Earnings Differences Between Men and Women in Rwanda Africa Region Working Paper Series No.81 January 2005 Abstract The paper examines wage and income differences between men and women in Rwanda, to determine the extent to which observed patterns are due to observed differences in education and experience. It examines this issue with particular attention to the agriculture and service sectors which together constitute roughly 80 percent of GDP and have been the key sectors of growth, over the past decade. The paper finds that a significant proportion of workers are engaged in unpaid work, and women appear to be over represented in this category. The paper also finds that there are significant differences in earnings between men and women and that these differences persist, even after controlling for socioeconomic factors. The paper discusses these results and offers policy recommendations to help reduce these differences. The Africa Region Working Paper Series expedites dissemination of applied research and policy studies with potential for improving economic performance and social condition In Sub-Saharan Africa. The series publishes papers at preliminary stages to stimulate timely discussions within the Region and among client countries, donors, and the policy research community. The editorial board for the series consists of representatives from professional families appointed by the Region s Sector Directors. For additional information, please contact Momar Gueye, (82220), Email: mgueye@worldbank.org or visit the Web Site: http://www.worldbank.org/afr/wps/index.htm. The findings, interpretations, and conclusions in this paper are those of the authors. They do not necessarily represent the views of the World Bank, its Executive Directors, or the countries that they represent and should not be attributed to them.
Earnings Differences Between Men and Women in Rwanda Kene Ezemenari Rui Wu January 2005
Authors Affiliation and Sponsorship Kene Ezemenari KEzemenari@worldbank.org, Rui Wu @worldbank.org February 2005 Acknowledgement The authors would like to thank Gaelle Pierre for discussions and suggestions during the preparation of this paper. Lucia Chuo and Momar Gueye provided logistical support."
Tables of Contents I. Introduction 1 II. Data Description 2 III. Measuring the Gender Earnings Gap 10 Econometric Estimation of Earnings Differentials 10 Data Summary 12 Regression Results 15 Summary and Conclusion 19 Table 1: Wage Earners Distribution by Gender (age 14 and older 4 Table 2: Average Hours Spent per Week on: Wage Labor, House Work, and Total 9 Table 3: Data Summary 14 Table 4: Gender Income/Wage Differences, t-statistics adjusted 16 Table 5: Income/Wage differences of men and women at each educational level when compared with primary (control) education 19 Figure 1: Average Annual Income Gap by Gender 6 Figure 2: Average Monthly Wage Rate Gap by Gender 6 Appendix 1: Average annual income and wage rate by gender 21 Appendix 2: Pooled Sample Regression 22 Appendix 3: Agriculture Sector Sub-Sample Regression 24 Appendix 4: Service Sector Sub-Sample Regression 26 Appendix 5: Other Sectors Sub-Sample Regression 28
I. Introduction The period from 1995-2002 was one of systematically high growth, for the Rwandan economy. During this period growth ranged from an average annual rate of 11.8 percent between 1996 and 1998 to 7.4 percent between 2000 and 2002, with low and stable inflation, making Rwanda one of the best performers in Sub-Saharan Africa. Much of this growth has been due to the reforms set in place by Government to reduce price distortions, and build infrastructure. However, the rate of growth is beginning to slow down, and to maintain the momentum, government will need to continue with planned reforms. Among the areas government will need to address is gender equality in labor market and earnings. Government has taken measures to reduce discrimination and gender inequality through labor market policy and legislation. In the political arena, half of elected local government members were women. Gender inequality has been significantly reduced in primary education, where girls are just as likely to be in school as boys, and net enrollment rates are now at 73%. Disadvantages in access to schooling emerge at higher levels of education because girls academic performance lags behind that of boys, starting from the primary level. However, despite the successes, inequalities still persist in terms of differences in earnings between men and women, which can pose a constraint to further growth in Rwanda. World Bank (2001) shows that gender inequality can have adverse effects on economic growth. In addition, there are numerous studies to show that higher incomes for women, leads not only to their increased voice and empowerment, but also to improved wellbeing for the household, and greater outcomes of nutritional status of children in household. The implication for Rwanda is that gender equality, particularly with respect to access to employment and earnings needs to be addressed as part of an overall strategy for sustained growth. This paper provides an initial step in that direction. It examines wage and income differences between men and women in Rwanda, to determine the extent to which the constraining factors are due to observed differences in education and experience. It examines this issue with particular attention to the agriculture and service sectors which together constitute roughly 80 percent of GDP and have been the key sectors of growth, over the past decade. The paper finds that a significant proportion of workers are engaged in unpaid work, and women appear to be over represented in this category. The paper also finds that there are significant differences in earnings between men and women and that these differences persist, even after controlling for socio-economic factors. The paper discusses these results and offers policy recommendations to help reduce these differences. The paper is organized as follows. Section 2 summarizes the data used in the analysis. Section 3 outlines the model used to estimate discrimination. Section 4 discusses the results, and Section 5 concludes the paper with some suggestions for policy in Rwanda. 1
