23_B_Kingdon_Unemployment in South Africa, 1995–2003 Causes Problem and Policies,

Unemployment in South Africa, 19952003: Causes,Problems and Policies

Geeta Kingdona* and John KnightbaCentre for the Study of African Economies,

University of Oxford, UKbDepartment of Economics, University of Oxford, UK

This paper examines an issue of overwhelming importance in SouthAfricaunemployment and its rise. It explains the factors behind thesharp rise in unemployment in the post-apartheid period, investigates therole of labour legislation and the system of labour market governance, evalu-ates the impact of the governments active labour market policies, identifiesthe knowledge gaps about the functioning of the labour market and drawssome policy prescriptions. It analyses unemployment using householdsurveys spanning 19952003 and explains the rise in unemployment bythe slow growth of the economy, and thus slow growth in the demand forlabour relative to the rapidly growing supply, together with labourmarket inflexibility. The paper argues that if unemployment is to betackled, it is crucial to pursue a set of policies that promote South Africasrate of economic growth to promote job-creation, and also that labourmarket regulations require reconsideration, giving greater weight to theconcerns of employers and investors, and to the interests of the unemployedand informally employed poor who are beyond the reach of the labour insti-tutions but can be hurt by them nevertheless. It highlights that lack of appro-priate data hinders analysis of important aspects such as entry into, exitfrom and duration of unemployment. Finally, the paper appeals for investi-gation of how active labour market policies to address unemploymentsuchas public works programmes, skills training programmes etc., formulatedlargely in the absence of local evidencehave performed.

JEL classification: J64, O17

# The author 2007. Published by Oxford University Press on behalf ofthe Centre for the Study of African Economies. All rights reserved.For permissions, please email: [email protected]

* Corresponding author: Geeta Kingdon; e-mail: [email protected]

JOURNAL OF AFRICAN ECONOMIES, VOLUME 16, NUMBER 5, PP. 813848doi:10.1093/jae/ejm016 online date 3 August 2007

1. Introduction

South Africa has one of the most interesting labour markets in theworld. Its sharp segmentation, high unemployment and low non-farm informal sector employment make it an international outlier.Moreover, developments in the labour market may well hold thekey to South African prosperity or penury. It is from the labourmarket that the income benefits from growing labour scarcity, orthe threat to social and political stability from growing unemploy-ment and underemployment, could emerge.

Our primary concern in this paper is with unemployment andthe informal employment that often disguises unemployment.However, in order to understand these phenomena it is necessaryto take a broader view and to examine a range of labour marketvariables and relationships. We make use of the official representa-tive household surveys that span the period under review. Theearliest such annual survey is the October Household Survey 1995(OHS95) and our end-date is provided by the Labour ForceSurvey, September 2003 (LFS03). Comparisons over time are com-plicated by subtle changes in questions, definitions and sampling,and by reweighting in the light of new census data.1 This hazardis somewhat diminished because the purpose of the paper is notto examine year-to-year changes but to take a long run view oflabour market trends.

In its starkest terms, the issue is depicted in Table 1. The problemis that the economy is unable to absorb productively all the currentlabour force or all the increment to the labour force. In the near-decade after the advent of democracy (19952003), the narrowlabour force grew by 4.6 million and the broad labour force (includ-ing the non-searching unemployed) by 6.3 million (by 4.2 and 4.8%per annum, respectively). By contrast, over the same period, wageemployment rose by only 1.3 million (1.8% per annum), self-employment grew by 0.7 million (5.1% per annum), and narrowand broad unemployment grew by 2.6 and 4.3 million, respectively(both above 9% per annum). Over that period the unemploymentrate rose from 17 to 28% on the narrow definition and from 29 to42% on the broad definition. South Africa now has one ofthe highest rates of unemployment in the world even on the

1 Such problems of the surveys are discussed in Altman (2002), Bhorat (2002),Casale and Posel (2002), Devey et al. (2003), and Klasen and Woolard (1999).

814 Geeta Kingdon and John Knight

official narrow (butas we shall explainpotentially misleading)definition.

The growing divergence between labour supply and demandinevitably had a depressing effect on market-determined realwages. However, the wage sector, and in particular the formalwage sector, was relatively protected, so pushing the burden ofadjustment onto the self-employment sector, especially that partof it which had relatively free entry. Whereas real wages fell by1.6% per annum over the period 19952003 (and formal sectorreal wages by 0.5% per annum over the period 19972003), self-employment incomes fell dramatically in real terms, by 11.4% perannum (and informal sector real wages by 7.8% per annum overthose 6 years). The growth of large parts of the informal sectorwith underemployed people eking out a livingis a sign oflabour market failure rather than of success.

The theoretical framework suggested by the evidence is that of atwo-sector model characterised by labour market segmentation.The formal sector wage is set exogenously at a level above themarket-clearing wage and the residual labour force is distributed

Table 1: Summary of Labour Market Outcomes, 19952003

1995OHS

2003LFS

Change000

Change% p.a.

Labour force, narrow (000) 11,628 16,192 4,564 4.2Labour force, broad (000) 13,667 19,954 6,287 4.8Wage employment (000) 8,231 9,509 1,278 1.8Self employment (000) 1,421 2,111 690 5.1Unemployment, narrow (000) 1,976 4,570 2,584 11.0Unemployment, broad (000) 4,015 8,332 4,317 9.6Unemployment rate, narrow (%) 17 28 11 Unemployment rate, broad (%) 29 42 13 Real earnings in wage employment,

2000 prices3,191 2,805 2386 21.6

Real earnings in self employment,2000 prices

6,866 2,610 24,256 211.4

Source: Tables 4 and 6 below.

Unemployment in South Africa 815

between informal sector employment and unemployment. We mustset this picture against the view that, given enough competition, thelabour market will clear at the natural rate of unemployment,voluntarily chosen, and so unemployment should not go on risingand rising. Why has it risen? The superficial answer is labourmarket rigiditieswhether they are due to trade unions, or bargain-ing councils, or profit-maximising interventions by employers (theefficiency wage and labour turnover arguments as to why wagesmay be set above the market-clearing levels). In reality, however,it is the sheer speed of divergence between the growth of thelabour force and the growth of formal sector employment (thedivergence over the period 19972003 being nearly 4% perannum) that is the underlying cause. This puts great strain on thenecessary downward wage adjustment process and inevitablymeets resistance. Nevertheless, the wage rigidities of the formalsector exacerbate the problem by narrowing the segment of thelabour market on which the burden of adjustment is placed.Given constraints in the opportunities for profitable informalsector activities, the result has been rising unemployment.

Our general argument is that the rise in unemployment is bestexplained by the rapid growth in the labour force relative to thegrowth of formal sector employment, as a result of which the bur-geoning residual labour force was absorbed either into the informalsector, which might disguise unemployment, or into open unem-ployment. Accordingly, we examine the growth of each of thesevariables in more detail: in Section 2 the labour force, in Section 3formal sector employment and in Section 4 informal sector employ-ment. Our main topic, unemployment, is covered in Section 5.Section 6 reviews the active labour market policies of the SouthAfrican government, and Section 7 draws conclusions.

2. Labour Force Growth

The South African labour force has grown remarkably rapidly: over4% per annum is extremely unusual in international terms (Table 2).There are three possible explanations: in-migration, rapid naturalincrease in the number of working-age people, and increasedlabour force participation. In-migration is difficult to quantify asmuch of it is informal or illegal. Only part of the explanation isa rapid rate of increase in the adult population, whether due to


natural increase or net in-migration. We see that this grew by 2.7%per annum over the period 19952003 whereas the labour forcegrew by 4.2% per annum on the narrow definition of unemploy-ment and by 4.8% on the broad definition. The labour force partici-pation rate rose from 48 to 54% (narrow) and from 56 to 67%(broad). We cannot rule out the possibility that labour forcegrowth is exaggerated by changing definitions, changing coverage,or sampling errors. Nevertheless, the labour force appears to havegrown at a daunting pace.

Table 3 shows that the growth of the labour force has been thegreatest among the African population group. The female partici-pation rate rose by fully 15% points and the male rate by over 5%points in a mere 8-year period.2 The increase in participation rates

Table 2: South African Population of Working Age, Not Economically Active, LabourForce and Labour Force Participation Rate, 19952003

1995 OHS 2003 LFS Change 19952003

000 % p.a.

