This Selected Issues paper highlights that aided by the easing of political uncertainties after the national elections in early 1994 and by the cautious stance of policies adopted by the new administration, economic performance and investor sentiment in South Africa strengthened markedly. Nonagricultural value-added grew by 4 percent in 1995, led by a sharp increase in real gross private fixed investment. In contrast, developments in 1996 were characterized by a shift in investor sentiment and unrest in the foreign exchange markets.


This Selected Issues paper highlights that aided by the easing of political uncertainties after the national elections in early 1994 and by the cautious stance of policies adopted by the new administration, economic performance and investor sentiment in South Africa strengthened markedly. Nonagricultural value-added grew by 4 percent in 1995, led by a sharp increase in real gross private fixed investment. In contrast, developments in 1996 were characterized by a shift in investor sentiment and unrest in the foreign exchange markets.


54. The Growth, Employment and Redistribution strategy called attention to “the unemployment crisis in the labor market.” It cited official data for 1994 showing that open unemployment stood at one-third of the labor force and, on the basis of official macroeconomic projections, it concluded that the unemployment rate would rise further in the medium-term without enhanced policy actions. It noted that those worst affected by this would include the rural poor, the young, and the unskilled, and that even the enhanced policy environment proposed in the GEAR—including wage moderation and reforms to wage-setting institutions—would only achieve a reduction in the unemployment rate after the turn of the century. More recently, the South African Reserve Bank has reported that employment in the private nonagricultural sector continued to decline in 1996, compounding the fall of over 10 percent between 1990 and 1995. In addition, the authorities have indicated that even their projections for the unemployment rate in 2000 under the GEAR policy framework (nearly one-third of the labor force) could prove to be overly sanguine. This forms the background to the current preparations of a “jobs summit” aimed at discussing ways to tackle the unemployment problem in late 1997.

55. Given the authorities’ concern with unemployment and employment and the emphasis given by the authorities in the GEAR to wage determination practices for employment growth, this section discusses some aspects of current wage determination processes and their role in employment creation. The discussion also identifies some gaps in current knowledge about these issues that remain to be filled. The section goes on to assess the reliability of the underlying data on these matters, suggesting that the data do not appear to be subject to major measurement or construction errors, and hence that the authorities are right to use them in developing labor market policy.

A. Unemployment and Wage Determination

56. Open unemployment rates are extremely high and the authorities anticipate that they may continue to increase for some time, even on the basis of the enhanced policy framework presented in the GEAR. In this light, an assessment of the roots of unemployment is critical to the design of policy to address it.

57. As the Department of Labor correctly points out, there are many elements that should be considered in this context.17 Skills training has been inadequate—compounding a skilled labor shortage—trade policies have discouraged export-oriented growth and the associated stimulus to employment, and multiple economic distortions in the allocation of rural land due to the policies of the former regime have all played their part in generating large unemployment in South Africa. All of these issues need to be addressed in conjunction with those related to “labor market flexibility,” if the “unemployment crisis” is to be confronted effectively. But even in this wider context, it is still appropriate to consider what contribution reform to wage setting practices might make to efforts to stimulate employment.

58. It should be noted at the outset that there is considerable “institutional flexibility” in South Africa. There is a great variety of institutional arrangements in its labor market, including (i) some sector level negotiating fora which sometimes exist alongside parallel plant- level fora for firms in the same sector; (ii) a wide range of labor contracts some of which are negotiated annually while others cover longer time periods including productivity-related clauses; (iii) a variety of different arrangements governing worker training; and (iv) many different trade union and employer organizations which interact with each other in diverse ways. In addition, there is no legal requirement forcing both employer and worker representatives to use any particular institution, procedure, or contractual arrangement.

