Is there a trade-off between the employment and wages of unskilled African South Africans?

ABSTRACT

The possible trade-off between employment and wages has characterised most of South Africa’s labour market debates, particularly with regards to decent wages versus unemployment. In this article we explore the relationship between labour market earnings and the level of employment among African birth cohorts using labour force data from 1997 to 2011. We find that the association between an increase in the proportion of unskilled employed in a birth cohort and earnings is mediated by the sector of employment. While some sectors exhibit the expected negative association, there is a robust positive relationship between the first two quartiles of the earnings distribution within birth cohorts and the proportion of the birth cohort who are employed in unskilled occupations in the manufacturing and trade sectors. Because a range of market forces determine this relationship, further research is needed to unpack the reasons for such varied outcomes in order to better inform the debates on labour market interventions like the proposed National Minimum Wage and to appreciate the potential impact of such policy interventions on wages and employment.

KEYWORDS:

• Earnings
• employment
• sector
• occupation
• birth cohort
• South Africa

1. Introduction

The visibly persistent inequality in income between white and African South Africans since the end of apartheid has provoked a large literature on the trends and potential causes of employment, unemployment and earnings levels among Africans in South Africa. This literature includes, among others, Kingdon & Knight (2007), Banerjee et al. (2008), Burger & Von Fintel (2009), Fourie (2011), Branson et al. (2013), Bhorat & Goga (2013) and Cichello et al. (2014). It suggests that the limited growth in employment and real wages among the majority of Africans in South Africa since the transition to democracy can be attributed to shifts in the economy from more unskilled labour-intensive sectors to sectors that are relatively more skill-intensive in labour. These shifts have exacerbated structural unemployment in South Africa and have disproportionally affected younger birth cohorts. At the same time, increased labour force participation, especially among females and younger cohorts since the end of apartheid, has elevated unemployment. This has placed downward pressure on real wages, particularly among workers who have not completed secondary education.

When taken together, the research on the labour market in South Africa implies that the only sustainable approach to simultaneously increasing both employment and earnings among African South Africans will require significant improvements to the education of workers so that the economy can respond to shifts in labour demand and expand into new markets. To achieve this policy-makers will, as Mariotti & Meinecke (2014) argue, have to find ways of increasing the returns to any given level of education. Spaull (2013), however, shows that the schooling system in South Africa is currently confronted with significant challenges in this regard. It seems unlikely that these constraints to targeting the structural unemployment problem will be confronted in the near future. In the interim, policy-makers will have to continue pursuing active labour market interventions that facilitate both transitions into employment among, and an increase in the productivity of, relatively unskilled workers. These include post-school training, direct employment-creation schemes and private-sector incentives such as subsidies, as well as fostering a macroeconomic posture that is conducive to both investments in unskilled labour-intensive production and public revenue. It is nevertheless unclear whether these employment interventions should be targeted at the economy as a whole or whether they should be targeted at particular sectors. As Bhorat & Goga (2013) show in their overview of the occupational and sectoral shifts in South Africa, employment in primary sectors (agriculture and mining) has decreased significantly, the manufacturing sector has remained stagnant and only tertiary sectors and tertiary occupations have experienced employment growth. This has been accompanied by an increase in earnings of skilled labour while wages of unskilled and semi-skilled workers have stagnated. This is politically problematic because the majority of the unemployed fall within the unskilled labour segment and households struggle to make ends meet. Thus, the call for decent wages is strongest for the most vulnerable segment of the South African labour force.

The literature on minimum wages, however, suggests that wage floors in developing countries may reduce employment (Neumark & Wascher, 2006). The reported relationships between wages and employment are nevertheless likely to be determined by the outcome of complex interactions of changes in educational attainment by the different cohorts, as well as sectoral shifts in the supply and demand for different types of labour. Especially, the relationship between wages and the level of employment in one particular sector can have more than one outcome: increased supply of labour to one sector should dampen wages through competition, but, at the same time, the increase in labour supply to the sector could be due to relatively higher wages in that particular sector compared with others. This is further complicated when the quality of education that a particular cohort is exposed to is changing over time. So far, the literature has not been able to unpack these linkages.

In this article we contribute to the literature on labour market outcomes in South Africa since the end of apartheid by showing that the direction of the relationship between employment and the earnings of unskilled labour in South Africa, among other workers, is also mediated by the sector of employment when controlling for cohort-specific effects. We use a one-year birth-cohort pseudo-panel to estimate the association between earnings quartiles within birth cohorts and the proportion of the birth cohort in particular occupations within sectors, relative to changes in the proportion of the cohort who is unemployed and searching for work. This allows us to eliminate the effect of any structural shifts in the economy that have had a permanent differential effect on the labour market outcomes of Africans in particular birth cohorts (such as access to primary and secondary education).¹ Our estimates confirm the varied nature of this relationship. While there is a negative relationship between earnings and an increase in the proportion of African South Africans who engage in elementary forms of self-employment, work in private households or work as unskilled and semi-skilled males in agriculture, there is evidence of a positive relationship between the first two quartiles of the earnings distribution within birth cohorts and the proportions of birth cohorts who are employed in unskilled occupations in the manufacturing sector and the retail and wholesale trade sector.

