South African Labor Markets Adjustment and Inequalities

DISCUSSION PAPER South African Labor Markets Adjustment and Inequalities Peter Fallon Robert Lucas THE WORLD BANK SOUTHERN AFRICA DEPARTMENT INFORMAL DISCUSSION PAPERS ON ASPECTS OF THE ECONOMY OF SOUTH AFRICA

FOREWORD This study is published by the World Bank in its informal series of Discussion Papers on the South African Economy. It draws on research supported by discussions with staff drawn from a wide range of South African institutions. The study focuses on labor market behavior in the light of disappointing employment growth and considerable wage inequality. Its main empirical conclusions are: i) that rising unemployment can only be partly explained by rising wages; ii) labor markets for skilled workers operate much better than for the unskilled; and iii) wage differentials associated with non-economic characteristics are large by international standards. The study reviews and is broadly supportive of recent labor legislation. It argues, however, that a broad policy framework that goes far beyond labor market interventions is needed to improve employment growth and reduce inequality. Pamela Cox Country Director for South Africa The World Bank November 1998 (revised) Copyright 1998 The World Bank 1818 H Street, N.W. Washington, DC 20433, U.S.A. This is an informal study by World Bank staff and contributing South Africans, published for discussion purposes. It is not an official World Bank document.

ACKNOWLEDGMENTS The authors are grateful to the staff of a number of organizations in South Africa for their advice and help. These include: Anglo-American; the Commission of Conciliation, Mediation and Arbitration; the Confederation of South African Trade Unions, the Department of Labour; the Department of Public Works; the Development Bank of Southern Africa; the National Economic Development and Labour Council; the South African Chamber of Mines; the South African Chamber of Business; the South African Labour and Development Research Unit; the University of Cape Town; and the University of Natal, Durban. We gratefully acknowledge useful comments received from various World Bank staff including Ataman Aksoy, Amit Dar, Peter Moll, Martin Rama, Yvonne Tsikata and Michael Walton.

Table of Contents Executive Summary...i I II Introduction...1 A Snapshot of Labor Allocation...3 III Employment, Unemployment and the Impact of Wages on Labor Demand...7 Poor Employment Performance...7 Rising Unemployment...8 Rising Wages Have Depressed Labor Demand...9 IV Labor Market Adjustment and Imperfections... 12 Markets Clear for Skilled Workers, But Not For Unskilled... 12 Large Wage Differentials by Non-Productive and Institutional Characteristics... 14 The Incidence of Unemployment Differs Considerably Between Different Workers... 19 V Legislative Developments... 22 The Basic Conditions of Employment Act... 22 The Labor Relations Act... 25 The Employment and Occupational Equity Statute... 26 VI Policy Directions... 28 Increase Human Capital... 29 Creating a More Competitive Environment... 30 Avoiding Excessive Wage Increases... 30 Encourage the Positive Aspects of Trade Unions... 31 Public Works and Labor-Intensive Construction... 32 Facilitating National Agreement... 32 References... 34 Annex A... A1 Annex B...B1

Executive Summary There are a number of extremely severe labor market problems in South Africa: one of the worst unemployment problems in the world with an unemployment rate that, at end- 1993, ranged between 12.3 and 29.8 percent depending on the definition used; wage inequalities are also extremely severe by international standards, and only partly explained by differences in productive characteristics between workers; a massive disparity in the incidence of unemployment by race - 33.6 percent among African males by the broadest definition as compared to only 3.6 percent among white males. There are two major issues for economic policy. How to create more jobs and reduce unemployment? How to remove or at least reduce labor market inequalities? The unemployment crisis has two origins: formal sector job creation since the mid-1960s was too slow to keep pace with labor supply growth; and small-scale agriculture and other informal activities were unable to fill the gap between labor supply and formal employment. Not surprisingly, therefore, the proportion of the labor force without formal jobs grew markedly. For historical reasons associated with the apartheid era and earlier periods, informal activities are an unusually low employment source, and, despite growth in urban areas, have been unable to prevent rising unemployment. Today, South Africa has the dubious distinction of being the only medium-sized country in the world in which the number of unemployed exceeds the number of informal workers. Labor market imperfections are a more major source of wage inequalities in South Africa than is common in other countries. Disparities in human capital endowments such as levels of education and job-related skills do matter, but even after standardizing for these, wage differentials are still heavily determined by influences associated with labor market discrimination, low mobility and institutional forces. This paper concentrates on six questions key to the policy debate: 1. How sensitive is employment to wages? 2. Do wage movements clear imbalances between worker supply and demand? 3. How large a role do labor market imperfections play in determining wage differences between otherwise similar workers? 4. How important are unions and wage-setting bodies in determining wages and employment? 5. Which worker characteristics most affect the incidence of unemployment? 6. What further labor reforms are needed in addition to forthcoming legislative changes? Page i