II. Data Description The data used for the analysis is taken from the Rwandan Integrated Household Living Conditions Survey (EICV) conducted between October 1999 and July 2001 1. The survey covers 32,153 households and is composed of two parts: Part A on Household Members and Housing; Part B on Agriculture, Budget and Consumption. Part A covers modules on household composition, education, health, employment and time use, migration and housing. Part B contains modules on agriculture and livestock, household expenditure, non-farm enterprises, income transfers and miscellaneous incomes and expenditures, credit, assets and savings. We focus on Section 4, Employment and Time Use, for the data used to examine wage structure and potential gender discrimination. We proceed to define those who are employed, based on the EICV by first limiting the sample to those aged 14 and over which reduces the sample down to 18,730 (from the total observation of 32,153). Within this sample, those who responded yes to any of the following four questions were then defined as being wage earners: During the past 12 months have you: 1) Worked for wages or any kind of remuneration? 2) Made any money including payment in kind through self-employment? 3) Worked on a farm, field, or done herding or fishing? 4) worked for an enterprise belonging to a member of the household, even if you were not paid? In addition to those who answered positively to these questions, we also included those who specified an occupation in response to the question What kind of work did you spend most of your time on? The above criteria led to a further reduction in the sample to 14,834 from 18,730. A significant proportion of workers in the sample of 18,730 are engaged in unpaid work. In total, 80 percent (11,958 observations), of the total sample of wage laborers are engaged in unpaid work. Of the total engaged in unpaid work, roughly 98 percent are engaged in the agriculture sector. Within the agriculture sector, the greater proportion of unpaid workers are female (58%) compared to males (40%). In addition, most of these workers have primary education or less males mainly have primary education (68%) and females are almost split between primary (56%) and no education (40%). Given the large number of individuals engaged in unpaid work, the estimation of the gender earnings gap is conducted with a sub-sample of workers that earn a positive wage. Individuals younger than 14 were also dropped from the sample, leading to a further reduction in sample size from 14,834 to 2,876. Based on the reduced sample of 2,876, an estimate of annual income was derived from wages. This variable was then adjusted for transfers and non-wage income (i.e. paid sick-leave, and non-monetary transfers in the form of food, crops, or animals). In terms of wage earnings, estimation of this variable is based on Question 11, Section 4, Part B of the EICV which asks Did you receive any money from this work? ; and What was the last amount received? 2. According to the data, 58.54% of wage-earners report that they were 1 This is cross-sectional data collected over the period October 1999 to July 2001. 2 In the case of wage income we focus only on those that were able to report a wage. For the income variable, however, a small sub-sample reported income in-kind (11 observations out of total of 14,830). These 11 observations were included in the estimation of annual income. 2
paid on a monthly basis. As a result, we adjust everyone s earnings to a monthly wage rate to minimize measurement errors. Table I, summarizes the percentage of wage earners (aged 14 years and older), by sector of employment. Agriculture and services the two dominant sectors. On average 22.86% and 37.15%, respectively, of wage or income earners, aged 14 and older are found in these two sectors. 3
Table 1: Wage Earners Distribution by Gender (age 14 and older Sector Male Female Total Agriculture 22.02 24.25 22.91 Mining 1.22 0 0.73 Manufacture 7.45 2.5 5.46 Electric, water, gas 1.28 0.17 0.83 Construction 7.45 1.21 4.94 Commerce, Hotels 15.43 25.88 19.65 Transport, Communic. 7.98 0.43 4.94 Banking 3.49 2.5 3.09 Services 32.73 42.54 36.68 Others 0.93 0.52 0.76 Total 1,717 1,159 2,876 Source: ECIV 2000/2001 Figure 1 and Figure 2 show the annual income gaps between male and female by age groups. The differences between males and females are less pronounced when the age cohort is 30 or younger. One begins to see larger differences after the age of 35 and these differences increase with age. The data suggest a decline in earnings differences between men and women (and potentially a reduction of past discriminatory labor policies and practices) with each new cohort of workers. Earnings differences, between men and women, are still present among the younger cohorts. An alternative interpretation arises when the earnings gap is represented graphically according to years of experience on the job. Figures 3 and 4 show the income gaps by level of experience. The same patterns are observed as for age groups. The differences are larger for those individuals with more than 6 years of experience. As mentioned in the preceding paragraph, this effect could be due to the fact that those who are older will tend to have more experience. Therefore, we are observing the same pattern reflected in the age cohort graphs. An additional interpretation is that women have less mobility than men. Men and women begin there careers, more or less equally, or with low gaps in earnings. With increased experience one begins to see larger earnings gaps due to the lower level of mobility of women. There are various reasons why women s mobility may be restricted. Part of this may be related to obligations related to house work and caring for children. As a result, one would see higher wage gaps during child bearing years, and as children grow older. This is supported by the data. Women spend more time on house work (on average 24 hours a week). Men spend on average 5.5 hours per week. Another reason for differences in wages may arise from the fact that women may engage more in part-time work or the informal sector. For example, the data show that women tend to be employed in smaller enterprises i.e. firms that employ 10.4 workers. This contrasts with men, who tend work in larger firms, or firms that employ on average 24.6 workers. Despite the above differences, the data suggest that the level of education may be an equalizing factor in earnings between men and women. Figures 5 and 6 show the earnings gap by level of education. As these figures show, the gender earnings gap declines with 4
