UNFPA COUNTRY SUPPORT TEAM Office for the South Pacific DISCUSSION PAPER NO. 23 Women’s Labour Market Status in Fiji: Are They Subjected to Discrimination? by William J House Adviser on Population Policies & Development Strategies UNFPA/CST, Suva 14 January 2000 The views and opinions contained in this Paper have not been officially cleared and thus do not necessarily represent the position of the United Nations Population Fund Preface The UNFPA Country Support Team for the South Pacific, based in Suva, Fiji, is one of eight regional technical support teams established by the United Nations Population Fund to provide countries with technical assistance and information to meet country needs in the population field. In fulfilling this function, apart from field missions, the Country Support Teams also try to foster active communication and open discussion with national experts to promote a more holistic approach to population programmes. This Discussion Papers series have been initiated by the CST (Suva) in an attempt to establish a dialogue among national population programme personnel on the integrated and co-ordinated multidisciplinary approach to population. Hence, CST Discussion Papers are not particularly addressed to academic audiences but to practitioners. This paper compares the status of women with that of men in Fiji's formal sector labour market, using data collected from over 8,000 employees. This analysis examines gender differences in relative earnings and occupational status as well as the returns to various kinds of human resource endowments, such as education, job characteristics and tenure with employer. It is anticipated that the results will be of practical interest to planners and policy makers in both Fiji and other Pacific Island countries. 14 January 2000 William House Officer-In-Charge Table of Contents EXECUTIVE SUMMARY .................................................................................................................................. I 1. INTRODUCTION ....................................................................................................................................... 1 2. THE COUNTRY CONTEXT .................................................................................................................... 2 LABOUR FORCE GROWTH AND THE LABOUR MARKET .................................................................................. 3 Males ............................................................................................................................................................. 4 Females ......................................................................................................................................................... 4 THE STRUCTURE OF EMPLOYMENT ................................................................................................................. 5 THE OCCUPATIONAL STRUCTURE OF EMPLOYMENT...................................................................................... 6 3. THE STRUCTURE OF OCCUPATIONS IN FIJI BY GENDER ......................................................... 7 THEORIES OF SEX INEQUALITIES IN THE LABOUR MARKET........................................................................... 7 THEORIES OF LABOUR MARKET SEGMENTATION........................................................................................... 9 OCCUPATION SEGREGATION AND CONCENTRATION .................................................................................... 10 THE SITUATION IN FIJI ................................................................................................................................... 11 ALTERNATIVE PRESENTATIONS: DISTRIBUTION OF THE MALE AND FEMALE LABOUR FORCE ................. 13 ANALYSIS OF THE EXTREMES ......................................................................................................................... 15 4. THE STRUCTURE OF EARNINGS BY SEX ....................................................................................... 20 THE SOURCE OF INFORMATION: THE FIJI EMPLOYMENT SURVEY .............................................................. 20 THE SURVEY METHODOLOGY ........................................................................................................................ 21 MALE-FEMALE PAY DIFFERENTIALS............................................................................................................. 22 5. A DECOMPOSITION OF MALE-FEMALE EARNING DIFFERENTIALS ................................... 26 THE METHODOLOGY ...................................................................................................................................... 26 6. CONCLUSIONS ....................................................................................................................................... 31 REFERENCES ................................................................................................................................................... 33 EXECUTIVE SUMMARY The Programme of Action of the International Conference on Population and Development (ICPD), held in Cairo in 1994, encourages countries to act to empower women and to take steps to eliminate inequalities between men and women. Unfortunately, in the Pacific Island countries, there is a dearth of information and data on the relative labour market status of women vis a vis men on which to build policy reform. A recent opportunity to break new ground in this area has come about through the activities of a UNFPA-sponsored initiative in Fiji. During 1995-1997, UNFPA funded a project in the National Planning Office entitled: “Assistance in Population, Workforce Planning and Human Resources Development Planning”. Since that time, UNDP has taken over the funding of the project, which is now, entitled: “Strategic Human Resources Planning in Fiji”. Both phases of the project have been executed by the International Labour Organization and received substantial technical input from the UNFPA Country Support Team (CST) based in Suva. One of the major activities of the project was to conduct a survey of employees in Fiji’s formal sector labour market, documenting the human resource endowments and other socioeconomic characteristics of respondents, including their earnings and hours of work in a recent reference period, and contributing to the design of policies to redress the seemingly disadvantaged position of female workers. One primary concern of the survey was to generate appropriate data from which to assess the relative labour market position of female compared with male employees and to attribute such differences to human resource endowments as well as to labour market “discrimination” against women. This paper has demonstrated that there was an apparent significant increase in the labour force participation of women in Fiji over the decade 1986-1996. The female share of the total labour force rose from 21% to 33%, but a large proportion of these women failed to find cash employment and large numbers entered the non-cash, subsistence, largely agricultural economy. By the mid-1990s women’s share of the two largest occupational groupings (agriculture and elementary occupations) had trebled to one-third and the gender concentration of the labour force seems to have become somewhat less extreme. Yet, their greater representation in the lowly paid occupations did little to raise women’s overall status in Fiji’s labour market. Women have made little headway into male occupational preserves. The analysis went on to examine how this extreme concentration has impacted on the relative earnings of women compared with men in the formal sector of the labour market. Using data collected from over 8,000 employees in a specially designed survey, analysis has shown that mean pay is consistently higher for men compared with women while women’s earnings profile across age groups is much flatter, especially in the private sector, perhaps reflecting the kinds of jobs to which they are assigned, lacking opportunities for much on-the-job skill acquisition and pay advance. This result may also reflect the effect of discontinuities in women’s attachment to the labour market and the loss of productivity and pay-enhancing continuous work experience. Male-female earnings differentials take a U-shape across education levels, being wider at lower and high levels of schooling. Again, since very few women with higher education g:\cstsuva\discuspa\bh_23.doc i attain the very senior, highly paid executive positions, gender differences in occupational assignment contribute to these differences. The paper concludes with a standard decomposition analysis of male/female pay differentials which has been widely utilised around the world. The method decomposes the pay differential between the sexes into a portion attributable to a composition effect – “endowments” – and a portion attributable to structural differences, or “discrimination”, reflecting differences in the returns to these endowment factors. The results show that roughly 40% of the pay difference can be attributed to the composition effects, reflecting differences between males and females in their education, general and firm experience, ethnic group, sector of employment, training received and occupation. Some of these attributes, including education and training, favour female workers. The remaining 60% of the pay difference is attributable to differences in the returns to various human capital and other characteristics of the workers. Again, some advantages in these returns favour women. The principal source of “discrimination” lies in an unexplained intercept term in the regression analysis, whereby mean male earnings lie above that of females after controlling for many of these human capital attributes and various types of insertion in the Fiji labour market. The analysis leaves open the possibility that occupational sorting may play a major role in explaining the differences in earnings between the sexes. By treating broad-based occupational groups as a set of dummy variables, the analysis has discounted both the possibility of different returns to human capital between occupations and the possibility that the formation of the occupational distributions for men and women may, themselves, be discriminatory. This paper has elaborated on the methodology – the “expanded method” – that would allow an estimate of occupational discrimination to be made, whereby women with the same human capital and other attributes as men, do not have equal access to all occupations, including the highly paid executive positions. A subsequent paper in this series will utilise the data collected in the special survey to test this model. g:\cstsuva\discuspa\bh_23.doc ii 1. INTRODUCTION The Programme of Action of the International Conference on Population and Development (ICPD), held in Cairo in 1994, encourages countries to act to empower women and to take steps to eliminate inequalities between men and women. The ICPD resolved: “All countries should make greater efforts to promulgate, implement and enforce national laws and international conventions to which they are party, such as the Convention on the Elimination of All Forms of Discrimination Against Women, that protect women from all types of economic discrimination. Governments and employers are urged to eliminate gender discrimination in hiring, wages, benefits, training and job security with a view to eliminating gender-based disparities in income” (United Nations, 1994, para. 4.5; 4.7). Unfortunately, in the Pacific Island countries, there is a dearth of information and data on the relative labour market status of women vis a vis men on which to build policy reform. A recent opportunity to break new ground in this area has come about through the activities of a UNFPA-sponsored initiative in Fiji. During 1995-1997, UNFPA funded a project in the National Planning Office entitled: “Assistance in Population, Workforce Planning and Human Resources Development Planning”. Since that time, UNDP has taken over the funding of the project which is now entitled: “Strategic Human Resources Planning in Fiji”. Both phases of the project have been executed by the International Labour Organization and received substantial technical input from the UNFPA Country Support Team (CST) based in Suva. One of the major activities of the project was to conduct a survey of employees in Fiji’s formal sector labour market, documenting the human resource endowments and other socioeconomic characteristics of respondents, including their earnings and hours of work in a recent reference period, and contributing to the design of policies to