The Impact of Rising Wages: Rising real product wages have had a substantial dampening effect on the demand for Black workers. To examine wage impacts on employment, labor demand equations were fitted to time series data for all sectors of the economy except agriculture. The estimated equations indicate an average long-run wage elasticity across the sectors of -0.71 (Table 3), i.e. a 10 percent increase in the real product wage would eventually lead to a 7.1 percent decrease in Black employment. Employment takes a significant period of time to adjust to wage changes. The weighted average mean lag of adjustment estimated for black employees is 2.8 years, or, in other words, after 2.8 years, only one half of the adjustment to a wage change will have taken place. Labor Market Adjustment: Labor markets clear well for skilled workers but not for the unskilled. First, using Africans and Whites as rough proxies for unskilled and skilled labor respectively, econometric analyses of wage equations indicate that unskilled wages have been less responsive to changes in unemployment levels than those of the skilled. Second, African wages do not adjust as quickly as those of Whites to changes in the inflation rate, thus suggesting greater rigidities in wagesetting. Third, within Black groups, although average unemployment probabilities are much higher than for Whites, these probabilities fall off very sharply at high education levels, thus indicating that markets clear for skilled Blacks. Labor Market Imperfections and Wage Differentials: Wage differentials in South Africa are heavily influenced by factors that should have little bearing in a well-functioning labor market. These may be divided into: a) discrimination by race and gender; b) barriers to mobility, i.e. location and informal or formal employment status; and c) job-related characteristics of those employed in the formal sector, i.e. union membership, public or non-public sector and coverage by Industrial ( now renamed Bargaining) Councils. Productive characteristics such as education and experience are important determinants of earnings, but non-productive characteristics are associated with larger differentials than usually found in studies of other countries. Unions and Industrial Councils: Union membership is associated with larger wage differentials than normally found in comparable studies of other countries. Time series analyses of manufacturing wages suggest that the African union/non-union earnings differential lay in the range of 25-35 percent in 1980-93. As union membership only covers 31 percent of the formal sector, this suggests that, given a union wage differential of 0.3, formal wages are increased on average by about 9 percent. Combining this with the estimated average long-run wage elasticity of -0.71 suggests that employment is reduced through union wage-raising effects by about 6.3 percent. The results also suggest that a one percent increase in the minimum wages set by Councils is associated with wage rises ranging between 0.21 to 0.33 percent for Africans, 0.19 to 0.23 percent for other Black groups, and 0.12 to 0.16 percent for Whites. These may be interpreted as the wage-raising effects of centralized bargaining over and above those associated with trade union activity. This is possible in the South African context as plant-level bargaining, often on a closed-shop basis, is permitted in many industries. The Incidence of Unemployment across Different Sections of the Labor Force: The group most at risk from unemployment are African women with no education living in rural areas. For this group, the probability of unemployment is very high at 0.71. In contrast, the likelihood of unemployment is negligible for a well-educated, white male living in a metropolitan area. The results confirm that females are generally at higher risk from unemployment than men even after standardizing for other relevant characteristics, and that Whites have lower unemployment probabilities than other groups. Unemployment is much higher among younger members of the labor force, even after standardizing for other influences. In most developing countries, unemployment is lower in rural areas, as subsistence agriculture tends to soak up excess labor supply,but this is not true in South Africa. Among Africans in particular, the probability of Page ii

unemployment is much higher in rural than in urban and metropolitan areas. Unemployment is not confined to the least educated. Although greater education reduces the probability of unemployment, it is still high among Blacks with 10 years of education. Policy Directions: In addition to the need to maintain macroeconomic stability, and increase employment outside the formal sector with expansion of small-to-medium enterprises and a marketbased land reform, a number the main policy directions are suggested. Human capital accumulation. Because of past policies that restricted the provision of highquality education and training to Whites, there is a scarcity of skilled labor that has acted as a brake on economic growth. This is reflected in the high returns to education received by all racial groups at almost all levels of education. Faster growth will require a more rapid expansion of skills across all racial groups. Such an expansion should also be redistributional. The two major priorities are an expansion of basic education to the disadvantaged majority and an upgrading of the skill of existing workers A more competitive environment. Discrimination between workers by non-productive characteristics tends to be less in a strongly competitive industry, as less-discriminatory employers gain an advantage over others. One reason why large wage differentials persist by gender and race in South Africa is that its industrial structure is very oligopolistic or even, in some cases, monopolistic. A more competitive environment should serve to lessen labor market discrimination. The current trade liberalization is a step in this direction. Avoid excessive wage increases. The Government has few policy instruments with which to intervene directly in wage-setting. There are through: 1) the Employment Conditions Commission (formerly the Wage Board); 2) a code to influence wage-setting in the new Bargaining Councils; and 3) wage policy regarding government employees. The Commission should try to ensure that wage increases do not increase unemployment, which, under present circumstances of very high unemployment would suggest that real wages be allowed to fall. In new areas of coverage, most notably, agriculture and domestic services, it may be wise for the Commission to restrict itself to trying to ensure that minimum non-wage standards as laid-out in the Basic Conditions of Employment Act are met, and avoid setting minimum wages in these sectors. The Government could issue a code that encouraged lower wages for young people and for all workers in areas of unusually high unemployment, and noted the maximum average pay increase consistent with national productivity growth. If necessary, the question of extending collective agreements to other parties could also be covered along with guidelines regarding the treatment of new entrants to the sector. With respect to government a different concern arises, as it is important that government pay to skilled employees does not fall far behind that of competing sectors. Encourage the positive aspects of trade unions. In conditions where unions are certain to continue to play a key role, it is very important that their positive aspects be allowed to develop fully. These aspects include their ability to raise productivity by countering arbitrary and unwise employer decisions, and providing an information channel through which employers learn of shop floor problems. The provisions of the Labour Relations Act regarding the establishment of workplace forums is a major step in the right direction. Page iii