The article proceeds as follows. First, we outline the Post-Apartheid Labour Market Survey (PALMS) data that we use to estimate the relationship between earnings and the level of employment within birth cohorts. We then proceed to present descriptions of the labour market in South Africa since apartheid and the results from our model. This is followed by a discussion and our conclusion.

2. The data

Kerr et al. (2013) have constructed a Post-Apartheid Labour Market Series 1994–2012 (PALMS) by stacking the cross-sectional datasets created from 39 labour-market surveys conducted by Statistics South Africa between 1994 and 2012. These include the October Household Survey (OHS) from 1994 to 1999, the bi-annual Labour Force Survey (LFS) from 2000 to 2007 – including the smaller LFS pilot survey from February 2000 – and the Quarterly Labour Force Survey (QLFS) from 2008 to 2012. The PALMS allows us to explore the relationship between labour-market earnings and employment sector, the education levels of workers in employment and their occupations (which we use as a proxy for their skill levels) over this period.

There are, however, several important considerations when using the PALMS. First, it includes cross-entropy weights which were constructed for the series because the weights presented in the original cross-sections have temporal inconsistencies and ‘there is no hierarchical consistency between the person and household weighted series until 2003’ (Kerr et al., 2013:6) Further, there are no data on the sector of employment prior to 1996 and there are also no earnings data prior to 1997 and from 2008 to 2009 in this particular dataset at the time of writing. It is also unclear how reliable the wage data are, primarily because of non-response. Further, for some of the individuals we only have an earnings band. Finally, the unequal spacing between the different cross-sections and the changes to the surveys place additional demands on a regression specification that pools the data from the PALMS. We consider these data concerns in our econometric specification.

The econometric specification is based on a one-year birth-cohort pseudo-panel. There are 52 birth cohorts for males and 52 birth cohorts for females, starting with the cohort born in the year 1940 up to the cohort born in 1991. Thus, the oldest cohort in our analysis (born 1940) was 57 years old in 1997, the first year of the PALMS with reported earnings data. The youngest cohort (born 1991) was only five years old in 1997 but 20 years of age in 2011, and therefore likely to have entered the labour market in some way.² Table 1 presents the mean number, first percentile and median number of observations that are used to calculate the aggregates for the outcomes and explanatory variables included in our specification. The table shows, for example, that the birth-cohort observations were calculated by, on average, 392 observations for males and 478 observations for females. We use these aggregates to investigate the relationship between the proportion of each birth cohort employed in an occupation in a particular sector and the first three quartiles of the earnings distribution for this birth cohort. As we will discuss further in the article, the (three) dependent variables will be the respective earnings quartiles (for the birth cohort–period combination), and the independent variables we use will include, among others, the proportions of each birth cohort employed in particular occupation and sector categories (which we will outline in the next section) in a particular period.

Table 1. Number of observations used to construct the one-year birth-cohort pseudo-panel.

Gender	Mean	First percentile	Median
Male	392	115	353
Female	478	134	447
Total	435	120	407

3. Descriptions of the South African labour market from 1996 to 2012

For the purposes of our analysis we will define employment sectors using the following definitions: self-employment, private households, agriculture, mining, manufacturing, construction and utilities, retail and wholesale trade, financial services and information technology (IT) (including financial services, real estate, research and development, and computer services), public administration, education, health and other services (not included under retail and wholesale trade, financial services and IT, education or health). The sectoral definitions we use are based on the industry classification code lists presented in the PALMS. It is important to note that some workers who are employed in service sectors may be outsourced to non-service sectors, although we are unable to establish where this is the case.

We follow Statistics South Africa in using the following occupations as proxies for the skill levels associated with a particular job: senior (including legislators, senior officials and managers), professional, technical (technical and associate professionals and skilled agricultural and fishery workers), clerk, service assistant (including shop and market sales workers), trade (craft and related trade workers), operator (plant and machine operators and assemblers) and elementary occupation (including domestic workers).

In the previous section we mentioned that there are some concerns about the reliability of the earnings data, that there is substantial non-response and that in some cases the datum only refers to an earnings band. This may have a bearing on any investigation of the movements of earnings among Africans in South Africa since apartheid. To mitigate the effect of this potential bias we will explore the movement of the three real (in year 2000 Rand) earnings quartiles as opposed to the mean. This has the advantages that it is less exposed to the effect of outliers and any systematic bias in the reporting of earnings, and we believe the three earnings quartiles may also be less affected by changes to the three surveys (the OHS, LFS and QLFS). However, as presented in Table 2, the differences in the real earnings quartiles from the LFS to the QLFS seem much larger than a linear trend would imply. The table presents the three earnings quartiles across all the surveys in each of the OHS, LFS and QLFS. There is a considerable increase in each of these quartiles from the LFS to the QLFS.

Table 2. Selected descriptions of the labour market in the OHS, LFS and QLFS.