redress the seemingly disadvantaged position of female workers. One primary concern of the survey was to generate appropriate data from which to assess the relative labour market position of female compared with male employees and to attribute such differences to human resource endowments as well as to labour market “discrimination” against women. The remainder of this paper utilises the results of recent Population Censuses in Fiji and the 1997 survey of employees to portray the changing position of females in the labour market over time and to estimate the nature and causes of women’s relative position, as reflected in their inferior pay and occupational assignment compared with male employees. Indeed, the survey indicates that the average female formal sector employee receives only 79% of the average weekly pay of males. In terms of the mean rate of hourly pay, the average woman earns 81% of the hourly pay of the average man. To what extent are these differences attributable to superior income-generating attributes of men, such as education, vocational training and occupational assignment? And to what extent are such differences in earnings attributable to “discrimination” against female workers in one from or another? Indeed, is the occupation assignment mechanism itself “discriminating” against female workers in its allocation of the most desired jobs? The remainder of this paper undertakes to explore some of these issues and concludes with some preliminary policy suggestions to redress women’s inferior labour market position in Fiji. g:\cstsuva\discuspa\bh_23.doc 1 2. THE COUNTRY CONTEXT Fiji has the largest and most developed economy of all the South Pacific island countries, with a per capita GDP of about US$2,300 in 1996 (UNDP, 1998). The country is favourably endowed with natural resources, a high level of human resource development, a dynamic class of entrepreneurs, and a strategic geographic situation at the hub of economic activities in the region. Furthermore, Fiji’s social indicators – life expectancy, infant mortality, adult literacy and school enrolment – reflect the country’s relatively high income and development status. According to the 1998 UNDP Human Development report, Fiji ranks 44th out of the 174 countries in its human development index. The closest South Pacific island country was Samoa, which ranks 94th, followed by Solomon Islands and Vanuatu at 123rd and 124th respectively. Set against this favourable portrait, the economy relies on a very narrow base and is overdependent of the tourism and sugar sectors. Perennially, the economy remains extremely vulnerable to dramatic changes in climatic conditions. Following the economic decline in the period after the coups of 1987, a series of economic reforms induced a rise in the economic growth rate, ranging between 1 and 5% per annum during the 1990s. In 1996 per capita GDP was 18% higher than in 1990. Despite these improvements, the generation of formal sector employment opportunities has failed to meet the needs of a burgeoning supply of new job-seekers. With a cap imposed on public sector employment, the private sector has been capable of absorbing no more than one-half of the new labour market entrants each year. This imbalance in the labour market – the gap between the number of new job seekers coming out of the school and vocational training institutions and the number of new job opportunities opening up – remains one of the major social and economic problems yet to be adequately addressed. At the time of the 1996 Census the total enumerated population of Fiji was 775,077 compared with 715,375 ten years earlier. The annual inter-censal growth rate was 0.8% which is significantly less than the 1.97% p.a. increase recorded over the 1976-86 decade. This fall reflects the continuing decline in fertility, for Fijians a fall in the total fertility rate (TFR) from 5.6 in 1966 to 3.9 in 1986; for Indians from 5.4 to 2.6. By 1996, the TFR for all women combined had fallen to 3.3. Another contributing factor to the decline in the rate of population growth was massive emigration, largely of the Indian ethnic group, soon after the 1987 military coups. While emigration levels have tended to stabilize, the country continues to suffer from the past and current loss of some of its more educated and trained citizens. The distribution of the population by broad age groups suggests that Fiji is slowly departing from its very youthful status in which, in 1976, 41% of the population was under 15 years old; by 1996 this had declined to 35%. Over the same period the share of the 65 years and over had grown from 2.4% to 3.2%. As a result, the share of those of working age and potential labour force members rose from 51% in 1966 to over 61% by 1996, a "demographic bonus" to the nation only if most are able to find productive employment. Among the most important challenges facing the government is the task of identifying development strategies which can generate new employment and income opportunities and reduce underemployment and unemployment. The urgent need to create employment opportunities is underscored by the higher rate of labour force growth than population growth, resulting from a period when fertility was much higher than today, and from recent g:\cstsuva\discuspa\bh_23.doc 2 increases in the rate of female labour force participation. Recent slow economic growth and consequent retarded growth in job opportunities, particularly in the public sector, have forced many frustrated school-leavers into marginal activities in small-scale agriculture and the urban informal economy. As noted earlier, Fiji’s population is still relatively young, with 35% of the total in 1996 less than 15 years of age. With the annual population growth rate of 0.8%, projections suggest that the total population will exceed 800 thousand by the end of this century, an increase of almost 13% over one and a half decades. The final outcome will depend on the scale of outmigration, which will be partly a function of the future performance of the economy. Labour Force Growth and the Labour Market In the meantime, the size of the labour force has been growing much faster than the overall growth rate of the population, as reflected in table 1. Over the 20 year period 1976-1996, the labour force grew at an annual average rate of 2.7% while the population rose at a rate of 1.4%. A major contributing factor has been the apparent and quite remarkable increase in the labour force participation rate of females, particularly in the most recent inter-censal period, rising from less than one-quarter to almost 40% overall. No doubt, some unknown part of this rise is attributable to a change in definition between censuses and a more complete enumeration of women’s economic contribution. It is also of interest to note the marginal decline in the age-specific labour force participation rates of males. While the fall in these rates for men aged 15-24 might be attributable to greater educational and training opportunities, some older men may have been pushed out of the labour force by younger newcomers as well as the large registered influx of women competing for similar jobs. The reported rates of unemployment in table 2 have tended to contract over time, particularly those for females since 1986. Thus, despite a rapid growth of labour supply, rates of unemployment appear to have fallen, indicating successful labour absorption into productive employment. On further examination in table 3, however, we find that much of the growth in labour supply over the period 1986-1996 failed to be absorbed into the cash economy. While total employment grew at an annual rate of 2.3% over the 1986-96 period, just exceeding the 2.1% average annual growth in labour supply, there was a marked difference in the labour market treatment of males and females. Male cash employment expanded at a rate (1.1% p.a.) which exceeded the growth in male labour supply (0.5% p.a.) with the result that the share of non-cash employment in total male employment shrank from 17% to 13%. In contrast, while female cash employment growth was impressive (4.2% p.a.), it was exceeded by the annual growth in female labour supply (6.7% p.a.) with the result that, as overall female unemployment fell from 15.3% in 1986 to 7.8% in 1996, the share of non-cash employment in total female employment expanded from 19% to 41% over the decade. g:\cstsuva\discuspa\bh_23.doc 3 Table 1 Rates of Labour Force Participation (%) by Age Group and Sex, 1976, 1986 and 1996 MALES 15-19 20-24 25-29 30-39 40-49 50-59 60+ Total Total Labour Force % Growth p.a. 1976-86 2.6 p.a. 1986-96 0.5 FEMALES 15-19 20-24 25-29 30-39 40-49 50-59 60+ Total Total Labour Force % Growth p.a. 1976-86 p.a. 1976 1986 1996 58.0 91.3 96.3 97.3 96.5 89.1 58.8 84.1 146,31 57.9 91.0 96.0 97.4 96.8 88.4 57.1 85.4 189,929 38.5 80.3 91.5 94.6 94.4 85.4 59.1 79.2 200,048 15.1 24.3 19.8 17.6 15.6 12.7 7.3 17.1 29,470 21.4 29.1 26.2 25.9 22.4 18.1 10.2 23.3 51,231 19.4 45.4 45.5 46.8 46.1 39.0 27.1 39.4 97,723 175,785 241,160 297,771 5.7 1986-96 6.7 National Labour Force (M&F) % Growth p.a. 1976-86 3.2 1986-96 2.1 Source: National Census Reports 1976 and 1986; Provisional Results of 1996 Census Table 2 Reported Rates of Unemployment by Sex, Selected Age Groups and Education Completed for Census Years Males Total Population Ages 15-19 20-24 25-29 Total 16.0 8.7 4.6 5.4 1976 Females 34.7 14.2 7.4 11.7 Total 20.5 10.1 5.2 6.6 Males 17.3 10.1 4.3 5.4 1986 Females 45.0 26.5 10.4 15.3 Total 24.7 14.1 5.6 7.5 Males 14.8 9.2 4.8 4.8 1996 Females 23.4 13.1 7.2 7.8 Total 17.6 10.5 5.6 5.8 Source: National Census Reports 1976 and 1986; Provisional Results of 1996 Census The extent to which these dramatic changes in the market for female labour are attributable to improved enumeration of female participation in non-cash economic activities in the most recent census remains unknown. At face value, however, it would appear that there has been a significant increase in the supply of female labour to the economy during a period when economic growth was disappointing. While many women have found low-wage employment in the buoyant garments sector, many more have had to be content with non-cash employment in the rural and urban subsistence and informal sectors. Their absorption in the g:\cstsuva\discuspa\bh_23.doc 4 subsistence and informal sector has contributed to the decline in their rate of unemployment. Furthermore, their presence as a potential labour supply to the cash economy has, no doubt, served to dampen wage increases in the low-wage segments of the labour market. Table 3 Employment for Cash and Non-Cash in Fiji 1986 and 1996 by Sex Cash Employment Non-Cash Employment Total Employment Males 1986 1996 % change per annum 1986-96 148,346 166,299 1.1 31,249 24,147 -3.5 179,595 190,446 0.6 Females 1986 1996 % change per annum 1986-96 35,278 53,015 4.2 8,098 37,045 16.4 43,376 90,060 7.6 Total 1986 1996 % change per annum 1986-96 183,624 219,314 1.8 39,347 61,192 4.5 222,971 280,506 2.3 Source: National Census Reports 1976 and 1986; Provisional Results of 1996 Census The Structure of Employment Very little recent analytical work has been undertaken on the operation of the labour market in Fiji1. If Fiji is to absorb a greater share of its burgeoning newcomers into productive employment, a greater understanding by planners and policy makers of how labour markets operate in the country is essential. Factors, which may serve to segment labour markets, to inhibit labour force absorption and to slow down growth in productivity and output, need to be identified. Especially important will be measures to redress the disadvantaged position of women, the poor and other vulnerable groups in the country’s labour market. At the same time a delicate balance needs to be struck between interventions which aim to protect women and other vulnerable groups and those which effectively inhibit their enhanced absorption into productive employment. Table 4 portrays the sectoral distribution of the total labour force for the most recent census year of 1996. The data confirm the concentration of economic enterprises in rural Fiji and in agricultural – related activities. Since the bulk of those classified in “subsistence" are, no doubt, engaged in agricultural-related activities, perhaps as many as 45% of the labour force are involved in the broad sector of agriculture, forestry and fishing. Even then, of the 219,314 employed for money, only 122,985 (or 41% of the total labour force) were employed for at least 5 days, perhaps constituting an estimate of the “formal sector”2. The government accounts for roughly 40% of total formal employment via central and local governments. 