Population 1565 (000) 24,232 29,917 5,685 2.7Narrow measureLabour force (000) 11,628 16,192 5,464 4.2Not economically active (000) 12,604 13,725 1,121 1.1Labour force participation rate (%) 48 54 6 Broad measureLabour force (000) 13,667 19,954 6,287 4.8Not economically active (000) 10,565 9,963 2602 20.7Labour force participation rate (%) 56 67 11

Note: Labour force participation rate (labour force)/(working age population).Source: Table 6 below.

2 The Labour Force Surveys made a greater effort to capture informal work than didthe previous October Household Surveys. Thus, some persons who worked veryinformally and who may previously have been classified as out of the labour forcemay now have counted as labour force participants. However, such a changewould not lead to an increase in unemployment. The fact that unemploymentrose so much suggests that the increase in labour force participation rates is notmainly due to the better capture of informal work. Moreover, as much as 13 of


is due partly to the lifting of apartheid restrictions on movement tourban areas and the new possibilities of employment that this wasperceived to open. The better occupational attainment then possiblefor non-white groups and for women (partly due to employmentequity legislation) is likely to have raised the expected returns totheir employment. Educational levels have risen and participationrates typically increase with educational level, particularly so forwomen. The significantly greater increase in the female than themale participation rate appears to be due to a decline in womensaccess to male income on account of increased unemploymentamong males, the HIV epidemic creating more single parent house-holds, and increased incidence of female headship resulting fromchanges in household structure.3

The figures for the recent period 20003 (using the LFSs) presenta less daunting picture. In this period, the narrow labour force grewby only 0.8%, and the broad labour force by 2.6%, both expressedper annum. At least part of this recent slow-down was due to HIV/AIDS. Looking to the future, it is necessary to consider the likely

Table 3: Labour Force Participation Rates (%), 19952003

Males Females Total

1995 2003 Change% points



African 59.9 69.2 9.3 44.3 63.1 18.8 51.8 65.9 14.1Coloured 72.1 75.7 3.6 55.4 64.8 9.4 63.5 69.8 6.3Indian 75.7 77.3 1.6 39.2 51.9 12.7 57.2 64.3 7.1White 72.1 79.8 7.7 47.9 61.5 13.6 59.9 70.4 10.5All 65.9 71.1 5.2 47.8 62.8 15.0 56.4 66.7 10.3

Source: OHS95, LFS03.

the 15 percentage point increase in the female participation rate occurred beforethe introduction of the LFS (Casale and Posel, 2002).

3 Casale and Posel (2002) show that, between 1995 and 2002, the percentage ofworking-age women living with at least one employed male in the householdfell from 53 to 44%, the incidence of female headship increased from 28 to34%, and the percentage of women reported to be married fell from 40 to 35%.


effect of the HIV/AIDS epidemic. By way of illustration, we draw onthe results of a particular study (Quattek, 2000). For the population asa whole, HIV infection rates are forecast to peak at 17% in 2006 andAIDS-related deaths about 5 years later, when the AIDS mortalityrate will be 1.8%. However, the peak HIV infection rate will behigher for economically active people, at 26%. In the absence ofAIDS, the growth of the labour force would be 1.9% per annumover the decade 200313, but in the AIDS-inclusive scenario it is fore-cast to be 1.1% per annum. Because the rate of growth of the economyis sure to be adversely affected in various complex ways by theepidemic, we cannot simply subtract the projected slower growthof the labour force in order to measure the effect on unemployment.Any projection would have to be based on numerous untestableassumptions. The study by Quattek, (2000), which projects employ-ment growth as slower by 1.2% per annum over the decade 200313on account of the disease, and unemployment to be 9.2% pointslower than it would otherwise be in 2013, suggests the possibleorders of magnitude. However, Nattrass (2003) cautions that thevarious models of the impact of HIV/AIDS in South Africa lead todiffering or even conflicting predictions for the labour market.

3. Formal Sector Employment

A distinction can be made between the formal sector (comprisingfirms that are formally registered) and the (unregistered) informalsector. The insider-outsider theory of labour economists is helpfulhere: formal sector employees can be regarded as insiders, andresidual workers, comprising those in the informal sector (whichserves as a residual labour sponge) and the unemployed, asoutsiders. South African insiders fall within the scope of the indus-trial relations regulations, including recognition of trade unions andcollective bargaining, the right to strike, protection against dismis-sal, and minimum standards concerning hours of normal and over-time work, minimum wages, and minimum leave provisions. Thereis provision for exemptions, generally granted on the basis of smallscale and inability to pay. The informal sector workers fall outsidethe labour regulation system, and generally receive much lowerincome. The simple ratio of monthly income per worker of theformal to the informal sector in 2003 was 3.4 to 1 (Table 4);Kingdon and Knight (2004a) report that in 1993 this ratio was


3.5:1 but that after standardising for the different average personalcharacteristics of workers in formal and informal sectors, theformal-informal earnings gap fell by between 50 and 64% depend-ing on whether OLS or selectivity-corrected earnings functions wereused, i.e., the ratio was roughly halved, to about 1.75:1. Inevitably, insome respects, the two groups have conflicting interests and are incompetition with each other.

The formal sector in South Africa possesses the characteristics of amiddle-income country, whereas the informal sector has those of apoor, less developed country. Like other middle-income countries,especially those with a mineral economy, the formal sector has acomparative advantage in natural resource-intensive, or capital-intensive, or skill-intensive manufacturing. It cannot compete withthe East Asian economies in low-wage unskilled-labour-intensivemanufacturing. Moreover, any successful labour-intensive manu-facturing by the informal or small-scale sector, leading to itsgrowth, is likely to lead to its formalisation.

Table 4: Real Earnings (Formal and Informal Sector; Wage and Self-employment; andTotal) 1995, 1997 and 2003

1995 1997 2003 Percentage change

Total Per annum

19952003

19972003

19952003

19972003

Non-agriculturalFormal 3,339 3,241 22.9 20.5Informal 1,523 941 238.2 27.7Large scale formala 15.4 10.8 1.8 1.7Wage employment 3,190 2,805 212.1 21.6Self employment 6,866 2,610 262.0 211.4Agriculture 690 608 211.9 21.6Total 3,014 2,360 221.7 23.0

Source: Adapted from Casale et al. (2005, Table 3).aFrom Annual Survey of Employment and Earnings.


There are two sources of information on formal sector employ-ment: one based on the OHS and LFS household surveys, and theother based on the enumeration of employees in enterprisesurveys. Both are prone to be inaccurate. The former can sufferfrom sampling error and the need to adjust figures periodically inthe light of new population census estimates. The latter sufferfrom non-reporting and may miss casual employees of registeredenterprises. Table 5 shows formal sector non-agricultural employ-ment in 1997 (the earliest year for which comparable figures areavailable) and for 2003. It is based on OHS and LFS sources.Formal sector employment increased by 702,000, implying annualaverage rate of growth of 1.6%. This figure is probably moreaccurate than the one obtained from the enterprise surveys, whichcapture employment only in the large- and medium-size businessesand show a fall in employment over that period. To some extentthe trends were the result of structural changes in the economy:the long run decline of mining with its high labour-intensity, thetrend towards capital-intensive and skill-intensive activities, andthe capital- and skill-using nature of technical progress in various

Table 5: Formal and Informal Sector Employment, Wage- and Self-Employment and TotalEmployment, 1995, 1997 and 2003 (000s)

1995OHS

1997OHS

2003LFS

Change000

Change% p.a.

Non-agriculturalFormal 6,839 7,541 702 1.6Informal 1,750 2,838 1,088 8.4Wage employment 7,845 8,684 839 1.3Self employment 703 1,694 991 11.2Agricultural 944 1,010 66 0.8Total 9,631 11,579 1,948 2.3

Notes: For 2003, the authors use the March round of the LFS data. Estimates for 1995and 1997 are weighted using the 1996 population weights and for 2003 using the2001 population weights. Domestic service work is treated as informal sectorwage employment.Source: Casale et al. (2004), Table 1.


industries (Alleyne and Subramanian, 2001; Seekings et al., 2004).Employment in the informal sector grew more rapidly than in theformal sector, by 1,088,000, equal to 8.4% per annum, but from amuch smaller base. Similar results are obtained by comparingwage employment and self-employment over the period 19952003. Wage employment grew by 839,000, or by 1.3% per annum.Self-employment grew more, and faster: by 991,000, or by 11.2%per annum (again, from a smaller base). Agricultural employmentless accurately recorded, especially in subsistence agricultureisshown separately; it rose very slowly. Overall, total employmentincreased by almost 2 million jobs, i.e., by 2.3% per annum, overthe 8 years. However, we see that the formal and wage-employingsectors grew disappointingly,4 and that they lost ground to thelower-paying informal and self-employment sectors which, intheir free-entry segments, served as a residual sponge.