Use public works schemes and labor-intensive methods in the provision of public infrastructure. Government can create productive jobs to help reduce unemployment. This can take the form of public works programs such as the Community Employment Program (CEP) or the adoption of more labor-intensive methods in the construction of public infrastructure as envisaged under the National Public Works Programme (NPWP). It is nevertheless important that such schemes be located in areas where there are substantial concentrations of surplus labor, otherwise local wage rates may be bid up to the detriment of employment outside the schemes. As unskilled labor is in excess supply in South Africa, it is appropriate that such labor be valued at its social opportunity cost (the shadow wage). The informal sector wage may be taken as a rough guide. There are major problems in paying wages below formal levels, which may best be solved at national level through the National Economic Development and Labour Council (NEDLAC). Encourage national agreements. The experience of other countries with adversarial industrial relations does not bode well for the success of a national social pact. Hopefully, NEDLAC may be used to reach national agreement on some issues, such as the extension of Bargaining Council (BC) agreements to other parties, a code of practice for wage agreements in BCs, the criteria to be used in wage-setting by the Employment Conditions Commission, and the circumstances in which low pay may be paid to enable greater labor-intensity in the provision of public infrastructure. Page iv

I Introduction As the employment crisis deepens, labor market issues in South Africa take on an increasingly central role in economic policy decisions. South Africa now has one of the worst unemployment problems in the world with an unemployment rate that, at end-1993, ranged between 12.3 and 29.8 percent depending on the definition used. This represents a tremendous underutilization of the country s human resources. Labor market inequalities are also extremely severe by international standards. White 1 workers in formal manufacturing earn on average about two-and-a-half times the wage paid to Africans, and much of this differential remains after adjusting for differences in skill levels and other relevant characteristics. Non-racial factors such as gender and location also generate large earnings differentials. There is also a massive disparity in the incidence of unemployment by race - 33.6 percent among African males by the broadest definition as compared to only 3.6 percent among white males. There are thus two major issues for economic policy. How to create more jobs and reduce unemployment? How to remove or at least reduce labor market inequalities? The employment or perhaps more accurately, unemployment crisis has two origins: formal sector job creation since the mid-1960s was too slow to keep pace with labor supply growth; and smallscale agriculture and other informal activities were unable to fill the gap between labor supply and formal employment. In the 1980s, average annual formal job growth was only about 0.6 percent compared with a labor force growth rate of 2.5 percent, and has been negative thus far in the 1990s. Not surprisingly, therefore, the proportion of the labor force without formal jobs grew markedly in the past fifteen years. For historical reasons associated with the apartheid era and earlier periods, informal activities are an unusually low employment source, and, despite growth in urban areas, have been unable to prevent rising unemployment. Today, South Africa has the dubious distinction of being the only medium-sized country in the world in which the number of unemployed exceeds the number of informal workers. Labor market imperfections are a more major source of wage inequalities in South Africa than is common in other countries. Disparities in human capital endowments such as levels of education and job-related skills do matter, but even after standardizing for these, wage differentials are still heavily determined by influences associated with labor market discrimination, low mobility and institutional forces. Promoting greater wage equality thus goes much further than improving access to education and training, but must also face the difficult task of ironing out major wrinkles labor market imperfections. It is tempting to think South African labor as divided between two labor markets: the one, a formal market with wages set by collective bargaining and other institutional mechanisms, and the other an informal market where traditional supply-and-demand considerations apply. This is too simple a view, as the evidence presented in subsequent Sections shows: while formal-informal considerations are important in influencing wage differentials between individuals, other factors also matter. The system is rather one of a set of loosely intertwined labor markets that nevertheless can be roughly divided along formal-informal lines. Much of the policy debate revolves around the role of unions and other wage-setting mechanisms as a source of weak employment performance and wage inequality. Despite growing unemployment, black formal sector wages grew significantly in real terms in the 1980s, although this trend dampened in the 1990s. Although the evidence is sparser, black informal wages seem to have fallen, thus suggesting a widening in the formal-informal differential. In contrast, white wages displayed little growth despite ongoing improvements in average educational attainment, while white 1 The four racial groups in South Africa are referred to as Africans, Asians, Coloreds and Whites. The term Blacks refers to all groups other than Whites. Page 1

