Assessing the Impact of Informality on Earnings in Tanzania: Is There a Penalty for Women?

Get Complete Project Material File(s) Now! »

Addressing the Gender Pay Gap in Ethiopia: How Crucial is the Quest for Education Parity?

Introduction

As part of its overall objective to reach the Millennium Development Goals (MDGs)6, the Government of Ethiopia has made remarkable efforts towards universal primary education, gender equality and women empowerment. While there are still large gender disparities in education, Ethiopia has seen an enormous and rapid increase in enrolment in primary education that has contributed to reduce the gender imbalance (MoFED, 2005). The emphasis given to education and gender equality reflects also its instrumental importance in fostering progress towards other goals, such as raising labour compensation and supporting women’s progress in the labour market. Research shows that women’s earnings can influence their status and decision-making power within the family, as well as their choices about labour force participation and fertility. Women’s wages are especially important for children, as they tend to spend their earnings directly on their needs (UNICEF, 1999). This raises important policy questions for Ethiopia, a country that has ratified the UN Convention on the Elimination of All Forms of Discrimination against Women. How significant is the gender pay gap? What lies behind the pay differentials between men and women? Are discrimination and other non-observable factors important and similar across the wage distribution and the types of employment? How likely will the achievement of the education MDGs translate into a significant reduction in wage disparities across gender? In contrast with the abundant literature of the gender pay gap in developed countries, and the growing number of studies for emerging countries, fewer studies have actually attempted to address these important questions in the case of Africa (Weichselbaumer and Winter-Ebmer, 2003).
Available evidence based on survey data confirms the presence of large gender pay gaps in several African countries. Some earlier studies estimate, for instance, that the ratio of female earnings to male earnings could range from 40 per cent in Kenya (Kabubo-Mariara, 2003), to 70 per cent in Cameroun (Lachaud, 1997), 80 per cent in Botswana (Siphambe and Thokweng-Bakwena, 2001) and 90 per cent in Burkina-Faso (Lachaud, 1997). In the case of Ethiopia, Temesgen (2006) finds that in the manufacturing sector in 2002 female hourly wages stood at 73 per cent of male wages. Similarly, in a study on the size and the determinants of the gender wage gaps in three African countries, Appleton et al. (1999) find that in urban Ethiopia in 1990, female earnings represented on average 78 per cent of male earnings. In most of these studies that attempt to explain the extent of the gender wage gap, the unexplained term, which is likely the result of discriminatory practices, gender specific preferences, cultural and other non-observable factors, along with differences in educational endowments, account for a non-negligible share of the pay gap.
Other more recent studies on Africa using matched employer-employee data indicate that the relative importance of the unexplained component decreases when other factors such as job tenure and job characteristics are included as controlled variables in the wage equations, and that much of the wage gap correlated with education can be explained by selection across occupations and firms (Fafchamps et al., 2006; Nordman and Wolff, 2008, 2009). In the case of Madagascar, using a better measure of female work experience obtained from matching a labour force survey with a biological survey also contributes to reduce substantially the size of the unexplained wage gap in the decompositions analysis (Nordman and Roubaud, 2005).
The aim of this Chapter is to cast new light on the gender pay gap in Ethiopia using the 2005 Ethiopian Labour Force Survey. A particular attention is drawn on the relative importance of education parity to mitigate the most pressing wage inequality and the role of job segregation. This Chapter complements the few available studies for Africa by adopting a comprehensive approach where the factors related to the gender pay gap in Ethiopia are analysed for different points in the wage distribution, different age cohorts and different type of wage employment. To this end, we start to estimate wage equations using two specifications and three different models, separately for men and women. We then apply decomposition procedures proposed by Neumark (1988) and Cotton (1988) to disentangle the effects on the pay gap of human capital and job characteristics from an unexplained component that captures the effect of discrimination and other non-observable factors.
The Chapter is organised as follows. Section 2 presents the data set, the concepts and some detailed summary statistics on gender disparities in employment, education and pay for different age cohorts, segments of the labour market and wage levels. The different methods chosen for estimating wage equations and decomposing gender wage gaps are explained in section 3. Section 4 presents the main results and section 5 concludes.

Data and concepts

In this section, we start off by presenting the data used for the analysis of the gender wage gap. We further provide an explanation of the definitions and measures of key relevant labour market indicators. We finally present some basic descriptive statistics on employment and education broken down by gender.