Table 4 shows how real earnings behaved over the relevantperiod. We see the real earnings of wage employees fell onlyslowly (by 1.6% per annum) and those of formal sector workerseven more gently (by 0.5% per annum). In fact, the formal sectorfall was confined to small employers: data obtained from VAT-registered firms (effectively large and medium size employers)show a rise of 1.7% per annum. We return to Table 4 for the analysisof the informal sector in Section 5.

Collective bargaining appears to produce real wage resistance.Kingdon et al. (2005) show that the union wage premium increasedin South Africa between 1993 and 1999: even though unionisedworkers wages fell a little in real terms, non-unionised wages fellmore sharply. Thus, unionised workers are more insulated thannon-unionised workers from the wage-depressing effect of labourmarket competition. This is consistent with another form of wageinflexibility noted by Kingdon and Knight (2006b). They examinedthe sensitivity of wages to unemployment across space by estimat-ing wage curves across the neighbourhoods in South Africa.Whereas union wages were found to be unresponsive, non-union

4 Although largely overlapping, wage employment does not equate with formalsector employment, nor self-employment with informal sector employment.For instance, casually employed workers may report themselves as workingfor a VAT-registered employer and thus count in the surveys as formal sectorworkers (Muller and Esselar, 2004).


wages were sensitive, negatively, to variations in local unemploy-ment rates using SALDRU 1993 survey.

It is apparent from Table 4, and also from estimated earningsfunctions, that much of the formal sector pays well above thelevel of competitively determined market wages. Do these relativelyhigh wages adversely affect the level of employment in the formalsector? The wage-employment elasticitythe proportionate fall inemployment that occurs in response to a unit proportionate risein the wage rateis therefore an important issue. It has attractedconsiderable research attention (surveyed in Bhorat et al., 2002).A World Bank study (Fallon and Lucas, 1998) showed a long-runweighted mean wage-employment elasticity for black workers inthe formal sector of 0.71 (after 3 years, when adjustment waslikely to be complete). This implies that a 10% increase in thewage rate results, in due course, in a 7% decrease in black employ-ment. Other researchers (Bowles and Heinz, 1996; Bhorat andLeibbrandt, 1998; Fields et al., 1999) have produced similar results.

The implication of these studies is that there is indeed a difficultpolicy trade-off between wages and employment. The real wageresistance shown within the formal sector over the last decade hashad a harmful effect on employment. However, it is also clear fromthe estimated elasticities that greater wage flexibility would havebeen unable to prevent the rise in unemployment. The problem isessentially the dynamic one of inadequate growth of formal sectorlabour demand in relation to the growth of labour supply. Thoseunable to find employment in the formal sector either remain inunemployment or enter informal sector employment.

4. Informal Sector Employment

Being the residual economic sector into which workers can in prin-ciple move, the informal sector can be predicted to have two fea-tures. First, it should have increased rapidly in response to thegrowing divergence between labour supply and formal sectorlabour demand. Secondly, its free-entry parts should contain agood deal of underemployment and poverty.

The informal sector, as defined by Statistics South Africa(StatsSA), absorbed a mere 16% of the labour force in 1997.Although the figures are unreliable (Devey et al., 2002), the informalsector shows growth in recent years, with the proportion rising to


19% in 2002. South Africa nevertheless remains an internationaloutlier in the size of its informal sector. With the unemploymentrate (on the narrow definition, for comparability) then 29%, itsratio of informal sector non-agricultural employment to unemploy-ment was a mere 0.7. By contrast, the averages for Sub-SaharanAfrica, Latin America and Asia are 4.7, 7.0 and 11.9, respectively(Kingdon and Knight, 2004a, Table 1, drawn from Charmes,2000). It is possible that the definition of the informal sector usedin South Africa is narrower than in other countries. StatsSA includesas informal worker persons in own-account activities and thoseworking for employers who are not VAT-registered. However,some employees of VAT-registered firms might also be informal inthe sense that they are not well protected from market forces.Defining formal employees as only those receiving good treatmentfrom employers (paid leave, pension rights and unemploymentinsurance) or, failing that information, a formal contract, Posel(private correspondence) produced a formal/informal employmentsplit of 4159%, to be compared with the StatsSA split of 7228%for LFS03. However, even with this adjustment, South Africawould still be an outlier: informal sector workers would still consti-tute only 34% of the 2003 labour force, and the ratio of informalemployment to narrow unemployment would be 34: 28, i.e., 1.2and to broad unemployment 34: 42, i.e., 0.8.

It is argued in Section 5 below that the bulk of unemployment isinvoluntary rather than voluntary, and that informal sector workershave both higher income and higher subjective well-being than dothe unemployed. The implication is that unemployed workersshould want to enter the informal sector. If they fail to do so, theimplication is that there are barriers to entry. What are these bar-riers? There is a paucity of evidence on whether the informalsector is a free-entry sector. In a survey of 500 informal sector opera-tors in the Johannesburg area in 1999, respondents listed crime, lackof access to credit, lack of access to infrastructure and services, andneed for training as the top four constraints on their businesses.Chandra et al. (2002, pp. 26, 30) find that the informal sectoroperators had required substantial start-up capital (averagingover 2.5 times the average monthly earnings in the sample). Newsmall businesses had to rely on their own financial resources:there was very little access to formal or even informal credit. The1999 survey suggests that government support continued to be


inadequate, particularly in relation to crime prevention, investmentin infrastructure, and the provision of credit and training facilities(Chandra et al., 2002, Table A2.6, pp. 18, 20, 445). Thirty percentof the informal businesses had been victims of crime in the previousyear, but the number of respondents expressing concern was doublethat figure. The Johannesburg survey found that 81% of all informalsector operators (and 90% of the self-employed non-employerswithin that group) had never received any business assistance ortraining. The lack of training reflected the high cost: the fewowners who had been trained had paid on average three timesthe average monthly earnings of the sample for their training.Sixty percent of the operators did not have access to the smallbusiness support centres that had been established by central andlocal government. Xaba et al. (2002, p. 25) argue that the SouthAfrican governments avowed support for small, medium andmicro-enterprises (SMMEs) is concentrated on the formal sectorand neglects the informal sector. However, such problems exist inmost developing countries and do not explain why the SouthAfrican informal sector is relatively so small.

One thing that might distinguish South Africa is the degree ofeffectiveness with which labour regulations are enforced. Labourmarket institutions such as Bargaining Councils and Wage Boardsset sectoral minimum wages and stipulate working conditions inmany industries, and these are applied to all firms in the industryand region, irrespective of size, via the extension provision.There are serious penalties for flouting the agreements of these insti-tutions (Moll, 1996; Nattrass, 2000). Such provisions impose aburden of high labour costs on small firms and it is likely thatthey do seriously inhibit the entry and growth of such firms. Thisis one explanation for the large average size of firms in South Africa.

These institutional features may inhibit small firms but theyshould not inhibit individual entrepreneurship, i.e., owner-operators. The lack of African self-employment is partly a legacyof apartheid. The apartheid system had repressed the informalactivities of black South Africans through such restrictive legislationas the Group Areas Act, harsh licensing, strict zoning regulations,and effective detection and prosecution of offenders. Bouts ofslum clearance and other periodic attacks on the illegal spaceswithin which informal enterprise thrived, served to rid SouthAfrican cities of informal sector niches that were construed as


hazardous to public health and stereotyped as unsightly and unsa-nitary (Rogerson, 1992). Although these restrictions have been pro-gressively lifted since the mid-1980s, there were lingering licensingcontrols and restrictive bye-laws in many urban centres in the late1990s. Moreover, repression and disempowerment of Africansunder apartheid would have inhibited the development of entrepre-neurial and social skills and of relevant social networks. Thesefactors are important for confidence in entering the self-employedsector and for success in it.