1 An earlier analytical study in 1984 was by the Fiji Employment and Development Mission: "Work and Income for the People of Fiji: A Strategy for More than Just Survival", which was financed by the then European Economic Community and commissioned to the Institute of Development Studies (IDS) at the University of Sussex, England. 2 This would be slightly more than the figure used in the official document of the Ministry of National Planning: Development Strategy for Fiji (1997). There, it is estimated that the formal sector, comprising “both the public and private establishments offering continuous wage and salary employment”, employs 109 thousand, or 36.2% of the labour force. g:\cstsuva\discuspa\bh_23.doc 5 Manufacturing employment is still responsible for less than 10% of the total labour force while trade and services, including government services, employ over 27% of the labour force. It is also revealing to note how over 1 in 5 of the labour force is still engaged in subsistence activities in Fiji. Men contribute two-thirds of the total labour force and tend to be concentrated in formal agriculture, construction and transport; women are over-represented in garment manufacturing, trade, services and the subsistence sector. Table 4 Sectoral Distribution of Employment 1996 Sector Agriculture, Forestry Fishing Mining and Quarrying Manufacturing Food and Drinks Clothes Wood Products Paper Chemicals Cement Metals Jewellery Other man. nec. Utilities Construction Trade Wholesale & Retail Restaurants Hotels Transport Finance Business Services Services Government Other Subsistence Unemployed Not Stated Total and Males (%) Females (%) Total 32.1 1.2 8.9 3.9 1.3 1.7 1.0 0.3 0.2 0.7 0.1 0.1 1.0 5.2 10.2 7.3 0.5 2.4 7.5 1.0 1.4 13.8 3.6 10.2 12.1 4.8 0.8 200,052(100.0) 7.4 0.2 11.5 1.8 8.6 0.3 0.3 0.2 0.0 0.1 0.1 0.0 0.1 0.3 12.1 7.6 1.3 3.1 1.8 1.7 1.0 17.3 2.9 14.4 37.9 7.8 0.9 97,718(100.0) 71,485 2,507 29,043 9,583 11,005 3,772 1,441 835 424 1,603 218 162 2,107 10,639 32,175 18,616 2,327 7,785 16,722 3,900 3,912 44,620 10,080 34,540 61,191 17,265 2,204 297,770 (%) 24.0 0.8 9.8 3.2 3.7 1.3 0.5 0.3 0.1 0.5 0.1 0.1 0.7 3.6 10.8 6.3 0.8 2.6 5.6 1.3 1.3 15.0 3.4 11.6 20.5 5.8 0.8 100.0 Source: Population Census, 1996, preliminary results The Occupational Structure of Employment The occupational distribution of those in the labour force is portrayed in table 5. Fiji’s dependence on the primary sector is again apparent with 21% of the paid labour force classified as agricultural and fisheries workers. Only 17% of the paid labour force are found in the top three occupational groups, of Legislators, Senior Officials and Managers; Professionals; and Technicians and Associates. Men dominate the senior occupations although women are well represented as Professionals in traditional 'female' occupations. g:\cstsuva\discuspa\bh_23.doc 6 Table 5 Occupational Distribution of Employment by Sex, Fiji, 1996 Occupation Males (%) Females Legislators, Senior Officials, Managers Professionals Technicians & Associates Clerks Service Workers Skilled Agric. & Fisheries Craft & Related Trades Plant & Machine Operators Elementary Occupations Unallocated occupations Subsistence Unemployed Total 6,477 10,398 8,350 7,440 14,405 40,231 23,997 16,190 36,712 2,099 24,151 9,602 200,052 3.2 5.2 4.2 3.7 7.2 20.1 12.0 8.1 18.4 1.0 12.1 4.8 100.0 1,403 7,586 3,228 9,061 7,734 5,414 3,218 6,289 9,045 37 37,040 7,663 97,718 (%) 1.4 7.8 3.3 9.3 7.9 5.5 3.3 6.4 9.3 0.1 37.9 7.8 100.0 Total 7,880 17,984 11,578 16,501 22,139 45,645 27,215 22,479 45,757 2,136 61,191 17,265 297,770 (%) 2.6 6.0 3.9 5.5 7.4 15.3 9.1 7.5 15.4 0.7 20.5 5.8 100.0 Source: Population Census, 1996, Preliminary Results 3. THE STRUCTURE OF OCCUPATIONS IN FIJI BY GENDER Theories of Sex Inequalities in the Labour Market3 Neo-classical economic theories emphasize sex differences in variables which affect labour productivity and labour supply such as family responsibilities, physical strength, education, training, hours of work, absenteeism and turnover, in order to explain why women generally earn less than men. A basic assumption of neo-classical economics is that workers in conditions of competition are paid the value of their marginal product; it then follows from this assumption that observed male-female differentials in earnings are either due to the lower productivity of women or to market imperfections. A separate assumption of the neoclassical 'new home economics' is that families allocate their resources (time and money) among family members in a rational manner which causes females to receive less human capital investments when young and to stay home to take care of the children when older. Neo-classical theory suggests that women earn less than men because they have lower levels of human capital - mainly education, training and on-the-job experience - and therefore lower labour productivity. For example, because some women interrupt their employment to marry, and then bear and rear children, employers are said to be reluctant to invest in the training of women. Also, for parents and the women themselves there is said to be less incentive to invest in education and training. Periods of withdrawal from the labour force mean that women accumulate less work experience than men and that their skills tend to depreciate more. According to the human capital approach, discrimination would be said to occur if employers pay different wages to persons with the same stock of human capital. Sex discrimination can 3 This section summarizes arguments presented in Anker and Hein (1986) and Anker (1997). g:\cstsuva\discuspa\bh_23.doc 7 then be measured by the amount of the wage gap between men and women which remains unexplained by male-female differences in human capital - mainly education, training and experience. A hypothesis which is useful in explaining why so much of the male-female wage gap cannot be explained by human capital differences between men and women is that of statistical discrimination - where average differences between population sub-groups are used as a basis for discriminating against all members in that subgroup. If the employer believes that women are on average more likely than men to be unstable, unreliable, etc. then he will tend to discriminate against women even though he knows he may be wrong in certain individual cases. The cost of obtaining detailed information on job candidates leads to the use of screening devices such as candidates' sex. In this way, negative stereotypes will penalize all women. The neo-classical approach has been very important in pointing out some of the productivityrelated differences between men and women which account, at least partially, for the fact that men earn more than women. Indeed, policies to improve the labour market position of women are often based on the need to improve women's educational levels and training - that is, their human capital. In developing countries - particularly in Africa, the Middle East and Asia - women are generally less well educated than men and this factor may, accordingly, be more important in these countries than in others. Recent cohorts of women in Fiji have gained much greater access to educational opportunities, although they remain underrepresented at the tertiary level, a factor which will continue to retard their progress to the top echelons of the occupational ladder. While it is clear that lower levels of education and skill can be an important handicap for women in the labour market, the human capital approach is based on several assumptions which have been questioned, notably by labour market segmentation and gender theorists. Two of these questionable implicit assumptions are: Women's labour force participation is of necessity intermittent because of their 'natural' child-rearing role. However, unlike pregnancy and breast-feeding, there is no biological reason why the child rearing role must be performed uniquely by the mother. Another underlying assumption is that men and women have equal access to job opportunities and compete on an equal basis in the labour market. This assumption ignores the sex segmentation of the labour market - which cannot be explained simply by sex differences in human capital. Evidence in the United States suggests that the rising educational levels of women are not related to any substantial decline in the extent of sex segregation of occupations. Institutional theories of labour market segmentation, which are discussed next, emphasize factors related to the structure of the labour market and how men and women get slotted into separate segments of the labour market in explaining sex inequalities. g:\cstsuva\discuspa\bh_23.doc 8 Theories of Labour Market Segmentation Theories of labour market segmentation can be considered as refinements of neo-classical theories, in that they view the labour market as stratified or segmented by institutional barriers. Within each segment, neo-classical principles are generally still assumed to be relevant. One of the best known theories of labour market segmentation is the dual labour market theory which distinguishes between two types of jobs: primary sector jobs which are relatively good in terms of pay, security and opportunities for advancement, and secondary sector jobs with low pay, low security and few opportunities for advancement. Primary sector jobs are those where workers' skills tend to be firm-specific and the resulting employers' need for stability in this part of the workforce leads them to offer high wages and good prospects for advancement. For jobs in the primary sector, worker stability is important for employers and the perceived high turnover of women means that they are more likely to be relegated to secondary jobs. Thus, even with equal pre-entry qualifications, men (being perceived as more stable) would be more likely than women to be recruited for primary jobs where their chances for subsequent progress within the firm in terms of wages, training and promotions would be greater. The main contribution of the dual labour market approach is that it emphasises the existence of segmented labour markets and analyses the different ways in which different labour market segments operate, thereby providing a refined alternative to the open competition between individuals assumed in neo-classical models. Two particularly important insights coming from this segmentation approach can be mentioned. First, it stresses the importance of the initial entry position into an organisation for determining future possibilities to acquire human capital (in terms of on-the-job training and experience) and for promotion. Second, it points out that worker behaviour is related to the characteristics of the jobs workers hold. Since absenteeism and turnover tend to be greater in low-level, dead-end jobs where women tend to be concentrated, reputedly higher turnover and absenteeism among women may be explained, at least partially, by sex differences in type of occupation rather than by inherent characteristics of women. While the dual labour market approach helps to explain the occupational distribution of men and women, it does not explain the sex segregation, which occurs within the primary and secondary sectors. There are both male and female occupations which involve lengthy schooling, such as engineers and nurses. There are female occupations which require firmspecific training such as executive secretaries and, at the same time, there are male occupations requiring relatively few skills and where stability is not an important factor, such as janitor or truck driver. The sex segregation of occupations within both the primary and secondary sectors has led some writers to suggest that sex also needs to be considered as one of the dimensions on which the labour market is segmented. The existence of two relatively separate labour markets for men and women is seen by some as an important determinant of the lower earnings of women. To the extent that women's occupational choices are restricted and there is an oversupply of candidates for women's jobs, women can be considered as 'crowded' into these occupations. According to such an 'overcrowding model', wages are lower for occupations which are highly feminised, since women must compete against themselves for g:\cstsuva\discuspa\bh_23.doc 9 relatively few jobs in what is essentially an artificially restricted, overcrowded segment of the labour market. Similarly, women do not compete with men for a large number of jobs considered to be 'male' jobs, which helps maintain the higher wages of these jobs. Although theories of labour market segmentation provide insights into sex