WIDER Working Paper 2020/122
Fiscal policy, labour market, and inequality
Diagnosing South Africa’s anomalies in the shadow of racial
discrimination
Giorgio d’Agostino,
1
Francesco Giuli,
1
Marco Lorusso,
2
and
Margherita Scarlato
1
October 2020
1
Roma Tre University, Rome, Italy;
2
Newcastle University, Newcastle, UK; corresponding author: margherita.scarlato@uniroma3.it
This study has been prepared within the UNU-WIDER project Southern Africa—Towards Inclusive Economic Development
(SA-TIED).
Copyright © UNU-WIDER 2020
Information and requests: publications@wider.unu.edu
ISSN 1798-7237 ISBN 978-92-9256-879-5
https://doi.org/10.35188/UNU-WIDER/2020/879-5
Typescript prepared by Gary Smith.
The United Nations University World Institute for Development Economics Research provides economic analysis and policy
advice with the aim of promoting sustainable and equitable development. The Institute began operations in 1985 in Helsinki,
Finland, as the first research and training centre of the United Nations University. Today it is a unique blend of think tank,
research institute, and UN agency—providing a range of services from policy advice to governments as well as freely available
original research.
The Institute is funded through income from an endowment fund with additional contributions to its work programme from
Finland, Sweden, and the United Kingdom as well as earmarked contributions for specific projects from a variety of donors.
Katajanokanlaituri 6 B, 00160 Helsinki, Finland
The views expressed in this paper are those of the author(s), and do not necessarily reflect the views of the Institute or the
United Nations University, nor the programme/project donors.
Abstract: Inequality in South Africa is the enduring legacy of racial discrimination. We use a
dynamic perspective to show the linkages between persistent effects of discrimination in the
labour market and the efficacy of redistributive fiscal policy in reducing inequality. We present a
machine-learning analysis based on household survey data in the Post-Apartheid Labour Market
Series to predict the main drivers of the relationship between workers’ heterogeneous
socioeconomic characteristics, the behaviour of variables related to labour market status, and
labour income inequality. The empirical investigation covers the period 2000–17. Drawing on
this preliminary evidence, we build a dynamic stochastic general equilibrium model with a dual
labour market and job search frictions that represents the structural features of South Africa’s
economy, which can be used to assess the effects of fiscal policy on inequality in the post-
apartheid period and to simulate the effects of alternative fiscal measures and labour market
reforms.
Key words: inequality, discrimination, job search, labour market, general equilibrium
JEL classification: D50, D63, J71, J46
1 Introduction
Extreme inequality is the most salient problem of post-apartheid South Africa, despite the widening of
social security guaranteed by the African National Congress (ANC) government, which won power in
1994 and is still the ruling party today. This issue has received much attention, and even if a number
of causes have been analysed in important studies,
1
an extensive literature shows that increasing unem-
ployment and wage differentials in the labour market are the main drivers of overall inequality in the
post-apartheid era (see Aguero et al. 2007; Finn and Leibbrandt 2018; Leibbrandt et al. 2018, among
many others).
This paper contributes to this literature, providing a broad narrative of a number of stylized facts re-
lated to inequality and proposing a theoretical model that can be useful for estimating the sensitivity to
economic policy shocks of labour market dynamics and distributional trends. This analysis exploits the
recent improvement of the household survey data integrated in the Post-Apartheid Labour Market Series
(PALMS), which allows a reliable comparison over time of inequality measures (Merrino 2020).
A body of literature emphasizes that in the post-apartheid period a mix of rising labour costs, skill
shortages, and weak international competitiveness has led to capital deepening in the manufacturing
sector and a transition to a service-led economy that is high-skill intensive (Bhorat et al. 2020a, 2020b;
Rodrik 2008). In this view, the main explanations of structural unemployment and rising wage gaps
are wage rigidities, centralized wage bargaining, strong union power at the service of the labour elite,
and weaknesses in the educational system that have not been addressed, notwithstanding the increasing
government spending in education. Other studies show that corruption and an expanding public sector
progressively put fiscal balances under strain, imposing austerity measures that have reduced the efficacy
of the redistribution implemented through the expansion of social grants (Bhorat et al. 2020b; Pons-
Vignon and Segatti 2013; Scarlato and d’Agostino 2019).
