How large is the wage penalty in the labour broker sector? Evidence for South Africa using administrative data

The use of temporary employment  has grown both globally and in South Africa (Deakin 2002; Benjamin et al. 2010). In part, this is related to firms requiring lower adjustment costs in certain economic environments, such as poor macroeconomic conditions (Holmlund and Storrie 2002) or when there is a need to become more competitive (Matsuura et al. 2011; Saha et al. 2013). Holmlund and Storrie (2002) find that poor macroeconomic conditions in Sweden in the 1990s resulted in employers offering more temporary contracts, and employees being more willing to accept this form of employment. In Japan, global competition in tradable goods led to a rapid increase in temporary employment, specifically in those sectors where the bulk of sales was to foreign markets (Matsuura et al. 2011). Similarly in India, both pro-worker labour institutions and increased import penetration led to greater use of contract labour in the Indian manufacturing sector (Saha et al. 2013). In South Africa, it has been suggested that trade liberalization led to firms externalizing employment because of the drive to lower wages in sectors where there has been increased competition (Theron 2005).

Given the context in which temporary employment grows, it is widely expected that there would be a wage differential between temporary workers and non-temporary workers (Lass and Wooden 2017). Indeed, a wage penalty for temporary workers has been found in a number of countries, including India (Saha et al. 2013), Portugal (Boeheim and Cardoso 2007), Germany (Pfeifer 2012), Britain (Brown and Sessions 2005), and the US (Segal and Sullivan 1997; Houseman 2001). International evidence on the size of the wage penalty for temporary workers, after adjusting for demographic factors and job characteristics or controlling for fixed effects, suggests a penalty ranging from 6 per cent in the UK (Booth et al. 2002) to around 20 per cent in France and the US (Segal and Sullivan 1997; Blanchard and Landier 2002). Picchio (2006) estimates a wage penalty for temporary workers of around 12–13 per cent in Italy, but this declines with the seniority of temporary workers, with a reduction in the wage gap of about 2.3 percentage points after one year of tenure.

While the wage gap tends to decline after controlling for certain characteristics, where the gap does persist for temporary workers is in terms of benefits, such as pension, medical aid, and unemployment insurance. Temporary workers have been found to have far lower levels of access to benefits than permanent workers, even after controlling for factors such as race, education, and location (Houseman 2001). This suggests that employers use labour brokers as a way to lower costs in terms of both the base wage and benefits.

In South Africa, the public debate on temporary employment services (TES), often referred to as the labour broker sector, has largely centred around the issue of decent work, and specifically the wages and benefits afforded to temporary workers (Bhorat et al. 2016). The focus on discrimination in this sector resulted in amendments being made in early 2015 to the part of the Labour Relations Act (LRA) that governs temporary employment. The new legislation attempted to better regulate the TES industry and offer greater protection to temporary workers. However, there is little empirical evidence on the extent of a penalty to temporary employment service sector workers in South Africa, mostly because current South African labour force surveys do not explicitly capture this sector.

Before they were replaced by the Quarterly Labour Force Surveys (QLFS), the earlier biannual Labour Force Surveys (LFS) for the years 2000 to 2007 did ask employees whether they were employed by a labour broker. The final LFS survey, conducted in September 2007, provided an estimate of 11 million employees in the country, of whom 37 000 (0.3 per cent) were reported as being employed by a labour broker, and 274 000 (2.5 per cent) by a contractor or agency. It has been suggested that this is too low an estimate for South Africa (Budlender 2013). Misreporting on sector of employment or nature of employment contract is a well-known problem in household surveys (Segal and Sullivan 1998), and particularly when there is proxy reporting as in the LFS.

The QLFS, which replaced the LFS in 2008, did not include a similar question. However, to try and identify TES workers, Benjamin et al. (2010) and Bhorat et al. (2016) used the standard industry classification code 889, Business Activities Not Elsewhere Classified, which falls under the broader category Finance and Business Services, and which includes, among a number of other activities, ‘labour recruitment and provision of staff; activities of employment agencies and recruiting organisations; hiring out of workers (labour broking activities)’.  Although it is not possible to separate out the TES sector from the other activities listed under the general code 889, Benjamin et al. (2010) attempted to estimate the size of the TES sector and arrived at a figure of just over 600 000 TES workers in 2008. Budlender (2013) undertook a similar exercise and found that between 2008 and 2012 the number tended to increase year on year, reaching over 865 000 in 2012. The only exception to the steady increase was for 2009, when the number recorded was closer to 883 000, suggesting that the global financial crisis may have resulted in an increased use of temporary employment services. Also cognizant of the limitations of the QLFS data, Bhorat et al. (2016) estimated that there were just under 1 million temporary jobs in 2014. 

Given the broad list of activities within the classification, Budlender (2013) suggests that the 889 code is not a good proxy for TES workers. According to her analysis for 2012, 44 per cent of the workers recorded in this industry are likely to be security guards and 15 per cent are likely to be cleaners in offices, hotels, and the like. These workers are outsourced,  not temporary agency workers. Of the rest, the bulk are likely to be employed internally by the company (rather than the TES firm). Budlender (2013) further noted that while over 93 per cent of the workers falling under this code are employees, 59 per cent of the employees are recorded as having permanent contracts, 22 per cent have contracts of limited duration, and 19 per cent have contracts of unspecified duration. Budlender (2013: 3) writes that ‘while there is widespread agreement that a large number of workers are employed by temporary employment agencies in South Africa, and that the number has grown over time, there is similarly widespread agreement that the available numbers are estimates based on various assumptions rather than more reliable “counts” of the phenomenon’.

In 2015, South Africa Revenue Services (SARS) and the National Treasury (NT) made company and employee income tax administrative data available for research purposes.  It is the first South African data set from the last decade that explicitly captures which firms are labour brokers and also contains individual employee wages. This paper makes use of the administrative panel data for the years 2011 to 2015 to explore whether there is a wage penalty for employees in the labour broker sector, examining both the base wage (the salary less contributions to medical aid, unemployment insurance, pension, etc.) and the total income received from a company. Although the data do not contain many demographic or job characteristics, the panel nature of the data allows us to control for time and individual fixed effects. In other words, we can examine variation in wages for employees who switched between TES and non-TES jobs over the period of the panel. In addition, we examine the temporary employee wage differentials before and after the temporary employment spell. The reason for this is that temporary workers often accept such jobs due to factory closure or after being laid off, and thus wage differentials may reflect the circumstances in which they accept the job rather than the job itself (Segal and Sullivan 1998). Providing empirical evidence on the earnings differential between labour broker workers and other workers in South Africa is an important first step to help inform debates on the role of this sector in the South African labour market.  

The rest of the paper is structured as follows. Section 2 describes the data and definitions used in the analysis. Section 3 presents the descriptive analysis. Section 4 explains the methodology. Section 5 presents the econometric analysis and Section 6 concludes.