increased level of education. For high levels of education, the figures suggest that the earnings gap disappears completely. 5
Figure 1: Average Annual Income Gap by Gender Average Annual Income by Age 1200000 1000000 800000 600000 male female 400000 200000 0 [14,20) [20,25) [25,30) [30,35) [35,40) [40,45) [45,55) [55,75) [75+,) Age Group Figure 2: Average Monthly Wage Rate Gap by Gender Average Wage Rate by Age 80000 70000 60000 50000 40000 male female 30000 20000 10000 0 [14,20) [20,25) [25,30) [30,35) [35,40) [40,45) [45,55) [55,75) [75+,) Age Group 6
Figure 3: Average Annual Income by Work Experiences Figure 4: Average Wage Rate by Work Experiences 1200000 60000 1000000 50000 800000 40000 600000 Male Female 30000 Male Female 400000 20000 200000 10000 0 0 <1 yr (1,3] yrs (3,6] yrs (6,10] yrs (10,20] yrs (20,30] yrs 30+ yrs Work Experiences (job tenure) <1 yr (1,3] yrs (3,6] yrs (6,10] yrs (10,20] yrs (20,30] yrs 30+ yrs Work Experience (job tenure) Figure 5: Average Annual Income by Education Level Figure 6: Average Monthly Wage Rate by Education Level 3000000 180000 160000 2500000 140000 2000000 120000 1500000 Male Female 100000 80000 Male Female 1000000 60000 40000 500000 20000 0 No education Primary Post primary Secondary Post secondary 0 No education Primary Post primary Secondary Post secondary Education Level Education Level 7
To determine the source of the observed differences in wages, we first examine total hours worked for men and women to see whether these differences are due to men working longer hours. Table 2 summarizes the average hours spent per week on wage labor, and house work. At first glance, it appears that women are working just as much as men. There are some variations across occupations, sectors, level of education and experience. Overall, men work more than women in the agricultural sector, and in agriculture occupations. Men also work longer hours (roughly 4 to 9 hours more) in the construction sector. Both men and women work the longest hours in the service sector. Finally, for both men and women, those with primary education, as well as those with the least years of experience, work the longest hours. However, when one examines time spent on house work, the results show that women work on average almost 20 hours more than men. Again, individuals engaged in the service sector work the longest hours, as do those with primary education, and the young. These results suggest that women are working longer hours, and at lower returns. To further examine the sources of differences in earnings between men and women, we derive two measures of the gender earnings gap. These measures are outlined in the following section. 8
Table 2: Average Hours Spent per Week on: Wage Labor, House Work, and Total Working Hours Housekeeping Hours Sum Average Male Female Male Female Male Female General: 44.43314 43.16869 5.430984 23.87726 49.8454 67.1068 Occupation: Tech. Professional 39.0368 40.1868 0.9223 14.2097 39.9741 54.6038 Admin. Director/Executive 47.6623 n.a. 0.5714 n.a. 48.2338 n.a. Admin. Staff/worker 39.5052 40.1992 0.5105 10.0885 40.0157 50.2877 Commercial personnel 44.3520 44.0142 2.2554 19.8392 46.6074 63.9631 Service Worker 57.8350 57.0170 16.6478 36.3766 74.7204 93.1771 Agricultural labor 36.1707 28.6400 4.7641 22.4339 40.7736 51.4416 Non-AG skilled labor 44.9635 40.1780 3.3195 18.4826 48.2893 58.6607 Industry Other 46.5057 31.7500 2.8750 10.1875 49.3807 41.9375 Agriculture 36.7768 28.7056 4.6228 22.3606 41.2871 51.4296 Mining 38.5238 n.a. 2.6190 n.a. 41.1429 n.a. Manufacture 38.8110 38.8025 3.5893 17.6356 42.4003 56.4381 Electric, water, gas 45.5723 42.72727 1.4280 0 (N=2) 47.0003 42.7273 Construction 45.5781 39.5195 3.1488 22.1036 48.7270 61.6231 Commerce, Hotels 45.7996 44.9542 2.5236 19.1781 48.3328 64.1603 Transport, Communic. 51.2106 49.0909 1.4830 0.7667 52.7044 49.8576 Banking 44.0591 39.5988 1.2875 11.6236 45.3466 51.2223 Services 48.4189 50.7380 9.9145 29.2241 58.4028 79.9197 Others 45.5682 35.5606 6.7240 11.2917 52.2921 46.8523 Education No education 41.1361 38.47406 7.0916 24.0522 48.1688 62.4850 Primary 45.3756 46.2576 7.1824 28.7798 52.5569 75.1426 Post primary 46.1442 40.9390 3.2458 17.3346 49.3899 58.3448 Secondary 44.5783 42.3354 0.9789 16.3399 45.5633 58.8138 Post secondary 40.4102 40.5682 0.5772 9.2097 40.9946 50.0849 Experience less than 1 year 49.1334 49.3946 7.2583 24.9051 56.3481 74.4626 (1,3] years 44.7995 47.2987 8.1371 27.6150 52.9579 75.1730 (3,6] years 45.8574 42.4253 5.8228 22.8414 51.6199 64.8620 (6,10] years 44.9786 38.0845 2.8087 23.1730 47.7794 61.3909 (10,20] years 41.1220 35.9849 2.8615 21.0072 43.9876 56.9921 (20,30] years 38.5563 34.0675 2.3085 17.3326 40.8467 51.4739 30+ years 34.8505 25.7361 1.3806 17.5919 36.1443 43.4422 Age [14,25) 46.1583 50.0572 11.1362 31.3758 57.3092 81.4905 [25,35) 45.9853 40.5407 3.5469 22.0075 49.5518 62.6629 [35,45) 42.2795 39.0669 2.4129 17.7165 44.7081 56.7892 [45,55) 41.7244 34.8431 1.8068 13.2214 43.4859 47.9442 55+ 37.6304 31.3352 1.3398 14.7632 38.4999 46.1087 9
III. Measuring the Gender Earnings Gap In addition to the above graphical summaries of the earnings gap, we adopt two methods to estimate the earnings gap between males and females: i) an index measure of the difference between male and female wages, relative to female the male wage or income: ( 1 ), as in Oostendorp (2004). This male indicator shows the earnings gap measured by male standards. ii) a measure of the difference between male and female wages adjusted for socioeconomic factors, based on OLS regression of the log of wages (or income) on dummy variables for age, education, marital status, rural/urban region, as well as variables on hours of house work, and work experience. The next section describes the econometric specification employed to estimate earnings differences. ECONOMETRIC ESTIMATION OF EARNINGS DIFFERENTIALS Estimating wage models usually requires a first stage correction that measure the probability a person will participate labor force, as elaborated by Heckman (1974). Nevertheless we are not able to apply the Heckman two-stage selection model to the data due to two major obstacles: 1) the questionnaire does not provide a clear indicator of employment, thus the first stage cannot be properly defined; and 2) among interviewees who report to have primary occupations, only 19% (2878 out of 14830) indicate that they receive a wage for the work. In other words, the majority of the working