inequalities, they still do not adequately explain why sex is such a persistent and important dimension for labour market segmentation. The basic cause is probably outside the economic realm and so it is not surprising that economic variables cannot explain it. Occupation Segregation and Concentration "Segregation concerns the tendency for men and women to be employed in different occupations from each other across the entire spectrum of occupations under analysis. It is a concept which is inherently symmetrical; the relationship of women workers to men workers is its key feature. In so far as women are separated from men, so are men separated from women in the labour force under consideration. Men can not be more 'segregated' than women, nor women more 'segregated' than men. Women and men are segregated in relationship to one another and, therefore, each are segregated to the same degree" (Siltanen, Jarman and Blackburn, 1993, p.4). Segregation refers to the extent to which a pattern occurs in which men and women are separated from each other in the overall occupational structure. Total segregation would exist when all occupations are staffed exclusively by one sex, with no occupation containing both men and women. A situation without segregation would be evident when the mix of men and women is the same in each occupation. “Concentration is concerned with the sex composition of the workforce in an occupation or set of occupations. Whereas segregation refers to the separation of the two sexes across occupations, concentration refers to the representation of one sex within occupations” (Siltanen, Jarman and Blackburn, 1993, p.5). One widely used measure of female concentration in labour market studies is the percentage of workers in an occupation who are women; male concentration would be reflected in the percentage of all workers in an occupation who are men4. Of course, the percentage of women in an occupation will partly depend on the share of the labour force which is female. The greater the female representation in the work force the more women there are likely to be in any single occupation. If female concentration were the same in all occupations, it would be equal to the overall female share of the total labour force. When attempting to compare levels of concentration by occupation it is useful to relate the gender composition of an occupation to the gender composition of overall employment. Therefore, one widely utilised ratio is the female percentage of a particular occupation divided by the female share of the labour force. A value greater than unity would signify over-representation of women in this occupation; a value less than unity indicates under-representation of women in the occupation. This approach to the measurement and analysis of concentration would answer some of the following types of questions: 4 Analogously, this concept of concentration can equally well apply to a group of occupations, an industry or a section of the labour force with a particular work status, e.g. self-employed workers or part-time workers. g:\cstsuva\discuspa\bh_23.doc 10 * Is a specific occupation, such as teaching, more likely to be staffed by men or women? If so, to what extent? (what percentage of teachers are female?) In which occupations are women more/less likely to be employed? In which occupations are men more/less likely to be employed? Is female employment well spread across the occupational structure, or is it restricted to a limited number of occupations? How well spread or restricted is male employment? In which occupations are women over-represented, and in which are they underrepresented? * * * * * The Situation in Fiji Attention now turns to examine the level of concentration of the male and female labour forces in Fiji, using the results of the 1986 and 1996 Censuses of Population, the most recent years available with this kind of data. The emphasis is on the production of visual displays of concentration patterns to help identify the broader features of the data reported in table 6(a), table 5 and table 6(b) and changes over time. Unfortunately, 1986 and 1996 are not directly comparable because, in the earlier year, the authorities used ISCO-68 to classify occupations; in 1996 they use ISCO-88. The sex composition of an occupation can be measured as the percentage of the total number of workers in an occupation who are female, or the percentage of the total workers in the occupation who are male, or the ratio of males to females. These are all essentially equivalent in terms of the information that they convey. Table 6 Economically Active Population by Occupation and Sex, Fiji (a) 1986 ISCO 1968 Occupation Total (1) 06/07/13 Medical Workers and Teachers Other Professionals Administrative and Managerial Clerical and Related Workers Sales workers Service Workers Agriculture etc. Workers Production and Related Workers Seeking Employment 10,343 4.3 4,366 2.2 5,977 11.8 Female Share % (7) 57.8 7,431 2,766 3.1 1.2 6,359 2,515 3.4 1.3 1,072 251 2.1 0.4 14.4 9.1 15,569 6.5 8,242 4.4 7,327 14.4 47.1 14,861 15,422 105,924 49,000 6.2 6.4 44.2 20.5 10,540 7,979 93,925 44,439 5.6 4.2 49.8 23.6 4,321 7,443 11,999 4,561 8.5 14.7 23.6 9.0 29.1 48.3 11.3 9.3 18,182 239,948 7.6 100.0 10,331 188,696 5.4 100.0 7,851 50,802 15.5 100.0 43.2 21.2 0-1 2 3 4 5 6 7-9 998 TOTAL % (2) Males (3) % (4) Females (5) % (6) Source: Fiji (1988), Census of Population, 1986, Bureau of Statistics, Suva g:\cstsuva\discuspa\bh_23.doc 11 (b) 1996 1988 1 2 3 4 5 6 7 8 9 0 Occupation Legislators, Senior Officials and Managers Professionals Technicians and Associate Professionals Clerks Service Workers and Shop and Market Sales Workers Skilled Agricultural and Fishery Workers Craft and Related Trades Workers Plant and Machine Operators and Assemblers Elementary Occupations Unallocated Subsistence Unemployed TOTAL Source: Census of Population, 1996, Preliminary Results Female Share % (7) 17.8 42.2 27.9 54.9 34.9 11.9 11.8 28.0 19.8 1.7 60.5 44.4 32.8 Figure 1: Sex Composition of Occupational Groups, Ordered by % Female g:\cstsuva\discuspa\bh_23.doc 12 Table 6 and Figure 1 display an overview of the distribution of the sexes in the labour force in 1986 and 1996. Although data for these years are not strictly comparable because of the changed occupational classifications, the structuring by gender is prominent in both years. Male-dominated occupations are on the left side of the figures and female-dominated occupations on the right. Interestingly, in both years, the female-dominated occupations all have larger representation of men in them, as compared with the share of women in maledominated occupations. Perhaps various cultural, social and political processes of employment underlie this pattern. Yet, since there are more men in the Fiji labour force, it is statistically more probable that men will be more widely distributed throughout the occupational structure. Or, since female-dominated occupations are broader than maledominated ones, they may contain pockets of 'male' jobs. Another important feature of Figure 1 is that, of the two largest occupations in 1986 Agricultural workers (with 106 thousand workers, or 44% of the total) and Production and Related Workers (with 49 thousand workers, or 21% of the 1986 total) - both are at the maledominated end of the spectrum. It is hardly surprising that women made up only one-in-five of the labour force when they are largely excluded (constituting only 11%) of the two largest occupations with 65% of the Fiji work force. During the intervening decade between 1986 and 1996, however, the female labour force participation rate rose significantly (from 23% to 39%) while the male rate declined (from 85% to 79%). The female share of the total labour force rose correspondingly, from 21% to 33%, but a large proportion of these women failed to find cash employment and large numbers entered the non-cash, subsistence (largely agriculture) economy. The result is reflected in the bottom part of figure 1 where Agriculture, including subsistence, is placed towards the 'female-dominated' end of the scale of occupations as opposed to its 'male-dominated' position in 1986. By 1996, therefore, women's share of the two largest occupations (Agricultural workers with 107 thousand workers or 36% of the total and Elementary Occupations with 46 thousand workers or 15% of the total) had become 33%, compared with 11% in 1986. Thus, the gender concentration of the labour force would appear to have become less extreme, yet women's greater representation in these lowly paid occupations would have done little to raise their overall status in Fiji's labour market. Alternative Presentations: Distribution of the Male and Female Labour Force An alternative way of examining male and female concentration is to present the distribution of employment by sex across occupations to identify its variability. This is calculated as the percentage of the male and female labour forces in each occupation5 Figure 2 illustrates how women are grossly under represented in the highly-paid, decision-making occupations Administrators and Managers, and Other Professionals, with only 2% or less of all women in these categories in 1986. With different occupational classifications, 1% and 7% of women were found as Legislators, Senior officials and Managers, and as Professionals, respectively in 1996. Women are a little better represented as Production workers, Sales workers, Clerical workers and Service workers, which contain between 9% and 15% of all female workers in 1986 and between 8%-9% in 1996. Twelve percent of all women work in the traditional female occupations as paramedics and teachers in 1986 and 7% in 1996, while figure 13 5 For example, (Number of women in occupation i/Number of women in the total labour force) *100. g:\cstsuva\discuspa\bh_23.doc 13 shows that almost 1 in 4 female workers were engaged in agriculture and fishing in 1986 but this had risen to almost 1 in 2 by 1996. Such an illustration is extremely useful for identifying those occupations where women are found, and are not found, and for estimating their relative sizes. Figure 2: Distribution of Males and Females Across Occupations 32% 1996 Females No. of Workers (thousands) Males 43% 18 % 12 % 8% 7% 5% 3% 7% 9% 4% 9% 8% 6% P ro fe s s io n a ls 8% 5% 3% 3% 1% Le g is la to rs , S e n io r O ffic ia ls , Ma n a g e rs 4% Te c h n ic ia n s & As s o c ia te s Cle rks S e rvic e Wo rke rs Ag ric u ltu ra l Wo rke rs Cra ft Wo rke rs P la n t & Ma c h in e O p s . E le me ta ry O c c u p a tio n s Un e mp lo ye d Sourc e: Censuses of Population Fiji, 1986 and 1996 An alternative method of presenting the material is portrayed in Figure 3 where all the occupations are grouped by their percentage female and then plotted against the total number of workers to be found in each occupation.6 On the horizontal axis, occupations are plotted according to their percentage female. In 1986, the minimum is 9.1% (Administrative and 6This technique has the advantage of allowing the researcher to employ a larger number of more detailed occupations (say ISCO 2-or 3digits). The data reported in Figure 3 maintains the same broad categories used earlier. g:\cstsuva\discuspa\bh_23.doc 14 Managerial) and the maximum is 57.8% (Medical workers and Teachers). Thus, we see that, in 1986, for occupations that are between 0 and 10% female, there are over 50 thousand workers; for occupations with between 41% and 50% female there are almost 50 thousand workers. Thus, most employment is found in occupations where the female share is less than 10% or exceeds 40%. In 1996, with different occupational classifications, the minimum is 11.8% (Craft workers) and the maximum is 54.9% (Clerks). We still find the great bulk of total employment in occupations where women’s share is relatively small or relatively large. Figure 3: Distribution of Workers in Occupations According to % Female 1986 Thousands 120 100 80 60 40 20 0 0-10 11-20 21-30 31-40 41-50 51-60 Percentage Points Female Source: Censuses of Population – Fiji, 1986 & 1996 1996 140 120 100 80 60 40 20 0 0-10 11-20 21-30 31-40 41-50 51-60 Percentage Points Female Source: Censuses of Population - Fiji, 1986 & 1996 Analysis of the Extremes The extreme ends of the distribution of these data will now be examined for additional insights on gender concentration. What proportion of women work in female dominated occupations? What proportion of women are engaged in occupations dominated by men? Clearly, the level of detail in the occupational scheme will be a major influence on the results. g:\cstsuva\discuspa\bh_23.doc 15 In this comparative analysis of the situation in Fiji in 1986 