages and less degree of flexibility to be able to carry out home production alongside market work.

The Neumark and Oaxaca-Ransom gender wage decomposition technique according to Appleton et al (1999) is attractive but they argue that its interpretation should be done with some caution. They assert that it is not clear that the pooled coefficients would be a good estimator of the nondiscriminatory wage structure considering the fact that there is no evidence of the validity of the zero-homogeneity restriction on employer preferences. Nonetheless, it provides some basis for estimating and assessing the extent to which favouritism or nepotism for males and discrimination against females affects male-female wage gap.

8.3

Estimation Procedure and Data Source

The main source of data in carrying out empirical analysis of the extent of gender wage discrimination is the GLSS5 dataset which was also used estimating wage differentials in chapter 7. The Oaxaca-Blinder wage decomposition and Neumark-Oaxaca-Ransom wage decomposition models under consideration are represented in equations (8.1d) and (8.2) respectively. In estimating the two models, we recognise the potential effect of underreporting of wages on the estimated results of the wage decomposition models. As indicated in chapter 7, using OLS technique in estimating the male and female wages functions may produce biased parameter estimates due to underreporting of wages. The significance of the inverse Mills ratio of estimated wage equation (7.1) for a sample of all ϭϲϳ

employed adults shown in table 7.2 in chapter 7 underscores the necessity of using Heckman two-step approach to correct for potential selectivity effect due to underreporting.

We therefore adjust equations (8.1d) and (8.2) to incorporate selection effects in estimating gender wage gap respectively as:

m

X

f

ln W i ln W i

m' i

f

X i ' E jm X if ' E jm E jf I jm Oim I jf Oif

(8.1G¶

and

ln W i ln W i m

f

X

m i

>

@ )@ I O I O


p

>

m

p

(E jp E jf ) X if (E 0p E 0f

m

m

m m j i

p

f f j i

¶

where Ȝm and Ȝf represents inverse Mills ratio for male and female model respectively and I m and I f denote the coefficients of the inverse Mills ratio for males and females wage functions. The estimated ȕ are the estimated parameters by the Heckman procedure separately for males and females while other variables remain as defined. The last part of the two equations is the selectivity correction factor while other parts of the models remain as explained already. One approach noted in the literature23 and adopted by Neumark and Oaxaca (2005) is to simply net out the selection effects as a result of underreporting from the wage gap as:

§¨ ln W im ln W if ·¸ I mOm I f O f j i j i © ¹

§¨ X im' X if ' ·¸E m X f ' E m E f i j j © ¹ j

(8.1G¶¶

23 See for example Duncan and Leigh (1980), Reimers (1983) and Boymond et al (1994) ϭϲϴ

and

lnW

m i

ln W i I jmOim I jf Oif f

X

m i

>


>

p

m

p

m

m

E j E j X i E0 E0 p

f

f

p

f

p

@

@

(8.2¶¶

This is consistent with the estimated gender wage decomposition results generated by STATA software using Oaxaca command developed by (Jann, 2008). The estimation of the two gender decomposition equations (8.1G¶¶ DQG (8.2¶¶ DERYH GRHV QRW JHQHUDOO\ WDNH LQWR DFFRXQW WKH UHOHYDQFH RI WKH differences in sectoral or employment structures between the two sexes. Many studies on gender wage gap in Africa have focused on the important distinction between public and private sectors (e.g. Appleton, 1999; Kabubo-Mariara, 2003). In this analysis, after controlling for occupation, we estimate equations (8.1G¶¶ DQG 8.2¶¶ IRU REVHUYDWLRQ RI ZRUNHUV LQ self-employment, paid-employment and full sample covering total employment.

8.4

Empirical Analysis of Gender Wage Decomposition

The results of the gender wage decomposition for paid-employment, selfemployment and full sample (i.e. all workers) using Blinder-Oaxaca and Neumark-Oaxaca-Ransom methods are shown in tables 8.1 and 8.2 respectively. The estimated results generally indicate a significant difference between uncorrected and selectivity corrected gender wage gap. While in paid employment and the full sample of employed people, the estimated gender wage gap drops after correcting for potential selectivity bias due to underreporting, the reverse is the case in self-employment. This appears to suggest that the potential selectivity effect due to underreporting ϭϲϵ

of wages is weaker in self-employment sample than in paid-employment and all employed adults in the labour market.

One important observation from the estimated results is the extent to which the alternative decomposition techniques produce quantitatively similar figures at least in terms of the a priori signs. From table 8.1, using either male or female wage structure, a more favourable characteristic for males than for females was reported without much difference in magnitude especially in the full sample. Similarly, in self-employment, differences in characteristics between the two sexes were more favourable to females, based on either male or female wage structure with marginal difference in terms of absolute values. This means that the Blinder-Oaxaca decomposition does not necessarily encounter any significant index number problem. Specifically, the results of Blinder-Oaxaca wage decomposition indicate that about 59 percent and 66 percent of the male-female mean log monthly wage gap of all workers (i.e. full sample of employed people) using male and female wage structure respectively is attributable to unexplained factors. By implication, about 41 percent and 34 percent of the gender wage gap using male and female wage structure respectively is directly linked to the differences in observable characteristics in favour of male workers.