	Male			Female
	OHS	LFS	QLFS	OHS	LFS	QLFS
Real earnings quartile for African birth cohort (in year 2000 Rand)^a
First	640	649	830	320	339	526
Second (median)	1281	1209	1457	682	639	961
Third	2093	2039	2803	1535	1293	2186
Percentage of African birth cohort
Employed	51.78	54.58	55.12	29.23	37.56	39.4
Not economically active	33.70	25.35	26.79	55.26	41.31	43.61
Searching unemployed	14.52	20.07	18.09	15.51	21.13	16.99
Percentage of African birth cohort employed in sector
Self-employed	4.01	8.72	7.90	3.25	9.50	5.98
Private household	1.82	2.30	2.51	7.24	8.59	8.33
Agriculture	5.77	4.71	3.28	2.04	1.82	1.51
Mining	5.13	4.92	2.57	0.11	0.12	0.26
Manufacturing	8.01	7.27	7.56	2.79	2.63	2.71
Construction and utilities	4.83	5.57	6.38	0.31	0.61	0.84
Retail and wholesale trade	6.07	6.63	7.76	3.96	4.70	6.26
Financial services and IT	3.10	3.93	5.61	1.14	1.66	3.16
Public administration	3.85	3.09	3.37	0.95	1.18	1.79
Education	2.38	2.16	2.13	3.18	3.16	3.59
Health	0.93	0.96	1.12	2.51	2.39	3.31
Other services	4.33	4.20	4.92	0.97	1.13	1.66
Percentage of African birth cohort employed in occupation
Senior	2.21	1.89	2.52	0.61	0.67	1.08
Professional	1.55	1.41	1.58	1.47	1.32	1.43
Technical	5.16	6.07	4.23	4.06	6.07	4.74
Clerk	2.60	2.53	2.93	2.44	3.14	4.62
Service assistant	6.62	7.28	8.37	3.50	4.95	6.34
Trade	10.27	11.44	11.56	1.80	2.01	1.65
Operator	9.61	9.86	8.60	0.83	1.14	1.14
Elementary occupation	12.10	13.98	15.33	13.89	18.20	18.40

Note: ^aEarnings figures are only for the OHS from 1997 to 1999, the LFS from 2000 to 2007 and the QLFS from 2010 to 2011.

Table 2 also presents the percentage of African males and females (born between 1940 and 1992) who are not economically active or searching unemployed, or are employed in particular sectors and occupations (an average for each of the OHS, LFS and QLFS). Here too we notice that there are considerable shifts between the OHS, LFS and QLFS. For example, the proportion of self-employed doubles from the OHS to the LFS. Consequently, at least some of the differences may be due to changes to the survey instrument, the spacing between surveys (yearly, bi-annually and then quarterly) and the sampling methodology.

Irrespective, the data are in line with the findings of other studies. The first three earnings quartiles show very low real earnings of African males and even lower real earnings of African females. While the share of African males in the different labour market statuses is relatively stable over time, females have experienced a significant decrease in the share of non-economically active participants. The employment sectors are also as expected. Males have a higher share of workers in mining, agriculture, construction and manufacturing, while female African workers have a higher share in domestic work and increasingly in the retail sector. Notable are developments over time. Self-employment and employment in the retail sector has increased as a share of employment for African men and women while the historical employment sectors of unskilled labour (mining and agriculture) have declined. However, as can be seen in Tables A1 and A2 in Appendix 1, there is significant variation not only between different birth cohorts but more importantly within the cohorts. Tables A1 and A2 present the between and within variations for African males and females.

4. Model and results

The descriptions outlined in the previous section show that there is substantial variation between the selected labour market outcomes, both across the different surveys that are used to construct the PALMS as well as between and within the birth cohorts that we use to construct the pseudo-panel. In this section we estimate the effect of the cohort’s movement in employment in occupations within the economic sectors on the cohort’s earnings after controlling for any permanent birth-cohort effects that may have played a role in their observed labour market outcomes.

Verbeek (2008) highlights the identifying restrictions for consistent estimation of the population parameters of interest in models that use the average value of both the dependent and explanatory variables for the individuals in a cohort. In particular, pseudo-panels are susceptible to measurement error. Each cohort in a pseudo-panel should therefore have a large number of individual observations.³ The cohorts should also be exogenous and stable over time. This excludes using province birth cohorts from the PALMS data, for example, and is why we will only use the one-year birth cohort. As presented in Table 1, 99% of the observations in the birth-cohort pseudo-panel we construct from the PALMS data separately for male and female African South Africans aged 20 to 60 born from 1940 to 1992 have at least 115 and 134 observations respectively in each time period.

Another constraint is that, because there are no wage data for the eight QLFS from 2008 and 2009 and the single QLFS for 2012 in the PALMS dataset, and because the different surveys are unequally spaced, the estimates from a standard fixed-effect estimator will be inconsistent. We therefore follow Baltagi & Wu’s (1999) approach to unequally spaced and unbalanced panel fixed-effects regressions with first-order autoregressive errors when we model the value of the first, second (i.e. the median) and third earnings quartiles within cohorts as a function of the proportion of these birth cohorts employed in an occupation in a particular employment sector. As Baltagi & Wu (1999:814) point out, this estimator ‘can handle a wide range of unequally spaced panel data patterns’.