and 1996, 2-digit ISCO-1968 occupations have been used for the earlier year and 3-digit ISCO-1988 occupations for the most recent year. This raises the likelihood of occupations being dominated by one sex or the other in contrast to when the broader groups of occupations are used above. Table 7 shows that, in 1986, there were 46 2-digit occupations, out of a total of 82, where women constituted only 15% or less of the workforce in those occupations. Almost 27% of all female workers were found in these overwhelmingly male-dominated jobs. On the other hand, there were only 8 occupations which were dominated with women, with more than 70% of employment being women. The patterns of concentration are fairly predictable, with the large number of male-dominated jobs being either high-skilled, or managerial/supervisory, or requiring artisan skills or a capacity to undertake heavy labouring work. The female-dominated jobs were traditional 'women's' occupations, making up 24% of all their employment opportunities. On this evidence, in 1986 one-half of all employed women in Fiji worked in an occupation heavily dominated by one or other of the sexes. By 1996, using a different set of occupation codes (ISCO-88), there were 45 3-digit occupations out of a total of 109 occupations where women made up 15% or less of the workforce (see Table 7). About 8% of all female workers were located in these maledominated jobs. At the other extreme there were only 9 female-dominated occupations with 60% or more of employment being women7. They held 55% of all female employment. Thus, in 1996, almost two-thirds of all employed women were engaged in an occupation heavily dominated by one or other of the sexes. The patterns of concentration are similar to those found in 1986. While it is difficult to make comparisons between 1986 and 1996 because of the changes in occupation codes, it seems reasonable to conclude that women made little headway in diversifying into male preserves, despite their significant rise in labour force participation. Figure 4 reflects the gender dominance of the distribution of occupations in both 1986 and 1996. In the earlier year, 80% of all workers were engaged in either male- or femaledominated occupations. Ten years later the relative situation had changed slightly whereby 72% of all workers were engaged in an occupation dominated by one sex or the other. Significantly, the absolute number of workers in the 'mixed' category had almost doubled, from 44 thousand to 83 thousand; the absolute number in strongly male-dominated occupations had fallen from 163 thousand to 123 thousand; and the absolute number in strongly female-dominated jobs had grown from 14 thousand to 92 thousand. 7 69% of total employment in these 9 occupations was found in one occupation, i.e. subsistence activities. g:\cstsuva\discuspa\bh_23.doc 16 Table 7 Occupations at the Extremes of the Distribution 1986-1996 1986 0-15% Female ISCO-68 01 Physical Scientists 02/03 Architects, Engineers 04 Aircraft, Ships Officers 05 Life Scientists 11 Accountants 12 Jurists 16 Creative Artists 17 Composers, Performers 18 Athletes, Sportsmen 20 Legislative Officials 21 Managers 31 Gov. Executive Officials 35 Transport/Comm. Superv. 36 Transport Conductors 37 Mail Distrib. Clerks 40 Managers in Trade 41 Working Propr. Trade 43 Technical Salesmen 44 Insurance Salesmen 49 Other Sales Workers 58 Protective Services 60 Farm Managers 61 Farmers 62 Farm Workers 63 Forestry Workers 70 Production Superv. 71 Miners Quarrymen 72 Metal Processors 73 Wood Prep. Paper Makers 80 Shoemakers, Leatherwork 81 Cabinet Makers, Woodwork 82 Stone Cutters 83 Blacksmiths, Toolmakers 84 Machine Fitters 85 Electrical Fitters 86 Sound Equip. Op. 87 Plumbers, Welders 90 Rubber, Plastic Products 92 Printers 93 Painters, Construction 95 Bricklayers, Carpenters 96 Stationary Engine Ops. 97 Material Handling 98 Transport Equip. Oper. 990 Other Labourers 991 Armed Forces TOTAL g:\cstsuva\discuspa\bh_23.doc 70% + Female ( 34 ) ( 70 ) ( 5 ) ( 68 ) (178) ( 11 ) ( 19 ) ( 54 ) ( 5 ) ( 40 ) (211) (129) ( 34 ) ( 10 ) ( 45 ) ( 87 ) (241) ( 25 ) ( 43 ) ( 18 ) (117) ( 22 ) (8657) (2077) ( 16 ) (109) ( 17 ) ( 2 ) ( 7 ) ( 13 ) ( 31 ) ( 0 ) ( 50 ) ( 56 ) ( 35) ( 9 ) ( 22 ) ( 17 ) ( 82 ) ( 5 ) ( 55 ) ( 12 ) (469) ( 68 ) (313) ( 11 ) (13599) 06/07 Medical, Dental, Veter 32 Stenographers 34 Computing Mch. Oper. 54 Maids 75 Spinners, Weavers 79 Tailors, Sewers 91 Paper Product Makers 94 Prod. Workers n.e.c. TOTAL (1838) (2991) ( 249 ) (4637) ( 321 ) (1845) ( 208 ) ( 229 ) (12318) 17 1996 0-15% Female ISCO-88 111 Legislators & Senior Officials 113 Traditional Chiefs & Heads 211 Physicists, Chemists etc. 214 Architects 221 Life Science Professionals 246 Religious Professionals 311 Physical etc. Eng. Tech 314 Ship/Aircraft Controllers 321 Life Science Technicians 345 Police Inspectors 348 Religious Assoc. Profs. 516 Protective Service Workers 611 Market Gardeners 612 Animal Producers 614 Forestry Workers 711 Miners etc. 712 Building Trade Workers 713 Building Finishers 714 Painters etc 721 Metal Workers 722 Blacksmiths etc 723 Machinery Mechanics 724 Electrical Mechanics 732 Potters 811 Mining Plant Ops 812 Metal-Processing Ops 814 Wood Processing Ops 815 Chem-Plant Ops 816 Power Production Ops 817 Auto-Ass. Ops 822 Chem-Products Mach. Op. 824 Wood-Product Op. 825 Printing Op. 831 Logo-Engine Ops 832 M.V. Ops 833 Agric. Ops 834 Ships Deck Crews 912 Shoe Cleaning 914 Building Caretakers 915 Messengers etc. 916 Garbage Collectors 921 Agric. Labourers 931 Mining/Const. Labourers 933 Transport Labourers Unallocated TOTAL 60% + Female (14) (1) (31) (71) (10) (117) (219) (46) (103) (7) (33) (337) (3,668) (37) (8) (27) (60) (53) (23) (39) (3) (60) (24) (36) (10) (6) (8) (7) (5) (5) (7) (14) (20) (7) (91) (38) (36) (19) (57) (155) (5) (1,692) (122) (28) (37) (7,388) 223 243 323 411 422 513 621 826 913 Nursing Profs. Archivists etc Nursing Assoc. Profs Secretaries Client Info. Clerks Personal Care Workers Subsistence Workers Textile Mach. Ops Domestic Workers (1,048) (169) (532) (2,844) (1,273) (287) (37,395) 5,598) (4,967) (54,113) Note: Figures in parentheses are numbers of women in these occupations Source: Fiji, 1988, Census of Population 1986 - and Fiji, 1988, Census of Population 1996, Bureau of Statistics, Suva g:\cstsuva\discuspa\bh_23.doc 18 Figure 4: Distribution of Gender Dominance of Occupations Thousands of Workers 1986 200 150 100 50 0 Strongly Male Dominated Mixed Strongly Female Dominated 1996 200 150 100 50 0 Strongly Male Dominated Mixed Strongly Female Dominated Source: Censuses of Population - Fiji 1986 & 1996 In the upper part of Figure 5, zero point on the vertical axis is where the percentage female of the occupation is 21%, equal to the percentage female in the total labour force in 1986. The most female-dominated occupations are plotted above the horizontal axis, those with the most male dominance lie below the horizontal axis. In Figure 5, the occupations are presented from the most male-dominated (Administrative and Managerial Workers) to the most femaledominated (Medics and Teachers) in 1986. In 1996, using 2-digit ISCO-88 occupations, the most male-dominated are construction jobs, drivers, machine operators and science professionals and managers. In contract but predictably, the most female-dominated jobs are in teaching, nursing and clerical jobs. The results of the analysis of the distribution of gender by occupations in Fiji confirms that it is extremely concentrated and has remained so through time. How has this concentration impacted on the relative earnings of women compared with men? It is to this question that we now turn. g:\cstsuva\discuspa\bh_23.doc 19 Figure 5: Difference Between % Female in Occupation & In Total 1986 40 30 20 10 0 Admin. & P roduc t ion Agric ult ura l Ma na ge ria l Worke rs Worke rs Ot he r P rofs. S a le s S e e king Cle ric a l S e rvic e Me dic s/ Te a c he rs Employme nt -10 -20 50 1996 40 30 20 10 0 -10 P hysic a l a nd Drive rs & Eng. S c i. P la nt . Op. Me t a l Tra de S t a t iona ry Worke rs P la nt Op. Agric . La bs. S c ie nc e Ge ne ra l Offic e Te a c hing S a le s & S ubsist e nc e Cust ome r Life S c ie nc e Ma c hine P rofs. Ma na ge rs Cle rks P rof. S e rvic e Fa rming S e rvic e & He a lt h Ops. Cle rks P rofs. Te c h Ele m. Oc c . -20 -30 -40 Source: Censuses of Population: Fiji, 1986 and 1996 4. THE STRUCTURE OF EARNINGS BY SEX The Source of Information: The Fiji Employment Survey One of the principal aims of this paper is to examine the relative position of female workers in Fiji’s formal sector labour market. After portraying the gender distribution of occupations and changes in recent years, it now turns to estimating the extent of “discrimination” against women in terms of their job assignments and relative pay. g:\cstsuva\discuspa\bh_23.doc 20 Most of the data in this particular analysis derive from a specially designed survey initiated by the Ministry of National Planning and Information of the Government of Fiji and financed under a UNFPA-funded project: “Assistance in Population, Workforce and Human Resources Development Planning” (FIJ/95/P02) and executed by the ILO. The survey data collection exercise was sub-contracted to a researcher from the School of Social and Economic Development of the University of the South Pacific (USP) while the UNFPA Country Support Team, Suva, provided overall guidance and technical advice. The survey was conducted in April 1997. The overall rationale for the establishment-based survey was to collect scarce information on Fiji’s formal sector labour market, including its gender dimensions, which could be analyzed and used in the preparation of a Strategic Human Resources Development Plan for the years through 2001 and beyond. Information collected included the establishment name, its location, and the public-private-parastatal nature of its operations. From each individual employee respondent, the information collected included: ethnicity, marital status, age, sex, highest level of formal education attained, type of vocational training undertaken, if any, its duration, hours of work in the last week and gross wages/salary, value of fringe benefits in past 12 months, occupation, years in the occupation, years of tenure with current employer and previous number of employers. The Survey Methodology According to the list of establishments of the Fiji National Provident Fund (FNPF), there were approximately 4221 firms and establishments paying FNPF contributions for their employees in 1997. The total number of employee contributors amounted to 110,091. While the payment of contributions by employers for those employees who work for more than two weeks is a legal requirement, it is believed that there are some firms in Fiji who are not included in the list because they are not registered with the FNPF. Aside from this unavoidable omission, this FNPF source is the most comprehensive list of firms and establishments in Fiji and, for this reason, this listing was used as the frame for selecting the sample of firms and their employees for this exercise. A total of 320 establishments (7.6% of the total) were selected for enumeration using a stratified random sampling technique. It was necessary that the sample of establishments was selected from a reasonable representation of the main economic activities and, in addition, that small, medium and large establishments were represented in the sample. Because of the concentration of formal economic activities on the two main islands of Viti Levu and Vanua Levu, and to minimise the costs of conducting the survey, only establishments located in the main urban areas were selected. Each of the selected establishments, including Government and semi-government or parastatal organisations, was visited by a trained enumerator from the USP. Contact with a senior administration officer had taken place in advance of the enumerator’s visit and the purpose of the exercise explained. From those establishments employing less than 200 employees, information was collected from each individual worker. For those with 200 or more employees, the enumerator made a stratified random selection of 200 names from a complete listing of employees provided by the firm. g:\cstsuva\discuspa\bh_23.doc 21 As a result, a total of 8153 individual employees were interviewed, representing a sample of just over 7% of the formal sector workforce. Since we do not have a complete profile of the population of formal sector workers, it is difficult to gauge how well the characteristics of our sample of employees fit those of workers in the formal sector. However, the age distributions from the 1996 Population Census