South Africa has very high unemployment. Depending on the definition used, the latest national unemployment rate is 31% or 42% (StatsSA, 2003). This may well be the result of a rapidly growing labor force, on the one hand, and declining formal employment, on the other: the labor market may lack the flexibility to cope with this divergence (Lewis, 2002; Moll, 1993; Boccara and Moll, 1997; Fallon and Lucas, 1998). Indeed, there are reasons to expect that the South African labor market is unusually inflexible. Based on surveys of company officials, the World Economic Forum's Global Competitiveness Report (1999: Tables 7.02-7.05, 7.09) ranks South Africa at the bottom of its fifty-nine nation comparison on whether labor regulations on wages, hours or dismissals favor flexibility. The trade union movement is apparently powerful and allied with the government. Moreover, there is an institutional framework - the Bargaining Councils and the Wage Boards - to set minimum wages in parts of the formal sector. However, recent research, examining the union - non-union and bargaining council . non-bargaining council wage differentials, has found this source of inflexibility to be minor (Schultz and Mwabu, 1998; Butcher and Rouse, 2001): these studies estimate the union premium to be similar to that found in other countries, including the US and the UK. Another potential test of labor market inflexibility concerns the responsiveness of wages to local labor market conditions. Our purpose is to measure this by examining the spatial relationship between wages and unemployment. The estimated elasticity of the wage with respect to the local unemployment rate can then be compared with the elasticities found in other countries.
There is accumulating evidence of a negative relationship between local unemployment and wages in various economies. This relationship has been called 'the wage curve' and claimed as an empirical 'law' of economics by Blanchflower and Oswald (1994). Using mesoeconomic analysis1, Blanchflower and Oswald present an impressive array of evidence in favor of the wage curve from the US, UK, and some other OECD countries. The evidence shows that wages are negatively related to contemporaneous unemployment, and that the unemployment elasticity of the wage is approximately -0.1, so that a 10% increase in the unemployment rate leads to a 1% decrease in wages.2 Carruth and Oswald (1987) and Blanchflower and Oswald (1990) provide theoretical models of the labor market consistent with the wage curve. In a searching review article, Card (1995) does not contest the validity of the wage curve, taking issue instead only with the interpretation of the results and the robustness of the elasticity.
In developing countries, by contrast, the relationship between wages and unemployment has generally been analyzed in the context of probabilistic models of migration (Harris and Todaro, 1970, and the vast literature, mainly theoretical, that followed). Unemployment is viewed as the equilibrating variable in a labor market characterized by segmentation and wage inflexibility: high urban wages attract high urban unemployment. The presumption is that unemployment is a function of the wage and not vice versa. There has been little testing of either causal relationship in developing countries using microeconomic datasets. There are two notable exceptions. Hoddinott (1996), contrary to the prediction of probabilistic migration models, found evidence of a wage curve in urban Cote d.Ivoire, with an elasticity similar to those in OECD economies, and Van der Meulen Rodgers and Nataraj (1999) adduced evidence of a long run wage curve in Taiwan.
The object of this paper is to test for the wage curve relationship in a developing country with very high unemployment, using data from South Africa where the average unemployment rate exceeds 30%. This is several times the typical OECD unemployment rate and twice that in urban Cote d.Ivoire. If a wage curve can be shown to exist in South Africa, this will throw further doubt on the importance in practice of probabilistic models in developing countries (Banerjee, 1983). Given that a wage curve is found, are wages as flexible at the high South African unemployment rate as in other countries?
There are two further reasons why the finding of a wage curve in South Africa is of interest. First, when unemployment is very high, its definition itself becomes an issue. In these conditions, the proportion of people actively searching for work is likely to depend on the number wanting work. If many unemployed people stop actively searching for work because they become discouraged, then it may be misleading to measure the unemployment rate by considering as unemployed only those who actively looked for work, i.e. by the narrow definition. The wage curve can assist choice among alternative definitions of unemployment by showing which measure of the unemployment rate is more important in explaining wages. If, as an ILO report on the South African labor market suggests (ILO, 1996, p. 104), those wanting work but not actively seeking (included in the 'broad' definition of unemployment) are outside the labor force, we would expect the wage curve to be steeper for narrow than for broad unemployment.
Secondly, the wage curve can have implications for poverty: a negative relationship may mean that high unemployment in a locality not only has a direct effect on poverty among its households but also an indirect effect via the lower wages of their employed members.
The paper is organized as follows. Section 2 presents a brief theoretical analysis of the relationships between wages and unemployment. Section 3 provides the South African context, and Section 4 discusses the data, model, and tests. The empirical results appear in Section 5, and Section 6 concludes.
Applying microeconomic data and methods to macroeconomic questions.
This finding is now corroborated by evidence from a large number of OECD countries.