The relatively large proportion of gender wage differentials in the entire labour market attributable to unexplained considerations is supported by Kabubo-Mariara (2003) who found 78 percent of gender wage gap attributable to unexplained factors in Kenya using either female or male wage structure. It is however in sharp contrast with the observations by $SSOHWRQ WKDW SHUFHQW RI JHQGHU ZDJH JDS LQ &RWH G¶,YRLUH LV attributable to returns to characteristics based on female wage structure and 312 percent based on male wage structure all in favour of women. ϭϳϬ

Table 8.1: Results of Blinder±Oaxaca Decomposition of Gender Wage Gap Full Sample 12.893** 12.442**

Mean log of Male monthly wage Mean log of Female monthly wage Wage gap Selectivity Corrected wage gap Using male wage structure Observed characteristics E m

X

m

X

f

Paid

Self-

Employed Employed 13.516** 12.606 13.202** 12.353

0.451** 0.233**

0.314** 0.226**

0.253 0.710

0.096

0.043**

±0.158**

(41.1) Productivity enhancing characteristics1 0.110*

(19.0)

(±22.3) 0.357** (50.3)

(47.1)

0.020** (8.9)

Others: age, location, occupation etc

±0.014 (±6.0)

0.023* (10.1)

±0.515 (±72.5)

Returns to characteristics E m E f X m

0.137**

0.183

0.894

Total Variation

(58.9) 0.233

(81.0) 0.226

(122.3) 0.710

100.0

100.0

100.0

0.079

0.069*

±0.184**

(33.9) 0.093** (39.7) ±0.014

(30.6) 0.062* (27.5) 0.007

(±25.9) 0.342** (48.9) ±0.526

Using female wage structure Observed characteristics E f

X

m

X

f

Productivity enhancing characteristics Others: age, location, occupation etc

(±5.8)

(3.1)

(±74.1)

Returns to characteristics E m E f X m

0.154**

0.157**

0.894

Total Variation

(66.1) 0.233 100.0

(69.4) 0.226 100.0

(125.9) 0.710 100.0

Note: Figures in parenthesis are percentages * 5% significant level ** 1% significant level; 1 Productive enhancing Characteristics comprises education, experience and effort

ϭϳϭ

In paid-employment, the Blinder-Oaxaca decomposition results suggest a quite huge percentage of wage gap attributable to unexplained factors WHUPHGDV³ZDJHUHPQDQW´6SHFLILFDOO\ percent of gender wage gap is found to be attributable to factors other than observed characteristics using male wage structure with less than one-fifth connected to differences in observable characteristics in favour of men. Similarly, based on female wage structure, the unexplained factors are reported to account for at least 69 percent of the gender wage gap while differences in observable characteristics account for less than one-third of the wage gap. This observation is in support of conclusion reached by Glick and Sahn (1997) in Guinea-Conakry to the effect that about a quarter of gender wage gap emanated from differences in observable characteristics. They however observed favourable gender wage gap for women in private wageemployment sector reflecting the superior schooling of women private sector employees.

In self-employment, female workers are reported to have more favourable observable characteristics than their male counterparts regardless of the decomposition technique used. This implies that over 100 percent of the male-female wage differentials in self-employment result from the differences in returns to characteristics between the two sexes using Blinder-Oaxaca approach (see table 8.1). The results show that about 122 percent and 126 percent of the gender wage gap is attributed to unexplained variation measured by differences in returns to characteristics using male and female wage structure respectively. This is in contrast with findings by Glick and Sahn (1997) that put the proportion of gender gap attributable to factors other than differences in observable characteristics in selfemployment at 56 percent.

ϭϳϮ

The

empirical

results

of

Neumark-Oaxaca-Ransom decomposition

technique reported in table 8.2 point to a stronger contribution of differences in observable characteristics to gender wage gap of all workers in the labour market compared with results of Blinder-Oaxaca decomposition technique in table 8.1. This is based on evidence that about 47 percent of gender wage gap is attributable to differences in observable characteristics based on the pooled wage structure compared with 41 ercent and 34 percent using male and female wage structure respectively. About 46 percent of gender wage gap based on Neumark and Oaxaca-Ransom decomposition technique is linked to deviation of male returns from pooled wage structure and 6 percent attributable to deviation of female returns from the pooled wage structure. This intuitively suggests a more SURQRXQFHG IDYRXULWLVP RU ³SXUH QHSRWLVP´ WRZDUGV PDOH ZRUNHUV DQG D minimal discrimination against women in the labour market as a whole.

This observation is however not consistent with the findings of Appleton (1999) and Kabubo-Mariara (2003) who found discrimination against women more pronounced than nepotism towards men in Uganda and Kenya respectively. Our findings do not also support the observation by Sebaggah (2007) that discrimination is more pronounced in Uganda with no evidence of nepotism towards men.