sample do not receive remuneration for their labor. Given the above limitations, we employ a model based on OLS regressions, to estimate gender earnings differentials, with the assumption that wages earned by an individual are a function of observed characteristics defined by : i) education level; ii) age or years of experience on the job; iii) demands on the individual s time from household production and obligations; iv) number of children in the household; v) whether the household or individual is located in an urban or rural setting; vi) sector or occupation of employment. The estimate of the outcomes of annual income and the monthly wage rate are based on a linear model, which uses dummy variables to capture the effects of qualitative indicators or characteristics: Y µ + αx + βd + ε (1) i = i i i where the subscript i refers to the ith individual. The dependent variable, Y i, represents either annual income or monthly wage rates. On the right hand side, µ i is a constant, X i is a matrix that contains continuous explanatory variables i.e., housekeeping hours, numbers of children living in households, and number of employees at the work place. D i represents a vector of qualitative dummy variables, i.e. gender, education levels, and geographic regions. 10
We therefore compare average earnings for males and females within a given education, and age group, after controlling for variables contained in X. Each component in X explains part of the difference in wages between men and women. Specifically, the responsibility in housekeeping and child-caring, as well as the tendency to be employed in smaller size firms, may lead women to become less competitive in the labor market, compared to men with similar background and characteristics. We write D i explicitly (the subscript is suppressed for simplicity) as: Y = µ + α1 X + α 2Fem + βedu + γ ( Fem* Edu) + δgeog + φage + λmarital + ε (2) in which Fem equals 1 if the individual is female, 0 if male; and Edu indicates the level of education, geographic area, Geog (Kigali, other urban area, rural area), and marital status, Marital (1 if married or cohabitant, 0 otherwise). The interaction term is the product of the gender and the education dummy variables, and allows for the estimation of differences between men and women for a given level of education. For example, assuming 5 education categories, equation (2) can be further expanded to allow estimation of gender differences in wages by level of education, as follows: Y = µ + α1x + α2fem + β2edu2 + β3edu3 + β4edu4 + β5edu5 (3) + γ Fem * Edu ) + γ ( Fem * Edu ) + γ ( Fem* Edu ) + γ ( Fem* Edu ) + ε 2( 2 3 3 4 4 5 5 The dummy variables Geog, Age, Marital are also included in the regression, to capture differences in location, age, and marital status. In the case of the variable Geog, the categories consist of people who live in Kigali (41.97% of the wage earners), other urban areas ( Geog 2 ), and rural areas 3 ( Geog 3 ). The age variable is divided into 5 cohorts, the age group 14 to 25 is used as the reference group, for the following reasons: a) this group exhibits the lowest income or wage gap, as seen in Figure 1 and Figure 2; and b) the number of individuals between the ages of 14 and 25 (34.84%) exceeds the number of individuals found in the other age categories. The marital status dummy variable is divided into two categories: a) individuals that are married or cohabitating; and b) all others (widowed, separated, and single). The coefficients for the three groups of dummy variables reflect income or wage differences, between the category in question and the reference group. For instance, the coefficient of Geog 3 captures the income gap between rural residents and Kigali residents. The remaining are continuous (as oppose to dichotomous) variables: i.e. hours spent on housekeeping tasks, numbers of children living in the household, number of years spent working at a particular job, and the number of employees at an individual s place of work which is a proxy for the size of the firm or enterprise. Hours spent on housekeeping captures the trade-off between work and home production, which is higher for women. The number of children in the household is also another measure of this variable. Number of years on the job captures individual experience. We split the sample into three sub-samples, to capture the 3 Note that the sub-sample is representative of those engaged in wage labor. An income earner is more likely to reside in the urban area than in the rural area, while the full survey sample shows that 81% (un-weighted) of the interviewees live in rural areas. 11
potential effects of sector of employment. This approach avoids the problem of endogeneity (Paternostro and Sahn, 1999). Gender differences in earnings can then be estimated as follows 4 : α 2, the coefficient associated with the gender dummy, estimates the difference in wages between males (reference group) and females within the primary education level. β 2, β3, β4, β5, respectively, estimate men s income or wage differences, between each of the four education levels (post-primary, secondary, post-secondary education, and no education), and males with primary education (reference group). β i + γ i (i=2,3,4,5) estimate women s income or wage differences, between each of the four education levels, respectively, and females with primary education (reference group). α 2 + γ i ( i = 2,3,4, 5 ) provides the measure of income or wage differences (between men and women), at each educational level?. Coefficient estimates taken directly from regression results have their associated t- statistics. T-statistics are derived for coefficients derived from regression estimates, (as described above). For example, the t-statistic for the coefficient α 2 +γ i is derived as follows: tˆ = = σ ( α + γ ) V ( α + γ ) n 2 α + σ 2 γ = σ 2 α n( α + γ ) + 2Corr( σ + σ α 2 γ, σ n( α + γ ) γ + 2Cov( σ Standard errors estimated based on equation (4) are used to test for significance of derived coefficient estimates. ) * σ α * σ γ α, σ γ ) (4) DATA SUMMARY Before going on to summarize the estimates of the earnings differential between men and women, we first provide an overview of the variables used in the regression. Table 3 provides a summary of the regression variables by gender, and also according to the age of the individuals. In order to avoid any potential problems of endogeneity, the sample is split into sectors in order to account for differences in earnings across various sectors. The sectors are defined by those engaged in the agriculture sector, service sector (which include services provided to the community and cultural social services i.e. services of sovereignty, safety and justice, administrative and financial management, economic promotion, social services 4 See Johnston 1984, p. 234 for more detailed explanation of derivation and interpretation of coefficients. 12