Ethiopia Labour Force Survey 2005

To explain the difference in earnings by gender and to analyse the factors related to the gender pay gap in Ethiopia, we draw upon the Labour Force Survey (LFS) collected in Ethiopia by the Central Statistics Agency (CSA) in March 2005. The LFS is a nationally representative household survey containing information on a large number of individuals. It is designed to monitor the social and economic situation of the economically active population7. Out of the total 230 680 individuals who were interviewed in the LFS, 50.5 per cent were located in urban areas8. Accounting for sampling weights, this figure declines to 14.2 per cent of overall population. The individual record includes a broad range of information about age, gender, education, employment status, wage and non-wage activities, job characteristics, and earnings, and thus represents a good opportunity for our study. However, it is important to emphasize two weaknesses of this data. Poor economies like Ethiopia are generally characterized by a preponderance of seasonal activities, especially in agriculture but also in small manufacturing9. Yet, the fact that data collection took place in a short period of time makes the data particularly sensitive to seasonality issues. Another weakness of the data is related with the survey questionnaire which does not collect interesting information like ethnicity, religion and language which may potentially affect labour outcomes10.
7 The LFS 2005 covers all parts of the country except the Gambela region (including Gambela town), and the non-sedentary population of three zones of Afar and six zones of Somali regions. For the purpose of the survey, the country was divided into three broad categories: rural areas, major urban centres and other urban centres. A stratified two-stage cluster sample design was used in the first two categories to select samples. The primary sampling units (PSUs) were enumeration areas (EAs). Households per sample EA were then selected as a second-stage sampling unit (SSU). As regards the third category, a stratified three-stage cluster sample design was adopted to select samples. PSUs were urban centres and SSUs were EAs. Households from each EA were selected at the third stage.
8 Urban households are oversampled.
9 According to Wodon and Beegle (2006), evidence for Malawi and other developing countries suggests the existence of labour shortages at the peak of the cropping season, and substantial underemployment for most of the year, especially in rural areas. As regards Ethiopia, Dercon and Krishnan (2000) use data from rural areas to show high levels of seasonal and year-to-year variability in consumption and poverty, with households also responding to changes in labour demand and prices.
10 According to CSA (2006), these variables were not included in the LFS because the number of cases obtained in this specific sample size was not found to be sufficient in order to provide reliable information.

Definitions and measurement issues

The 2005 LFS contains self-classification information on productive activities such as work for payment, family gain or profit for own consumption performed in the last 7 days by individuals of age 5 and above. In this study, our measure of the labour force refers to all persons aged 15 or above either engaged in, or available to undertake, productive activities11. We further include under the label wage employment all individuals engaged in productive activities that worked as paid employees at least four hours in the last seven days. We also include all those who were working less than four hours or were not working the last seven days, and who were paid while on temporary leave or who had an assurance or an agreement for returning to work.
Wage employment, which represents a minor share of total employment in Ethiopia (only 8.7 per cent), is the basis for our analysis of the gender pay gap. Our coverage of wage employment is broader than what is usually found in wage studies for Ethiopia as it includes rural labour markets. Rural wage employment represents a significant share of total wage employment (37.8 per cent).

Formal and informal wage employment

This is unfortunate because, for instance, ethnicity is likely to play an important role in wage determination and may shed light on gender differences in the labour market. Barr and Oduro (2002) provide evidence that the Ghanaian labour market is ethnically fractionalized and that this is leading to quite considerable earnings differentials between ethnic groups. Based on an analysis of data from a survey of owner-managed manufacturing businesses in Ethiopia, Mengistae (2001) find that an indigenous minority group, namely, the Gurage, happens to have a far higher rate of business ownership than other major ethnic groups as well as minorities, partly because Gurage-run businesses perform better.
11 See the World Bank (2007) Ethiopia’s report for a detailed discussion on the various concepts used by the CSA and the World Bank to classify employment. Our definition of the labour force is that of the World Bank, while the CSA uses a different age threshold (10 and above).
employment. In this Chapter, we follow the recommendation of the World Bank (2007) and use a broad concept of the informal sector. In the data, individuals who work for a wage or a salary are asked to describe the employment status of their main occupation. Those who report working in the public sector as government or parastatal employees, as well as NGO employees and other employees working in private organizations which have ten or more employees, or which have a license or a book account, are classified as formally wage employed.
In contrast, informal wage employment includes paid employees who are domestics or who work in a private organization which has less than ten employees, is not licensed and has not a book account. It also includes employees for which this information is missing and who are only paid in kind or doing casual work. The latter are in fact very likely to be located in the unregulated sector.
Following this classification, we further decompose wage employment in three components: public formal wage employment (government and parastatal employees), formal private wage employment (employees in formal private organizations and NGOs) and informal private wage employment (employees in informal private organizations and domestic employees).

Earnings

Unlike the 2001 survey, the 2005 LFS provides a good opportunity to analyse the gender pay gap as it provides information on the amount paid to wage employees in their main occupation during the last pay period and the number of times they were paid during the last month. To account for the impact of the duration of work on wages, we calculate hourly earnings from the main occupation for each worker in wage employment by dividing the monthly earnings by the monthly hours of work in the main occupation. The latter is indirectly calculated by subtracting the number of hours worked on additional activities from the total number of hours worked at all jobs in the last seven days, and multiplying the result by four12. The use of earnings as a proxy for the returns to work is not exempt of problems, however. As earnings are available exclusively for the wage employed and from their main occupation only, this leaves aside the possibility to analyse the returns from self-employment. Moreover, it does not allow taking into account the returns of secondary employment13. Finally, non-wage benefits may be important in some cases (in particular, the government and the parastatal might offer additional benefits in terms of pension benefits or job security) and since they are not imputed this may underestimate the true level of earnings. Notwithstanding these issues, earnings data remain essential to understand the gender pay gap in Ethiopia.