59. However, there is little “wage flexibility” in South Africa, as it is heavily restricted by a multiplicity of floors on wage rates, many of which are enforceable in the courts. Although there is no national minimum wage—and the GEAR emphatically rejects the introduction of such a minimum wage—legally enforceable wage floors are set by Wage Boards and Labor Orders in sectors with limited trade union representation, and by Bargaining Council (BC)18 agreements between employer and labor representatives at sector level. Under the recent Labor Relations Act, these wage floors have been made applicable to all firms in the sector if the BC represents over half of the employees in the sector and if the representatives at the BC request this extension, whereas previously, this extension was at the discretion of the Minister of Labor. In addition, plant level negotiations, including in firms covered by sector level BC agreements, may set wage floors above the rates determined by the BC agreement, although these do not have the force of law. Finally, there are employer-employee negotiations outside any of these legal frameworks, which establish negotiated floors on remuneration and conditions of employment.

60. An assessment of wage flexibility thus must consider a variety of wage-floor mechanisms and the interaction between them. In this process, consideration should be given to the evidence that the Wage Boards and Labor Orders have a limited and somewhat arbitrary coverage, and that the wage floors they have set have been low and infrequently revised, despite inflation. Furthermore, there were only 19 wage floors in operation in mid- 1995 applying to some 730,000 workers, with the bulk of these being covered by one determination. It is also important to consider that in 1993, only 86 BC agreements were in force covering about 10 percent of the total employed labor force, they were limited in their geographical coverage, and they provided exemptions for small firms.

61. In considering these issues and their implications for an assessment of how these institutions affect employment and unemployment, three issues are critical: whether the data cited above is appropriate to assess the coverage of the institutions, whether coverage is a useful means of assessing the role of any wage setting mechanism in discouraging employment, and whether the wage setting institutions can sensibly be considered in isolation from the other institutions, rather than as a group.

62. First, consider the data cited. BC agreements may cover some 10 percent of the employed labor force. But it is not clear that comparing coverage to the entire employed labor force is helpful in assessing its effects on employment creation. It would be more useful to compare coverage with employment in the private, nonmining, nonagricultural sector; government employment should be excluded because the aim is to discern the role of wage flexibility in determining private employment. Similarly, employment in mining and agriculture should be excluded because wage floors in the mines have always been set in bilateral negotiations between unions and firms outside these structures, and because agriculture acts as one of the residual sectors for surplus labor. In a sample of 340 manufacturing firms in Gauteng, Cape Town, Durban and Port Elizabeth studied by the ILO,19 just under 60 percent of firms (and therefore a higher share of workers in firms covered by the survey) were subject to a BC agreement, and a little over 20 percent of those firms were non-parties to the agreement. It would also appear that the data cited from the Labor Department for coverage of BCs, which underlies the number of 10 percent coverage mentioned above, excludes workers in firms that are not represented on the BCs, and therefore understates coverage for this reason.

63. Furthermore, the concentration of unskilled and young people among the unemployed suggests that from the perspective of assessing the causes of unemployment, the relevant coverage that researchers should be attempting to measure is that of unskilled workers, possibly excluding unskilled but experienced workers. Coverage measured on this basis could be at a completely different level and could be evolving quite differently from coverage measured against either total employment or private nonmining employment.

64. Second, coverage itself may not be a useful means of assessing the role of these mechanisms in generating unemployment. If covered workers represent a small portion of all (relevant) workers, this could reflect two quite different circumstances: that the mechanism is setting wages below market-clearing rates and therefore covers few workers because few have any incentive to negotiate with employers within that apparatus; or alternatively, that the mechanism is setting wages above market-clearing wage rates, shrinking the affected sectors relative to others, and therefore inducing low coverage compared to total employment. In the first instance, the BCs are economically irrelevant, while in the second, they are critical to determining employment patterns, but in both cases, coverage is low.

65. Third, the above suggests that to assess the contribution of these wage-setting mechanisms to unemployment, other measures than coverage should be used. At the sector level, a relevant measure would be wage differentials across firms or sectors for similar workers. If it could be shown that wage-setting institutions did not generate wage differentials compared to uncovered sectors, this would suggest that they had no harmful effects on employment, and could even be stimulating it if the institutions contributed to more harmonious industrial relations. However, evidence from many sources indicates that council wage structures set wages above those applied in firms outside council sectors, even allowing for factors such as firm size and unionization, and that they are certainly well above wage rates in the informal sector.20

66. Additional insight would be gained into these issues by new research on cross section data on the distribution of wage rates relative to the applicable wage floors. This research would indicate whether the floors set in these institutional arrangements were binding, and in what circumstances. Therefore, it would be important to start collecting these data, as they could be highly revealing about the impact of the wage floors on actual wage rates.