Several authors note that many activities in the so-called informalsector of developing countries are highly stratified, requiring skills,experience and contacts, with identifiable barriers to entry. Forexample, petty trading often has highly structured labour andproduct markets with considerable costs of entry. Banerjee (1986)found that even in urban India, with its large self-employmentsector, entry is not easy. Even when skill and capital are not required,entry can be difficult because of the presence of cohesive networks,which exercise control over location and zone of operation.Support for these ideas also comes from Latin America. For instance,Maloney (1999, 2002) argues that the informal sector workers tend tobe older and to enter from the formal sector after they have accumu-lated knowledge, capital and contacts, and that lack of experienceand capital are barriers to entry that deter participation in the infor-mal sector. If such barriers are greater in South Africa, this mighthelp to explain why the South African informal sector is so small.

There is considerable scope for the further development of a pro-ductive informal sectorcomprising medium and small-scaleenterprises (SMMEs). These have the advantage of often beingmore labour-intensive than large-scale enterprises, and of promot-ing black economic empowerment. The existing policy support toSMMEs in South Africa has focused on the provision of financeand facilities. However, while finance removes one obstacle, theSMME sector faces another, probably more important, obstacle toits development, namely unfair competition. The extension pro-vision, which requires each Bargaining Council agreement aboutminimum wages and working conditions to be extended to allemployers in the industry and area, puts SMMEs at a disadvantage,with crippling labour market burdens. Removing this provisionwould provide a boost to the development of the productive partof the self-employment sector in South Africa.


Informal sector non-agricultural employment grew rapidly inrecent yearsinformal sector employment rose by 8.4% and (over-lapping) self-employment by 11.2% per annum, though from asmall base (Table 5). This was mainly due to the residual labourforce crowding into activities such as petty trading and crafts,which are relatively free to enter. Support comes from Table 4: inthe non-agricultural sector, real earnings in informal employmentfell by 7.7% and in self-employment by no less than 11.4% perannum. Part of the recorded increase in employment and of thedecrease in real earnings in the informal sector is the result of theattempt made in the LFS to extend the coverage of the informalsector to more marginal economic activities. For the most part,however, the burgeoning of informal activities in the samplesurveys should be taken as evidence of the expansion of that partof the informal sector which is relatively easy for the unskilled toenter, and thus as a sign of economic failure rather than success.

5. Unemployment

5.1. The Definition of Unemployment

In South Africa, two different concepts of unemployment are usedroutinely: the strict (narrow) and the expanded (broad) definition.The narrow definition applies a job-search test whereas the broaddefinition accepts as unemployed those who did not search forwork in a 4-week reference period but who report being availablefor work and say they would accept the offer of a suitable job. In1998, the narrow concept was declared the official definition ofunemployment and it is now the one generally used. Yet it hasbeen argued that the broad measure of unemployment is a moreaccurate reflection of joblessness than the narrow measure inSouth African conditions.

Kingdon and Knight (2006a), investigating the issue, use threenew approaches to test whether, in conditions of high unemploy-ment, the searching and non-searching unemployed states are dis-tinct. They find, first, that in South Africa the non-searchingunemployed are, on average, significantly more deprived than thesearching. The fact that they are not better-off casts doubt onthe interpretation based on tastes (lack of desire for employment)and favours the interpretation that active search is discouraged


(low prospective returns to search). This view is supported byevidence from a job-search logit model, which suggests that thesearch is hampered by poverty, by the cost of job-search fromremote rural areas (almost uniquely in the world, South Africa hasa higher rural than urban unemployment rate), and by high localunemployment. Secondly, the non-searching unemployed are notanymore happier than the searching unemployed: their unemploy-ment depresses their subjective well-being to the same extent as isthe case for the searching unemployed. Thirdly, evidence on thewage-unemployment relationship indicates that local wage deter-mination takes non-searching workers into account as genuinelabour force participants. The searching and the non-searchingunemployed are very close in terms of potential labour supply.These findings indicate that lack of search is due to discouragementand constraints such as poverty. The de-emphasis of the broadmeasure in policy circles may be because even narrowly measuredunemployment is a large enough problem in itself, or because thesearching unemployed are viewed as more deserving of policyconcern.

5.2. The Nature of Unemployment

The high level of joblessness begs the question: why do the mass ofthe unemployed not join the informal sector, as in most other deve-loping and middle-income countries. If the informal sector is anopen-entry sector then, in principle, persons not entering it maybe considered as voluntarily unemployed. The dominant view ofunemployment in developing countries is indeed that much openunemployment is due to search and is voluntary (Harris andTodaro, 1970), i.e., people choose to remain jobless while theysearch for a good job. If search for a formal sector job from theunemployed state is more efficient than from informal employment,those able to afford unemployment remain openly unemployed andsearch. However, the poor cannot afford it. Thus, if most unemploy-ment in the economy is of the voluntary search variety, the relation-ship between unemployment and household income is likely to bepositive because the well off will choose search unemployment butthe poor will enter informal sector employment.

Kingdon and Knight (2004a) find little support for the idea thatSouth Africans choose to be unemployed. The unemployed are,


on average, substantially worse off than the informally employedboth in terms of income and expenditure and in terms of a range ofindicators of well-being. This contradicts the search, or luxury,unemployment interpretation, whereby higher income raises theincentive to remain searching and reduces the incentive to obtaininformal employment. Moreover, they find that the unemployedare substantially and significantly less happy than informallyemployed people, suggesting that their unemployment is not dueto choice. Finally, the average duration of uncompleted spells ofunemployment (2.2 years) is too long to sustain a person insearch unemployment. The fact that the unemployed are signifi-cantly poorer and unhappier than the informally employedsuggests that the failure of unemployed people to enter self employ-ment is due to some barriers to entry into the informal sector.

Several papers have examined the labour supply response ofadults in South African households that receive state old age pen-sions, pensions being an important source of income in manypoor households. While Bertrand et al. (2003) find that pensionincome is associated with reduced labour supply of prime-agemembers of the household, Posel et al. (2006), using the same dataand methods but widening the definition of the household toinclude non-resident members, find that the pension is positivelyassociated with labour supply. They conclude that pensionincome facilitates the migration of household members to placesof employment. Moreover, Edmonds et al. (2005) find that whenan older woman becomes eligible for a pension, there is a fall inthe number of co-resident women in their thirties. Thus, thepension enables the household to send out members to search for,and possibly find, employment. These results are consistent withthe findings and interpretation of Kingdon and Knight (2006a)that unwillingness to search reflects a privately rational decisionbased on the perception of the benefits of searching in relation tothe costs, and the ability to fund the search activity. Overall, theresearch on pensions does not provide support for the view thatnon-labour income encourages voluntary unemployment or with-drawal from the labour force. The fact that prime-age males tendto attach themselves to households in receipt of pension income(Klasen and Woolard, 2000) may simply reflect a propensity forendogenous household formation in conditions of hardship andinvoluntary unemployment.


It is arguable that some workers would be willing to accept a jobonly at an unrealistically high wage, and that this would suggestthat they are voluntarily unemployed. Nattrass and Walker (2005)examine the proposition that the unemployed price themselvesout of employment by requiring high wages, i.e., that their reser-vation wages exceed the wages on offer. In a study of a workingclass area of Cape Town they find that workers reservation wagesare not out of line with the wages that they are predicted to earnon the basis of their characteristics: the ratio of reservation wagesto predicted wages averaged 0.85 for unemployed persons and itexceeded 1.0 for only 1.3% of them. This is further evidenceagainst the view that unemployment is voluntary.

5.3. Is High Unemployment a Mirage?

Commenting on the official estimate of narrow unemployment,which then exceeded four million (giving an unemployment rateof 27%), President Mbeki (2005) has stated that this is such alarge number of people that nobody could possibly have missedthe millions that would be in the streets and village paths activelylooking for work in all likely places of employment. It, there-fore, seems quite unlikely that the StatsSA figure is correct. Apossible explanation is that employment is underestimated, forinstance because it excludes irregular or illegal work. In fact, avery broad definition of employment is used. The LFS03 asked allhousehold members aged 15 or over whether in the last 7 days,even for only 1 h, they did any of a wide range of activities, includ-ing running any kind of business, working for a wage, even as adomestic worker and even in kind, helping unpaid in a householdbusiness, working on the household land or food garden, lookingafter livestock, etc. An affirmative answer to any of these questionswas taken as employment.