In paid-employment, a more pronounced discrimination against women and favouritism for men were recorded. Specifically, an estimated 33 percent of wage gap is attributable to deviation of female returns from the pooled wage structure and about 38 percent wage gap explained by deviation of male returns from the pooled wage structure. In effect, from Neumark (1988) interpretation, both nepotism towards men and discrimination against women are important in paid employment with the former being marginally more pronounced. Discrimination against women in paidϭϳϯ

employment may take the form of unwillingness of employers to hire women due to sheer prejudice or an attempt to avoid potentially higher labour cost associated with hiring women as a result of their higher degree of absenteeism and intermittent breaks in market work related to childbearing. Favouritism for men may also be related to the advantage men have due to education and cultural norms that often get them to the top hierarchy on the job ladder. Table 8.2: Results of Nuemark-Oaxaca-Ransom Decomposition of Gender Wage Gap (based on non-discriminatory wage structure) All Mean log of Male monthly wage

12.893**

Paid Employed 13.516**

Mean log of Female monthly wage Wage gap

12.442** 0.451**

13.202** 0.314**

12.353 0.253

Selectivity Corrected wage gap

0.233**

0.226**

0.710

0.110

0.065

±0.164

(47.3)

(28.9)

(±23.2)

0.108

0.085

0.366

Observed Characteristics E

p

X

m

X

Male advantage E m E p X m

f

SelfEmployed 12.606

(46.4)

(37.7)

(51.6)

Female disadvantage E p E f X f

0.015

0.076

0.508

Total Variation

(6.3) 0.233

(33.4) 0.226

(71.6) 0.710

(100.0)

(100.0)

(100.0)

Note: Figures in parenthesis are percentages * 5% significant level ** 1% significant level; 1

Productive enhancing Characteristics comprises education, experience and effort

In

self-employment,

the

outcome

of

Neumark-Oaxaca-Ransom

decomposition approach show that about 72 percent of gender wage gap (see table 8.2) is connected to disadvantage of ZRPHQ WHUPHG DV ³SXUH ϭϳϰ

GLVFULPLQDWLRQ´ DJDLQVW ZRPHQ. On the other hand, advantage of men WHUPHG DV ³SXUH QHSRWLVP´ IRU PHQ XQGHUVFRUHV SHUFHQW of wage gap among the self-employed. By implication, nepotism towards men seems to be less important in the determination of gender wage differences than discrimination against women in self-employment while the reverse is the case in paid-employment and total employment in the labour market. Discrimination against women in self-employment may stem from LQVWLWXWLRQDODQGFXOWXUDOEDUULHUVWKDWLPSHGHZRPHQ¶VDFFHVVWRDIIRUGDEOH credit, land and other productive assets relative to their male counterparts to facilitate and promote their economic activities.

Overall, the reported differences in productivity enhancing characteristics largely influenced by labour market endowment between men and women is found to be an important contributory factor in determining male-female wage gap in Ghana. In both approaches, the contribution of productivity enhancing factors is favourable for male workers in total employment (full sample) and in self-employment as well as in paid-employment. Evidently, the contribution of productivity enhancing characteristics is relatively high in self-employment than in paid-employment.

From table 8.1, differences in productivity enhancing characteristics in favour of men accounted for 47 percent (two-third of which stems from differences in education) of the male-female wage gap of all workers based on male wage structure. Similarly, about 34 percent (70 percent from differences in education) of the gender wage gap of all employed people in the labour market was accounted for by favourable productivity enhancing characteristics for men using female wage structure. In both approaches however, the differences in other characteristics in favour of women caused a narrowing effect on the overall favourable differences in observed characteristics for men. ϭϳϱ

In paid employment, the contribution of differences in observed characteristics to gender wage gap significantly varied between the use of male and female wage structures. Based on male wage structure, only 19 percent of gender wage gap could be explained by differences in observable characteristics compared with 31 percent using female wage structure (table 8.1). In terms of the detailed classification of variation in observed characteristics, only 9 percent (education accounting for less than one-tenth) of gender wage gap is linked to differences in productivity enhancing characteristics using male wage structure. A strong contribution of productivity enhancing characteristics to gender wage gap of 28 percent (education accounting for less than one-twentieth) is however reported based on female wage structure. Although, there are inherent difficulties in making conclusion about the contribution of variation in productivity enhancing characteristics to gender wage gap among paid-employees, one issue is quite clear. Whichever way one looks at the results, differences in educational attainment does not seem to have significant effect on malefemale wage gap among paid-employees.

The role of differences in productivity enhancing characteristics, especially education appears to be very high in the determination of gender wage gap among workers in self-employment. This is based on the evidence that using male wage structure, at least 50 percent (123 percent24 of which comes from better educational attainment of men than women) of gender wage gap is accounted for by differences in productivity enhancing characteristics in favour of men (see table 8.1). Similarly, difference in productivity enhancing characteristics is reported to account for over 48 percent (education constituting 39 percent) of gender wage gap using 24 The effect of other productivity enhancing characteristics on gender wage gap was negative towards males and combining with 123 percent from education in favour of men brings the the Component of wage gap emanating from the overall productivity enhancing characteristics to about 50 percent. ϭϳϲ

female wage structure. However, the substantially more favourable differences in other characteristics (especially differences arising from sales, service and production occupations relative to agriculture) for women outweighed the more favourable differences in productivity enhancing characteristics for men resulting in a net differences in observed characteristics in favour of women. In effect, the important role of the gender differences in productive enhancing characteristics through education and experience in particular in favour of men in the determination of gender wage gap gives credence to the prevalence of premarket discrimination in Ghana.