and cultural, public bodies, research centers, international organizations, religious organizations; entertaining services and leisure i.e. tourism, sports, leisure, and craftsman arts; services provided to households and to private individuals i.e. domestic services, services of entretien and of repair, laundry/dyeing/hairdressing salon, and other personal services), and non-agriculture sectors (which include mining, manufacturing, electricity, gas, water, construction, commerce, hotels, transport and communication, and banking). Table 3 provides a summary of the variables used in the regression by sector. Average wages are lowest in the agriculture sector, and highest in the non-agriculture sectors, and on average, women earn less than men. The difference in earnings is larger for the over 35 age cohort. As the table shows, those individuals less than 35 spend more time on house work. In particular, women less than 35 spend more time on house work than men of the same age. To support this, women reside in households that have slightly more children. The table also shows that most of the wage earners are in Kigali, particularly those less than 35 years of age. 13
Table 3: Data Summary Full Sample: Agriculture Sample Male Female Male Female Age<35 Age>=35 Age<35 Age>=35 Age<35 Age>=35 Age<35 Age>=35 Monthly Wage Rates 28739.53 52346.45 20976.75 28523.96 9075.313 12471.96 5520.462 4618.724 Annual Income 409627.1 728690.6 294291.8 398606.4 137890.9 154125.6 98760.89 53572.94 Education (%) No Education 13.71 17.86 17.59 33.33 32.48 42.36 40 60.31 Primary 58.31 45.62 53.27 34.99 64.53 51.39 54.67 36.64 Post-Primary 8.45 10.39 7.16 9.09 0.85 2.08 2 1.53 Secondary 16.44 18.34 19.85 18.73 2.14 2.78 3.33 1.53 Post-Secondary 3.09 7.79 2.14 3.86 n.a. 1.39 n.a. n.a Region (%) Kigali 39.42 37.99 48.99 41.05 4.7 14.58 13.33 16.03 Other Urban 20.44 19.64 25.13 23.42 11.97 12.5 19.33 19.85 Rural 40.15 42.37 25.88 35.54 83.33 72.92 67.33 64.12 # of kids 1.66 2.75 2.27 2.58 2.05 2.61 2.33 2.40 Ratio of related kids 0.67 0.84 0.85 0.88 0.81 0.84 0.88 0.89 # of People at Work 23.70 26.21 11.02 9.13 17.54 21.01 12.68 6.11 Experiences (tenure) 4.40 15.19 3.79 15.02 6.11 22.04 7.45 23.89 Weekly Housekeeping 7.30 2.08 27.39 16.18 4.72 4.47 25.24 19.06 Service Sample Other Sectors Sample Male Female Male Female Age<35 Age>=35 Age<35 Age>=35 Age<35 Age>=35 Age<35 Age>=35 Monthly Wage Rates 24498.75 43718.61 18164.69 41839.98 41893.68 75661.08 35329.58 41636.9 Annual Income 399535.6 671734.6 294014.4 666305.6 553114.3 1029972 416967.9 549706.9 Education (%) No Education 9.85 6.63 15.27 11.49 7.64 12.42 7.5 22.07 Primary 51.52 36.14 56.65 11.49 60.93 48.04 46.67 47.59 Post-Primary 8.33 11.45 5.42 17.24 12.31 13.73 13.33 11.03 Secondary 25.25 28.31 19.95 47.13 16.14 20.26 30 17.24 Post-Secondary 5.05 17.47 2.71 12.64 2.97 5.56 2.5 2.07 Region (%) Kigali 46.46 31.93 56.9 37.93 50.74 52.29 57.92 65.52 Other Urban 27.53 24.1 25.62 24.14 18.68 20.59 27.92 26.21 Rural 26.01 43.98 17.49 37.93 30.57 27.12 14.17 8.28 # of kids 1.65 2.67 2.25 2.43 1.48 2.85 2.28 2.82 Ratio of related kids 0.65 0.83 0.84 0.84 0.62 0.85 0.85 0.88 # of People at Work 34.50 42.56 6.36 14.30 17.79 19.85 17.85 8.77 Experiences (tenure) 3.54 11.54 2.61 13.98 4.27 13.95 3.51 7.57 Weekly Housekeeping 13.62 1.07 32.57 13.62 3.28 1.51 19.96 15.12 Note: Other sectors includes, mining, manufacturing, electricity/gas/water, construction, commerce, hotels, transport and communication, and banking. 14
This contrasts with the agriculture sector where most people reside in the rural area. As summarized in Pierre (2004), wage or income earners are more likely to live in urban areas and work in the tertiary sectors (e.g. commerce, banking, and social/private services). They are less likely to be in the agriculture industry or work as agricultural laborers, and are likely to be better educated (around 50% of them have completed primary schools). Table 3 also shows that most people in this sample have primary education, and that men tend to work in larger firms. This may also be an indication of women being more engaged in part-time work and the informal sectors compared to men. REGRESSION RESULTS female Estimates of wage gap i.e. ( 1 ) vary considerably across occupations, male sector and education level, and are summarized in Appendix I. Overall, men earn 37% more than women. Except for three categories ( administrative staff/worker in occupations, electricity/water/gas and other in industry), the gender gap varies from 5.67% (for the wage rate in the manufacturing sector) to 92.8% (income, more than 30 years experience in the current job). Overall, the gender gap appears to be significant, and is largest between men and women who have been engaged at the same job for more than 20 years. If one looks at occupations, gender earnings differences are largest for those engaged in non-agriculture skilled labor. Across sectors, it is highest for the commerce and hotels sector, followed by the agriculture sector. In terms of education, the greatest difference is seen between men and women with primary education. The results of the regressions are summarized in Appendix II to V. We estimate the coefficients for (the log of) both income and wage rates, regressed against demographic and socio-economic characteristics described in the previous section. As a result, differences in income between men and women are measured according to levels of education, marital status, and geographical location. Three regressions are estimated to reflect these measures, for each sector (agriculture, service, non-agriculture), and pooled sample. These are summarized in Appendix II to Appendix V. For each table, three models are estimated which account for interaction terms between the dummy denoting gender and the dummies for education level (Model I in the tables), marital status (Model 2), and geographic region (Model 3). Appendix II summarizes the pooled regressions for both income and wages. Most of the coefficients in the regression are significant. In the case of earnings by age cohort, the coefficients show that earnings are significantly lower for the age group 14 to 25 compared to all other age groups. The regressions also show that average earnings are significantly higher with increased schooling. Earnings are higher in Kigali compared to other urban areas and the rural areas. Females earn significantly less than men. Finally, longer hours of house work lead to lower average earning. This effect is supported by the result that an increase in the proportion of children in the household that are related to the head of household, leads to a reduction in the average earnings. Work experience increases average earnings at a 15