In contrast, it can be argued that narrow unemployment is under-estimated, both because some of the workers recorded as employedare underemployed and because StatsSAs definition of search is toorestrictive. First, according to the LFS03, 4% of employed workersworked for no more than 10 h a week, and 10% worked for nomore than 25 h a week. Secondly, in both the OHS and LFS theunemployed are asked whether they have taken any action tolook for work in the relevant period, and only if the answer is yes


are they asked about the method. A common method of job search isto wait to be called by friends or relatives: a respondent who doesnot regard this as action to find work will not be asked aboutthe method, and will be classified as a non-searcher. The existenceof very high unemployment cannot be doubted, even if many ofthe unemployed are not publicly and visibly searching much ofthe time.

5.4. Unemployment, Employment and Gender

Table 6 shows the increase in unemployment rates from 1995 to2003. By the broad definition, which includes the discouragedworkers, the unemployment rate rose from 29 to 42%, an increaseof more than 12% points. By the narrow definition, it rose by 11%points, from 17 to 28%. Thus, the extent of both searching andnon-searching unemployment rose similarly: the problem is notonly due to growth in the number of non-searchers.

The broad unemployment rate rose more for men (from 23 to36%) than for women (from 38 to 48%). This differential rise (by3% points) is striking given that the labour force participation rateincreased far more for women than for men. It is to be explainedby the fact that, of the 1.97 million increase in the total number ofemployees, no fewer than 1.43 million were women (Table 6).Thus, the remarkable increase in female labour force participationwas at least in part a response to perceived new opportunities foremployment. Assisted by the employment equity legislation,women were able to increase their share of (combined farm andnon-farm) wage employment, from 35% in 1995 to 44% in 2003.Indeed, for men all net new employment during that period wasin self-employment and none in wage employment. Women main-tained their share of total unemployment over the 8 years, at 57% ofthe total.

5.5. Changes in the Incidence of Unemployment

Although the absolute level of unemployment may be determinedlargely by macroeconomic forces, it is nevertheless relevant toknow the characteristics of the workers who end up unemployed.Table 7 presents, for both 1995 and 2003, the actual proportions ofpeople in various categories who are unemployed, the coefficient


Table 6: Unemployment and Employment, by Alternative Definitions and by Gender

1995 (OHS) 2003 (LFS) 19952003 (change)

Men Women Total Men Women Total Men Women Total

Broad unemploymentRate % 22.5 38.0 29.4 35.7 47.8 41.8 13.2 9.8 12.4000 1,710 2,305 4,015 3,529 4,753 8,332 1,869 2,448 4,317Share % 42.6 57.4 100.0 43.0 57.0 100.0 43.3 56.7 100.0Narrow unemploymentRate % 13.0 22.6 17.0 25.4 31.5 28.2 12.4 8.9 11.2000 879 1,097 1,976 2,187 2,382 4,570 1,309 1,285 2,594Share % 44.5 55.5 100.0 47.9 52.1 100.0 50.5 49.5 100.0Total employment000 5,892 3,760 9,652 6,436 5,187 11,622 544 1,427 1,970Share % 61.0 39.0 100.0 55.4 44.6 100.0 27.6 72.4 100.0Wage employment000 5,379 2,852 8,231 5,302 4,207 9,509 277 1,355 1,278Share % 65.4 34.6 100.0 55.8 44.2 100.0 26.0 106.0 100.0Self employment000 513 908 1,421 1,134 977 2,111 621 69 690Share % 36.1 63.9 100.0 53.7 46.3 100.0 90.0 10.0 100.0

Note: Agricultural workers are included in wage employment.Source: October Household Survey, 1995; Labour Force Survey, September 2003.

832G

eetaK

ingdon

and

John

Kn

ight

Table 7: Binary Probits Predicting Unemployment, 1995 and 2003

1995 (OHS) 2003 (LFS) 19952003 (change)

U rate Marginal effect t-value U rate Marginal effect t-value U rate % point Marginal effect

RaceAfrican 36.9 0.296 35.8 48.8 0.361 33.7 11.9 6.5Coloured 22.2 0.211 17.3 29.4 0.275 17.6 7.2 6.4Indian 13.2 0.076 4.5 20.7 0.163 7.3 7.5 8.7White (base) 5.3 7.6 2.3 GenderMale 22.5 20.144 233.2 35.7 20.129 225.1 13.2 1.5Female (base) 38.0 47.8 9.7 EducationNone (base) 35.0 33.8 21.2 Primary 36.3 20.001 20.1 44.7 0.057 5.4 8.4 5.8Junior 32.8 0.005 0.6 47.4 0.068 6.3 14.6 6.3Secondary 27.3 20.047 25.6 44.9 0.030 2.8 17.6 7.7Higher 6.0 20.225 225.5 13.0 20.211 216.6 7.0 1.4Age1620 years 59.4 0.450 38.9 76.7 0.476 43.9 17.3 2.62125 years 45.8 0.305 38.4 66.2 0.386 46.4 20.4 8.12635 years 30.4 0.118 19.7 43.1 0.174 24.8 12.7 5.63645 years (base) 20.6 28.3 7.7

(continued on next page )

Un

emp

loym

ent

inS

outh

Africa

833

Table 7: Continued

1995 (OHS) 2003 (LFS) 19952003 (change)

U rate Marginal effect t-value U rate Marginal effect t-value U rate % point Marginal effect

4655 years 17.6 2 0.028 23.8 23.4 20.056 26.5 5.8 22.85664 years 11.9 20.089 28.2 15.8 20.124 29.2 3.9 23.5RegionRural (base) 37.9 49.7 0.018 3.1 11.8 23.5Urban 24.1 0.053 10.5 36.8 12.7

Note: The dependent variable is being broadly unemployed, with the alternative as being employed.

834G

eetaK

ingdon

and

John

Kn

ight

on each characteristic in an estimated probit equation predictingunemployment as opposed to employment, and the marginaleffect of that characteristic. For instance, we learn from the firstrow that 36.9% of African workers were unemployed in 1995, thatthe incidence of unemployment was significantly higher than forwhites, and that, holding all other observed characteristics constant,Africans had a higher probability of being unemployed than whites,of 29.6%. The actual increase in the African unemployment rate(11.9% points) represented a 9.6% point increase over the whiterate. This was due partly (3.1% points) to the white possession ofdifferent average characteristics and partly (6.5% points) to thepure effect of race or of unobserved characteristics correlatedwith race.

In 1995, possession of secondary and of higher education pro-vided strong protection against unemployment: they reduced thechances of unemployment by 4.7 and 22.5% points, respectively,vis-a`-vis those with no education. However, by 2003 secondary edu-cation actually raised the relative probability of unemployment, by3.0% points. Relative to those without education, the chances ofunemployment of workers with primary, junior and secondary edu-cation had all increased by some 68% points. Although the actualunemployment rate of graduates from higher education rose from6.0 to 13.0% over the 8 years, the standardised increase (giventhat the unemployment rate of the tiny group without educationfell by 1.2%) was merely 0.2%.

Given that the omitted category in the dummy variable analysisis the age group 3645, we see that, in both years, there is a mono-tonic fall in the standardised probability of unemployment as ageincreases. Moreover, relative to the reference group, between 1995and 2003 the probability of unemployment rose for those aged 35and under, and fell for those aged 46 and over. The problem ofunemployment is worst for the young, and has become more so.Whereas older workers are often incumbents whose jobs are pro-tected by firm-specific human capital or by legislation, new entrantsto the labour market bear the brunt of labour market competition.Province dummies are included but not shown. Over 8 years, theunemployment rate rose most rapidly in Gauteng, probably reflec-ting its attraction to rural-urban migrants. Relative to Gauteng, thestandardised unemployment rate fell in almost all of the otherprovinces.