ϭϳϳ

Chapter 9

Summary, Conclusion and Policy Implications 9.1

Summary of Empirical Findings

This piece of work basically sought to analyse gender differences in the Ghanaian labour market, identify the sources and establish whether these differences smack of gender discrimination. Model formulation, empirical estimation and empirical analysis have essentially been driven by the nature and characteristics of the Ghanaian labour market. The labour market in Ghana like many other developing countries is dominated by self-employment with paid employment accounting for less than one-fifth of total employment. This gives little room for the application of traditional economic models discrimination in explaining of gender discrimination in the Ghanaian labour market.

Economic models of discrimination largely blame the employer for the existence of employment and wage discrimination. This kind of employer discriminatory behaviour is more likely to occur in paid-employment which is a common type of employment in developed countries rather than selfemployment which considers the employer and the worker as one and the same individual. The discriminatory behaviour of the employer may take WKHIRUPRISUHMXGLFHDJDLQVWZRPHQRUVXEMHFWLYHDVVHVVPHQWRIZRPHQ¶V probable capability in their hiring exercise. Based on the evidence that paid-employment accounts for less than one-fifth of total employment in Ghana, it therefore means that economic models of discrimination are largely applicable to a limited segment of the labour market in Ghana. ϭϳϴ

Clearly, it is quite difficult to imagine the existence of employer discrimination in self-employment which is a dominant type of employment in Ghana and many developing countries. By implication, gender discrimination among self-employed workers is more appropriately explained by institutional and non-economic theories of discrimination. &XVWRPHUGLVFULPLQDWLRQXQGHU%HFNHU¶VSUHMXGLFHK\SRWKHVLV and to some extent human capital theory of discrimination, are perhaps the only economic theory of discrimination that can be used explain discrimination among the self-employed. These observations form the basis of analysing gender differences from the perspective of self-employment and paidemployment in this book.

A walk through the indicators of the gender dimension in the Ghanaian labour market suggests a considerable reduction in gender gap in terms of labour force participation and employment rates due to improvement in the participation and employment of females in the labour market. The quality of employment of females has seen remarkable improvement on account of improved representation of women in paid and/or formal sector employment as well as highly skilled occupations. These labour market outcomes have largely emanated from enhanced educational attainment of females which has also reflected in the decline in their illiteracy rate. Indeed, the observed narrowing gender gap in education and employment is a partial outcome of gender advocacy and implementation of domestic policies and international conventions and treaties that seek to promote economic empowerment of women and facilitate the realisation of gender equality.

However, these positive outcomes from gender perspective

appear to have failed to translate into a narrowing gender wage gap considering the observed widening gender wage gap in Ghana since 1991. In addition, female-male unemployment rate gap has been seen to be widening in recent times. The book DOVRIRXQGVRPHLPSURYHPHQWLQPHQ¶V ϭϳϵ

education and employment although not as remarkable as the performance of their female counterparts.

Empirical analysis of the extent and changing pattern of sex segregation of occupation reveals a generally low and declining trend in the degree of sex segregation of occupation in the Ghanaian labour market since 1960. The empirical findings also confirms the positive relationship between the level of disaggregation of occupations and the degree of sex segregation with the evidence that the more disaggregated the occupation, the higher the degree of sex segregation of occupation. A higher degree of sex segregation of occupation among self-employed workers than among paid-employed workers in the 1990s and before is also established. In addition, while a declining pattern of sex segregation of occupation is found in paidemployment, the reverse is the case among self-employed workers. The decomposition of changes in segregation overtime attributed much of the declining segregation to residual interactive effect. Generally, sex composition effect is found to be relatively stronger between 1970 and 2000 while occupational structure effect contributed much more to the decline in segregation in the 1960s and over the 1991±2006 periods.

Gender differences in the Ghanaian labour market are also empirically reflected in wages. The book provides empirical evidence to show that female workers earn significantly lower than male workers in all segments of the labour market. Lower wages of females than males are empirically established within five of the seven broad occupations and within both male and female occupations. The empirical analysis also suggests a larger gender wage differential in favour of men in self-employment than in paid employment.

ϭϴϬ

The analysis of gender wage differentials also links lower female wages or earnings to the declining female occupational composition effect on HDUQLQJV LQ VXSSRUW RI WKH SUHGLFWLRQ RI ³FURZGLQJ´ K\SRWKHVLV RI JHQGHU discrimination. A negative female composition effect on wages or earnings is empirically established with the observed composition effect reported to be stronger on female wages than male wages. This observation largely underscores the observed lower wages of female dominated occupation relative to male dominated occupation. The analysis further establishes that, contrary to expectation, women engaged in better-paid male occupations earn less than women in lowly-paid female dominated occupations. On the other hand, as expected, men in low-remunerated female occupations are observed to earn marginally less that men who work in better paid male occupations.

The results of wage decomposition technique also provides empirical evidence to indicate the existence of gender wage discrimination in Ghana based on the estimated high proportion of wage gap attributable to differences in unobserved characteristics termed ³wage remnant´. The analysis saw the proportion of gender wage gap emanating from discrimination measured by differences in unobserved characteristics to be higher among self-employed workers than among paid-employed workers based on Oaxaca-Blinder decomposition technique. This suggests that wage discrimination is empirically higher in self-employment than in paidemployment.