are reasonably close to those of our survey, particularly that representing cash workers in the urban economy, a concept coming closest to that of the formal sector as represented in our survey. According to this criterion, that is, the age distribution of workers in our sample, we can be reasonably satisfied that our sample of workers is very similar to the age distribution of formal sector workers in Fiji. In addition, the public sector, including public corporations or parastatals, is believed to account for roughly 40% of total formal sector employment; our survey contains a total of 38.6% of public and parastatal employees. Male-Female Pay Differentials The following section examines the structure of earnings from our employment survey according to various personal human capital and other institutional characteristics. Table 8 reports the mean level of weekly earnings, including overtime and annual fringe benefits converted to a weekly basis, by age group, sector and sex. It demonstrates that mean pay is consistently higher for men compared with women and is higher in the public and parastatal sectors compared with the private sector. Table 8 Mean Weekly Earnings by Age Group, Sector and Sex (F$) Age Group 14-19 20-24 25-34 35-44 45-54 55+ Total No. of Obs. Public Sector Males Females 112+ 140 174 206 247 223 199 1147 111+ 135 156 183 225 140+ 166 638 Parastatal Sector Males Females 141+ 169 224 245 221 214 223 1009 86+ 149 208 214 251 102+ 199 247 Private Sector Males Females 63 99 142 163 192 190 136 3279 58 87 117 121 137 124+ 107 1688 Males Total Females 67 110 161 194 219 206 165 5435 61 104 136 146 185 131 130 2573 Note: + less than 20 cases Source: Fiji Employment Survey, 1997 Earnings are higher for younger, new recruits in the public and parastatal sectors and rise fairly consistently in all sectors with age, proxying for labour market experience, up to age 55, after which earnings tend to decline. The earnings profiles are much steeper in the public and parastatal sectors, perhaps reflecting the structured internal labour market in these sectors and the formalised procedures allowing for job promotion along with lengthening job tenure. It is also revealing to see the much flatter profile of female earnings as age increases, particularly in the private sector, perhaps reflecting the kinds of occupations to which they are assigned and the lack of opportunity for much on-the-job skill acquisition and productivity and pay advance. It may also reflect, to some extent, the effect of certain discontinuities in women’s attachment to the labour market and the loss of productivity and pay-enhancing continuous work experience. g:\cstsuva\discuspa\bh_23.doc 22 Table 9 Mean Weekly Earnings by Educational Attainments, Sector, and Sex (F$) Public Sector Total Males Males Education None Classes 1-6 Forms 1-2 Forms 3-5 Complete Sec. (F67) Certificate/Diploma University Degree Total No. of Obs. 153+ 118 135 173 190 284 367 199 1147 Parastatal Private Sector Sector Females Males Females Males Females 126+ 121+ 120+ 152 158 212+ 290 166 638 133+ 147 152 213 248 355 456 223 1009 129+ 131+ 198 187 255+ 341+ 199 247 83 119 115 113 146 356 427 136 3279 48 59 72 86 135 223+ 255 107 1688 Females 106 129 127 137 175 316 408 165 5435 66 66 78 105 148 222 292 130 2573 Note: + less than 20 cases Source: Fiji Employment Survey, 1997 Mean weekly earnings in table 9 rise consistently with school attainments, but especially after completing at least 8 years of education. The public and parastatal sectors pay a significant premium to workers with schooling up to the completed secondary level, for both men and women, and thus appear as wage leaders in the market for the lower skilled. The private sector is much more competitive in its earnings scale in the market for the most educated labour and exceeds the mean pay of the public sector for university level males and both the public and parastatal sectors for males with a post-secondary certificate or diploma. Male-female earnings differentials are significant in all three sectors in table 9 and generally take a U-shape, being wider at lower and higher levels of education compared with completed secondary education. This is also the case in the private sector where the U-shape pattern is exaggerated. No doubt, one possible explanation would be that there are differences in the occupational assignment of males and females at lower and higher levels of education. For example, very few females with higher education attain the very senior executive positions with correspondingly high pay. These positions remain almost exclusively male preserves. The same is also true in the public and parastatal sectors except that the male-female pay differences are not nearly so marked as in the private sector and the pattern is one of low pay differences in line with the lower the level of education. It would appear that the public and parastatal sectors are much less discriminating in their treatment of the sexes. Even then, male university graduates earn about one-third more than female graduates, so complacency on the part of public sector officials is not warranted. These patterns are more clearly demonstrated in figures 6 to 7. Figure 6 demonstrates how the earnings profile of more educated males has a much steeper slope than for the less educated whose earnings profile is quite flat as age progresses; the pattern is broadly similar for females in figure 7. The less educated seem to be assigned to occupations where opportunities for productivity–enhancing, on-the-job skill acquisition are severely limited. g:\cstsuva\discuspa\bh_23.doc 23 Figure 6: Mean Weekly Earnings by Education - Male Workers 800 600 400 None Years 1-6 200 Forms 1-2 Forms 3-5 0 20-24 14-19 25-34 35-44 45-54 Forms 6-7 55+ Cert. Age Group University Source: Fiji Employment Survey 1997 Figure 7: Mean Weekly Earnings by Education Female Workers 400 300 200 100 0 None Years 1-6 Forms 1-2 Forms 3-5 Forms 6-7 14-19 20-24 25-34 35-44 45-54 Certificate 55+ University Age Group Source: Fiji Employment Survey 1997 Table 10 Mean Weekly Earnings by Occupation, Sector and Sex (F$) Public Sector Age Group Males Females 318 234 184 139 166 139 131 199 1147 247 195 145 190+ 133 161+ 110 166 638 Senior Professionals Middle Professionals Clerical Sales Service Artisans Garment Workers Other Blue Collar Total No. of Obs. Parastatal Sector Males Females 467 293 208 213+ 149 238 140 223 1009 343+ 304+ 185 201+ 182+ 243+ 121+ 199 247 Private Sector Males 380 236 164 112 107 109 75 100 136 3279 Females 300 217 160 87 96 82 55 93 107 1688 Total Males Females 374 251 179 118 133 130 75 117 165 5435 283 205 161 97 111 86 55 98 130 2573 Note: + Less than 20 cases Source: Fiji Employment Survey, 1997 g:\cstsuva\discuspa\bh_23.doc 24 Table 10 demonstrates that occupational pay differences are large in Fiji, with earnings being five times greater for the highest paying occupation group (Senior Professionals) over the lowest (Garment Workers). Senior professionals in the parastatal and private sectors earn more than their peers in the public sector; lower skilled workers in the public and parastatal sectors have a pay advantage over those in the private sector. For example, male blue-collar public sector workers receive a 31% premium over their counterparts in the private sector; female public sector workers in this broad occupational category have an 18% advantage over their private sector equivalents. Male-female pay differentials are significant after controlling for these broad occupational groupings, particularly at the higher skill level. Rather than suggest that male and female employees receive different rewards from performing the same job side-by-side, it is much more likely that more specific gender-based occupational assignments explain much of these pay differences. These issues will be investigated in more detail using multivariate techniques in later papers. Table 11 reports the mean level of weekly earnings by industrial sector and sex. Earnings are highest in the Finance and Insurance and Utilities sectors. The lowest mean pay arises in the sectors of Agriculture, Manufacturing – especially garment making – and Trade and Hotels, all traditionally employers of relatively low-skilled, cheap labour. Predictably, in most of the sectors, average female earnings are exceeded by male earnings, with the exception of Construction and Transport and Communications, where perhaps, women are performing office work and men are engaged in more manual tasks. However, only in a few cases is the overall significant earnings difference between the sexes replicated at the sectoral level. Table 11 Mean Weekly Earnings by Sector of Activity and Sex (F$) Sector Agriculture, For., Fishing Food Manufacturing Garments and footwear Other Manufacturing Electricity/Water Construction Trade & Hotels Transport & Communications Finance & Insurance Services Total No. of Obs. Males Females 124 178 80 110 253 145 132 162 252 202 165 5435 144+ 124 60 79 210 155 105 173 238 170 130 2573 Total 124 170 65 107 245 147 124 164 246 192 154 8008 Male/ Female 0.9 1.4* 1.3* 1.4 1.2 0.9 1.3* 0.9 1.1 1.2* 1.3* Note: + less than 20 cases; * significant at the 5% level Source: Fiji Employment Survey, 1997 g:\cstsuva\discuspa\bh_23.doc 25 5. A DECOMPOSITION OF MALE-FEMALE EARNING DIFFERENTIALS Women’s earnings are inferior to men’s throughout the world where average female-male pay ratios are roughly 70-75%, based on daily and weekly reference periods, and 75-80% based on an hourly reference period (Anker, 1997). Ratios are especially low in east and south-east Asian and some OECD countries where, for all non-agricultural earnings, the ratio for hourly pay is as low as 68% in Luxembourg and Switzerland and as high as 88% in Australia and 91% in Sri Lanka. On a daily or weekly basis the ratio is low in Hong Kong (70%) and Cyprus (59%) and higher in Sri Lanka (90%) and Turkey (85%). Unweighted world averages are 77.8% on an hourly basis, 76.7% on a daily/weekly basis and 71.6% on a monthly basis (Anker, 1997). In Fiji, for the whole of the formal sector, the ratio is 78.9% on a weekly basis and 82.2% on an hourly basis. This section of the paper presents a description of the methodology to be employed to explain these gender differences in pay in the Fiji formal sector labour market, followed by a report of the in-depth analysis undertaken. The Methodology (a) The Standard Method The standard decomposition of male/female earnings differentials has been widely utilised in many studies around the world. The method requires separating the male/female earnings differential into two parts: the portion attributable to the ‘composition’ or ‘endowment’ effect which accounts for the unequal distribution of human capital throughout the population, and the portion reflecting ‘discrimination’ – indicating that women obtain a lower wage despite comparable qualifications and potential productivity. More formally, men and women are assumed to have separate earnings functions described by: Wim = Xijmßjm + Єim i = 1,. . .,Nm; j = 1,. . ., Nj (1) Wif = Xijfßjf + Єif i = 1, . . .,Nf; j = 1,. . .,Nj (2) Where Wim [Wif] denotes the earnings of the i-th male [or female] worker, Xim[Xif] is a 1xNj row vector of j individual characteristics that impact the wage determination process, ßjm [ßjf] is a set of regression coefficients, and Єim and Єif are well-behaved error terms with zero means and constant variances. Recalling a property of ordinary least squares estimation, that the fitted regression must pass through the vector of means, the difference between male and female earnings can be expressed as: m - f = j (ßjm – ßjf) jm + Σj ßjf (jm - jf) (3) where a bar over a variable denotes its mean value. This decomposition/standardisation of earnings differentials is attributable to Blinder [1973] and Oaxaca [1973]. In equation (3) the difference between male and female earnings has been divided into two components: (1) the difference in wages which results from differences in male and female pay structures, {Σj [ßjm-ßjf] jm} and (2) the difference which results from a compositional effect due to differences in endowments, {Σj ßjf [jm-jf]}. The greater the first term, the higher the degree of labour market discrimination.8 While this type of decomposition has been applied 8 Note that the decomposition can also proceed as: g:\cstsuva\discuspa\bh_23.doc 26 successfully in many previous studies, it has primarily been employed in the developed country literature.9 (b) The Expanded method The above strategy is subject to the criticism that it does not distinguish adequately between wage discrimination and occupational segregation; and