We use the following specification:(1) where is the log earnings (real, in year 2000 Rand) quartile q for the cohort c in period t in Rand; is the proportion of the cohort c population ()employed () in an occupation in a sector in period t; includes the age of the cohort, the mean years of education of the employed in the cohort and the proportion of the cohort c not economically active in period t; is the fixed effect for cohort c; and ε_ct is the error term for the cohort c in period t.

We collapse those occupations in with only a small number of observations (fewer than 6000 individual datum in total for all periods) into an ‘Other’ sector–occupations category, and exclude the proportion of the cohort population who is unemployed in the period as the comparison group.

The specification in Equation (1) allows us to interpret the coefficients as the proportionate change in quartile earnings, for a percentage-point increase in employment relative to a percentage-point decrease in the proportion of the cohort who is searching unemployed (holding the proportion employed in other sector–occupation combinations and the proportion of the cohort who is not economically active constant).

Further, the autoregressive component of the specification captures the first-order effects of macroeconomic fluctuations to the first three quartiles of the earnings distribution within cohorts over time. It should be noted, however, that there are no tables for the Baltagi & Wu (1999) locally best invariant test statistic (for ), although it is unlikely that there is no serial autocorrelation.

The results are presented in Table 3.

Table 3. Estimates from birth-cohort pseudo-panel fixed-effects regression with autoregressive error to account for unequal spacing and unbalanced number of observations in the panel.