Using Neumark-Oaxaca-Ransom decomposition approach, the book further establishes high proportion of gender wage gap resulting from favouritism for males while a lower proportion of the wage gap emanated from discrimination against females in total employment in the labour market. In paid-employment,

the

contribution ϭϴϭ

of

favouritism

for

men

and

discrimination against women to male-female wage gap are almost equally matched. In self-employment however, discrimination against women is observed to account for a larger proportion of gender wage gap than favouritism for men. Overall, a greater proportion of gender wage gap is accounted for by returns to characteristics referred to as ³wage remnant´ or the combination of favouritism for males and discrimination against women.

The essential role of human capital in the determination of gender wage gap is also empirically established in this study. A substantial proportion of gender wage gap in total employment is observed to be accounted for by differences in education and experience in favour of men. Contribution of differences in education and experience to gender wage gap is reportedly high in self-employment but low in paid employment. Notwithstanding, the limited labour market endowment of women, the study provides evidence to suggest higher returns to education and experience for women than men. By implication therefore, pre-market discrimination that undermines female education does not only influence their occupational choice decision and eventually feed into sex segregation of occupation, but more importantly contribute to widening gender wage gap.

9.2

Conclusion and Policy Recommendations

The main conclusions drawn from this study have a number of important implications for policy. The observed higher degree of sex segregation of occupation and wider gender wage differentials among self-employed workers than paid-employed workers requires different policy approaches in addressing the issue of gender differences in the Ghanaian labour market. There is a limited or no basis for self-employed women to face discrimination in hiring from economic perspective of labour market discrimination. Consequently, gender discrimination in self-employment ϭϴϮ

cannot be tackled effectively within the economic framework of discrimination.

More appropriately, the phenomenon of sex segregation of occupation and gender wage discrimination among self-employed workers might better be addressed

from

institutional

and

non-economic

perspective

of

discrimination. Performance of women in self-employment could be improved through policies that largely focus on addressing barriers faced by women in that type of employment estimated to be a dominant form of employment in Ghana. Some of such barriers that tend to impede the efficient operations of self-employed women include limited access to affordable and collateral-free credit and traditional norms in some communities that make it difficult for women to own assets. As indicated by Oduro et al (2011), women in Ghana form majority (70 percent) of owners of businesses but their share of total value of businesses is only 38 percent. The reason is that, most of these businesses owned by women are small and remain small because RIZRPHQ¶VDFFHVVWRFUHGLW is very limited.

Deliberate action in the form policy and advocacy by government and civil society respectively to encourage banks and other financial institutions to improve credit advances to women entrepreneurs is one key approach to promoting growth of women businesses. Female owners of businesses can also come together as a credible and formidable association and approach micro finance institutions for credit under flexible terms to grow their businesses.

In paid-employment where gender discrimination is largely explained by economic models of discrimination, the approach to dealing with the phenomenon should appropriately emphasise economic factors. Women may face discrimination in hiring in paid-employment which is ϭϴϯ

predominantly a formal form of employment, Our gender gap decomposition results show the existence of discrimination against women which could take the form of less willingness of employers to hire women than men or readiness to hire women only in traditionally female occupations in which prospects for higher wages and career advancement are poor.

Employers¶ dislike for women may be based on sheer prejudices or stereotyping and sometimes objective assessment of cost associated with hiring women on account of their high absenteeism, disruptions, and cost in relation to maternity and childbearing. In an informal interaction with a woman employer on the issue of discrimination against women in hiring decision by employers, she stated that, based on profit maximisation objective, she would choose a young male graduate over his female counterpart with similar characteristics and equal performance at the job interview. Her reason is that while the focus of the female at that young age is marriage and raising family with its accompanying cost to the firm in terms of pay maternity leave, the ambition of the young graduate is to work harder to make money to be able to afford a wife. This confirms the human capital model of discrimination which perceives women as higher cost workers compared to their male counterparts on account of their high absenteeism and labour turnover rate at work. ,Q *KDQD ZRPHQ¶V HQWLWOHPHQW RI WKUHH-month maternity leave enshrined in the Labour Act, Act 651 may constitute an unjustifiable basis for employers to show preference for men over women in their recruitment decision or at best place them at lower echelon of the job ladder. However, employers may ³justifiably´ indulge in the act of discrimination from the perspective of maximising profit. Gender discrimination in paid employment could be addressed through enforcement of antidiscrimination ϭϴϰ

laws, provision of child care services by government and policy measures such as financial and other forms of incentives (e.g. tax incentives) for private employers to minimise the high cost associated with employment of women.