Descriptive statistics

Basic labour market indicators are reported in Table 2.1 separately for males and females. According to all these indicators, the situation of women appears less favourable than that of men. The participation rate and the employment ratio are lower for women, while female unemployment is similar to male unemployment. In addition, a higher proportion of the male population is in wage employment.
12 Hours of work include overtime and exclude lunch and journey time. Here, we multiply weekly hours of work by 4 to obtain monthly hours of work, and not by 4.2 or 4.3 as it is usually done, because the last month preceding the interview was February 2005 and included only 28 days.
13 This should not be a serious problem, however, since the LFS 2005 reveals that less than 9 per cent of Ethiopian wage employees hold multiple jobs. Moreover, according to data from the Addis Labour Market Survey (ALMS), wages from secondary jobs do not appear to be an important element of overall earnings (World Bank, 2007).
The characteristics of wage employment are shown in Table 2.2. There are large gender variations in the nature and the terms of wage employment. For men, public formal wage employment and private formal wage employment constitute altogether the biggest share of the wage employed. Only 15 per cent of wage employed men are in informal private jobs. For women, however, the proportion of the wage employed in private informal jobs represents the second most frequent form of wage employment (32 per cent) after public employment (40 per cent). In relative terms, women are more likely than men to work in informal jobs, as temporary or casual employees, and less likely to work in permanent or contract employment, suggesting that the conditions of work among wage employed women are less favourable than for men.

Table of contents :

Chapter 1 Introduction
1.1 Why looking at gender disparities in Africa’s labour markets?
1.2 Objective and structure of this thesis
1.3 Countries selected
1.4 Overview of main findings
Chapter 2 Addressing the Gender Pay Gap in Ethiopia: How Crucial is the Quest for Education Parity?
2.1 Introduction
2.2 Data and concepts
2.2.1 Ethiopia Labour Force Survey 2005
2.2.2 Definitions and measurement issues
2.2.3 Descriptive statistics
Gender disparities in the labour force and in employment status
The unadjusted gender pay gap
Gender disparities in education characteristics among the wage employed
Gender disparities across sectors of activity and occupations
2.3 Methodology
2.3.1 Estimation of wage equations
Heckman’s two-step estimation procedure
Bourguignon-Fournier-Gurgand two-step estimation procedure
Quantile regression analysis
2.3.2 Decomposition of the gender wage gap
Neumark and Cotton decomposition procedures
Treatment of the sample selection correction
2.4 Results
2.4.1 Estimations of the wage equations
2.4.2 Wage decompositions
2.5 Conclusion
Appendix A. Generalized Lorenz curves for hourly earnings
Appendix B. Earnings equations
Appendix C. Gender earnings gap decompositions
Chapter 3 Analysing the Nature and Extent of Gender Inequalities in Time Use: New Insights from Ethiopia
3.1 Introduction
3.2 Data and concepts
3.2.1 Brief theoretical literature review
3.2.2 Ethiopia Labour Force Survey 2005
3.2.3 Definitions and measurement issues
3.3 Methodology
3.3.1 Decomposition of total work time
3.3.2 The determinants of market and household work time
A generalized Tobit model for market work time
A standard Tobit model for housework time
Decomposition of the total marginal effect
3.4 Results
3.4.1 Decomposition of total work time
3.4.2 Further disaggregations of time use estimates
3.4.3 The determinants of market and household work time
Preliminary considerations on models’ specification and data quality issues
Ancillary parameters of the generalized Tobit models for market work time
Sex and area of residence
Complementarity/substitutability of time allocation decisions
Human capital and other individual characteristics
Composition of the household
3.5 Conclusion
Appendix A. Generalized Lorenz curves for the market and household work time
Appendix B. Housework time equations
Appendix C. Market work time equations
Chapter 4 Assessing the Impact of Informality on Earnings in Tanzania: Is There a Penalty for Women?
4.1 Introduction
4.2 Data, concepts and descriptive statistics
4.2.1 Tanzania Integrated Labour Force Survey 2006
4.2.2 Key concepts
Employment
Informal employment
Labour income
4.2.3 Descriptive statistics on gender-differentiated employment patterns
4.3 Methodology
4.3.1 Informal employment effects under the assumption of homogeneity
4.3.2 Informal employment effects under the assumption of partial heterogeneity
4.3.3 Informal employment effects under the assumption of full heterogeneity
4.4 Results
4.4.1 Informal employment effects under the assumption of homogeneity
4.4.2 Informal employment effects under the assumption of partial heterogeneity
4.4.3 Informal employment effects under the assumption of full heterogeneity
4.5 Conclusion

GET THE COMPLETE PROJECT