	African male log earnings quartile (in year 2000 Rand)			African female log earnings quartile (in year 2000 Rand)
	First	Second (median)	Third	First	Second (median)	Third
Percentage of birth cohort in occupations within sector
Self-employed
Senior	0.010 (0.006)	0.013** (0.005)	0.019*** (0.005)	0.013 (0.013)	0.006 (0.013)	0.030* (0.018)
Technical	0.001 (0.004)	0.001 (0.004)	−0.004 (0.004)	0.016*** (0.003)	0.017*** (0.003)	0.019*** (0.004)
Service assistant	−0.020*** (0.006)	−0.007 (0.005)	0.001 (0.005)	0.020*** (0.006)	0.018*** (0.006)	0.003 (0.007)
Trade	−0.011** (0.005)	−0.019*** (0.004)	−0.013*** (0.004)	−0.004 (0.008)	−0.007 (0.008)	−0.014 (0.010)
Elementary	−0.015*** (0.005)	−0.006 (0.004)	−0.008* (0.004)	−0.013*** (0.004)	−0.012*** (0.004)	−0.015*** (0.005)
Private household
Elementary	−0.018*** (0.005)	−0.014*** (0.004)	−0.012*** (0.004)	0.004 (0.003)	−0.004 (0.003)	−0.023*** (0.004)
Agriculture
Operator	−0.023*** (0.008)	−0.025*** (0.007)	−0.021*** (0.007)	0.033 (0.051)	0.004 (0.052)	−0.021 (0.069)
Elementary	−0.026*** (0.005)	−0.015*** (0.005)	−0.014*** (0.005)	0.005 (0.007)	−0.011 (0.007)	−0.012 (0.010)
Mining
Trade	0.031*** (0.006)	0.019*** (0.006)	0.004 (0.005)	0.072 (0.067)	0.080 (0.067)	0.147 (0.089)
Operator	0.022*** (0.005)	0.011** (0.004)	0.001 (0.004)	−0.040 (0.067)	0.002 (0.067)	0.052 (0.090)
Elementary	−0.007 (0.008)	0.001 (0.007)	−0.005 (0.007)	0.050 (0.038)	−0.020 (0.038)	0.026 (0.051)
Manufacturing
Trade	0.014*** (0.005)	0.007 (0.005)	0.003 (0.005)	0.009 (0.010)	0.019* (0.010)	0.002 (0.014)
Operator	0.009** (0.004)	0.009** (0.004)	0.002 (0.004)	0.007 (0.009)	0.002 (0.009)	0.003 (0.012)
Elementary	0.026*** (0.006)	0.011** (0.005)	−0.000 (0.005)	0.026*** (0.009)	0.026*** (0.009)	0.019 (0.012)
Construction and utilities
Trade	0.012*** (0.004)	0.008** (0.003)	0.001 (0.003)	0.009 (0.020)	−0.001 (0.020)	−0.009 (0.026)
Elementary	0.008 (0.006)	0.002 (0.005)	−0.008 (0.005)	−0.021* (0.013)	0.021 (0.013)	0.040** (0.017)
Retail and wholesale trade
Clerk	0.001 (0.009)	0.005 (0.009)	0.011 (0.008)	0.020*** (0.008)	0.028*** (0.008)	0.018* (0.011)
Service assistant	0.011* (0.006)	0.008 (0.005)	0.007 (0.005)	0.022*** (0.006)	0.026*** (0.006)	0.021** (0.009)
Trade	0.012* (0.006)	−0.004 (0.006)	−0.009* (0.006)	0.046*** (0.014)	0.047*** (0.014)	0.017 (0.018)
Elementary	0.018*** (0.006)	0.003 (0.006)	0.002 (0.006)	0.023*** (0.008)	0.026*** (0.008)	0.019* (0.010)
Financial services and IT
Service assistant	0.008 (0.005)	0.001 (0.005)	−0.008* (0.005)	0.006 (0.019)	0.020 (0.019)	0.035 (0.026)
Elementary	0.019** (0.009)	−0.005 (0.008)	−0.007 (0.007)	0.004 (0.009)	0.009 (0.009)	−0.007 (0.012)
Public administration
Clerk	−0.008 (0.011)	0.007 (0.010)	0.005 (0.009)	0.010 (0.013)	0.010 (0.013)	0.053*** (0.018)
Service assistant	0.012* (0.007)	0.013** (0.006)	0.027*** (0.006)	0.007 (0.020)	0.006 (0.020)	0.043* (0.026)
Elementary	0.026*** (0.007)	0.018*** (0.006)	0.004 (0.006)	0.031** (0.013)	0.024* (0.013)	0.035** (0.017)
Education
Professional	0.018 (0.011)	0.013 (0.010)	0.014 (0.009)	0.001 (0.008)	0.009 (0.008)	0.036*** (0.011)
Technical	−0.009 (0.008)	−0.005 (0.007)	0.007 (0.007)	−0.008 (0.006)	−0.002 (0.006)	0.033*** (0.008)
Elementary	−0.013 (0.014)	−0.030** (0.013)	−0.025** (0.012)	−0.003 (0.014)	0.016 (0.014)	−0.007 (0.019)
Health
Technical	0.013 (0.015)	0.012 (0.014)	0.010 (0.013)	−0.003 (0.008)	0.003 (0.008)	0.028*** (0.011)
Elementary	−0.002 (0.015)	−0.002 (0.013)	−0.008 (0.013)	−0.011 (0.009)	0.004 (0.009)	0.027** (0.012)
Other services
Operator	0.013** (0.005)	0.010** (0.005)	−0.000 (0.004)	0.038 (0.028)	0.066** (0.028)	0.071* (0.037)
Elementary	0.010 (0.007)	0.003 (0.007)	−0.005 (0.007)	0.014 (0.012)	0.010 (0.012)	−0.002 (0.016)
Other sector–occupations
Senior	−0.004 (0.005)	−0.001 (0.005)	0.001 (0.005)	−0.000 (0.010)	−0.000 (0.010)	0.020 (0.013)
Professional	0.010 (0.007)	0.003 (0.006)	0.020*** (0.006)	−0.003 (0.007)	0.006 (0.008)	0.003 (0.010)
Technical	−0.003 (0.004)	−0.003 (0.004)	0.005 (0.004)	0.014 (0.009)	0.019** (0.009)	0.037*** (0.012)
Clerk	0.013** (0.007)	0.014** (0.006)	0.014** (0.006)	0.006 (0.007)	0.020*** (0.007)	0.031*** (0.009)
Service assistant	0.023*** (0.006)	0.015*** (0.005)	0.002 (0.005)	−0.005 (0.007)	0.011 (0.007)	0.013 (0.009)
Trade	0.026*** (0.007)	0.016** (0.006)	0.019*** (0.006)	0.011 (0.027)	0.033 (0.027)	0.117*** (0.036)
Operator	0.008* (0.004)	0.003 (0.004)	0.007* (0.004)	0.033** (0.015)	0.019 (0.015)	−0.014 (0.020)
Percentage of the birth cohort not economically active	0.002 (0.002)	0.001 (0.001)	0.002 (0.001)	0.007*** (0.002)	0.008*** (0.002)	0.007*** (0.002)
Mean years of education among employed workers in birth cohort	0.441*** (0.049)	0.553*** (0.040)	0.503*** (0.038)	0.353*** (0.041)	0.436*** (0.040)	0.442*** (0.053)
Mean years of education among employed workers in birth cohort squared	−0.023*** (0.003)	−0.030*** (0.003)	−0.028*** (0.003)	−0.012*** (0.003)	−0.019*** (0.003)	−0.020*** (0.004)
Age of birth cohort	0.099*** (0.012)	0.133*** (0.010)	0.162*** (0.010)	0.104*** (0.010)	0.116*** (0.010)	0.129*** (0.013)
Age of birth cohort squared	−0.001*** (0.000)	−0.001*** (0.000)	−0.002*** (0.000)	−0.001*** (0.000)	−0.001*** (0.000)	−0.001*** (0.000)
Constant	1.838*** (0.294)	1.638*** (0.216)	1.669*** (0.181)	0.625*** (0.174)	1.007*** (0.158)	1.793*** (0.180)
Observations	1 049	1 049	1 049	1 049	1 049	1 049
Number of cohorts	52	52	52	52	52	52

Notes: Standard errors in parentheses. The proportion () of cohort c that is unemployed (U) in period t is omitted.

***p < 0.01.

**p < 0.05.

*p < 0.1.

5. Discussion

The estimates presented in the previous section suggest that the effects of the employment sector and skill level on earnings within birth cohorts are less pronounced than those for age and, in the case of females, education. However, there is (generally) a positive correlation between earnings and an increase in skilled employment (in particular among clerks) even after we consider the effects of age and education. This is in line with the results presented in Bhorat & Goga (2013).