The declining female occupational composition effect on individual wages particularly that of women which conform to the prediction of crowding hypothesis of discrimination requires policy initiative that minimises the concentration of women into certain occupations. Thus, policy measures aimed at eliminating barriers that tenG WR REVWUXFW ZRPHQ¶V HQWU\ LQWR certain occupations perceived to be male occupations are very important in addressing gender differences in labour market outcomes. Clearly, education, training and skill development of women that promote their employability in all occupations and employment type is one such initiative that can reduce the concentration of women into certain occupations. (GXFDWLRQ SROLFLHV WR IDFLOLWDWH WKH H[SDQVLRQ RI JLUOV¶ DFFHVV WR traditionally non-female disciplines such as engineering, architectural design, agriculture, economics and other physical sciences are critical for the promotion of female entry into potentially more lucrative, but male dominated occupations. Such intervention should be in tandem with measures like public education, advocacy and media campaign to address WKHQHJDWLYHDWWLWXGHRIHPSOR\HUVWRZDUGVZRPHQHPSOR\PHQWZRPHQ¶V own preference and societal perception about women in the labour market.

One major conclusion drawn from the empirical analysis is the important role of human capital in the form of education and accumulation of labour market experience in the determination of labour market outcomes of both men and women. The empirical results of wage differences in chapters 7 and 8 suggest that education is particularly important in addressing gender ϭϴϱ

wage gap. A deliberate government policy towards reducing gender inequality in access to education is one key step to minimising gender differences in employment and wages or earnings in the Ghanaian labour market. The reported improvement in gender parity index (GPI) largely on account of increased enrolment of girl child education without compromising boy child education is a positive sign of enhancing female education and must be sustained.

The eYLGHQFHLQWKLVVWXG\VXJJHVWLQJKLJKHUUHWXUQVWRZRPHQ¶VHGXFDWLRQ and labour market experience relative to men makes the promotion of female education more imperative. Promoting female education could be done through policy initiatives that target equal access to education for boys and girls and reducing dropout rate of children particularly girls to the barest minimum. 'DWD IURP *KDQD¶V 0LQLVWU\ RI (GXFDWLRQ LQGLFDWH percent completion rate for boys against 83.2 percent for girls in primary school and 79.7 percent for boys and 70.1 percent for girls in JHS in 2009/10. In the face of poverty, parents may express their preference for ER\V¶ HGXFDWLRQ 7KLV FRXOG EH PLQLPLVHG WKURXJK JRYHUQPHQW¶V commitment to providing absolute free basic education in public schools beyond abolishing of payment of fees to cover books, uniforms and other essential materials to facilitate much more improvement in school enrolment for both boys and girls.

The implication of education for occupational distribution and gender wage differentials vis à vis unequal access of boys and girls to education point to the role of pre-market discrimination in the determination of labour market outcomes of men and women. Low representation of women in highly skilled occupations has been blamed on relatively low education of women which arises as a result of pre-market discrimination in the form of SUHIHUHQFH RI ER\V¶ HGXFDWLRQ WR JLUOV¶ HGXFDWLRQ WKURXJK VRFLDOLVDWLRQ ϭϴϲ

process. The skill development of boys and girls at the basic level of education are often influenced by action of teachers to the extent of shaping their future career in line with their biological characteristics.

Consequently, it is more appropriate to address gender differences in occupational distribution and wages by resolving gender differences at the initial developmental stage of the children. This means that affirmative action that seeks to address economic inequality at the top is rendered ineffective if indeed, the major source of gender differences is identified to be pre-market in nature. An offer of employment and job position for women through affirmative action without recourse to their competence and capability could promote inefficiency which is observed to be one of the fundamental issues associated with labour market discrimination. In effect, anti-discriminatory policy measures that are essentially targeted at resolving gender discrimination at the pre-market level is strongly recommended.

One noticeable limitation of this research work has been the difficulty in controlling for some important variables such as trade union activity and LQGLYLGXDO¶VUDQNRUSRVLWLRQRQWKHMREODGGHULQWKHwage equation. This is largely due to the fact that the GLSS dataset did not have information on such key variables or in some cases where some of such information were captured, there were not enough observations to facilitate its inclusion in the analysis(YHQWKRXJKLWLVH[SHFWHGWKDWWKHHIIHFWRILQGLYLGXDO¶V job position or rank on wages would be captured by experience, it may not often be the case since the position of some workers on the job may not be determined by experience. These observations underscore the need for further studies to provide explanation to some of these unanswered questions. Essentially, this piece of research should be considered as one step towards a comprehensive assessment of the presence of gender ϭϴϳ

discrimination in the Ghanaian labour market and its potentially adverse effect on the efficiency of the labour market and the economy as a whole.