67. Issues of geographical coverage and exemptions should also be considered with care. A BC may cover only part of the country, but may nevertheless cover the industry in the entire country. Given that the geographical coverage can be readily extended, BCs or any other legally binding wage floor mechanism may discourage potential entry by new firms, even in regions not currently covered by the arrangement. Furthermore, this reduces competition in the goods markets, and thus may generate monopoly rents while discouraging employment creation. At the same time, it may maintain the current geographical concentration of industry to the detriment of the high unemployment and peripheral regions. Therefore, reference to current geographic coverage of BC agreements will tend to understate their economic impact.

68. On a similar basis, the observation that small firms are often exempted cannot be interpreted to imply that these mechanisms are generally unimportant to the small firm sector. Exemptions involve considerable administrative effort on the part of the applicants to submit, which can explain why applications come disproportionately from larger firms that can afford to make the effort. Furthermore, the exemptions that are approved are overwhelmingly partial—often relating to matters which have no bearing on labor cost. In addition, exemptions may subsequently be withdrawn, or lapse after their usual term of one year. If entry by new firms involves any sunk costs, the fact that exemptions are required at all, at any stage of the prospective life cycle of the new firm, will discourage entry by new firms. This disincentive does not depend on the degree of discretion in granting exemptions because, even if exemptions were granted automatically for “small” firms defined in some clear way, the disincentive to entry would remain for those firms that could only operate efficiently above the size considered “small.” Moreover, these mechanisms would continue to discourage existing small firms from growing—which can explain the rather limited number of medium size enterprises in South Africa—as well as potential larger firms, including foreign direct investment.

69. The impact of these mechanisms—wage boards, labor orders, BCs, and business and trade union negotiations—on wage flexibility and on employment patterns cannot be determined by consideration of each wage-setting mechanism in isolation from the others. If one such mechanism set above market-clearing wage rates in one sector but all other sectors set market-clearing wage rates, it would lead to a reduction of employment in that sector. But this would not necessarily cause unemployment because workers shed by the sector with above market-clearing wage rates would shift into the other sectors at lower wage rates. However, if all major sectors are faced with above market-clearing wage rates, then the workers shed by them would have few sectors with market-clearing wage rates in which to find employment. Thus, the combined effect of these wage-setting mechanisms would be to generate very sharp wage inequality between the above market-clearing sectors and the market-clearing sectors, as is observed in the South African data, and it would lead to considerable unemployment if the market-clearing sectors were unable to absorb all the labor released by the other sectors.

70. In practice, it is probable that the relative importance of the different wage-setting mechanisms to the aggregate black unskilled labor market has evolved considerably in the last 20 years. ICs probably became more important to the determination of the wage rates of black South Africans in the early 1980s, and plant level negotiations probably increased in importance thereafter. But both may have induced unskilled labor to shift increasingly to the market-clearing “informal” and agricultural sectors.

71. A development that has taken place in recent years and that has attracted some attention has been the introduction of explicit productivity-related clauses in remuneration agreements. While these clauses have helped ease industrial tensions in the negotiating process, and have protected the share of profits in income—both of these are useful achievements—they may have had an unintended negative impact on employment creation by inducing layoffs on the grounds that average labor productivity—and hence the wages of the remaining workforce—can be raised by doing so. These contracts have not yet spread widely, but they could compound the process of rapid labor productivity growth stemming in part from the shift toward capital intensive technologies, which has been associated with large falls in employment since 1990.

72. Finally, and as described in previous occasions, poor training and an inward oriented trade regime have contributed to the anti-employment bias in the South African economy. However, these factors can only account for unemployment if accompanied by wage rigidities. Therefore, even though the authorities are right to emphasize initiatives in these broader areas to stimulate employment, increased wage flexibility would contribute significantly to the early impact of these efforts to stimulate the growth of employment.

