Unemployment: causes, measurement and labour market institutions¶
Is South African unemployment structural or cyclical, caused by wage rigidity and institutions or by demand and structural factors, and how should it be measured?
Works engaging the debate: 21
Positions¶
- Wage rigidity and labour costs (Fedderke, Fallon and Lucas, Nattrass and Seekings)
- Structural, demand and capital-intensity explanations (Kaplinsky, Maree)
- Broad vs narrow definitions; Simkins's estimates vs Sadie's
- Bargaining councils and institutional wage floors (Moll)
Evidence base¶
The debate rests on two bodies of evidence that rarely meet. The first is the measurement literature of the 1970s and early 1980s. It reconstructed African unemployment indirectly, as labour supply minus recorded employment. Its inputs were the 1946, 1951, 1960 and 1970 population censuses, the Department of Statistics' employer-reported non-farm employment series (South African Statistics, Bulletin of Statistics), agricultural censuses for farm and casual labour, and Tomlinson Commission norms used to impute employment in Reserve or Homeland agriculture from output. Knight (1976a) and Knight (1976b) did this most carefully. They cross-checked census coding instructions and Department of Statistics memos, and applied Sadie's undercount adjustments (about 10% before 1970 and 3.9% in 1970). They bracketed 'residual labour' under alternative assumptions and concluded that the data could not support firm trends. Simkins' full-time-equivalent series, known through (Bromberger (1978)), (Hindson (1978)) and Bell and Padayachee (1984), put unemployment and underemployment at about 18.3% in 1960 and 22.4% in 1977. The first direct measure, the October 1977 Current Population Survey, gave 12.4% (634,000) on a narrow definition (Bromberger (1978)). Small local surveys supplemented these aggregates: Maree and Cornell's Crossroads sample, the Market Research Africa Omnireef survey, and Hofmeyr's Vulindlela survey of 55 unemployed men among some 270 respondents (Hofmeyr (1984)). Greenberg (1987) added interviews with labour bureau and TEBA officials in 1982. Firm-level evidence in this period consists of small purposive samples. These are the 1973 manufacturing survey of 282 respondents (Nattrass (1977)), 14 Durban firms (Sitas et al. (1984)), eleven British firms (Harvey (1974)), Knight's three company case studies, and a single colliery cost comparison (Spandau (1980b)).
The second body is the causal literature of the 1990s and 2000s, built on sectoral panels and the first national microdata. Fedderke and Mariotti (2002a), (Fedderke and Mariotti (2002b)) and Fedderke (2006) use the TIPS panel of 48 three-digit SIC sectors (28 in manufacturing), annual 1970-97, covering employment, remuneration, value added, capital stock and skill composition. They estimate a long-run manufacturing wage elasticity of about -0.5 to -0.55 by pooled mean group, and add WEFA and Chamber of Mines mining series and Customs-based effective protection rates. Kaplinsky (1995) works instead from the IDC 76-sector manufacturing database, UNIDO cross-country data on 44 countries and World Bank tables to examine capital intensity and net investment for 1972-90. Heintz (2000) combines Reserve Bank sectoral capital stock and profit-rate series for 1965-95 with an unrest index constructed from prison population, detentions and strikes. Kingdon and Knight (1999) uses the 1993 SALDRU/World Bank household survey (6,498 wage workers in 360 clusters) to estimate a wage curve. The institutional strand is thinner. Moll (1996a) draws on two industrial councils' minimum-wage schedules, a 1985 coverage estimate of 64% and Mincerian regressions on the 1994 October Household Survey. Nattrass and Seekings (2012a) combines the 2004 Labour Force Survey, NBC registration and exemption records, and secondary wage-differential studies for clothing. Nattrass and Seekings (2010b) is a synthetic essay resting only on aggregate SARB and Stats SA GDP series.
