High Unemployment Yet Few Small Firms: Appendix: Model, Maps, and Auxilliary Tables¶

Literature › Works

Jeremy R. Magruder (2000). High Unemployment Yet Few Small Firms: Appendix: Model, Maps, and Auxilliary Tables. Online appendix to the paper 'High Unemployment Yet Few Small Firms'.


Type	working paper
Scope	Post-1994 South African economy
Method	econometric
Period studied	2000–2003
Regions	South Africa (magisterial districts), Furniture manufacturing, construction, retail trade, hotels and restaurants
Themes	labour markets wages, inequality poverty unemployment, unions strikes, methods data sources, transition post apartheid
Citation key	`magruder2000`

Argument¶

The appendix supports the main paper's claim that industry-level bargaining council agreements in post-apartheid South Africa raise wages and reduce employment, especially in small firms. A partial-equilibrium model has bargaining council wages set above the market-clearing wage but below the private union wage, so small firms and non-unionized large firms cut employment. Large unionized firms may expand, and total employment effects depend on firm composition. The appendix describes how bargaining council coverage is mapped and instrumented with district-council eligibility. It then reports robustness checks: worker composition, spatial-industry-year fixed effects, and weighted spatial fixed effects. These show that the employment and wage patterns hold across a wide range of assumptions about spatial heterogeneity.

Evidence and method¶

Magisterial-district-level coding of bargaining council coverage from labour law for four industries over 2000-2003. Labour force survey data on wages, employment, tenure, worker characteristics and unionization rates. The appendix does not name the survey series beyond 'labour force surveys'. Firm-size categories and population controls are also used.

Findings¶

The instrument (bargaining council eligibility) strongly predicts actual bargaining council status. (p. 2)

In all speciﬁcations, the t-statistic of Bargaining council status on bargaining council eligibility is at least 9.8.
The model predicts that small firms will employ fewer workers under a bargaining council, while the net effect on large firms and total employment is ambiguous. (p. 6)

small ﬁrms will overall employ fewer workers under the bargaining council regime, while the direction of the large ﬁrm employment eﬀect will depend on λ and the production function.
Controlling for worker composition reduces the estimated wage effect somewhat, and the clearest compositional change is a higher male share of the workforce. (p. 6)

controlling for these observable characteristics attenuates the eﬀect of bargaining councils on wages somewhat, with estimated eﬀects dropping by about 5 percentage points.
Bargaining councils are associated with longer average tenure, consistent with hiring and firing regulation. (p. 7)

Mean tenures are increasing by about 18-19% in response to bargaining council regulations
Using only spatial variation (spatial-industry-year FEs) gives much larger employment effects than the main specification. (p. 8)

using only spatial variation, we would conclude that employment eﬀects were 10-23% and small ﬁrm employment eﬀects were 26-37% rather than the 8-12% estimates using the main speciﬁcation.
Wage effects under spatial-industry-year FEs are also larger and concentrated among small-firm employees, with small or no effects on large firms. (p. 9)

large wage eﬀects overall and on small ﬁrm employees, and small or nonexistent wage eﬀects on large ﬁrm employees.
Employment effect estimates are stable across spatial weighting schemes, but significance depends on the weight placed on local comparisons. (p. 11)

coeﬃcient estimates of the eﬀect of bargaining councils on employment are very stable to spatial heterogeneity, ranging from about a 7% to an 8% eﬀect
Small-firm employment effects remain significant under all spatial weights. (p. 12)

point estimates range from about 7% to 12% depending on the spatial heterogeneity, and point estimates are always statistically signiﬁcant.
Coverage changes over time are limited in some industries; hotels and restaurants barely change. (p. 14)

Magisterial Districts with a BC 31 32 32 32