In this paper, we approach this subject stressing a different perspective: the lingering effects of racial
discrimination in the labour market. Indeed, a sharp social stratification along racial lines has been
sustained over time through a labour market in which members of the middle class and elite are formally
employed with a permanent work contract and union coverage, whereas vulnerable workers, who are
overwhelmingly African individuals, are employed in precarious relationships or more often are either
unemployed or economically inactive (Schotte et al. 2018; Zizzamia et al. 2019). Note that, even
if the share of Africans in the middle class has increased from 47 per cent in 2008 to 64 per cent in
2017, only about one-fifth of the population can be considered middle class and Whites are represented
disproportionately highly in this class relative to their population share. In addition, the elite is almost
homogeneously White (Zizzamia et al. 2019). Similarly, Assouad et al. (2018) show that the legacy of
the apartheid system in South Africa is a society characterized by a dualistic structure and absence of a
broad middle class comparable in size to that in high-income countries.
A related aspect is the combination of high levels of unemployment for Black workers and worker dis-
couragement (Burns et al. 2010; Kingdon and Knight 2004, 2006). According to Bhorat et al. (2020b),
the dualistic nature of the labour market might shape expectations of discouraged work-seekers, re-
inforcing the advantages of a small portion of highly-skilled workers who easily obtain secure and
well-paid jobs in the formal economy while the larger portion of the labour supply have to compete for
low-security and low-paid jobs in the informal economy. Zizzamia (2020) complements this evidence,
showing that Black, urban youths turn down or quit wage work because they face non-negligible dis-
1
For example, Alvaredo and Atkinson (2010) explain the evolution of top income shares in South Africa with the former
colonial legacy, the effects of apartheid, and the concentration of its natural resource wealth. For a review, see Bhorat et al.
(2020a).
1
incentives to accept low-quality jobs in the formal labour market, given alternative sources of support
in the informal economy that increase their outside options. This means that, for disadvantaged South
Africans workers, the benefits of employment are not significantly greater than the costs. Interestingly,
this study also shows that for urban, unskilled, Black workers with relatively strong outside options (i.e.
social networks and state grants), wage employment is often deliberately temporary. On this basis, ‘bad’
forms of wage employment can be reconceptualized as ‘survivalist’ just as certain forms of informal
self-employment are considered (Zizzamia 2020).
As dualism is reflected in the historical racial segregation in the labour market, and this is the key channel
through which inequality persists in South Africa, the theoretical framework of our analysis distinguishes
two sides of the economy: the formal economy that is predominantly populated by White individuals;
and the informal economy that is populated by Black individuals. What is peculiar in the South African
case is that the informal economy integrates informal employment
2
and systems of informal social pro-
tection based on the redistribution within households of income earned through precarious wage work
in the formal sector and social grants (Burns et al. 2010; Scarlato and d’Agostino 2019).
3
To underpin these features of South Africa’s economy in the theoretical model, we first present a pre-
liminary empirical analysis that uses household survey data and applies age–period cohort analysis and
machine-learning methods to highlight the relationship between workers’ socioeconomic characteris-
tics, labour market status of Black and White workers in the two segments of the economy, and the
correlation of these variables with labour income inequality.
Drawing from this empirical evidence, we then build a dynamic stochastic general equilibrium (DSGE)
model with a dual labour market and search frictions that can be used to assess the effects of alternative
fiscal policies in such a segmented context. In further research, this model can be applied to test whether
a redistributive policy focused on the extension of social grants to the Black population represents an
efficient tool to mitigate inequality or may exacerbate the dual structure of the labour market and, hence,
indirectly contribute to worsening the cumulative disadvantages of the Black population.
The paper is structured as follows. Section 2 presents the dataset and provides the preliminary empirical
analysis. Section 3 proposes a DSGE model for South Africa’s dual economy. Section 4 concludes the
paper and suggests some policy implications and possible extensions of the paper.
2 Race, hidden employment, and inequality: a data-driven analysis
2.1 Data
We use the PALMS dataset provided by the DataFirst research unit based at the University of Cape Town.