decreasing rate, and individuals working in larger firms earn more. These results are generally repeated for the service and non-agriculture sectors (Appendix IV and V). In the case of the agriculture sector (Appendix III), earnings differences between the 14 to 25 age cohort and all other cohorts are less significant. Also, unlike the pooled and service sector sample, the higher the total number of children in the household, the higher the average earnings (and this result is significant for the income measure). Years of experience and size of the firm does not significantly affect earnings. The differences here between agriculture and the other sectors is a reflection of the differences in the type of work, which allows for more flexibility with respect to the trade-off with house work, and the ability for children lend a hand around the house and with work in the sector. In addition, the education effect is less strong for the agriculture sector, compared to the service and non-agriculture sectors. Table 4 below summarizes the gender differences in wages according to education level, marital status, and region. These differences are derived from regression coefficients summarized in Appendix II to V, using the model outlined in Section III. Table 4 therefore, is a summary of the estimates of α 2 + γ i. The Table also indicates levels of significance based on the adjusted t-statistics. Table 4: Gender Income/Wage Differences, t-statistics adjusted Pooled Sample Agriculture Sample Service Sample Non-Agriculture Sample income Wage income wage income wage income wage Education Primary -0.4378-0.2901-0.3704-0.1038-0.5219-0.5099-0.0612 0.0015 Post-primary -0.372-0.1907 0.2312-0.024-0.06 0.1495-0.4403-0.3576 Secondary -0.1507-0.0236-0.6377-0.6796-0.0158** 0.0276-0.2594-0.1467 Post-secondary 0.0758 0.1109 n.a. n.a. 0.0465 0.2287-0.0578-0.2232 No education -0.1014-0.0281-0.1332-0.0727 0.0772 0.0762-0.0612-0.114 Marital status Single/other -0.2671-0.1382** -0.1002 0.143-0.2372-0.2140-0.2894** -0.1391 Married/cohabitant -0.3127-0.1859-0.435-0.365-0.2107-0.114-0.0375-0.0742 Region Kigali -0.4358-0.3181-0.8010-0.9840-0.3533-0.2748-0.2842** -0.2152** Other urban -0.1704-0.0533 0.0518** 0.0858-0.2434-0.1805 0.0120 0.0325 Rural -0.1989-0.0523-0.2634-0.0504** 0.0041-0.0156 0.0174** 0.0742 Note: Underlined estimates are insignificant; ** denotes significance at 5% level; all other estimates are significant at 1% level. There are no female in post-secondary education in the agriculture sample. The results show that aside from those with post-secondary education, women earn less than men, across all categories; and the differences are all significant. The gender income/wage gaps fall as the level of schooling increases from primary level to postsecondary. Moreover, women earn 7.58% in income and 11.09% in wage rates more than men at post-secondary level, suggesting a strong incremental reward of schooling for women. There may be two factors influencing these results. First, the higher one s education level is, the less likely that the individual will work in the agriculture sector and the more likely that the individual will join the secondary or tertiary sectors, such as the service industry 5. 5 See Table A. 16
Second, in the non-agriculture sector, since there are few women who have high education (secondary and higher), the sample is small, and therefore there may be some degree of bias in the estimates. Individuals with primary education make up the largest sub-sample (53.76% for men and 47.54% for women; total average 51.25%), and they appear to be the lowest paid workers. For the no-education group, the data does not provide sufficient information to disentangle the effect of the concentration in the agriculture sector from the effect of the absence of schooling. The gender wage gaps are greater in the income model compared to the wage model. For example, the annual income of married or cohabitant women is 31.27% less than that of married men, yet the gap shrinks to 18.59% in terms of wage rates; while rural women s annual income is 20% less than men, they earn 5.23% less than rural men in monthly wage rates. Women s greater responsibility for house work and child care may contribute to this tendency. In addition, with respect to work related transfers, women s lack of mobility may inhibit their access to benefits and transfers i.e. such as paid sick leave, etc. The education estimates in this sample follow a different pattern from the pooled regression. In the agriculture sector there is no available female observation at the postsecondary level, thus we cannot measure gender gaps at this educational level. The gender gap between women and men with secondary education appears to be more than 60%. The gender wage gap is reduced and even reversed when people have post-primary education. The insignificant estimates of marital status demonstrate that single persons (including widowed and separated) in the agriculture sector earn similar income and wage rates, but the gap between married men and women increases to around 40%. This may be a result of the general availability of agricultural jobs and single persons lack of housework responsibility. The biggest geographic gender gap is shown in Kigali, and men s incomes are less than women s among other urban residents. Further examination of the data found that the there are more women in this sample who are engaged in self-employment relative to men (80 percent for women compared to 45 percent for men). In addition, women were slightly younger than men in the sub-sample. The gender income/wage differences by educational levels are positive for those with higher education and no education. Women earn either significantly more than or slightly less than men do. Within the four educated groups, we observe similar increasing returns to schooling. As Table 5 shows, there are positive and significant returns to education for both men and women, particularly for the agriculture and service sector. In terms of marital status and regions, the service sub-sample replicates the pooled sample results except for the minor differences in scale. The rural area gender comparison, however, demonstrates minor gender inequalities in income and earnings. 17