B. Youth Unemployment

73. The heavy concentration of unemployment among young people is one of the more striking aspects of the composition of unemployment in South Africa, and this is increasingly reflected in the discussion of unemployment issues. The causes and nature of youth unemployment thus require particular attention in labor market issues.

74. In a context of slow employment growth, employers might prefer to recruit older people, implicitly placing young people in a queue. But it seems improbable that this could account for the extremely high rates of youth unemployment observed in South Africa (see Figure 5). Even if employers were to prefer older workers, high involuntary youth unemployment would not occur if the differentials between youth and adult wages adjusted to offset this behavior of employers. One possible explanation is that the wage-setting institutions in South Africa are having a particularly marked effect on youth if they are impeding the necessary wage adjustment in response to the employers’ behavior.

Figure 5
Figure 5


Citation: IMF Staff Country Reports 1997, 082; 10.5089/9781451840933.002.A002

Source: October Household Survey 1995; and staff estimates.1/ Including those wanting work but not actively seeking it.2/ Excluding those not actively seeking work.

75. There are three indirect pieces of evidence consistent with this proposition. First, preliminary research undertaken by the University of Witwatersrand suggests that high reservation wages do not account for high youth unemployment. Second, the recent ILO survey finds “that employers’ preferred age for recruits is unusually high—between 26 and 35—considerably higher than is the case in the Philippines or Malaysia, with very few indicating a preference for youth between the ages of 16 and 25.”21 Third, the ILO survey also finds that many firms provide some training to new recruits, but only 7 percent of the firms provide industry-accredited training, and that training is overwhelmingly focussed on high- level workers. New recruits therefore do not receive anything like adequate training, despite the obvious need for it. In addition to simply impeding the employment of young people, wage floors may impede the provision of training in the private sector by preventing young recruits from accepting remuneration in the form of training.

76. Although firms do pay young workers less than adults with similar skills, the Department of Labor reports that such differentiation is limited, to at most 10 percent, which is considerably lower than in other countries. In view of its importance, it would be critical to collect evidence of the circumstances under which differentiation is applied.

C. The Unemployment Rate

77. The official estimate of unemployment is derived from the October Household Survey (OHS) that is conducted by the Central Statistical Service (CSS). These unemployment data for black South African people are shown in Figure 5. The household survey is conducted annually, on the basis of direct interviews with a large sample of households, and the survey covers a range of issues in addition to unemployment. The survey is used for the official estimate of unemployment because few of the unemployed are entitled to unemployment insurance benefits, so an estimate based on the latter will not be an appropriate measure of unemployment.

78. The survey defines the employed as those over 15 years old who, during the week prior to the survey, worked for five or more hours for a wage, profit, or family gain in cash or kind. This is a very comprehensive definition, as it includes many underemployed people and those engaged in informal sector activities. The official measure of unemployment defines the labor force to include the unemployed who want work, even if they take no action to seek it. Measurement and definitional issues arise in the assessment of this measure of unemployment. A number of these problems were raised by South African commentators and by researchers from the ILO, and are considered below.

Measurement issues

  • Some respondents that are absent from work, on leave, or on strike may have been misenumerated as unemployed.

  • Mine workers in hostels, workers in defense force camps, hospitals, hotels, and those working in old age homes were excluded from the survey.

  • Respondents may mistrust the enumerators and therefore under-declare employment.

79. The concern, and corresponding bias, that people on leave or sick-leave might have been included as unemployed seems unlikely to be serious because subsequent questions on the questionnaire about activities would have clarified their status.

80. The under-enumeration of those resident in mining hostels is a well-known feature of the OHS surveys, to which the CSS has drawn attention, but results in a very small bias. It arises because enumerators have faced considerable personal risks in working in the mining compounds. But CSS calculations using the employment data reported by the mines show that adjusting for this gap would at most reduce the rate of unemployment by 1 percentage point. Similarly, the under-enumeration of those employed in army bases, hospitals, and old-age homes introduce insignificantly small biases.