The dataset consists of high-quality microdata from 69 household surveys conducted by Statistics South
Africa over the period 1993–2019. The harmonized data include the Household Surveys from 1994
to 1999, the Labour Force Surveys from 2000 to 2007, and the Quarterly Labour Force Surveys from
2008 to 2019 (see Finn and Leibbrandt 2018; Kerr and Wittenberg 2019; Merrino 2020). We restrict the
analysis to the period 2000–17 for reasons related to data availability.
2
On the features and size of the informal labour market in South Africa, see Rogan and Skinner (2018).
3
Social grants benefit about 31 per cent of the country population (SASSA 2019). Over 50 per cent of South Africans live in
households with grant income, and grants represent the most important source of income for half of these households (Zizzamia
2020).
2
In order to analyse the dynamic of inequality, we first transform the data and set-up a pseudo-panel
framework that is useful for analysing the time behaviour of the variables of interest when only indepen-
dent repeated cross-sectional data are available or, as in this case, when we collect data from different
sources. Pseudo-panels observe cohorts (i.e. stable groups of individuals) instead of observing individ-
uals over time, and individual variables are replaced by their intra-cohort means. When the aggregation
of different data sources introduces a measurement error in the used data, intra-cohort mean values are
still consistent if the used survey data are representative of the studied population.
The main concern in applying the pseudo-panel is related to the trade-off between bias and variance in
the formation of the cohort. Indeed, the cohort must be large enough to limit the extent of measurement
errors in intra-cohort variable means that generate biased and imprecise estimators of the model parame-
ters. Table A1 in the Appendix reports the frequency of observations for each birth cohort (from 1929 to
1999) and calendar year of our sample covering the period 2000–17. The table reveals that the number
of individual observations in each year/cohort is large enough to set-up a proper cohort analysis.
4
After excluding self-employed from the sample, we extracted relevant data on labour income deflated
with the 2015 CPI.
5
We restrict the analysis to individuals who are aged between 18 and 65 and exclude
individuals in the range 15–17 to avoid problems related to intra-cohort averaging. To further reduce the
measurement error in the used data, we use weights to correct for bracket responses and take account of
changes in the population distribution, so as to allow for comparisons over time.
The upper part of Table 1 reports some descriptive statistics of monthly labour income and the Gini index
by cohort and year. Furthermore, from the PALMS dataset, we extract several variables describing the
socioeconomic characteristics of the individuals. In particular, we define the informal sector by using
three measures reported in the PALMS dataset, indicating: when the worker has a written contract, when
the job is formal, and when the firm operates in the formal market and is VAT registered. Table 1 reports
the selected variables along with some descriptive statistics.
Several papers warn that methodological changes in the construction of wages and earnings data oc-
curred from the third quarter of 2012 onwards and discuss the shortcomings of the inequality measures
obtained by using the PALMS dataset (Finn and Leibbrandt 2018; Kerr and Wittenberg 2016, 2017).
For this reason, our results should be taken with caution.
2.2 Labour income distribution: a cohor t analysis
Our starting hypothesis is that racial discrimination and cumulated disadvantages for the Black popu-
lation have underpinned the division of the labour market in two segments, which we name the formal
and informal economy, characterized by different workers’ socioeconomic characteristics and structural
dynamics. White workers are mainly employed in the formal economy, whereas most Black workers
are segregated in the informal economy. The formal economy is regulated by centralized bargaining
and high employment protection legislation (EPL) standards, and is characterized by the traditional di-
vision between work and non-work, employment and unemployment. Differently, we conceptualize the
informal economy referring to the stylized view of labour underutilization that includes labour under-
utilization according to hours available to work, marginal labour market attachment, and people outside
any form of paid employment—the unemployed and the hidden unemployed or discouraged workers
(Baum and Mitchell 2010). We coalesce these different labour force states into the informal economy to
emphasize the fluid picture of the broad labour market outcomes of Africans.
4
Following Deaton (2018), the frequency of observations has to be not less than 100 in each year/cohort.
5
Following the definitions of the PALMS dataset, labour income is measured by the variable ‘monthly earnings’ generated
from data across all waves where earnings amounts were asked and data have been released.
3
