Skip to content

Union Wage Premium in South Africa - an instrumental variable approach

Literature › Works

Øystein Jørgensen (2009). Union Wage Premium in South Africa - an instrumental variable approach. Unpublished working paper (work in progress, version of June 12, 2009).
Type working paper
Scope Post-1994 South African economy
Method econometric
Period studied 2001–2004
Regions South Africa
Themes unions strikes, labour markets wages, inequality poverty unemployment, methods data sources, transition post apartheid
Citation key jrgensen2009

Argument

The paper estimates the causal union wage premium for African and white workers in post-apartheid South Africa. It uses a novel instrument, the relative union density of other household members (net of local sector or occupation density), as a proxy for social and political motivation to join a union, estimated by 2SLS instead of a Heckman selection model. Instrumenting cuts the African premium from about 47 to 37 percent. The average conceals large differences: premiums are substantial in non-traded, public-sector and expanding industries, but near zero in traded and declining industries. The author reads this as evidence of insider power constrained by international competition and job security.

Evidence and method

Statistics South Africa Labour Force Survey panel, six waves from September 2001 to March 2004. The sample is formal-sector African and white workers aged 25-65 in households with at least two workers. The paper also uses the COSATU 2005 workers survey and earlier South African studies (Schultz and Mwabu, Butcher and Rouse, Hofmeyr and Lucas, Banerjee et al., Armstrong and Steenkamp).

Findings

  • Instrumenting for union membership lowers the African union premium from about 47 to 37 percent, still high by international standards. (p. 3)

  • Large raw pay gap between unionized and non-unionized African workers in the LFS panel. (p. 4)

    The average hourly wage was 6.8 Rand1 for non-unionized African workers and 18.8 Rand an hour for union members (Table 1).

  • The instrument is a strong predictor of African union membership, supporting its relevance. (p. 11)

    The instruments have substantial explanatory power for Africans.

  • The OLS premium for Africans is 47 percent, while the 2SLS estimate is 37 percent but not statistically distinguishable. (p. 14)

    When I instrument for union membership the estimate for the union wage premium drops by 10 percentage points to 37 % (exp(0, 313) −1)

  • Evidence suggests a larger union impact in the public sector once endogeneity is controlled. (p. 19)

    Although the evidence is not conclusive it seems plausible that unions have a larger impact in the public sector

  • In declining industries the union premium disappears, while expanding industries show a high premium, which links union power to job security. (p. 21)

    The union wage premium in the sectors with highest net loss of jobs has practically disappeared.

  • The paper finds little support for the claim that union wage gains cause job loss. (p. 22)

    The argument that unions raise wages and that this leads to the loss of jobs get little support.

  • Policy conclusion: unions are constrained by trade competition in traded sectors and show insider-outsider behaviour; the author recommends traded-sector wage leadership and more centralization. (p. 22)

    making the traded sector wage leader and strive for more centralization rather than less, seems particularly important to reduce insider power in the labor market