unemployment remained at modest levels. Given growth in black trade unionism and the ongoing operation of other wage-setting mechanisms, it would seem plausible that these factors may have pushed up black formal wages, depressed employment, and widened the gap between formal and informal. This view remains, however, hotly contested. This paper builds on, and extends, earlier work 2 to address six questions: 1. How sensitive is employment to wages? 2. Do wage movements clear imbalances between worker supply and demand? 3. How large a role do labor market imperfections play in determining wage differences between otherwise similar workers? 4. How important are unions and wage-setting bodies in determining wages and employment? 5. Which worker characteristics most affect the incidence of unemployment? 6. What further labor reforms are needed in addition to forthcoming legislative changes? Some of these questions were addressed in previous studies, but whereas earlier work relied heavily on the estimation and simulation of macroeconomic models using highly aggregate data, the empirical sections of this paper concentrate on analyses at the sectoral level and on results derived from household data collected in the South Africa Living Standards and Development Survey (LSDS) 3 in late 1993. While this approach does not use macroeconomic simulations, it offers potentially richer and more accurate analyses of key labor market relationships. The rest of the paper is divided into six sections. Section II sets the scene with a snapshot of the South African labor market using some detailed household information. Employment trends are then briefly summarized in Section III followed by a presentation of our analysis of the determinants of labor demand in the formal sector. This section gives wage elasticities of demand by sector and estimates of the impact of past wage growth on employment. This leads to a discussion of unemployment behavior in Section IV which extensively discusses the incidence of unemployment on different sections of the labor force. The evidence on labor market malfunction is presented in Section V. This section examines in turn wage unemployment linkages, evidence on the wage impact of institutional factors, and the relative strengths of a range of other labor market imperfections. Forthcoming developments as incorporated in the Labor Relations and Employment Standards Acts are described and evaluated in Section VI. Some further policy directions are discussed in Section VII. Annexes to the paper provide descriptions of data and estimation techniques. 2 See Fallon (1992) and Fallon and Pereira da Silva (1994). 3 This is a national survey carried out by the Southern African Labour Development Research Unit at the University of Cape Town in October-December, 1993. See Annex B for further details. Page 2

II. A Snapshot of Labor Allocation There are a number of features of the South African labor market that stand out as being rather unusual for a country at its present stage of economic development: the level of urbanization is relatively high, and agriculture accounts for a low share of the labor force; formal wage employment is by far the largest source of productive labor absorption; the urban informal sector is a relatively small employer; unemployment is extremely high. In contrast to most developing countries, South African labor is mostly urban as only 38.5 percent of the labor force are to be found in rural areas. Despite attempts to restrict Africans to the homeland areas during the apartheid era, 49 percent of African workers are urban. Agriculture accounts for only about 10.7 percent of the labor force. This is a rather low proportion for a country at South Africa s level of development that reflects in part the systematic encouragement of capital-intensive agriculture by previous governments and the eradication of subsistence farming in much of the country in earlier decades by the Land Acts. In many other countries in sub-saharan Africa, agriculture has acted as a valuable buffer that absorbed what would otherwise have been excess labor, but South Africa has not enjoyed this advantage. The formal sector is by far the largest employer. In 1995, estimates 4 of labor force and employment suggest that about 49.9 percent of the labor force held formal jobs, and although this proportion varies across racial groups -- Africans: 42.6 percent, Asians: 64.1 percent, Coloreds: 66.8 percent, and Whites: 71.6 percent -- it remains true that the formal sector generates the most employment for every group. The proportions of the labor force holding regular wage jobs as calculated from the Poverty Survey give similar estimates. The Survey also shows (Table 1), that the formal sector is the biggest employer for both men and women, as regular wage jobs provide the bulk of employment for all workers regardless of gender or racial group. Formal employment is quite highly unionized by developing country standards. While the rights of African workers to organize in trade unions were heavily restricted prior to 1979, such restrictions were lifted in that year resulting in rapid growth in union membership. After reaching a peak of 2.9 million members in 1992, membership dropped to 2.69 million in 1995, but still accounted for 35.1 percent of formal workers. Unions are especially strong in mining and much of manufacturing, but as in most countries, are weak in agriculture and most service sectors. Whites are the least unionized racial group among regular employees (Table 1), and, consistent with experience in other countries, women are much less unionized than men. 4 Labor force is calculated from demographic projections based on Sadie (1991), while the Standard Employment Series (SES) published by the Central Statistical Services is the measure of formal employment. The SES data include domestic servants and cover the whole of present-day South Africa. Page 3

TABLE 1: EMPLOYMENT STATUS BY GENDER AND POPULATION GROUP (Adults Aged 16-70, Percent) EMPLOYMENT STATUS Males Females Total AFR WHT COL ASN AFR WHT COL ASN Regular Wage Employment... Public Sector 8.4 16.8 12.9 16.7 4.7 16.3 8.1 8.1 8.3 Private Sector 26.6 49.9 41.2 41.7 13.3 29.5 28.2 21.7 23.8 Informal Employment... Casual Wage Employment 4.3 2.8 6.0 3.8 2.8 4.7 5.9 1.2 3.7 Self-Employed 4.1 8.3 2.7 9.8 4.7 5.1 2.2 3.7 4.6 Employed NEC... 0.1 0.1 0.0 0.1 0.1 0.1 0.0 0.0 0.1 Total Employed 43.6 77.9 62.7 71.6 25.5 55.7 44.5 34.6 40.5 Unemployed... Looking For Work 7.2 2.1 8.5 5.2 5.3 1.9 9.9 3.4 5.7 Discouraged 14.9 0.8 4.4 0.8 14.7 1.1 4.5 3.4 11.5 Total Unemployed (Broad Definition) 22.1 2.9 12.9 6.0 20.0 3.0 14.4 6.8 17.2 Out of Labor Force... In School 24.1 10.5 13.2 12.9 20.7 9.6 11.8 13.8 19.4 Ill/Disabled 3.9 0.8 6.0 4.9 3.2 0.7 3.9 3.0 3.2 Retired 4.7 7.3 3.8 3.6 8.5 9.2 5.2 4.9 6.7 Others 1.6 0.6 1.4 1.0 22.1 21.8 20.2 36.9 13.0 Total Out of LF 34.3 19.2 24.4 22.4 54.5 41.3 41.1 58.6 42.3 TOTAL 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Union Members as % of Regular Employment 34.9 18.7 40.9 39.4 20.6 10.7 39.7 34.6 26.6 % of Labor Force (Broad Definition) Regular Employment 53.2 82.5 71.6 75.3 39.6 76.7 50.6 72.0 55.6 Informal Employment 13.2 13.9 10.8 17.0 16.4 18.2 25.0 11.6 14.6 Unemployment 33.6 3.6 17.6 7.7 44.0 5.1 24.4 16.4 29.8 TOTAL 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Source: Staff Calculations from LSDS (1993). Page 4