This sample is complementary to the agriculture sample and the service sample it contains all industries other than the two sectors addressed above: commerce, banking, electricity/water/gas, manufacture, mining, construction, transportation. We compare the estimates as an averaged group with the estimates in the other two sectors. Such a mixed sample contains outliers who may earn higher/lower incomes in sectors that are not popularized. For example, only male workers participate the mining sector (21 observations), yet they earn less than the pooled sample average; both male and female transportation workers receive higher work payments than their respective sample average, however the participants are not evenly distributed: 137 male but only 5 female; although female workers in electricity/water/gas have higher income than male in the same sector, there are 22 men and only 2 women. Comparing the estimates of this sample with those of agriculture and service samples, we observe that in some demographic categories female agriculture and service laborers receive far less income and wage rates than men, yet such inequalities are lower for other sectors. Unlike Pierre (2004), who concludes that the agriculture sector is among the sectors with the least inequalities (page 14), we detect that the agriculture workers do not always enjoy income/wage equality once we control for education level, marital status, and regional categories. The comparison between Agriculture Sector and Non-Agriculture Sector indicate that female agricultural laborers with primary and secondary education fare worse than women in other sectors, when both are compared with men with the same level of schooling. The same conclusion is observed in married/cohabitant people, Kigali residents, and rural residents. 18
Table 5: Income/Wage differences of men and women at each educational level when compared with primary (control) education Male Female Income Wage Income Wage Pooled Sample: No Education -0.6537-0.6288-0.3713-0.3668 Post-Primary 0.4796 0.5211 0.5454 0.6205 Secondary 0.8631 0.8046 1.1502 1.0711 Post-Secondary 1.6928 1.2954 2.2064 1.6964 Agriculture Sample: No Education -0.2927** -0.2146* -0.0555-0.1835 Post-Primary -0.4751 0.1337 0.1265 0.2135 Secondary 0.7088* 0.8976 0.4415 0.3218 Post-Secondary 3.5260 3.2381 n.a. n.a. Service Sample: No Education -0.5843-0.5931 0.0148-0.07 Post-Primary 0.3496** 0.4437 0.8115 1.1031 Secondary 0.8085 0.9182 1.3146 1.4557 Post-Secondary 1.7946 1.3408 2.363 2.0794 Other Sectors Sample: No Education -0.4355-0.3879-0.4355-0.5034 Post-Primary 0.2951** 0.3317-0.084-0.0274 Secondary 0.7289 0.5787 0.5307 0.4305 Post-Secondary 1.4682 1.2739 1.4648 1.0492 Note: Underlined estimates are insignificant;* denotes significance at 10% level; ** denotes significance at 5% level; all other estimates are significant at 1% level. SUMMARY AND CONCLUSION This paper has examined the wage differences between men and women using an measure of percent difference in wages and also a regression based approach which adjusts for socio-economic differences. The results of the analysis show that more women than men are engaged in unpaid work. There is definitely a positive return to education, and the results here suggest these returns may be higher for women compared to men. Hours spent on house work and responsibility for care of children has a dampening effect on wages, particularly for women. This latter factor may also affect the mobility of women, and explain the larger differences in wages observed for the older or more experienced cohorts of the sample. However, the evidence strongly suggests that education is an equalizing factor for wages. In addition, the wage differences are lower (and even favor women) for individuals engaged in self-employed activities, particularly for those less than 35 years of age. Given these results, policies to increase access of women to higher education (beyond primary and particularly secondary education) will be key to reducing these differences in earnings. Ideally these policies should be coupled with measures to reduce wage discrimination, as well as the provision of child care (or early childhood education), to help reduce the cost to women for home and child care. 19
References: Heckman, J. Shadow Prices, Market Wages, and Labor Supply, Econometrica, 42, 679-693 Hersch, J. Male-Female Differences in Hourly-Wages: The Role of Human Capital, Working Conditions, and Housework, Industrial and Labor Relations Review, Vol. 44, No. 4 (Jul. 1991), 746-759 Oostendorp, R. H. Globalization and The Gender Wage Gap, The World Bank, Policy Research Working Paper, No. 3256, April 2004 Pierre, G. Descriptive Analysis of the Rwandan Labor Market, The World Bank, Draft 4, Feb 2004 Paternostro, S. and D. E. Sahn, Wage Determination and Gender Discrimination in a Transition Economy: The Case of Romania, The World Bank, Policy Research Working Paper, No. 2113, May 1999 20