The relationship between the different skill levels of employment and earnings, however, is mediated by the sector of employment. There is evidence of a positive relationship between less skilled employment in the private and public sectors (again generally) and the first quartile of the earnings distribution within these one-year birth cohorts. We observe such a positive relationship between both the first and second quartiles of the earnings distribution for unskilled and semi-skilled employment in the manufacturing sector and in the retail and wholesale trade sector.⁴ In contrast, we find that there is a negative relationship between earnings for unskilled labour in self-employment, particularly among females, and for both unskilled and semi-skilled male labour in agriculture.

Another important finding is that there appears to be very little scope for both employment and earnings growth among unskilled Africans South Africans in self-employment. The coefficients associated with the proportion of the cohort in elementary self-employment are all negative for both males and females in all three quartile specifications. While we are unable to distinguish between formal and informal employment, a large proportion of the unskilled self-employed are probably in informal enterprises. Consequently it appears that any attempts to expand self-employment among the unskilled and uneducated, other than to address chronic poverty and other externalities, may be misguided.⁵ Yet there is evidence to suggest that increases in the proportion of skilled African South Africans female entrepreneurs are associated with increases in earnings.

The coefficients associated with elementary-occupation public employment are also all positive. This highlights the vital role that public employment has played in redistribution since the end of apartheid. It should be noted, however, that the coefficients presented in Table 3 are not large. For example, a one percentage point increase in elementary occupation employment in the manufacturing sector is associated with a 2.6% increase in the first quartile of the earnings distribution within birth cohorts. We must also emphasise that these results may be driven by individual-level heterogeneity, where for example particular individuals in elementary occupations within these one-year birth cohorts are matched to particular employment sectors (including public administration). Further, any analysis from these results is limited in important ways. In particular we do not explicitly trace the interdependence between employment across birth cohorts and sectors, occupations and for particular levels of education in employment. Secondly, it is not clear why the directions of the relationships between levels of employment in particular occupations in sectors and the earnings quartiles follow the forms presented in the previous section. Finally, the existence of a positive relationship between earnings and employment in elementary occupations in particular sectors may also be indicative of the effect of lay-offs in these sectors on the earnings distribution of African South African birth cohorts since the end of apartheid.

6. Conclusion

Since the transition to democracy in 1994 unemployment has generally been increasing among unskilled African South Africans while real wages paid to unskilled African workers in this country have not grown substantially. These trends can be attributed to structural changes in the economy, and overcoming the structural barriers that have an effect on the labour market in this country will require improvements to the quality of the education that formerly disadvantaged South Africans are exposed to. In the interim, the government will have to address this inequality through active labour market policies that attempt to promote both employment and an increase in the real wages of unskilled African South Africans.

While there are reasons to believe that simultaneously achieving these two objectives (higher levels of employment and higher wages) is unlikely, we contribute to the literature on the labour market outcomes of African South Africans since the end of apartheid by showing that there is a positive correlation between the proportion of any given birth cohort employed in elementary occupations in the manufacturing and the retail and wholesale trade sectors and increases in the first quartile of the earnings distribution within African male and female birth cohorts since 1997. In contrast, we find that there is a negative relationship between this quartile and the proportion of the birth cohort employed in elementary occupations in agriculture, private households and self-employment.

These are important findings because they suggest that the relationship between earnings and the level of employment among unskilled Africans in South Africa does not necessarily require a trade-off. Thus it is possible that, for example, minimum wage regulation may not necessarily reduce employment in particular economic sectors. Adelzadeh & Alvillar (2016) also highlight the potential sectoral differences in the employment effect of the proposed National Minimum Wage in South Africa. It should nevertheless be noted that the estimates we present in this article show that the magnitude of the positive relationship is not substantial. More importantly, we cannot identify the mechanisms that lead to the relationships between the proportions of unskilled Africans who are employed in particular economic sectors and the earnings distribution for these workers.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Swiss Agency for Development and Cooperation and the Swiss National Science Foundation under the Swiss Programme for Research on Global Issues for Development (project Nr. 400340_147836).

Notes

1 It is important to note that we do not explicitly explore the differences in labour market outcomes between birth cohorts. The purpose of using a birth-cohort pseudo-panel is to eliminate the effects of these differences from the analysis.

2 We also include those born in 1992 (i.e. those aged 20 in 2012), but they are not used in the sample that is used to estimate the specifications we outline in Section 4 because we only have data for the first quarter of 2012.

3 Although there is neither consensus on what sufficiently large is nor on the asymptotics that should be used to demonstrate the consistency of these models.

4 Indeed, the retail and wholesale trade sector appears to have contributed significantly to the earnings of African females across quartiles.

5 Unless these interventions raise the productivity associated with this employment by, for example, allowing unskilled firm owners to expand into different markets or up the value chain, as opposed to driving other firms from existing markets and thus further exacerbating the dire situation of any losers.

Appendix 1

Table A1. Selected descriptions decomposing the variation between and within African male birth cohorts in the PALMS.