5.6. Changes in the Duration of Unemployment

There are effectively no panel data available on labour force parti-cipants in South Africa. While the Labour Force Surveys wereintended to provide a rotating panel, the person identifiers are notcoded in such a way that they can enable the user to achieve reliablematching of individuals from one round to another (Hertz, 2005).Consequently, there are no satisfactory data on unemployment dur-ation in South Africa. What is available is the length of uncompletedspells of unemployment of those currently narrowly unemployed.While they understate the true duration of unemployment, uncom-pleted spells are suggestive. However, even this information isdeficient: data are in categorised rather than continuous form andthe categories are not sensibly distributed. We merely note that thepercentage of persons whose uncompleted duration of unemploy-ment was greater than 3 years rose by 4.6% points between 1995and 2003, from 32.5 to 37.1% of the total. The growth of long-termunemployment meant that the unemployment problem was exacer-bated in this additional dimension.

6. Policies to Tackle Unemployment Directly

As the scale of unemployment and its rise sank in, it became amatter of pragmatism that priorities be set for targeting the unem-ployed. It is potentially important to distinguish among the unem-ployed according to their employability: for those less employable(owing to their lack of skills, any previous employment experienceand their long duration of unemployment), the appropriate policymight be poverty alleviation, whereas for those who have agreater probability of finding employment it might be skills develop-ment and job creation (Poswell, 2002). The South African govern-ment has pursued various active labour market policies toalleviate unemployment. We review two broad types of policies:public works programmes and skill development programmes.While systematic impact evaluations of these policies are yet toappear, there are some case studies based on small samples, inthe former case, and some labour market surveys, in the latter,which shed light on their effectiveness.

Public works programmes (PWPs) are an important part of theSouth African governments social protection framework. A study


by McCord (2004) examines the impact of PWPs on the labourmarket outcomes of programme participants. It is based on a 2003survey of some 700 households that contain current and recent par-ticipants in two different PWPs. The difference-in-differencemethod is used, i.e., the income and employment outcomes ofhouseholds of the treatment group and the control group are com-pared, before, during, and after PWP employment, to discoverthe net effect of a programme. The impact of the programmes waslimited because they failed to take the majority of householdseven temporarily out of poverty. Even with their PWP income,99% of the households in one case and 87% in the other still fellbelow the poverty line: participation merely reduced the depth ofpoverty. Moreover, the evidence suggests that programme partici-pants tend normally to return to unemployment. The author con-cludes that such interventions have limited potential to addressunemployment and poverty in South Africa.

The idea that PWPs should provide sustained employment wasruled out when the Code of Good Conduct for PWPs were agreedin tripartite negotiations between the union movement, the stateand the public sector. The code permitted the payment of lowerthan minimum wages for PWP employees only if the employmentoffered under government PWP schemes would be of short-termduration and if workers would be given training to compensatefor the reduced wage. This condition rules out employment guaran-tees and highlights the tension between the protection of therights of the employees and effective unemployment alleviationprogrammes. Whereas policy documents characterise the publicworks as a work experience and training programme to improvelabour market access and performance, at the end of whichworkers will graduate to employment under normal conditions(e.g., Department of Public Works, 2004), they might be betterregarded as a social protection response to the challenge of unem-ployable working age people with no skills and geographical iso-lation (Abedian, 2004).

Skills training has become a central plank of labour marketpolicy: in 1998 the government enacted the National Skills Actand in 1999 it created Sector Education Training Authorities(SETAs) which charge a skills levy from firms, to be repaid on theproduction of evidence by the firm that it is undertaking approvedtraining for its workers. SETAs are also mandated to encourage.


Further Education and Training (FET) institutions in the provisionof education and skills for work. The rationale for introducing theSETAs is skills shortages that, according to employing firms, actas a constraint on their growth.

Apart from a few studies (Felton, 2003; Lundall, 2003), relativelylittle is known or systematically documented about workplace skillsin South Africa. What is known is based mainly on surveys of firms,which provide information on their perceptions of skills shortages,their training activities and expenditures, the types of training insti-tutions they consider most useful, and their views about theemployment effects of the legislation on skills.5 There are apparentcontradictions and gaps in the evidence. On the one hand, firmsreport in the surveys that they have difficulty in recruiting pro-fessional, managerial and technical staff. Reasons mentionedinclude the emigration of skilled people and the employment-equity requirements of racial balance in each occupation groupwithin a firm. On the other hand, the observed behaviour of firmsappears inconsistent with the existence of serious skills gaps. Forinstance, the surveys show that vacancy rates are low: three-quarters of the Johannesburg respondents in 1999 reported novacancies, and only a third of the workers in manufacturing firmswith more than 50 employees had received some form of training,whether in-house or contracted-out (Chandra et al., 2000, pp. 41, 43).Evidence from a smaller but more probing qualitative survey(Felton, 2003) suggests that firms find it hard to identify, or to articu-late, their skill needs, and that they respond weakly if at all to thegovernments skills policy.

The skills that are lacking may be of a general and thus marketablenaturesuch as the desired degree of numeracy, literacy, communi-cation skills, problem-solving, initiative-taking etc. Firms well-documented reluctance to undertake training themselves might bebecause they are unable sufficiently to recoup the expense of invest-ment in such skills if their trained workers quit not long after train-ing. In principle, since they stand to benefit from their training,workers have an incentive to pay for investment in marketableskills, for instance in the form of lower wages during the period

5 These include the World Bank surveys of firms in Johannesburg and Durban in1999 and 2003, and a national survey of 1,400 firms commissioned by the Office ofthe President in 2000 (Bhorat and Lundall, 2002; Chandra et al., 2002 and Deveyet al., 2003).


of training. However, this solution is not possible if workers areunable to afford the costs of training. It is precisely the poachingexternalities problem that the grant-levy system is designed toovercome. However, the SETAs have found that their skill-levy reven-ues remain largely unclaimed by firms (Felton, 2003). This would bedifficult to explain if firms did indeed face serious shortages ofskills.

Skills development might alleviate unemployment in two ways.First, it can relax a skills constraint on the growth of the economyand so indirectly expand the employment of the unskilled.Secondly, it can improve the skills of the unemployed who werepreviously without skills, and so make them directly more employ-able. To investigate the latter possibility there is a need for impactevaluation studies of the effects of such training. However, suchstudies cannot provide a social cost-benefit analysis, as opposedto a private one, of schemes to develop the skills of the unemployedinsofar as they are unable to separate the redistribution of jobsamong workers from the creation of additional net employment.

7. Conclusions

The legacy of history continues to have long-term effects on thelabour market. Economic stagnation in the two decades beforedemocracy, the extreme economic inequality and the developmentof a powerful trade union movement, all continue to have theirimpact today. Since it took office, the ANC government has hadlittle political and economic room to manoeuvre in its economicpolicies (Gelb, 2004). The two basic policy choices were: whatpoints on the trade-off between economic efficiency and socialequity, and between long and short run benefits, to choose? Thegovernment has resisted the calls of populism and has tended tostress efficiency and long-term considerations in most areas ofeconomic policy (such as trade liberalisation, deregulation andthe encouragement of the private sector) that underlie its generalmacroeconomic strategythe GEAR. Nevertheless, the labourmarket outcomes of this set of policies have been unsuccessful.That raises the broad question: to what extent are the generaleconomic policies and to what extent are labour market policiesto blame for this? Our argument is that these two are intertwined:labour market policies can have adverse effects in themselves but


they can also reduce the effectiveness of the GEAR policies. Thereis a potential for any economy to enter a virtuous circle of econ-omic growth or get caught in a vicious circle of relative stagnation.In the last decade, South Africa has been, and still remains, on aknife-edge: the economy could go either way. The danger is thatlow business confidence and inadequate investment make thingsworse in the labour market, which in turn by various processesof cumulative causation feeds through into self-fulfilling pessi-mism about the economy.

In explaining the rise in unemployment and its possible cures,we placed our emphasis on the slow growth of the economy, andthus slow growth in the demand for labour relative to the rapidlygrowing supply, and the need above all else to pursue economicpolicies to help raise the economic growth rate. The rapid diver-gence in the supply and demand for labour places a great burdenof adjustment on the labour market, which would put a greatstrain on even the most flexible of labour markets. In fact, itappears that major segments of the South African labour marketare not flexible. This imposes an excessive burden on the more flex-ible segments, the result of which has been rising unemployment.