Table A: Level and percentage of agriculture and service sector income/wage earners Agriculture sector workers Male Male (N) Percent Female (N) Female Percent No Education 137 0.3624 139 0.4947 Primary 225 0.5952 130 0.4626 Post Primary 5 0.0132 5 0.0178 Secondary 9 0.0238 7 0.0249 Post Secondary 2 0.0053 0 0.0000 Total 378 281 Service sector workers Male (N) Male Percent Female (N) Female Percent No Education 50 0.0890 72 0.1460 Primary 264 0.4698 240 0.4868 Post Primary 52 0.0925 37 0.0751 Secondary 147 0.2616 122 0.2475 Post Secondary 49 0.0872 22 0.0446 Total 562 493 Appendix 1: Average annual income and wage rate by gender Annual Income Monthly Wage Rate Male Female Gap Male Female Gap General 524096 326963.2 0.3761 37229.74 23329.12 0.3734 Number of observation 1717 1159 1710 1155 Occupation: Tech. Professional 899336.7 665595.5 0.2599 58664.55 42895.7 0.2688 Admin. Director/Executive 1591910 n.a. n.a. 97873.43 n.a. n.a. Admin. Staff/worker 1163856 1203063-0.0337 55046.25 76670.38-0.3928 Commercial personnel 998340.9 456431.3 0.5428 77587.26 37611.18 0.5152 Service Worker 260758.1 147356.7 0.4349 17596.34 9544.277 0.4576 Agricultural labor 119687.2 82107.82 0.3140 8625.58 5019.005 0.4181 Non-AG skilled labor 517146.2 186052.4 0.6402 39761.6 17958.75 0.5483 Other 132075 122850 0.0698 10775 8750 0.1879 Industry Agriculture 144075.5 77694.62 0.4607 10381.01 5103.774 0.5084 Mining 205673.8 n.a. n.a. 18922.22 n.a. n.a. Manufacture 432252.6 332395.7 0.2310 29316.29 27653.45 0.0567 Electric, water, gas 926540.9 960000-0.0361 54784.09 64000-0.1682 Construction 410437.3 325750 0.2063 38066.8 21492.86 0.4354 Commerce, Hotels 1040115 453734.7 0.5638 79839.67 37539.47 0.5298 Transport, Communic. 729764.5 457500 0.3731 45940.44 40000 0.1293 Banking 1018372 758637.6 0.2550 79247.58 55819.31 0.2956 Services 479936 359712.8 0.2505 30175.79 22342.68 0.2596 Others 400650 541900-0.3526 28400 34166.67-0.2031 21
Education Primary 321510.7 182819.4 0.4314 24776.9 14195.27 0.4271 Post primary 612875.1 384493.7 0.3726 52096.11 27676.34 0.4687 secondary 905214.4 565926.2 0.3748 62608.59 43008.39 0.3131 superieur 2438216 2503401-0.0267 136736.3 156318.7-0.1432 No education 156437.9 146006.7 0.0667 12043.51 8136.711 0.3244 Experience less than 1 year 210998.2 144191.7 0.3166 20827.88 15384.16 0.2614 (1,3] years 633425.8 446576.9 0.2950 40499.24 29635.23 0.2683 (3,6] years 495183.9 421161.5 0.1495 38256.29 26807.6 0.2993 (6,10] years 529432.9 373692.5 0.2942 36917.5 21380.24 0.4209 (10,20] years 464091.2 316975.1 0.3170 35861.5 26407.35 0.2636 (20,30] years 722863.3 283522.3 0.6078 55075.38 22375.99 0.5937 30+ years 1077310 77124 0.9284 47715.95 6844.667 0.8566 Notes: 1. There is neither female observation in Administration Director/Executive nor in Mining. Therefore the corresponding Gap returns no value. 2. Negative gaps are the results of very small samples, relative to the full sample(n=2876): occupation Administrative Staff/Worker, N_male=79, N_female=68; industry Electricity/Water/Gas, N_male=22, N_female=2; industry Other, N_male=16, N_female=6; education Superier, N_male=82, N_female=31 3. A summary of the income or wage gaps between males and females are presented in Appendix I, which measures gender gaps across occupations, industries, education levels, and years of experience. The Male and Female columns represent the average income or wage rates of each gender in the corresponding categories, e.g. male technical professionals average income is FRw 899,366 per year, and female technical professionals earn about FRw 665,595 annually. The column Gap is an equivalent measure of gender wage inequality. Appendix 2: Pooled Sample Regression Y=ln(income/wage rate) INCOME WAGE Control: age [14,25) (1) (2) (3) (1) (2) (3) Age [25,35) 0.5725*** 0.5723*** 0.5793*** 0.5599*** 0.5595*** 0.5673*** (0.0644) (0.0647) (0.0645) (0.0544) (0.0546) (0.0544) Age [35,45) 0.5547*** 0.5490*** 0.5572*** 0.6355*** 0.6299*** 0.6391*** (0.0796) (0.0803) (0.0796) (0.0672) (0.0678) (0.0671) Age [45,55) 0.6484*** 0.6354*** 0.6501*** 0.7074*** 0.6950*** 0.7113*** (0.1002) (0.1010) (0.1001) (0.0844) (0.0851) (0.0843) Age 55+ 0.7600*** 0.7617*** 0.7939*** 0.6332*** 0.6311*** 0.6662*** (0.1447) (0.1452) (0.1449) (0.1220) (0.1224) (0.1220) Control: primary education Post primary 0.4796*** 0.5033*** 0.4978*** 0.5211*** 0.5573*** 0.5506*** (0.1040) (0.0838) (0.0837) (0.0876) (0.0705) (0.0704) Secondary 0.8631*** 0.9786*** 0.9741*** 0.8046*** 0.9126*** 0.9072*** (0.0812) (0.0637) (0.0635) (0.0684) (0.0536) (0.0534) Superier 1.6928*** 1.8468*** 1.8384*** 1.2954*** 1.4168*** 1.4063*** (0.1396) (0.1199) (0.1200) (0.1175) (0.1010) (0.1009) No education -0.6537*** -0.5062*** -0.5152*** -0.6288*** -0.5152*** -0.5257*** (0.0834) (0.0624) (0.0624) (0.0704) (0.0526) (0.0527) Control: unmarried other Married/Cohabitant 0.0852 0.1085 0.0825 0.2039*** 0.2285*** 0.2019*** 22
(0.0556) (0.0708) (0.0555) (0.0469) (0.0597) (0.0468) Control: Kigali Other urban -0.3316*** -0.3275*** -0.4481*** -0.3261*** -0.3227*** -0.4429*** (0.0581) (0.0583) (0.0784) (0.0489) (0.0491) (0.0659) Rural -0.8869*** -0.8761*** -0.9686*** -0.9361*** -0.9261*** -1.0287*** (0.0547) (0.0547) (0.0670) (0.0461) (0.0461) (0.0565) Housework: Housekeeping hours* -0.0003*** -0.0003*** -0.0003*** -0.0035*** -0.0036*** -0.0035*** 0.0000 0.0000 0.0000 (0.0003) (0.0003) (0.0003) # of kids 0.0269 0.0271 0.0259-0.0006-0.0001-0.0014 (0.0170) (0.0170) (0.0170) (0.0143) (0.0143) (0.0143) # of related kids/# of kids** -0.1705** -0.1825** -0.1598** -0.1532** -0.1654*** -0.1408** (0.0709) (0.0715) (0.0711) (0.0597) (0.0602) (0.0599) Work Experiences: Tenure 0.0256*** 0.0262*** 0.0261*** 0.0077 0.0081 0.008 (0.0066) (0.0066) (0.0066) (0.0055) (0.0055) (0.0055) Tenure^2-0.0008*** -0.0008*** -0.0008*** -0.0004*** -0.0004*** -0.0004*** (0.0002) (0.0002) (0.0002) (0.0001) (0.0001) (0.0001) # of people at work 0.0004* 0.0004* 0.0004* 0.0001 0.0001 0.0001 (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) Control: male Female -0.4378*** -0.2671*** -0.4358*** -0.2901*** -0.1382** -0.3181*** (0.0722) (0.0691) (0.0772) (0.0609) (0.0583) (0.0650) Female*Post primary 0.0658 0.0994 (0.1686) (0.1420) Female*Secondary 0.2871** 0.2665** (0.1228) (0.1034) Female*Superier 0.5136** 0.4010* (0.2566) (0.2160) Female*No education 0.3364*** 0.2620** (0.1213) (0.1023) Female*Married/Cohabitant -0.0456-0.0477 (0.0957) (0.0807) Female*Other Urban 0.2654** 0.2648*** (0.1167) (0.0981) Female*Rural 0.2369** 0.2658*** (0.1058) (0.0893) Constant 11.8803*** 11.8200*** 11.8795*** 9.5309*** 9.4776*** 9.5402*** (0.0701) (0.0701) (0.0708) (0.0591) (0.0591) (0.0596) Observations 2868 2868 2868 2857 2857 2857 R-squared 0.413 0.4104 0.4119 0.4912 0.4891 0.4912 Standard errors in parentheses; * significant at 10%; ** significant at 5%; *** significant at 1% Notes: outcome = ln(annual income) or ln(monthly wage rate) (1) controls for interaction=female*education (2) controls for interaction=female*marital status (3) controls for interaction=female*regions *Housekeeping hours: first we calculate the weekly housekeeping hours (fetching wood/water, shopping, cooking, cleaning, and childcare), then use annualized (weekly hours*37) in income regression; use monthly (weekly hours*4) in wage regression **Number of children: people younger than 14 (excluded) living in a household 23