		Mean	Standard deviation	Minimum	Maximum	Observations
Real earnings quartile for African birth cohort (in year 2000 Rand)
First	Overall	700	175	179	1315	N = 1 101
	Between		85	480	825	n = 52
	Within		156	241	1190	T-bar = 21.1731
Second (median)	Overall	1322	326	390	2394	N = 1 101
	Between		205	872	1597	n = 52
	Within		263	591	2118	T-bar = 21.1731
Third	Overall	2304	713	653	6160	N = 1 101
	Between		394	1439	2888	n = 52
	Within		609	1070	5576	T-bar = 21.1731
Percentage of African birth cohort
Unemployed	Overall	15.80	7.39	0.50	40.27	N = 1 507
	Between		6.57	4.59	26.24	n = 53
	Within		4.26	1.81	36.11	T-bar = 28.434
Not economically active	Overall	26.14	13.82	8.44	84.13	N = 1 507
	Between		13.78	15.19	72.69	n = 53
	Within		10.00	1.58	86.37	T-bar = 28.434
Percentage of African birth cohort employed in sector
Agriculture	Overall	4.48	2.03	0.00	17.65	N = 1 507
	Between		1.60	0.84	8.34	n = 53
	Within		1.70	−1.79	13.79	T-bar = 28.434
Self-employed	Overall	9.13	4.23	0.10	29.71	N = 1 507
	Between		3.46	0.76	14.33	n = 53
	Within		3.29	−2.47	25.96	T-bar = 28.434
Mining	Overall	4.31	3.07	0.00	17.56	N = 1 507
	Between		2.16	0.00	7.64	n = 53
	Within		2.30	−1.75	15.56	T-bar = 28.434
Manufacturing	Overall	7.97	3.09	0.00	18.75	N = 1 507
	Between		2.31	1.01	10.24	n = 53
	Within		2.55	0.94	17.30	T-bar = 28.434
Construction and utilities	Overall	5.90	2.31	0.00	16.85	N = 1 507
	Between		1.34	1.34	7.07	n = 53
	Within		2.14	−0.10	17.03	T-bar = 28.434
Retail and wholesale trade	Overall	6.73	2.61	0.31	15.81	N = 1 507
	Between		1.73	2.15	8.58	n = 53
	Within		2.20	−0.64	16.86	T-bar = 28.434
Financial services and IT	Overall	4.26	2.59	0.00	14.34	N = 1 507
	Between		1.79	1.11	7.02	n = 53
	Within		1.96	−2.40	12.50	T-bar = 28.434
Private household	Overall	2.63	1.49	0.00	14.77	N = 1 507
	Between		0.84	0.42	3.99	n = 53
	Within		1.33	−0.53	14.12	T-bar = 28.434
Public administration	Overall	3.81	2.24	0.00	17.22	N = 1 507
	Between		1.74	0.09	6.75	n = 53
	Within		1.60	−0.89	15.05	T-bar = 28.434
Education	Overall	2.56	1.78	0.00	8.32	N = 1 507
	Between		1.44	0.00	5.03	n = 53
	Within		1.12	−0.38	8.65	T-bar = 28.434
Health	Overall	1.18	0.98	0.00	11.38	N = 1 507
	Between		0.49	0.00	1.91	n = 53
	Within		0.90	−0.73	11.34	T-bar = 28.434
Other services	Overall	4.85	2.05	0.07	17.35	N = 1 507
	Between		1.31	0.55	6.28	n = 53
	Within		1.82	0.25	16.68	T-bar = 28.434
Percentage of African birth cohort employed in occupation
Senior	Overall	2.71	1.88	0.00	17.86	N = 1 507
	Between		1.27	0.00	4.25	n = 53
	Within		1.51	−1.02	16.84	T-bar = 28.434
Professional	Overall	1.54	1.20	0.00	8.25	N = 1 507
	Between		0.70	0.08	2.70	n = 53
	Within		1.04	−0.93	7.30	T-bar = 28.434
Technical	Overall	5.73	3.04	0.00	25.33	N = 1 507
	Between		2.48	0.54	12.25	n = 53
	Within		2.44	−3.26	18.81	T-bar = 28.434
Clerk	Overall	2.77	1.33	0.00	7.40	N = 1 507
	Between		0.78	0.46	3.75	n = 53
	Within		1.19	0.00	7.55	T-bar = 28.434
Service assistant	Overall	7.43	3.28	0.37	20.97	N = 1 507
	Between		2.45	0.64	10.47	n = 53
	Within		2.46	−1.98	19.91	T-bar = 28.434
Trade	Overall	11.94	4.17	0.97	26.09	N = 1 507
	Between		3.34	1.85	14.88	n = 53
	Within		3.38	0.84	24.62	T-bar = 28.434
Operator	Overall	10.54	4.49	0.17	26.11	N = 1 507
	Between		4.03	0.74	14.78	n = 53
	Within		2.94	−0.95	26.03	T-bar = 28.434
Elementary	Overall	15.15	3.99	2.32	29.89	N = 1 507
	Between		2.98	3.76	17.96	n = 53
	Within		3.50	2.62	28.12	T-bar = 28.434

Table A2. Selected descriptions of the variation between and within African female birth cohorts in the PALMS.