The system of labour market governance that has been put inplace over the last decade has been built on consensus and co-operation. However, it has involved largely government, formalsector employees and employers. Another party, which is poten-tially affected but not adequately represented in developing theconsensus on governance, is the growing number of informalsector poor and the unemployed poor. The system embodied inthe various pieces of legislationsuch as the Labour RelationsAct of 1995, the Basic Conditions of Employment Act of 1997 andthe Employment Equity Act of 1999establishes a framework,which would be appropriate, indeed admirable, in a fully employedeconomy with little labour market segmentation. It is arguably lessappropriate in South African conditions of unemployment andextreme labour market segmentation. The danger is that in protect-ing the rights of formal sector workers, the legislation and itsimplementation harms the interests of those outside the formalsector. Well-intentioned labour laws can have unfortunate unin-tended labour market consequences. In particular, they may dis-courage employment (a static effect) and discourage investment(a dynamic effect). It is an uncomfortable fact that a government,


which espouses trade liberalisation must be prepared when neces-sary to espouse also a good degree of labour market liberalisation.

The measurement of potential ill-effects of labour legislation isdifficult, and well-established research results on this issue are notyet available. It is certainly the case, however, that many employerscomplain about various constraints on their decisions such asthe legal and procedural requirements with respect to hiring andfiring of workers, the extent of union power to raise wages ordisrupt production, the extension of bargaining agreements, andthe hassle factor arising from the labour relations legislation.While it is natural for survey respondents to stress the importanceof things they dislike, complaints of this sort should not beignored. If acted upon, these complaints can result in hiring fewerworkers, substituting capital for labour, diverting managerialresources from the main task, depressing business optimism anddiscouraging investment. In this way, the inflexibilities of thelabour market may exacerbate not only income inequalities butalso the underlying problem of inadequate employment creation.It is true that the efficient managers can be expected to try to circum-vent the problems that they perceive, for instance by hiring casualworkers and employing sub-contactors. This behaviour canreduce the problems but not remove them.

It is beyond the scope of this paper to examine many determinantsof poverty including the effects on poverty of trends in the labourmarket. Nevertheless, the evidence is consistent with our viewthat the remarkable increase in unemployment in both open and dis-guised form, and the associated fall in real earnings in the lessprotected parts of the labour market, have raised the number ofhouseholds in poverty. For instance, between 1995 and 2003 theaverage real earnings of all workers fell by 22% (Table 4).Although the magnitudes of the estimates differ, there is a generalconsensus that poverty rose over the post-apartheid period. This isreflected in a number of studies using different surveys, questions,levels of aggregation, time periods and methods (Hoogeveen andOzler, 2004; Casale et al., 2005; Leibbrandt et al., 2005; Van de Ruitand May, 2003; Meth and Dias, 2004; van der Berg and Louw, 2003).

There is empirical evidence that crime has increased in SouthAfrica at the same time as unemployment has risen (SAPS, 2004;Masuku, 2003). For instance, the SAPS website shows thatbetween 19945 and 20034 reported common robbery was up


by 193% and reported burglary of residential premises by 29%. Therise in unemployment and in poverty may force more people intoillegal ways of getting income. A World Bank survey of 600 firmsin Durban in 20023 (Devey et al., 2003) found that crime andtheft were listed as the biggest constraint on firm growth. Thismight produce another vicious circle: a higher crime rate canreduce prospective profits, sap business confidence and lead toa brain drain. Economic growth, and thus the labour marketoutturn, can therefore suffer.

Our main policy conclusions are fourfold. First, if labour marketstatistics are to be used effectively for policy purposes, theirdeficiencies and problems need to be addressed by means ofthorough interaction among StatsSA, policy-makers, and indepen-dent researchers. Despite the statistical progress of the lastdecade, the lack of appropriate data hinders the analysis of import-ant issues that impinge on unemployment.6 Secondly, there is needfor investigation of how active labour market policies have per-formed. For instance, the policy on skills promotion is based onthe supposition that training raises employment chances but, toour knowledge, this has not been rigorously tested in SouthAfrica. Similarly, it is assumed that public works programmesimprove the employability of rural unemployed persons but,again, this has not been adequately tested nor alternative schemescompared. Impact evaluation studies would assist policy design.Thirdly, labour market regulations require reconsideration, givinggreater weight to the concerns of employers and investors, andalso to the interests of thosethe unemployed and informallyemployed poorwho are beyond the reach of the labour insti-tutions but can be hurt by them nevertheless. For instance, theremoval of the provision for the extension of Bargaining Councilagreements to all (including small) employers might well promotethe SMME sector by removing the burden of high labour costs onsmall firms. Lastly but most importantly, it is crucial to pursue aset of policies that promote the rate of economic growth.

6 For instance, at present we know little about entry into, exit from, and duration ofunemployment: these and many other issues are best analysed by means of apanel survey. Unemployment duration data are currently collected by categoryof duration, with inappropriate categories, and not at all for the non-searchingunemployed. The questions about job search, and their sequencing, can lead totoo restrictive a definition of search. There are no reliable data on reservationwages.


Government should bear constantly in mind the growth impli-cations of all its policieswhether these are general economic poli-cies, or labour market policies, or even policies on internationalrelations (such as those on relations with Zimbabwe), crime orhealth (e.g., on HIV/AIDS). It is the growth of the economy that,above all else, will determine the future of unemployment inSouth Africa.

Acknowledgements

We are grateful to Daniela Casale, Haroon Bhorat, Stephan Klasen,Murray Leibbrandt, Nicoli Nattrass and Dori Posel for commentson this paper. Geeta Kingdons time on this research was fundedby an ESRC grant under the Global Poverty Research Group atthe Centre for the Study of African Economies. The views andopinions expressed are those of the authors alone.

References

Abedian, I. (2004) Beyond Budget 2004: Job Creation and PovertyAlleviation in South Africa: What Should We be Doing?, PublicSymposium, Institute for Justice and Reconciliation, BreakwaterLodge/UCT Graduate School of Business, Cape Town, March.

Alleyne, T. and Subramanian, A. (2001) What Does South AfricasPattern of Trade Say about its Labor Market?, IMF WorkingPaper No. 01/148, October.

Altman, M. (2002) Progress in Addressing Employment andUnemployment, Prepared for Presidency, Policy Coordinationand Advisory Services, as part of the 10 year policy review andstrategic planning.

Banerjee, B. (1986) Rural to Urban Migration and the Urban LabourMarket: A Case Study of Delhi, Bombay: Himalaya PublishingHouse.

Bertrand, M., Mullainathan, S. and Miller, D. (2003) Public Policyand Extended Families: Evidence from Pensions in SouthAfrica, World Bank Economic Review, 17 (1): 2750.


Bhorat, H. (2002) Employment and Unemployment Trends in Post-Apartheid South Africa, Paper Prepared for the Presidency10-Year Cabinet Review Process, 2nd Draft.

Bhorat, H. and Leibbrandt, M. (1998) Estimates of Wage-Employment Elasticities for South Africa, unpublished.

Bhorat, H., Lundall, P. and Rospabe, S. (2002) The South AfricanLabour Market in a Globalizing World: Economic and LegislativeConsiderations, Employment Paper 2002/32, Geneva: InternationalLabour Organization.

Bowles, S. and Heinz, J. (1996) Wages and Jobs in the South AfricanEconomy: an Econometric Investigation, Department ofEconomics, University of Massachusetts.

Casale, D. and Posel, D. (2002) The Continued Feminisation of theLabour Force in South Africa: An Analysis of Recent Data andTrends, South African Journal of Economics, 70 (1): 15684.

Casale, D., Muller, C. and Posel, D. (2004) Two Million Net NewJobs: A Reconsideration of the Rise in Employment in SouthAfrica, 19952003, South African Journal of Economics, 72 (5):9781002.

Chandra, V., Moorty, L., Rajaratnam, B. and Schaefer, K. (2000)Constraints to Growth and Employment in South Africa,Report No. 1: Statistics from the Large Manufacturing FirmSurvey, Discussion Paper 14, Informal Discussion Papers onAspects of the Economy of South Africa, World Bank.

Chandra, V., Nganou, J. and Noel, C. (2002) Constraints to Growthin Johannesburgs Black Informal Sector: Evidence from the1999 Informal Sector Survey, World Bank Report No. 24449-ZA, June.