we have observed earlier the importance of the latter phenomenon in Fiji. Wage discrimination arises when workers with exactly the same endowments are paid differently. Occupational segregation or job discrimination arises when workers with the same endowments do not have equal access to all occupations. We would suspect that differences in the occupational distributions of men and women can play a major role in explaining differences in earnings between the sexes. Brown et al. [1980] have proposed an alternative decomposition which can be thought of as a straightforward extension of the basic methodology outlined above. The expected mean wage (for either males or females, [m; f] can be thought of as the sum of each occupational wage (hm; hf; h = 1, . . ., Nh) times the sex-specific probability of attaining that occupation (Phm; Phf; h = 1, . . ., Nh). More formally: m = h Phm hm (h = 1, . . ., Nh) f = h Phf hf (h = 1, . . ., Nh) h Phm = h Phf = 1 (4) The probability of the I-th worker of either sex having occupation h, is a function of his/her characteristics (Zikm; Zikf), written as: Pihm = exp(Zihkm hkm)/t=1,.,h,.,Nh exp(Zitkm tkm) = Hhm(Zikm) Pihf = exp(Zihkf hkf)/t=1,.,h,.,Nh exp(Zitkf) tkf) = Hhf(Zikf) (5) where Z is the vector of individual characteristics, and is the set of (Njx(Nh-1) parameters to be estimated. The proportion of men in occupation h is written: Phm = Hhm (hm) (6) The proportion of women in occupation h, if their occupational attainment is determined the same way as men, would be: Phf = Hhm (hf) (7) By assuming that the errors in the occupational choice process are independent of the errors in the wage equations within occupations, and using logic identical to that above, it is possible to decompose the differences in mean earnings in the following manner: m - f = h Phm hm - h Phm hm - h Phf hm + h Phf hm - h Phf hf m - f = j (ßjm – ßjf) jm + Σj ßjf (jm - jf) (3) :reflecting the familiar ‘index-number’ problem whenever heterogeneous goods (X’s) can be aggregated using two different sets of prices (ß’s) [Cain, 1986]. 9 An exception is the application for Khartoum, Sudan by Cohen and House (1994). g:\cstsuva\discuspa\bh_23.doc 27 = (h (Phm – Phf) hm + h Phf (hm - hf) (8) The first term in equation (8) captures differences in earnings attributable to differences in the distribution of jobs by sex, while the second captures differences in earnings attributable to each occupation having a sex-specific wage. Both of these terms can be further decomposed. Finally, we have: m - f = h(Phm - Phf) hm + h Phf (hm - hf) = h hm (Hhm(hf) – Phf) + h hm (Phm - Hhm(hf)) + h Phf ((ßhm – ßhf)hm) + h Phf (ßhf (hm - hf) (9) All expressions are evaluated at their means. The first term represents occupational segregation. It captures how differences in personal characteristics affect the probability of attaining a given occupation. The second term represents differences in the occupation distribution of men and women due to differences in their personal characteristics. Although, if the entry criteria for certain jobs includes prior on-the-job or other firm specific training, differences in personal characteristics can reflect unequal access to training opportunities. The third term provides an alternative measure of wage discrimination than the one derived from the standard method presented earlier. It isolates the effect of sex differences in wage structures controlling for levels of endowment and occupation. The fourth term is the proportion of differences in wages that can be explained by differences in worker characteristics. Calculation of the extended model requires estimating a model of occupational attainment as well as separate earnings equations for each sex-occupation cell. Only the results of estimating the standard method are reported here; the empirical results of the expanded method will be reported in a subsequent paper. The Results (a) The Standard Method Earnings functions were estimated separately for males and females using a variant of the model in equation (10) is derived from equations (1) and (2) above. The individuals and sex subscripts have been dropped: Ln(HRPAY) = o + u * EDu + 7 *EXP + 8EXP² + 9FIRMEXP + 10FIRMEXP² + v*ETHNICv + 14PUBLIC + 15PARSTAT + 16TRAIN + o*OCCUPo + s*SECTORs (10) Equation (10) contains a set of measures of formal school attainments (EDu; U = 1 …. 6) years of potential labour market experience and its square (EXP, EXP²), years of firm specific experience and its square (FIRMEXP, FIRMEXP²), ethnic group (ETHNICv; v = 11, 12, 13), variables denoting whether the respondent works in the public sector (PUBLIC) or for a public corporation (PARSTAT), an indicator variable for whether the employee has ever received formal vocational training (TRAIN), and additional control variables relating to the respondent’s occupation (OCCUPo; o = 17 …23). The dependent variable, Ln(HRPAY) is the natural logarithm of the hourly wage rate, which is believed to be a superior measure to annual, monthly or weekly earnings, since it does not rely on a restricted and fixed quantity of labour supply. g:\cstsuva\discuspa\bh_23.doc 28 The detailed equations, including variable definitions and mean values, are reported in the Appendix. The model is very successful, explaining over 60% of the variance of the natural logarithm of hourly earnings for males and over 70% for females. A Chow-test on the model was performed to test for parameter or structural differences in the male and female equations. The null hypothesis, that the slopes are equal, was rejected at the one percent level. In other words, the labour market returns to the various human capital and other attributes of male and female workers are significantly different. Taken together, the dummy variables denoting educational attainments are highly significant although positive returns do not appear until at least 9 years of education are attained. Women perform better than men in achieving higher returns to senior secondary over junior secondary education and to certificate/diploma-level education over senior secondary schooling. At the highest level of university education, however, the returns relative to the next lower level are slightly higher for men, perhaps reflecting the lower occupational attainments, especially lower professional and managerial positions, of more educated women. Men receive a higher return to general labour market experience than families; there is a much higher rate of return to firm or current employer experience with a negligible advantage to women.10 Using the ‘standard’ method described above, the earnings differential is decomposed in Table 12 into (1) a portion attributable to a composition effect (the column labeled ‘Endowments’), and (2) a portion attributable to structural differences in the two earnings functions, (the column labeled ‘Discrimination’). Referring to the summary at the bottom of the table, approximately 40% of the difference in earnings can be attributed to a compositional effect, while 60% can be attributed to differences in the structural parameters. It should be mentioned that no attempt has been made to account for workers’ innate ability, degree of motivation or commitment to the labour force, quality of education, or union effects. As the first part of the table reveals, the advantage of male endowments in the total labour market comes mainly in the form of greater labour market and firm specific experience being employed in the public sector and being engaged in the high-wage industrial sectors. As shown above, men do not have an advantage in formal education; this is confirmed by a negative coefficient on education in the endowment and discrimination columns. With respect to the cause of discrimination, the principal source lies in the size of the differential in the intercept term in favour of men, which accounts for 74% of the total hourly earnings differential between the sexes. The returns to potential experience and vocational training also favour men. Our measure of general experience is only potential experience (Age minus Years of Education-6) and does not control for protracted absences from the labour market which may result in still obsolescence, obsences which are more common among women because of their maternal and domestic duties. 10 g:\cstsuva\discuspa\bh_23.doc 29 Table 12: Decomposition of Earnings Differentials between male and Female Formal Sector Employees in Fiji Explanatory Variable Education Potential Experience Firm Experience Ethnic Group Public Sector Employee Training Received Occupation Sector Total Summary of Results: Contribution due to different endowments Contribution due to differences in returns to explanatory variables Intercept differential Total differential due to structural parameters Overall earnings differential (Difference in log earnings) bm(m-f) ‘Endowments’ -.016 .013 .048 -.007 .027 -.007 -.018 .045 .085 f(bm-bf) ‘Discrimination’ -.034 .068 -.009 .010 -.017 .021 -.007 -.062 -.030 bm(m-f) ‘Endowments’ -.077 .026 .080 .006 -.011 -.050 -.019 -.045 f(bm-bf) ‘Discrimination’ -.120 -.001 -.030 .003 .012 .021 -.021 -.136 bm(m-f) ‘Endowments’ .008 .007 .007 -.012 -.003 .006 .052 .065 f(bm-bf) ‘Discrimination’ -.024 .109 .004 .018 .009 -.041 -.028 .047 .085 (40.7%) -.045 (-80.4%) .065 (26.6%) -.030 -.136 .047 .154 .124 (59.3%) .237 .101 (180.4%) .132 .179 (73.4%) .209 (100.0%) .056 (100.0%) .244 (100.0%) The combined effect of the occupational dummy variables is minimal even though, as documented earlier, there is extreme occupational segregation. It is not possible, however, to conclude from this analysis that occupational sorting does not play a major role in explaining differences in mean earnings between men and women. By treating broad-based occupational groups as a set of dummy variables the analysis has discounted implicitly both the possibility of different returns to human capital between occupations and the possibility that the formation of the occupational distributions for men and women may, themselves, be discriminatory. The conclusion that occupational differences make little difference would, therefore, be premature. More in-depth analysis of the expanded method, to be reported in a later paper, will shed greater light on the extent of sex discrimination in access to occupational assignments. Earnings functions are decomposed separately for the combined public/parastatal or public corporations sector and the private sector and the results are reported in table 12. In the public sector the female-male pay differential (F$1.06) is relatively small compared with the private sector (F$1.28). Overall endowments favour female workers in the public sector as do the returns to these human capital attributes; but again the interecept term is relatively large and in favour of men such that the net effect is a small male pay advantage. In the private sector all the causal factors, endowments, returns to endowments and the intercept term, all favour men. However, the intercept differential is the largest contributor to the male pay advantage.11 Interestingly, the private sector appears to attract men with Reflecting the importance of the intercept term as an indicator of discrimination, it is revealing to examine the male-female weekly pay differential across the public-private sectors for one of the youngest age groups in table 8. The male advantage in pay is apparent in the parastatal and private sectors for 20-24 year olds, where experience differences would be negligible. Also in this age group, males with 3-5 years of education receive a premium in the hourly rate of pay of 17% over similarly educated females; for those with university education the premium is 24%. For those with completed secondary education there is no pay advantage for males. Again, it could well be that discriminatory practices in occupational assignment may underlie these pay advantages of young males, which are then perpetuated throughout their future working lives. 11 g:\cstsuva\discuspa\bh_23.doc 30 superior human capital attributes – education, potential experience and firm experience – and they are more likely to be located in the most desirable occupations and sectors. In addition, the returns to these endowments very often favour male workers. 