81. The third measurement concern, arising from mistrust of enumerators, again appears unlikely to be a major source of bias in the official estimate of the unemployment rate. There are a number of ways in which mistrust could affect the data which have different implications for the extent and even the direction of the bias in the measure of unemployment, including (i) employed mistrusters could report themselves as wanting work but unemployed—implying that the OHS overstates unemployment rates but correctly estimates participation rates; (ii) employed mistrusters could report themselves as outside the labor force altogether—implying a less significant overestimate of the unemployment rate than under the circumstances noted above, and an underestimate of participation rates; and (iii) unemployed mistrusters might also report themselves as outside the labor force, (to avoid attracting official attention as job-seekers)—implying that the OHS underestimates both unemployment rates and participation rates.

82. Moreover, if mistrust were a major factor affecting the results, one would expect both nonsensical results and substantial inconsistencies when comparing the various surveys of unemployment. However, the detailed results of the unemployment surveys in 1993, 1994, and 1995 are consistent. For example, the broadly defined unemployment rate for South African black males was estimated at 32 percent in 1993, 34 percent in 1994, and 29 percent in 1995. These results are also consistent with the 1993 survey conducted by a consortium of independent institutions, which estimated the unemployment rate (defined in the same way as the OHS) at 34 percent. These surveys also consistently show higher unemployment rates among women, in rural areas, and among young poorly-educated people, and they indicate that unemployment rates are most severe in the Eastern Cape, Kwazulu Natal, and in Northern Province, corresponding correctly with anecdotal evidence from commentators with direct experience regarding conditions in rural areas and townships. Moreover, some of these commentators suggest that the unemployment rates in these areas are understated in the official data. Accordingly, it is not clear that mistrust fundamentally affects the official surveys, nor is it clear that it would necessarily introduce an upward bias in the unemployment rate. Finally, even if mistrust played a part in weakening survey results in the past, mistrust has likely eased since 1994, with the coming of the new political dispensation.

Definitional issues

  • Only those readily available for work, meaning those able to take up employment within one week, should count as unemployed, but the OHS puts no time limit on this.

  • Those not engaging in active job search should be regarded as outside the labor force, not as unemployed.

83. Although surveys of unemployment in OECD countries generally define the labor force to exclude those not available for work within one week, even if they express a desire to work and are seeking it, it is not clear that the OHS approach to this issue is incorrect. If it were too lax, it would tend to generate estimates of participation rates—the proportion of people of working age in the population who are in the labor force—that are implausibly high. But as discussed below, the estimated participation rates are, if anything, very low, and restricting the definition of availability for work would reduce the estimated participation rates even further.

84. In addition, the definition of availability for work should reflect the particular features of the labor market in which it is applied. In South Africa, the spatial dislocation brought about by past segregation policies means that many unemployed still live at considerable distances from job opportunities. As a result, a greater proportion of the unemployed would have to move residence to take up such opportunities than would be the case in, for example, OECD countries. This suggests that it might be appropriate to define availability for work over a longer time period. However, the way that the data are currently collected does not allow a study of the impact on the participation and unemployment rates of different definitions of availability for work.

85. Three issues should be considered in determining the appropriate statistical treatment of those who declare a desire to work but who do not engage in active search for work: (i) the impact of different treatments of participation rates; (ii) the nature of the incentives for the unemployed (who are willing to work) to engage in job search; and (iii) evidence of recruitment practices of employers.

86. The current treatment of non-searchers—to include them in the labor force estimates—seems adequate. If non-searchers were to be excluded from the labor force estimates, then participation rates, estimated at a little over 60 percent for black African males of working age, would be even lower. Compared with international participation rates among males of over 80 percent, South Africa’s rate is puzzlingly low. The high incidence of students repeating years in school could partly account for South Africa’s low participation rates relative to OECD countries, but the relatively low South African enrollment rates in tertiary institutions would tend to offset this.

87. Consideration of the incentives for active job search suggests that unemployed South Africans are less likely to engage in active job-search than their OECD counterparts, even with an identical willingness to work. Unemployment benefits in OECD countries are often contingent on active job search, creating a direct incentive to do so, but there is no such an incentive in South Africa. In addition, active search is more likely to be successful and lead to employment in the OECD—and to be remunerated in that way also—simply because the ratio of applicants to vacancies is substantially lower in OECD countries than in South Africa. This suggests that definitions of the labor force in South Africa should include non-searchers to a greater extent than would be appropriate in the OECD, and this supports the OHS approach to this issue.