Employment other than in regular wage jobs, i.e. casual work and self employment in agriculture and other activities, is a minor part of the labor market, and absorbs only 14.6 percent of the labor force (Table 1). For Whites, such activities are often of a relatively formal nature, but for Blacks, this gives a useful measure of informal activities. Only 14.7 percent of the African labor force is engaged in informal pursuits (22 percent of employment), with broadly similar proportions applying to Asians and Coloreds -- 15.2 and 17.4 percent respectively. As in other developing countries, women are more likely to engage in informal activities than men. These figures calculated from the LSDS are consistent with the conclusions drawn by the CSS 5 for 1989, where it was estimated that 19.9 percent of the Black labor force were involved in statistically unrecorded activities, of whom 12.3 percent worked on a full-time basis. The CSS Household Survey of October, 1994 (OHS) suggests a lower estimate for black informal participation of only 6.5 percent, but all sources agree that informal participation is low. As discussed above, subsistence agriculture is underdeveloped because of past land policies. In urban areas, low informal participation also reflects the legacy of past Apartheid laws that restricted the access of Blacks in a whole range of urban activities. Urban informal activities are heavily based upon trading activities 6. A synthesis of various studies of the informal sector suggests that only 23 percent of Blacks participating in the informal sector in 1985 worked in production or trading activities. The rise of black taxi services in the 1980s has been the most striking phenomenon, and, by the end of the decade, it is estimated that this activity employed nearly 300,000 workers. The most striking feature of labor allocation is very high levels of unemployment 7 among blacks (Table 1). Two unemployment definitions are used here. First, by the narrow definition, a person is unemployed if not working, but actively seeking work. This is essentially the same as that used by the International Labour Office (ILO). Second, by the broad definition, a person is unemployed if not working, but either seeking work or available for work if offered. This adds discouraged workers to the narrow definition. The broad definition seems more appropriate in countries like South Africa where many jobseekers face such a low probability of finding employment that active job search may seem pointless, although the narrow definition has the advantage of being based solely on factual, as opposed to subjective, information. The broad concept is almost certainly the more useful when measuring unemployment among males, as international experience shows that proportionately few males in the prime age groups opt out of the labor force unless sick or disabled, or attending educational institutions. An indication of being willing to work thus probably does mean that work would be accepted if offered. For females, however, the narrow definition may be the more appropriate as labor force participation rates are usually much lower, which makes the identification of discouraged workers more subject to subjective response. Even by the narrow definition, the unemployment rate estimated from the LSDS for the entire labor force is high at 12.3 percent, but it reaches the staggering level of 29.8 percent when discouraged workers are included. There is great contrast between the incidence of unemployment among Blacks and Whites. White unemployment is very modest by international standards -- only 2.6 and 3.6 percent among males by the two definitions -- but Black unemployment is very high, particularly among Africans. By the broad definition, the unemployment rate among African males (44 percent) is one of the highest in the world, and perhaps the highest if compared with rates in 5 This source excluded the formerly independent states of Transkei, Bophutatswana, Venda and Ciskei. 6 See Kirsten (1988). 7 The recent ILO country review by Standing, Sender and Weeks is rightly critical of existing unemployment estimates. But even they suggest that the unemployment rate by the narrow definition may be around 20 percent -- a lower figure than found in the LSDS, the principal source of much of the present paper. Page 5