6. CONCLUSIONS This paper has demonstrated that there was an apparent significant increase in the labour force participation of women in Fiji over the decade 1986-1996. The female share of the total labour force rose from 21% to 33%, but a large proportion of these women failed to find cash employment and large numbers entered the non-cash, subsistence, largely agricultural economy. By the mid-1990s women’s share of the two largest occupational groupings (agriculture and elementary occupations) had trebled to one-third and the gender concentration of the labour force seems to have become somewhat less extreme. Yet, their greater representation in the lowly paid occupations did little to raise women’s overall status in Fiji’s labour market. Women have made little headway into male occupational preserves. The analysis went on to examine how this extreme concentration has impacted on the relative earnings of women compared with men in the formal sector of the labour market. Using data collected from over 8,000 employees in a specially designed survey, analysis has shown that mean pay is consistently higher for men compared with women while women’s earnings profile across age groups is much flatter, especially in the private sector, perhaps reflecting the kinds of jobs to which they are assigned, lacking opportunities for much on-the-job skill acquisition and pay advance. This result may also reflect the effect of discontinuities in women’s attachment to the labour market and the loss of productivity and pay-enhancing continuous work experience. Male-female earnings differentials take a U-shape across education levels, being wider at lower and high levels of schooling. Again, since very few women with higher education attain the very senior, highly paid executive positions, gender differences in occupational assignment contribute to these differences. The paper concludes with a standard decomposition analysis of male/female pay differentials which has been widely utilised around the world. The method decomposes the pay differential between the sexes into a portion attributable to a composition effect – “endowments” – and a portion attributable to structural differences, or “discrimination”, reflecting differences in the returns to these endowment factors. The results show that roughly 40% of the pay difference can be attributed to the composition effects, reflecting differences between males and females in their education, general and firm experience, ethnic group, sector of employment, training received and occupation. Some of these attributes, including education and training, favour female workers. The remaining 60% of the pay difference is attributable to differences in the returns to various human capital and other characteristics of the workers. Again, some advantages in these returns favour women. The principal source of “discrimination” lies in an unexplained intercept term in the regression analysis, whereby mean male earnings lie above that of females after controlling for many of these human capital attributes and various types of insertion in the Fiji labour market. g:\cstsuva\discuspa\bh_23.doc 31 The analysis leaves open the possibility that occupational sorting may play a major role in explaining the differences in earnings between the sexes. By treating broad-based occupational groups as a set of dummy variables, the analysis has discounted both the possibility of different returns to human capital between occupations and the possibility that the formation of the occupational distributions for men and women may, themselves, be discriminatory. This paper has elaborated on the methodology – the “expanded method” – that would allow an estimate of occupational discrimination to be made, whereby women with the same human capital and other attributes as men, do not have equal access to all occupations, including the highly paid executive positions. A subsequent paper in this series will utilise the data collected in the special survey to test this model. g:\cstsuva\discuspa\bh_23.doc 32 REFERENCES Anker, R. (1997), “Theories of Occupational Segregation by Sex: An Overview”, International Labour Review, Vol. 136, No. 3 Blinder, A.S. (1973), “Wage Discrimination: Reduced Form and Structural Estimates”, Journal of Human Resources, Vol.8, No.4 Brown, R.S., Moon, M. and B.S. Zoloth (1980), “Incorporating Occupational Attainment in Studies of Male-Female Earnings Differentials”, Journal of Human Resources, Vol. 15, No.1 Cain, G.G. (1986), “The Economic Analysis of Labour Market Discrimination: A Survey”, in O. Ashenfelter and R. Layard (eds); Handbook of Labour Economics, Vol. 1, Amsterdam; North Holland Cohen, B. and House, W.J. (1994), “Education, Experience and Earnings in the Labour Market of a Developing Country: The Case of Khartoum”, World Development, Vol. 22, No. 10 Fiji (1984), “Fiji Employment and Development Mission: Final Report to the Government of Fiji”, Parliamentary Paper No. 66 Fiji, Censuses of Population, 1976, 1986 and 1996, Bureau of Statistics, Suva Fiji, (1997), Development Strategy for Fiji: Policies and Programmes for Sustainable Growth, Ministry of National Planning, Suva Oaxaca, R.L. (1973), “Male-Female Wage Differentials in Urban Labour Markets”, International Labour Review, Vol. 14, No. 3 Siltanen, J., Jarman, J. and Blackburn, R.M. (1993) Gender Inequality in the Labour Market: Occupational Concentration and Segregation: A Manual on Methodology, Interdepartmental Project on Equality for Women in Employment, International Labour Office, Geneva United Nations (1994), Programme of Action Adopted at the International Conference on Population and Development, New York United Nations Development Programme (UNDP) (1998), Human Development Report, 1998, UNDP, New York g:\cstsuva\discuspa\bh_23.doc 33 APPENDIX Table A1: Variable Definitions in Regression Analysis Variable Ln(HRPAY) ED1 ED2 ED3 ED4 ED5 ED6 EXP EXP² FIRMEXP FIRMEXP² ETHNIC11 ETHNIC12 ETHNIC13 PUBLIC PARSTAT TRAIN OCCUP17 OCCUP18 OCCUP19 OCCUP20 OCCUP21 OCCUP22 OCCUP23 SECTOR24 SECTOR25 SECTOR26 SECTOR27 SECTOR28 SECTOR29 SECTOR30 SECTOR31 SECTOR32 Definition Natural logarithm of hourly pay D.V: One if 1-3 years of schooling D.V: One if 4-6 years of schooling D.V: One if 7-9 years of schooling D.V: One if 10-12 years of schooling D.V: One if post-secondary schooling (diploma/certificate) D.V: One if university level education Years of potential experience other than with current employer The square of EXP Years with current employer The square of FIRMEXP D.V: One if respondent is Indian D.V: One if respondent is Rotuman D.V: One if respondent is not Fijian, Indian or Rotuman D.V: One if respondent works in the public sector D.V: One if respondent works in a public corporation D.V: One if respondent has received vocational training D.V: One if employed as a lower level professional D.V: One if employed in a clerical occupation D.V: One if employed in a sales occupation D.V: One if employed in a service occupation D.V: One if employed in an artisan occupation D.V: One if employed as a garment worker D.V: One if employed in any other blue collar occupation D.V: One if engaged in the agricultural sector D.V: One if engaged in the utilities sector D.V: One if engaged in the construction sector D.V: One if engaged in the transport sector D.V: One if engaged in the finance sector D.V: One if engaged in the services sector D.V: One if engaged in the food manufacturing sector D.V: One if engaged in the garment manufacturing sector D.V: One if engaged in other sector g:\cstsuva\discuspa\bh_23.doc Males Mean S.D. 1.128 0.637 0.017 0.131 0.038 0.191 0.210 0.408 0.640 0.480 0.105 0.307 Females Mean S.D. 0.920 0.689 0.007 0.081 0.024 0.153 0.119 0.324 0.728 0.445 0.114 0.318 0.049 8.714 0.217 8.669 0.031 7.725 0.175 8.096 151.076 8.218 138.216 0.647 0.012 0.032 272.654 8.408 247.654 0.478 0.110 0.177 125.189 6.344 87.533 0.516 0.016 0.049 228.250 6.878 178.598 0.500 0.125 0.216 0.211 0.186 0.307 0.408 0.389 0.461 0.248 0.096 0.370 0.432 0.295 0.483 0.112 0.162 0.098 0.136 0.240 0.025 0.154 0.315 0.368 0.297 0.342 0.427 0.157 0.361 0.110 0.356 0.092 0.069 0.116 0.156 0.071 0.313 0.479 0.289 0.253 0.321 0.363 0.256 0.027 0.037 0.094 0.132 0.052 0.255 0.072 0.050 0.161 0.188 0.292 0.339 0.222 0.436 0.259 0.218 0.001 0.017 0.045 0.068 0.094 0.246 0.028 0.272 0.028 0.130 0.208 0.251 0.292 0.431 0.164 0.445 0.043 0.203 0.007 0.086 34 Table A.2: Estimated Earnings Equations: Males and Females – All sectors Independent Dependent Variable = Ln(HRPAY) Variable Males Females (EDO) (None) ED1 (1-3 years) -.326(.049)* -.100(.096) ED2 (4-6 years) -.239(.038)* -.017(.060) ED3 (7-9 years) -.134(.022)* -.060(.036) ED4 (10-12 years) -.100(.024)* .129(.039)* ED5 (13-15 years) .266(.032)* .257(.047)* ED6 (16+ years) .772(.038)* .726(.059)* EXP .017(.002)* .008(.003)* EXP² x 10-4 -1.409(.541) -1.352(.879) FIRMEXP .043(.002)* .047(.003)* FIRMEXP² x 10-4 -6.327(.673)* -8.745(1.168)* (ETHNICO) (Fijian) ETHNIC11 (Indian) -.029(.013)* -.042(.016)* ETHNIC12 (Rotuman) .192(.050)* .033(.057) ETHNIC13 (Others) .174(.032) .168(.034)* (PRIVATE) PUBLIC .047(.024)** .098(.039)** PARSTAT (Parastatal) .323(.020)* .369(.039)* TRAIN (Vocational Training) .112(.013)* .055(.017)* (OCCUPO) OCCUP17 (Lower Prof.) -.262(.026)* -.267(.047)* OCCUP18 (Clerical) -.436(.025)* -.406(.043)* OCCUP19 (Sales) -.614(.029)* -.623(.049)* OCCUP20 (Service) -.704(.027)* -.627(.050)* OCCUP21 (Artisan) -.628(.025)* -.569(.052)* OCCUP22 (Garments) -.671(.045)* -.252(.032)* OCCUP23 (Other Blue Collar) -.737(.027)* -.762(.051)* (SECTORO) SECTOR24 (Agriculture) .071(.036) .048(.253) SECTOR25 (Utilities) .140(.034)* .311(.065)* SECTOR26 (Construction) .010(.027) .222(.052)* SECTOR27 (Transport) .031(.020) .129(.041)* SECTOR28 (Finance) .441(.027)* .648(.030)* SECTOR29 (Services) .102(.024)* .199(.041)* Constant 1.040 .885 Adjusted R .620 .735 F-Statistics 278.0* 223.8 Sample Size 5433 2572 *, ** Significant at 1% and 5% respectively Numbers in parentheses are standard errors Source: Fiji Employment Survey g:\cstsuva\discuspa\bh_23.doc 35 Table A.2: Estimated Earnings Equations: Males and Females – Public & Parastatal Sectors Independent Variable (EDO) (None) ED1 (1-3 years) ED2 (4-6 years) ED3 (7-9 years) ED4 (10-12 years) ED5 (13-15 years) ED6 (16+ years) EXP EXP² x 10-4 FIRMEXP FIRMEXP² x 10-4 (ETHNICO) (Fijian) ETHNIC11 (Indian) ETHNIC12 (Rotuman) ETHNIC13 (Others) (PRIVATE) PUBLIC PARSTAT (Parastatal) TRAIN (Vocational Training) (OCCUPO) OCCUP17 (Lower Prof.) OCCUP18 (Clerical) OCCUP19 (Sales) OCCUP20 (Service) OCCUP21 (Artisan) OCCUP22 (Garments) OCCUP23 (Other Blue Collar) (SECTORO) SECTOR24 (Agriculture) SECTOR25 (Utilities) SECTOR26 (Construction) SECTOR27 (Transport) SECTOR28 (Finance) SECTOR29 (Services) Constant Adjusted R F-Statistics Sample Size Dependent Variable = Ln(HRPAY) Males Females -.383(.065)* -.266(.057)* -.164(.037)* -.181(.039)* .288(.049)* .727(.053)* .009(.003)* 1.391(.901) .021(.0030* .001(.891)* -.135(.276) -.002(.180) -.004(.097) .298(.101)* .389(.107)* .878(.112)* .015(.005)* -2.495(1.860) .033(.005)* -5.125(1.682)* .051(.017)* .062(.069) .196(.045)* .052(.026)** -.050(.084) .171(.063)* .068(.019)* .047(.027) -.228(.032)* -.352(.034)* -.496(.056)* -.655(.035)* -.431(.037)* -.539(.039)* -.273(.060)* -.371(.057)* -.324(.094)* -.584(.072)* -.292(.129)** -.721(.084)* .215(.141) .207(.033)* .-.235(.023)* .162(.030)* -.014(.083) 1.310 .540 98.2 2155 .251(.350) .280(.060)* -.034(.038) .117(.038)* -.029(.087) 1.073 .392 22.1 884 Source: Fiji Employment Survey g:\cstsuva\discuspa\bh_23.doc 36 Table A.2: Estimated Earnings Equations: Males and Females – Private Sector Independent Variable (EDO) (None) ED1 (1-3 years) ED2 (4-6 years) ED3 (7-9 years) ED4 (10-12 years) ED5 (13-15 years) ED6 (16+ years) EXP EXP² x 10-4 FIRMEXP FIRMEXP² x 10-4 (ETHNICO) (Fijian) ETHNIC11 (Indian) ETHNIC12 (Rotuman) ETHNIC13 (Others) (PRIVATE) PUBLIC PARSTAT (Parastatal) TRAIN (Vocational Training) (OCCUPO) OCCUP17 (Lower Prof.) OCCUP18 (Clerical) OCCUP19 (Sales) OCCUP20 (Service) OCCUP21 (Artisan) OCCUP22 (Garments) OCCUP23 (Other Blue Collar) (SECTORO) SECTOR24 (Agriculture) SECTOR25 (Utilities) SECTOR26 (Construction) SECTOR27 (Transport) SECTOR28 (Finance) SECTOR29 (Services) Constant Adjusted R F-Statistics Sample Size Dependent Variable = Ln(HRPAY) Males Females -.152(.079) -.178(.050)* -.102(.027)* .065(.031)** .256(.042)* .863(.056)* .021(.002)* -2.845(.673)* .055(.003)* -10.011(1.048) -.084(.102) -.020(.064) -.065(.038) .087(.043)** .229(.055)* .771(.094)* .005(.003) -.759(.997) .052(.004)* -8.704(1.722)* -.055(.018)* .298(.070)* .178(.044)* -.078(.020)* .125(.079) .134(.041)* .099(.018)* .062(.022)* -.309(.045)* -.471(.037)* -.655(.038)* -.742(.040)* -.675(.035)* -.724(.051)* -.796(.039)* -.087(.084) -.393(.062)* -.627(.067)* -.649(.072)* -.586(.069)* -.212(.033)* -.784(.070)* .033(.039) .034(.062) .029(.035) .005(.020) .442(.028)* .013(.043) 1.062 .551 135.0* 3278 .298(.355) .235(.182) .052(.135) .127(.051)** .631(.032)* -.038(.060) .930 .711 139.5* 1688 Source: Fiji Employment Survey g:\cstsuva\discuspa\bh_23.doc 37