ϭϴϴ

Bibliography Agboli M. (ed.) 2008. ³*HQGHU DQG (FRQRPLF *URZWK $VVHVVPHQW IRU *KDQD´,QWHUQDWLRQDO)LQDQFH&RUSRUDWLRQ:RUOG%DQN*URXS and ministry for Women and Children Affairs Aigner D. J and G. G Cain, 1977. ³6WDWLVWLFDO7KHRULHVRI'LVFULPLQDWLRQLQ WKH /DERXU 0DUNHW´ ,QGXVWULDO DQG /DERXU 5HODWLRQV 5HYLHZ 30:175-187 Allah-Mensah B., 2005. ³:RPHQ LQ 3ROLWLFV DQG 3XEOLF /LIH LQ *KDQD´ Friedrich-Ebert Foundation, Accra Altonji G. J and R. Blank, 1999. ³5DFHDQG*HQGHULQWKH/DERXU0DUNHW´ Handbook of Labour Economics, Vol. 3C Anker R., 1998. ³*HQGHUDQG-REV6H[6HJUHJDWLRQRI2FFXSDWLRQVLQWKH :RUOG´,QWHUQDWLRQDO/DERXU2IILFH -----------, 1995. ³0HDVXULQJIHPDOHODERXUIRUFHZLWKHPSKDVLVRQ(J\SW´ iQ $QNHU 5 ³*HQGHU DQG -REV 6H[ 6HJUHJDWLRQ RI 2FFXSDWLRQVLQWKH:RUOG´,QWHUQDWLRQDO/DERXU2IILFH Appleton S., Hoddinott J., and P. Krishnan, 1999. ³7KH*HQGHU:DJH*DS LQ 7KUHH $IULFDQ &RXQWULHV´ (FRQRPLF 'HYHORSPHQW DQG &XOWXUDO Change, 47 (2), 289-312 ---------, et al., 1990. ³7KH:RUOG%DQN6RFLDO'LPHQVLRQVRI$GMXVWPHQW in Sub-6DKDUDQ$IULFD´:RUNLQJ3DSHU1R3ROLF\$QDO\VLV Arrow K., 1973. ³7KH 7KHRU\ RI 'LVFULPLQDWLRQ´ LQ &DLQ * * ³7KH (FRQRPLF $QDO\VLV RI /DERXU 0DUNHW Discrimination: A 6XUYH\´LQ+DQGERRNRI/DERXU(FRQRPLFV9RO Aryeetey, B-D E., 2000. ³7KH 3DUWLFLSDWLRQ RI :RPHQ LQ WKH *KDQDLDQ (FRQRP\´ LQ $U\HHWH\ ( +DUULJDQ - DQG 0 1LVVDQNH HG ³(FRQRPLF5HIRUPVLQ*KDQD7KH0LUDFOHDQGWKH0LUDJH´-DPHV Currey Ltd. Publication

ϭϴϵ

Ashraf J., 1996. ³,V *HQGHU 3D\ 'LVFULPLQDWLRQ RQ WKH ZDQH"´ (YLGHQFH from Panel data (1968-´ ,QGXVWULDO DQG /DERXU 5HODWLRQV review Vol. 49, No. 3, 537-546 Atieno R. and F. Teal, 2006. ³*HQGHU (GXFDWLRQ DQG 2FFXSDWLRQDO OuWFRPHV .HQ\D¶V ,QIRUPDO 6HFWRU LQ WKH V´ Global Poverty Research

Group,

&

Economic

and

research

Council:

http://www.gprg.org Baah-%RDWHQJ : ³*HQGHU 3HUVSHFWLYH RI /DERXU 0DUNHW 'LVFULPLQDWLRQ LQ *KDQD´ $Q

Unpublished

PhD

Thesis,

Department of Economics, University of Ghana -----------., 2007. ³0HDVXULQJ WKH ([WHQW RI *HQGHU 6HJUHJDWLRQ LQ WKH /DERXU 0DUNHW (YLGHQFH IURP *KDQD´ -RXUQDO RI /HDGHUVKLS Management and Administration Volume 5, Number 1, June 2007, pp. 57-81; GIMPA Press ---------- and F. E., 2005. ³(PSOR\PHQW´LQ*OREDOLVDWLRQ(PSOR\PHQWDQG 3RYHUW\ 5HGXFWLRQ $ &DVH 6WXG\ RI *KDQD´ HGLWHG E\ (UQHVW Aryeetey, ISSER publication Baden S et al., 1994. ³%DFNJURXQG 3DSHU RQ *HQGHU ,VVXHV LQ *KDQD´ Report prepared for the West and North Africa Department, Department of Overseas Development (DFID) Barr A. and Oduro A., 2000. ³(WKQLFLW\ DQG :DJH 'HWHUPLQDWLRQ LQ *KDQD´ $ 3DSHU SUHSDUHG IRU WKH &6$(¶V &RQIHrence on µ2SSRUWXQLWLHV LQ $IULFD 0LFUR (YLGHQFH IURP )LUPV DQG Households April 9-10 Beaudry P. and Sowa N.K., 1994. ³*KDQD´LQHGV6+RUWRQHWDO³/DERXU 0DUNHWLQDQ(UDRI$GMXVWPHQW´(FRQRPLF'HYHORSPHQW,QVWLWXWH Vol. 2, pp. 357-403 Washington DC. Becker G. S., 1991. ³$7UHDWLVHRQWKH)DPLO\´+DUYDUG8QLYHUVLW\3UHVV Cambridge, MA ϭϵϬ