88. Finally, there is survey evidence that suggests that recruitment practices by South African firms rely heavily on informal methods, buttressing the case against a requirement for active job-search in the definition of the labor force. For example, most firms that recruit production level workers use informal means of doing so: over 40 percent relying on friends and relatives of existing workers and a further 13 percent calling on former workers.22

89. These considerations suggest that the authorities are right to focus on the definition of unemployment including non-searchers. There is, nevertheless, an important gap in technical knowledge about search behavior in South Africa that needs to be closed, including through formal studies using available data sets. For instance, if non-search is correlated with the rate of unemployment defined excluding non-searchers, controlling for other factors, then this would corroborate the inclusion of the non-searchers in the labor force because it would imply that non-search behavior is caused by the low prospective returns to search activity.

90. In addition to the measurement and definitional issues considered above, there are other factors that suggest that the official OHS estimates of the unemployment rates are understated—apart from the anecdotal evidence to this effect.

91. First, as explained above, the OHS defines as employed all those who worked in any form for at least 5 hours in the preceding week, thereby including underemployed people in the employment estimate. This number of hours is below the floor that is implicit in OECD surveys of unemployment because the unemployed are unlikely to accept employment with limited work hours if the associated earnings do not compensate for the loss of unemployment benefits. This puts an implicit floor on the number of hours of work that they will accept, which will generally substantially exceed 5 hours per week. Given the way the data are currently collected in South Africa, it is not possible to assess the extent to which the estimated unemployment rate would change by raising the minimum number of hours of work per week, but it is clear that the official measure of unemployment rate understates actual unemployment.

92. Second, as the CSS staff has noted, informal settlements and squatter camps, in which unemployment rates are thought to be extremely high, are under-enumerated. This is because the sampling frames for the OHS surveys were based on the 1991 census, and many of the informal settlements mushroomed after that time. The CSS statisticians consider that this downward bias is thought to be about 1 percentage point, roughly equal the opposite bias from undercounting in the mines, army bases, etc.

93. On balance, therefore, taking into account the institutional characteristics of the South African labor market, the officially estimated unemployment rate does not appear to be subject to significant upward measurement biases. If anything, there are both measurement and definitional concerns which suggest that the official estimates of unemployment could be too low.

D. Formal Employment

94. The principal formal employment series published by the SARB, based on data collected by the CSS, are shown in Figure 6. As with the unemployment data discussed above, a number of concerns with these estimates have been raised, including:

Figure 6
Figure 6


Citation: IMF Staff Country Reports 1997, 082; 10.5089/9781451840933.002.A002

Source: South Africa Remove Bank.
  • Adjustments to the data were not applied retroactively, rendering time series inconsistent.

  • The monthly data are drawn from firms that reported to the most recent industrial census, so new entrants are undercounted, the samples used for the industrial censuses have been updated irregularly, data from the TBVC states have been doubtful,23 and the response rate to the census has been around 70 percent with “busy” firms least likely to respond.

  • There are discrepancies between the sectoral decomposition of employment between the OHS and the CSS employment series, with the estimates of the latter being lower for all sectors than those of the former.

  • Illegal immigrants are likely to be excluded from the employment series.

95. The SARB data exclude agriculture. This avoids a number of difficulties with the inconsistent adjustments to long time-series data on agricultural employment, particularly regarding female employment. However, this means that total employment is understated and that the overall employment growth trends may be under- or over-stated in the employment series, depending on whether employment in agriculture is rising more or less rapidly than employment elsewhere in the economy.

96. It is not clear that the sample of firms being used by the CSS to estimate employment is outdated. For the industrial censuses—which are conducted every three years—the CSS updates its sample of firms by using its own business register, the Registrar of businesses, VAT lists, telephone books, Industrial Councils (IC), and Federations, etc. For the inter-census data, the CSS continually updates the sample of firms surveyed on the basis of new information on company registrations. Non-respondents are pursued, ultimately through the courts, which normally yields the necessary returns. The comprehensive list of sources being used, and the continuous updating of the sample suggests that, at the least, the methodology underlying the data is not systematically liable to induce undercoverage of new firms.