countries of similar or larger population size. Similar results were obtained by the OHS which found estimates of 33 and 29 percent for the overall unemployment rate by the broad definition in 1994 and 1995 respectively 8. In general, women are more susceptible to unemployment than men. This is a common phenomenon in developing countries, as women are often more restricted to work at or near to the home and may face negative discrimination when looking for work. TABLE 2: UNEMPLOYMENT RATES BY GENDER, POPULATION AND AGE GROUPS (Percent) AGE GROUP Males Females Total AFR WHT COL ASN AFR WHT COL ASN Narrow Definition... All 14.2 2.6 11.9 6.8 17.2 3.3 18.2 8.9 12.3 16-19 29.0 24.8 40.8 0.0 31.0 9.3 38.9 33.6 20-24 27.9 2.0 19.8 15.2 37.1 6.7 33.3 18.0 Broad Definition... All 33.6 3.6 17.6 7.7 44.0 5.1 24.4 16.4 29.8 16-19 66.9 40.0 54.3 0.0 74.3 9.3 48.0 43.0 20-24 57.0 3.8 29.6 18.8 68.9 10.1 44.2 25.7 Source: Staff calculations based on LSDS (1993). Youth unemployment is particularly severe (Table 2). Over two-thirds of African males aged 16-19 years are unemployed under the broad definition, and very high unemployment rates prevail among teenagers in other racial groups. Even by the narrow definition, unemployment is very high among young people. While in virtually every country, youth unemployment is more severe than the average, as first-time jobseekers often experience a period of unemployment before securing a job, the South African problem is amongst the worst in the world. Fortunately, unemployment rates are lower in the 20-24 age group, but this is only strongly marked among White males. To some degree, labor allocation by race can be reinterpreted according to skill levels. Although, there has been significant narrowing of education and skill attainment between the racial groups over the years 9, large gaps still remain. Predominately, whites have much greater education and skills than other groups, among which Africans lag well behind Asians and Coloreds. Labor allocation could then be reinterpreted as: 1) skilled workers are more likely to work in the formal sector than semi-skilled or unskilled; 2) unskilled workers are more likely to end up in the informal sector than skilled; and 3) the unskilled experience much higher unemployment rates. This interpretation is too simple, however, as the analyses of subsequent Sections suggest that the probabilities of being unemployed or employed in the formal sector depend not only on skill levels, but also on other characteristics including race. 8 The OHS results are not strictly comparable in these two years given different survey methodologies. 9 See Fallon (1992). Page 6

III Employment, Unemployment and the Impact of Wages on Labor Demand Despite significant growth in the number of government employees, employment performance has been disappointing in the formal sector, and this has worsened progressively since 1970, although there are very recent signs of improvement. Given ongoing growth in black labor supply, the proportion of blacks without formal jobs increased considerably leading to growing unemployment. Rising black wages contributed significantly to the unemployment problem, although other factors played a major role. Poor Employment Performance Formal employment growth peaked in the 1960s at an annual rate of about 2.9 percent per year. This declined to 1.9 percent in the 1970s, and dropped further to 0.7 percent in the 1980s. So far, employment growth has been negative in the 1990s, and by end-1996 had reached a level 6.9 percent below its average level for 1990. Black and White employment have behaved similarly since 1970 (Figure1), although this masks a marked shift in the distribution of employment across sectors as both groups moved away from primary producing activities. Government has been the only consistent performer over the past 20 years with a yearly employment growth rate of 4.5 percent in the 1970s and 3.4 percent in the 1980s. Employment growth would have been negative in 1976-90 if there had been no growth in government employment. Figure 1. Employment Index (1970-1995) 240 220 Index 200 180 160 140 120 Govt Whites Blacks Non-Gov 100 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 Year Source: Standardized Employment Series. South African Labour Statistics (various issues) and revised estimates. There are some recent signs of improvement. Total employment in the non-agricultural sectors at end-1996 10 was slightly below its level in mid-1993, thus signaling an end to the continuous decline of the previous three years. But, the economic revival that began in the latter half of 1994, only triggered very modest employment growth in 1995 with an offsetting decline in 1996. This is 10 Data on total employment are more up-to-date than data on employment by race as used in Figure 1. Page 7

worrying as government employment continued to grow in 1994-96 while private employment fell by about 2 percent. Rising Unemployment The employment slowdown was not matched by any major reduction in labor supply growth, and there was an increasing imbalance in recent years between levels of Black labor supply and formal employment. Labor supply estimates 11 indicate that the African labor force growth increased from about 2.5 percent per year in the early 1960s to about 2.8 percent in the 1980s, while growth of other Black groups declined from 2.8 to 2.6 percent. The imbalance between black labor supply and formal sector demand grew substantially in 1970-94, and now stands at over 50 percent (Figure 2). In sharp contrast, the proportion of Whites without formal jobs remained reasonably constant in the 1970s and 1980s, although it increased appreciably in the 1990s. This illustrates a stark historical contrast in the degree of labor market imbalance between Blacks and Whites. Africans were and still, to a large extent, are an outsider group in excess supply providing predominately unskilled labor to the formal wage economy. Whites were much more of an insider group supplying mostly skilled labor with the patronage of the apartheid regime. The relative stability of the "not-wage-employed" ratio for Whites corresponds to a situation of near full employment most of the period. Evidence from the Population Censuses 12 also support these conclusions. The Censuses use the narrow definition of unemployment, and may therefore tend seriously to underestimate unemployment among groups with a high unemployment incidence, as numbers of discouraged workers may grow disproportionately with excess labor supply. Changes in unemployment rates for South Africa exclusive of the TBVC states, as measured in the Censuses of 1980, 1985 and 1991, are roughly consistent with the movements in the 'not-wage-employed' proportions shown in Figure 2. These show that that Black unemployment rates more than doubled in 1980-91, that of Africans rising from 10.9 to 24.5 percent, and those of Asians and Coloreds from 4.5 and 7.7 percent to 12.8 and 16.6 percent respectively. In contrast, White unemployment only rose from 1.3 to 4.0 percent. There is also evidence from the censuses of wide disparity in African unemployment rates across different regions of the country. The 1991 Census indicates that the unemployment rate among Africans in South Africa exclusive of TBVC was 24.5 percent. African unemployment was much higher in the TBVC countries where the unemployment rate was consistently greater than 30 percent. In all of the homelands, both TBVC and self-governing territories, the unemployment rate rose consistently, and in most cases sharply, over the past decade. 11 These are taken from Sadie (1991). 12 Although some preliminary results are available from the 1996 Census, they do not include information on unemployment. Page 8