---------, 1983. ³$ 7KHRU\ of Competition among Pressure Groups for 3ROLWLFDO,QIOXHQFH´4XDUWHUO\-RXUQDORI(FRQRPLFV-400 ---------, 1971. ³7KH (FRQRPLFV RI 'LVFULPLQDWLRQ´ nd edition, Chicago: University of Chicago ---------, 1957. ³7KH (FRQRPLFV RI 'LVFULPLQDWLRQ´ &hicago; The University of chicag Press Beller, H. A., 1985. ³&KDQJHVLQWKH6H[&RPSRVLWLRQRI862FFXSDWLRQV 1960-´-RXUQDORI+XPDQ5HVRXUFHV -50 ---------, 1981. ³7UHQGV LQ 2FFXSDWLRQDO 6HJUHJDWLRQ E\ 6H[ DQG 5DFH 1960-´ LQ %DUEDUD 5HVNLQ HG ³6H[ VHJUHJDWLRQ LQ WKH :RUNSODFH 7UHQGV ([SODQDWLRQ 5HPHGLHV´ :DVKLQJWRQ '& National Academy Press, 11-26 Bell-Lowther E., 1985. ³:RUOG )HPLQLVDWLRQ RI 3RYHUW\ $ FRQIHUHQFH 5HSRUW´7KH6RFLDO:RUNHU 6XPPHU Benabou, R. J., 1994. ³+XPDQ &DSLWDO ,QHTXDOLW\ DQG *URZWK D ORFDO 3HUVSHFWLYH´(XURSHDQ(FRQRPLF5HYLHZ-826 Bergmann, B.R., 1989. ³'RHVWKH0DUNHWIRU)HPDOH/DERXU1HHG)L[LQJ" Journal of Economic Inquiry Vol. 3, No. 1: 43-60 ---------, 1974. ³2FFXSDWLRQDO 6egregation, Wages and Profits when (PSOR\HUV'LVFULPLQDWHE\UDFHRUVH[´(DVWHUQ(FRQRPLF-RXUQDO 1:103-110 Bertaux N. E., 1991. ³7KH5RRWVRI 7RGD\¶V µ:RPHQ¶VMREV¶DQGµ0HQ¶V -REV¶ 8VLQJ WKH ,QGH[ RI 'LVVLPLODULW\ WR PHDVXUH 2FFXSDWLRQDO segregation by Gender: Exploration in Economic History 28(4): 433-459 Bianchi S. M and N. Rytina, 1986. ³7KH 'HFOLQH LQ 2FFXSDWLRQDO SegregatiRQ GXULQJ WKH V &HQVXV DQG &36 &RPSDULVRQV´ Demographic Vol. 23 No. 1: 79-86 Black D., 1995. ³'LVFULPLQDWLRQLQDQ(TXLOLEULXP6HDUFK0RGHO´-RXUQDO of Labour Economics Vol. 12, No. 2, 309-334 ϭϵϭ

Black S. E. and E. Brainerd, 2004. ³7KHLPSDFWRI*OREDOisation on Gender 'LVFULPLQDWLRQ´ ,QGXVWULDO DQG /DERXU 5HODWLRQV 5HYLHZ 9RO No. 4: 540-559 Blackburn M. R, Jarman J, and J. Sitanen, 1993. ³7KH $QDO\VLV RI Occupational

Gender

Segregation

over

Time

and

Place:

Considerations of Measurement and Some NHZ (YLGHQFH´ :RUN Employment and Society, SAGE publication Blau F. D., 1996. ³:KHUHDUHZHLQ(FRQRPLFVRI*HQGHU"7KH*HQGHU3D\ *DS´1%(5:RUNLQJ3DSHU ---------- and L. M. Kahn, 2000. ³*HQGHU 'LIIHUHQFHV LQ 3D\´ 1%(5 Working Paper 7732 ----------, Ferber M and Winkler, A., 1998. ³(FRQRPLFV RI :RPHQ 0HQ DQG:RUN´ Engelwood Cliffs, NJ: Prentice Hall ---------- and L. M. Khan, 1997. ³6ZLPPLQJ XSVWUHDP 7UHQGV LQ WKH *HQGHU :DJH 'LIIHUHQWLDOV LQ WKH V´ -RXUQDO RI /DERXU Economics, Vol. 15, No. 1 part 1, 1-42 ----------, and M Ferber, 1992. ³7KH (FRQRPLFV RI :RPHQ 0HQ DQG :RUN´(QJOHZRRG&OLIIV1HZ-HUVH\3UHQWLFH-Hall ----------, and M. Ferber, 1987. ³'LVFULPLQDWLRQ(PSLULFDO(YLGHQFHIURP WKH8QLWHG6WDWHV´$PHULFDQ(FRQRPLF5HYLHZ Vol. 77, No. 2 ----------, and W. Hendricks, 1979. ³2FFXSDWLRQDO 6HJUHJDWLRQ E\ 6H[ 7UHQGVDQG3URVSHFWV´-RXUQDORI+XPDQUHVRXUFHV-210 ----------, P. Simpson and D. Anderson, 1998. ³&RQWLQXLQJ 3URJUHVV" Trends in Occupational segregation in the United States over the VDQGV´)HPLQLVW(FRQRPLFV-71 Blinder, A.S., 1973. ³:DJH'LVFULPLQDWLRQ5HGXFHG)RUPDQG6WUXFWXUDO (VWLPDWHV´The Journal of Human Resources 8: 436-455 Boateng K., 2001. ³,PSDFW RI 6WUXFWXUDO $GMXVWPHQW RQ (PSOR\PHQW DQG ,QFRPHV LQ *KDQD´ LQ %DDK