		Mean	Standard deviation	Minimum	Maximum	Observations
Real earnings quartile for African birth cohort (in year 2000 Rand)
First	Overall	387	118	154	782	N = 1 101
	Between		77	228	567	n = 52
	Within		101	56	654	T-bar = 21.1731
Second (median)	Overall	716	205	285	1770	N = 1 101
	Between		114	460	963	n = 52
	Within		184	242	1663	T-bar = 21.1731
Third	Overall	1561	603	414	4162	N = 1 101
	Between		254	917	2069	n = 52
	Within		558	320	3880	T-bar = 21.1731
Percentage of African birth cohort
Unemployed	Overall	15.42	9.27	0.00	42.55	N = 1 507
	Between		8.46	2.55	27.40	n = 53
	Within		4.83	−0.30	33.48	T-bar = 28.434
Not economically active	Overall	44.19	12.86	20.05	87.73	N = 1 507
	Between		11.76	33.32	74.73	n = 53
	Within		9.18	8.87	91.12	T-bar = 28.434
Percentage of African birth cohort employed in sector
Agriculture	Overall	1.71	0.92	0.00	5.70	N = 1 507
	Between		0.48	0.12	2.32	n = 53
	Within		0.83	−0.18	5.70	T-bar = 28.434
Self-employed	Overall	8.38	4.93	0.00	31.28	N = 1 507
	Between		3.76	0.43	12.56	n = 53
	Within		3.78	−3.43	28.74	T-bar = 28.434
Mining	Overall	0.18	0.25	0.00	1.83	N = 1 507
	Between		0.09	0.00	0.35	n = 53
	Within		0.23	−0.18	1.78	T-bar = 28.434
Manufacturing	Overall	2.69	1.38	0.00	7.92	N = 1 507
	Between		0.90	0.56	3.85	n = 53
	Within		1.19	−0.19	8.37	T-bar = 28.434
Construction and utilities	Overall	0.72	0.62	0.00	3.97	N = 1 507
	Between		0.23	0.11	1.08	n = 53
	Within		0.59	−0.37	4.11	T-bar = 28.434
Retail and wholesale trade	Overall	4.90	2.45	0.00	12.99	N = 1 507
	Between		1.83	1.61	7.20	n = 53
	Within		1.83	−1.51	12.21	T-bar = 28.434
Financial services and IT	Overall	2.14	1.54	0.00	7.75	N = 1 507
	Between		0.97	0.27	3.38	n =  53
	Within		1.26	−1.24	6.62	T-bar = 28.434
Private household	Overall	9.73	4.81	0.34	29.61	N = 1 507
	Between		4.38	0.83	15.01	n = 53
	Within		2.74	1.48	24.33	T-bar = 28.434
Public administration	Overall	1.47	0.99	0.00	6.43	N = 1 507
	Between		0.53	0.00	2.31	n = 53
	Within		0.89	−0.63	5.59	T-bar = 28.434
Education	Overall	3.84	2.49	0.00	13.36	N = 1 507
	Between		2.15	0.00	6.90	n = 53
	Within		1.44	−1.59	11.20	T-bar = 28.434
Health	Overall	3.23	1.75	0.00	9.00	N = 1 507
	Between		1.26	0.10	4.74	n = 53
	Within		1.39	−0.23	7.70	T-bar = 28.434
Other services	Overall	1.30	0.89	0.00	11.17	N = 1 507
	Between		0.43	0.44	1.93	n = 53
	Within		0.81	−0.56	10.77	T-bar = 28.434
Percentage of African birth cohort employed in occupation
Senior	Overall	0.92	0.78	0.00	6.76	N = 1 507
	Between		0.39	0.00	1.44	n = 53
	Within		0.71	−0.45	6.30	T-bar = 28.434
Professional	Overall	1.42	1.11	0.00	11.64	N = 1 507
	Between		0.59	0.00	2.41	n = 53
	Within		0.99	−0.77	11.46	T-bar = 28.434
Technical	Overall	5.91	3.38	0.00	28.66	N = 1 507
	Between		2.55	0.23	8.38	n = 53
	Within		2.59	−0.71	27.17	T-bar = 28.434
Clerk	Overall	3.39	2.09	0.00	10.18	N = 1 507
	Between		1.61	0.44	5.49	n = 53
	Within		1.50	−1.98	8.76	T-bar = 28.434
Service assistant	Overall	5.39	2.19	0.00	12.59	N = 1 507
	Between		1.32	1.86	6.95	n = 53
	Within		1.95	−0.44	11.79	T-bar = 28.434
Trade	Overall	1.98	1.16	0.00	10.85	N = 1 507
	Between		0.70	0.11	2.96	n = 53
	Within		1.01	−0.76	9.88	T-bar = 28.434
Operator	Overall	1.16	0.81	0.00	5.31	N = 1 507
	Between		0.44	0.00	1.71	n = 53
	Within		0.73	−0.40	5.15	T-bar = 28.434
Elementary	Overall	20.12	8.00	1.55	40.52	N = 1 507
	Between		7.50	2.74	27.90	n = 53
	Within		4.74	4.13	37.76	T-bar = 28.434