97. There could still be concerns about the coverage of the so-called informal sector in the non-agricultural employment time series data. But this series understates overall employment growth only if employment in the unmeasured (informal) sector is rising faster than the measured sector or vice-versa. In the absence of compelling evidence concerning which of these biases prevails over any given time period, the nonagricultural employment series remains the best available measure of long-term overall employment trends. It is therefore important to support the CSS’s efforts to ensure that the coverage of firms is as comprehensive as possible.

98. Another important issue is related to the smaller unmeasured firms because they may have more marked pro-cyclical employment behavior, as international experience would suggest. In these circumstances, even though the employment series correctly reflect the underlying secular trend in employment, the cyclicality of employment could be understated by this series. On this basis, employment may have fallen even faster in the early 1990s than is suggested by the nonagricultural employment series, while employment growth may have been stronger than recorded since 1994.

99. The SARB has made considerable efforts to adjust the CSS employment series to include employment in the TBVC states (Transkei, Bophuthatswana, Venda, and Ciskei) on a consistent basis in their data. In contrast, employment in these areas has only been included in the CSS estimates of South African employment in recent periods (giving rise to a step jump in their employment series). Accordingly, using the employment series of the SARB to establish short-term trends in the growth of employment would seem more appropriate than using the data of the CSS because they avoid the step-jump generated by the CSS treatment of the TBVC states. Nevertheless, the very considerable weaknesses in the employment data for the former TBVC states make such adjustments by the SARB problematic, and it is unclear if the “ghost workers” in the public sector in these states offset the underenumeration of employment elsewhere in these states.

100. Finally, it would appear probable that the discrepancies between the stock of employment from the OHS data and those from the employment series of the SARB can largely be accounted for by so-called “informal” activities as opposed to unenumerated “formal” employment. These discrepancies can be large, however. So while it may be appropriate to identify overall employment growth rates from those employment series—on the basis that employment growth (or contraction) in the sectors excluded from the employment data does not deviate substantially from that in the included sector—the employment series should not be used to estimate unemployment rates at any point in time.24 The latter would substantially overstate unemployment. As discussed above, the official measurement of unemployment rates derived from the OHS would be more appropriate for this purpose.

E. Summary and Conclusions

101. Many factors appear to explain the high rates of unemployment observed in South Africa, and wage inflexibilities appear to play a significant role. Further basic statistical research is required to clarify the role of wage inflexibilities more precisely, and the Department of Labor may be best placed to establish the necessary reporting mechanisms.

102. The particular role of wage inflexibilities in youth unemployment also requires further research, to better establish its importance. The available evidence—on reservation wages, employer’s hiring preferences, and the composition of training provision by the private sector—points in this direction, and the GEAR proposal to scale down trainee wages may be a useful means of addressing this particular cause of youth unemployment.

103. The suggestion that the official data on unemployment rates substantially overstate actual unemployment is hard to defend. Given the very high unemployment rate among black South Africans, the authorities have appropriately defined job-creation as a key goal of economic policy.

104. The methodology underlying the nonagricultural employment time series also appears sound, and it is widely recognized that they do not cover the informal sector. They therefore provide a meaningful indication of secular employment trends although they may understate the cyclicality of employment.


  • Boccara, Bruno and Peter Moll, 1997, “Labor Market Flexibility and Industrial Councils in South Africa,” African Region, World Bank, work in progress.

    • Search Google Scholar
    • Export Citation
  • Standing, Guy, John Sender and John Weeks, 1996. Restructuring the Labor Market: The South African Challenge. An ILO Country Review. Geneva: International Labor Office.

    • Search Google Scholar
    • Export Citation

See Mr. S. Pityana, quoted in Business Day, 24 January, 1997.


These were known as Industrial Councils prior to the recent Labor Relations Act.


Under the former political dispensation, employment data in these states were excluded from CSS estimates of South African employment.


For this reason, an earlier study of employment issues in South Africa (SM/95/9) described the difference between formal employment and the labor force as the “formal employment gap,” not as unemployment.

South Africa: Selected Issues
Author: International Monetary Fund