Figure 2. Percent of Labor Force Outside Formal Employment (1970-1995) 60.00 50.00 Percent 40.00 30.00 20.00 Whites Blacks 10.00 0.00 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 Year Source: Staff calculations based on projections of labor supply and Standardized Employment Series taken from South Africa Labour Statistics (1994). Rising Wages Have Depressed Labor Demand Earlier work 13 suggested that, while Black wage growth played a role in depressing formal employment levels, most of the decline in employment growth was attributable to other factors. The main non-wage influences identified were: rising capital-intensity linked to major parastatal investments; insufficient skill acquisition by the work force that acted as a constraint on growth and employment; declining private investment that led to slow growth in the non-government capital stock. The same study found, on the basis of aggregate data, a wage elasticity of demand for Black labor of -0.28, i.e. a 10 percent increase in the real product wage of blacks would, other things being equal, lead to a 2.8 percent decrease in their employment. 13 See Fallon (1992). An ILO-supported study by Standing, Sender and Weeks (1996) presents the conclusions through selective quotations as though wages were identified as the major factor inhibiting employment growth. This is untrue. Page 9

Much controversy remains, however, about the true sensitivity of employment to wages. This is a very important issue as it is central to virtually every question regarding wage policies. The purpose of this sub-section is to give a more detailed analysis of wage-employment relationships. Earlier work used a macro labor market model, which, while useful for analyzing and simulating the impact of changes in the macroeconomic environment on employment, necessarily suffers from the potential biases inherent in making estimates from aggregate data. The relationships between wages and employment have now been re-examined using data on a sectoral basis. The conclusion is that employment is more sensitive to wage levels than previously believed. Despite growing unemployment among black workers, their average real wages rose considerably in the 1970s and 1980s. This inevitably meant that real product wages (wages deflated by the price of output) paid by employers also increased (Figure 3.). While there is little doubt that earnings-related characteristics of blacks such as educational and occupational attainment have improved over time, there is also evidence that other factors have combined to push up black wages. These may include: reduced discrimination, including positive effects stemming from the abolition of job reservation practices; the abolition of influx controls; and the pressures of growing African trade unionism. Figure 3. Real Manufacturing Wage Index (1970=100) 180 160 140 Index 120 100 80 60 Africans Other White 40 20 0 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 Year Source: Staff calculations. The most striking feature of real product wage behavior since 1970 (Figure 3.) is that White wages were near stationary while Black wages grew in a somewhat erratic manner. Wage differentials thus narrowed considerably between Whites and other groups of workers. Growth in real African wages largely ceased after the economic recession deepened in 1989, but Asians and Coloreds continued to make some gains, although at a slower pace than before. Although Figure 3 is based only on manufacturing, the picture is broadly similar in other major sectors. Rising real product wages have had a substantial dampening effect on the demand for Black workers. To examine wage impacts on employment, labor demand equations were fitted to time series data for all sectors of the economy except agriculture. The estimated equations indicate an Page 10

average long-run wage elasticity across the sectors of -0.71 (Table 3) 14, i.e. a 10 percent increase in the real product wage would eventually lead to a 7.1 percent decrease in Black employment. This result is in line with estimates of wage elasticities for other countries 15. Mining has a rather low elasticity of -0.15, while not unexpectedly, the elasticity in services is much higher at -.95. In manufacturing as a whole, the average elasticity is -1.0, although this estimate may be a little on the high side, as an implausibly large wage elasticity was obtained in at least one manufacturing subsector. Employment takes a significant period of time to adjust to wage changes. The impact elasticities (the elasticity of employment over a one-year period to the wage) are mostly much lower than in the long run. The weighted average mean lag of adjustment estimated for black employees is 2.8 years, or, in other words, after 2.8 years, only one half of the adjustment to a wage change will have taken place. In manufacturing, the corresponding period is 2.25 years. The employment gains from a real wage reduction would this take a number of years to be realized fully. TABLE 3. WAGE ELASTICITIES FOR BLACK FORMAL SECTOR EMPLOYEES Beverages Tobacco Textiles Wearing Apparel Wood Products Furniture Chemicals Rubber and Plastic Non-Met Minerals Basic Metals Fabricated Metals Non-Elec Machinery Transport Equipment Long Run -0.184-0.057-0.984-2.508-0.196-0.364-1.166-0.243-2.929-0.758-0.466-0.632-0.440 Impact -0.095-0.018-0.346-0.709-0.603-0.139-0.344-0.153-0.451-0.166-0.175-0.408-0.201 Mining Construction Services -0.146-0.554-0.948-0.118-0.360-0.147 WEIGHTED MEAN -0.709-0.156 Source: Derived from labor demand equations. See Annex A. These results suggest significant job loss associated with wage increases. If Black real product wages had remained constant in 1970-94, Black formal employment would have grown at 1.48 percent per year instead of the 0.9 percent actually achieved. This implies that Black employment would have been around 13.3 percent higher than its current level. As about 45 percent of the Black labor force currently hold formal jobs, this indicates that black unemployment rates would have been about 6 percentage points lower than at present. 14 An elasticity of a similar value is quoted in the Presidential Labour Commission Report (1996). 15 See Hamermesh (1992) and OECD (1994). Page 11