Charmes, J. (2000) Informal Sector, Poverty and Gender: A Reviewof Empirical Evidence: Background paper for WorldDevelopment Report 2001. Centre of Economics and Ethics forEnvironment and Development, University of Versailles-St Quentin en Yvelines.

Devey, R., Skinner, C. and Valodia, I. (2002) The InformalEconomy in South Africa: Who, Where, What and HowMuch?, Paper presented to the DPRU Conference on LabourMarkets and Poverty in South Africa, Johannesburg, 2224October, 2002.


Devey, R., I. Valodia, B. Rajaratnam and M. Velia (2003) Constraintsto growth and employment: Evidence from the Greater DurbanMetropolitan Area. Trade and Industry Policy Secretariat (TIPS)and Development Policy Research Unit (DPRU) Conference,Pretoria.

Department of Public Works (2004) Expanded Public WorksProgramme, presentation by Dr. Sean Phillips, Director of theEPWP, Midrand, 25th February.

Edmonds, E., K. Mammen and D. Miller (2005) Rearranging theFamily? Household Composition Response to Large PensionReceipts, Journal of Human Resources, 40 (1): 186207.

Fallon, P. and R. Lucas (1998) South Africa: Labor Markets,Adjustment, and Inequalities, Discussion Paper no. 12,Informal Discussion Papers on Aspects of the Economy ofSouth Africa, Southern Africa Department, World Bank,Washington D.C.

Felton, M. (2003) The Skills Gap: Research into Workplace Skills ofFurther Education Students and Labour Market Requirementsin Todays South Africa, DFID Research Report.

Fields, G., M. Leibbrandt and J. Wakefield (1999) Key labour marketelasticities in South Africa, report commissioned by the SouthAfrican Department of Finance.

Gelb, S. (2004) Inequality in South Africa: Nature Causes and Responses,Johannesburg: The Edge Institute.

Harris, J. and Todaro, M. (1970) Migration, Unemployment, andDevelopment: A Two-Sector Analysis, American EconomicReview, 60 (1): 12642.

Hertz, T. (2005) The Effect of Minimum Wages on the Employmentand Earnings of South Africas Domestic Service Workers,Upjohn Institute Working Paper 05-120. American University,August.

Hoogeveen, J. and B. Ozler (2004) Not Separate, Not Equal: Povertyand Inequality in Post-Apartheid South Africa, mimeo: WorldBank, presented to the NEUDC Conference, Montreal, October.

Kingdon, G. and J. Knight (2001) Unemployment in South Africa:The Nature of the Beast, WPS/2001.15, Centre for the Study ofAfrican Economies, Department of Economics, University ofOxford.


Kingdon, G. and J. Knight (2004a) Unemployment in SouthAfrica: The Nature of the Beast, World Development, 32 (3):391408.

Kingdon, G. and J. Knight (2004b) Race and the Incidence ofUnemployment in South Africa, Review of DevelopmentEconomics, 8 (3): 198222.

Kingdon, G. and J. Knight (2006a) The Measurement ofUnemployment when Unemployment is High, Labour Economics,13 (3): 291315.

Kingdon, G. and J. Knight (2006b) How Flexible are Wages inResponse to Local Unemployment in South Africa?, Industrialand Labor Relations Review, 59, 3: 47195.

Kingdon, G., J. Sandefur and F. Teal (2005) Labour MarketFlexibility, Wages and Incomes in Sub-Saharan Africa in the1990s, GPRG-WPS-030, Centre for the Study of AfricanEconomies, University of Oxford. November.

Klasen, S. and I. Woolard (1999) Levels, Trends, and Consistency ofEmployment and Unemployment Figures in South Africa,Development Southern Africa, 16: 336.

Klasen, S. and I. Woolard (2000) Surviving Unemployment WithoutState Support: Unemployment and Household Formation inSouth Africa, IZA Discussion Paper No. 237, Institute for theStudy of Labor, Cologne, Germany.

Leibbrandt, M., J. Levinsohn and J. McCrary (2005) Incomes inSouth Africa Since the Fall of Apartheid, draft mimeo, May.University of Cape Town and University of Michigan.

Lundall, P. (2003) Sector Education Training Authorities andthe Delivery of Training: Preliminary Remarks on the NewSkills Dispensation in South Africa, Working Paper 03/79,Development Policy Research Unit, University of Cape Town,August.

Maloney, W. (1999) Does Informality Imply Segmentation in UrbanLabor Markets? Evidence from Sectoral Transitions in Mexico,World Bank Economic Review, 13: 275302.

Maloney, W. (2002) Informality Revisited, Mimeo: World Bank.

Masuku, S. (2003) For Better and for Worse: South African CrimeTrends in 2002, South Africa Crime Quarterly, (3) March.


May, J. and N. Hunter (2005) Poverty, Inequality and the Social Wage:Balancing the Books in Post-Apartheid South Africa?, mimeo: Schoolof Development Studies, University of KwaZulu-Natal.

Mbeki, T. (2005) DRC: Forward Ever, Backward Never!: Letter fromthe President, ANC Today, 5 (20) May: pp. 206.

McCord, A. (2004) Policy Expectations and Programme Reality:The Poverty Reduction and Labour Market Impact of TwoPublic Works Programmes in South Africa, ESAU WorkingPaper No. 8, Overseas Development Institute, London.

Meth, C. and R. Dias (2004) Increases in Poverty in South Africa:19992002, Development Southern Africa, 21 (1): 5986.

Moll, P. (1996) Compulsory Centralisation of Collective Bargainingin South Africa, American Economic Review, 86 (2): 32629.

Muller, C. and J. Esselar (2004) Documenting the informalisation ofworking South Africa: Evidence from national household surveys,19972001, Journal of Interdisciplinary Economics, 15: 22949.

Nattrass, N. (2000) Inequality, Unemployment and Wage-SettingInstitutions in South Africa, Studies in Economics andEconometrics, 24 (3): 12942.

Nattrass, N. (2003) Aids, Economic Growth and Distribution inSouth Africa, South African Journal of Economics, 71 (3): 42854.

Nattrass, N. and R. Walker (2005) Unemployment and ReservationWages in Working Class Cape Town, South African Journal ofEconomics, 73 (3): 498509.

Posel, D., J. Fairburn and F. Lund (2006) Labour Migration andHouseholds: A Reconsideration of the Effects of the SocialPension on Labour Supply in South Africa, Economic Modelling,23 (4): 83653.

Poswell, L. (2002) The Post-Apartheid South African Labour Market:A Status Report, Development Policy Research Unit, Universityof Cape Town, February.

Quattek, K. (2000) The economic impact of AIDS in South Africa: adark cloud on the horizon, reprinted with permission of INGBarings.

Rogerson, C.M. (1992) The Absorptive Capacity of the InformalSector in the South African City, in D.M. Smith (eds), TheApartheid City and Beyond, London: Routledge.


SAPS (2004) Crime Statistics Per Category, South African PoliceServices, http://www.saps.gov.za/statistics/reports/crimestats/2004/crime_stats.htm

Seekings, J., M. Leibbrandt and N. Nattrass (2004) Income inequal-ity after apartheid, CSSR Working Paper No. 75, Centre for SocialScience Research, University of Cape Town.

Van der Berg, S. (2001) Education and Skills Constraint, extract from atalk delivered to a conference of the Bureau for EconomicResearch, Cape Town, November.

Van der Berg, S. and Louw, M. (2003) Changing patterns of SouthAfrican income distribution: Towards time series estimates of dis-tribution and poverty, paper to the Conference of the EconomicSociety of South Africa, Stellenbosch, 1719 September, 2003.

Van de Ruit, C. and J. May (2003) Triangulating Qualitative andQuantitative Approaches to the Measurement of Poverty:A Case study in the Limpopo Province, South Africa, IDSBulletin, 34 (4): 2133.

Xaba, J., P. Horn and S. Motala (2002) The Informal Sector in Sub-Saharan Africa, ILO Working Paper on the Informal Economy,Employment Sector, ILO, Geneva.


Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

23_B_Kingdon_Unemployment in South Africa, 1995–2003 Causes Problem and Policies,

Documents

labour forcesurvey

growing labour scarcity

high unemployment

duration of unemployment

therole of labour legislation

interesting labour markets

labour institutions

unemployment bythe slow