IV. Labor Market Adjustment and Imperfections The high rate of unemployment is sufficient evidence in itself that labor markets in South Africa do not work smoothly. But some sections of the labor market work better than others, and some groups are more adversely affected than others. There are three main conclusions explored here: wage movements clear markets for skilled labor much more efficiently than for unskilled; wage differentials arising from non-productive differences between individuals and institutional factors are very high by international standards; probabilities of being unemployed differ very widely across individuals, and are heavily related to non-productive characteristics. Markets Clear for Skilled Workers but not for the Unskilled The evidence here is threefold. First, using Africans and Whites as rough proxies 16 for unskilled and skilled labor respectively, econometric analyses of wage equations indicate that unskilled wages have been much less responsive to unemployment levels than those of the skilled. Second, African wages do not adjust as quickly as those of Whites to changes in the inflation rate, thus suggesting greater rigidities in wage-setting. Third, within Black groups, although average unemployment probabilities are much higher than for Whites, these probabilities fall off very sharply at high education levels, thus indicating that markets clear for skilled Blacks. Rather different considerations apply when considering the behavior of Black and White wages. Within the black category, it is also useful to distinguish between the behavior of African wages, and those of Asians/Coloreds. The central conclusion of the econometric analysis of real consumption wages in 18 manufacturing sub-sectors in 1960-93 (Annex A) is that white wages adjusted much more strongly to market forces than those of Africans, and apparently maintained white employment close to its full employment level. The behavior of Asians/Coloreds wages seems to have been a mixture of that of Whites and Africans. The wages of all racial groups tend to fall in the face of increased unemployment, but given existing labor market conditions in South Africa, the effect of a given proportionate fall in the number of workers employed is much stronger for Whites. Suppose that the wage employment of Africans and Whites were each to fall by 1 percent. Since, at the present time, around 50 percent of the African labor force but only 20 percent of the White labor force are outside wage employment, this would increase these "not-wage-employed" ratios by about 0.5 percent and 0.8 percent points for Africans and Whites respectively. Applying these changes to each comparable wage equation for Africans and Whites (Annex A, Tables A2 and A3) reveals much stronger induced real wage decreases for Whites than for Africans. White wages are thus much more sensitive to a given proportionate change in white employment than is the case for Africans. This conclusion suggests that one reason why real White wages have shown a greater tendency to fall in the face of slower economic growth than those of Africans is that they are more responsive to lower growth in labor demand. The picture underlying the behavior of the African real wage is one in which conflicting forces have operated -- rising unemployment exerted downward pressure on 16 Given renewed skill accumulation among Blacks, racial characteristics are progressively worse proxies over time for skills. Nevertheless, the skills gap between Whites and Blacks remains very large. Page 12

wages, but other forces more than countered this. In 1975-85, the decline and eventual abolition of the system of influx controls seems to have had a positive effect on African real wage rates, and a correspondingly negative impact on White wages (Annex A). But, as discussed below, another force supporting African wages was the growth in African trade unionism. Rigidities in the determination of African real wages are also illustrated by their vulnerability to unanticipated levels of inflation 17 (Annex A). The results suggest that an unforeseen one percentage point increase in the inflation rate temporarily reduced the real African wage by about -0.27 to -0.42 percent, thus suggesting that nominal wage changes do not immediately adjust to inflation. This effect is absent among Whites, thus indicating that white nominal wages adjust smoothly in line with changes in the inflation rate, while among Asians/Coloreds, there is only weak evidence of insufficient wage indexation. The other main indicator of labor-market clearing among skilled groups is that, after standardizing for other relevant individual characteristics, the probability of being unemployed is much lower for well-educated workers irrespective of racial group (Figure 4). This is clearest among Africans where the probability of being unemployed by the broad definition falls drastically from over 30 percent to around 1 percent when workers with 14 years of education Figure 4. Unemployment Probabilities For Males (by education level) 0.5 Probability 0.4 0.3 0.2 0.1 0 years 5 years 10 years 14 years 0 Africans White Race Source: Staff calculations from LSDS. are compared with those with 10 years. Increased schooling at lower levels of education has much less impact on unemployment, probably because the quality of primary and secondary education received by Africans is very poor. 17 Unanticipated inflation is estimated in Annex A as the residuals from a Box-Jenkins autoregression of actual inflation rates. Since such residuals must average to zero, unanticipated changes in the inflation rate can only have temporary effects on the real wage. Page 13