Contested evidence¶
Measurement disputes turn on how the pre-survey aggregates are read. Bromberger (1978) argues that Sadie's and van der Merwe's estimates are anchored on the 1970 census self-classification, which undercounted the unemployed. Knight (1976b) shows that changing census definitions make inter-census comparisons of the economically active worthless. Simkins' residual series is therefore the main alternative, but its sensitivity to agricultural employment assumptions is acknowledged by Hindson (1978) and by Loots' variations as reported in Bromberger (1978). Bell and Padayachee (1984) defends Simkins against the charge of cyclical insensitivity. It does so by reading yearly movements off a graph and by visual co-movement with the UK rate, with no formal test. The broad-versus-narrow question has one systematic test: Kingdon and Knight (1999) finds an elasticity of about -0.11 using broad unemployment and near zero using narrow. It reads this as evidence that wage-setters count discouraged workers.
On causation, the sides often rest on different data rather than reading the same data differently. The wage-rigidity case, in Fedderke and Mariotti (2002a) and Fedderke (2006), relies on sectoral employment-real wage correlations and a manufacturing PMG labour-demand equation. Kaplinsky (1995) argues from IDC sectoral investment shares that high aggregate capital intensity reflects underinvestment in labour-intensive branches rather than distorted factor prices, but his political explanation is untested. The user cost of capital shows the clearest conflict on overlapping SARB-derived series. Heintz (2000) finds no significant effect on investment. Fedderke's labour-demand estimates find a negative user-cost effect, and Bell and Padayachee (1984) uses five-year averages of real interest rates to reject capital-labour substitution. Heintz and Fedderke both use political-instability indices (Fedderke, de Kadt and Luiz) yet interpret them differently: Heintz as unresolved distributive conflict, Fedderke as property-rights uncertainty favouring a liberal reading. Firm evidence also diverges. In Nattrass (1977), 74% of firms reported no employment change after 1970-75 wage rises. Knight (1976a) and Spandau (1980b) document wage- and strike-induced mechanisation in sugar, timber and coal, but from a handful of cases. On institutions, (Moll (1996a)) infers exclusionary effects from covered-uncovered wage gaps of 77% and 131% and from union premia regressions. Nattrass and Seekings (2012a) argues that studies finding negligible bargaining-council effects (Godfrey et al., Bhorat et al.) understate them because they omit unregistered firms. Its evidence is a comparison of LFS clothing employment with 56,000 NBC registrations. Neither side has data linking council coverage to employment at firm level. Finally, (Fedderke and Mariotti (2002a)) stresses a long-run structural problem dating from the 1970s, yet its own panel shows a sharp 1990 break, with 31 of 48 sectors shrinking in 1990-97.
Underused and missing sources¶
Several sources that could settle points are absent from the evidence ledgers. The first is manufacturing census plant-level microdata, with consistent establishment panels of employment, wages and capital; Kaplinsky (1995), (Fedderke (2006)) and Moll (1996a) all note its absence. The second is disaggregated industrial council and Wage Board wage schedules linked to registered and unregistered firms. Others are labour bureau registration and placement series and influx-control arrest data, which would test the breakdown and job-queue claims of Greenberg (1987) and Hofmeyr (1984), and Chamber of Mines occupational wage and employment records. No evidence ledger traces the original sources behind the TIPS series, so coverage of informal, homeland and farm employment cannot be assessed (Fedderke and Mariotti (2002b)).
Strength of key claims¶
| Basis | Share of key claims |
|---|---|
| Systematic data | 18% |
| Partial data | 60% |
| Illustrative or anecdotal | 15% |
| Asserted | 7% |
Comparative figures for all debates are on the Literature overview.
Best-evidenced works¶
Kingdon and Knight (1999), Fedderke and Mariotti (2002a), Fedderke (2006), Kaplinsky (1995), Heintz (2000), Knight (1976b)
Most-used primary sources¶
Holdings are matched by machine against the archive's file manifest; a dash means no match was found, not that the project certainly lacks the source.
Works and their positions¶
| Work | Position |
|---|---|
| Bell and Padayachee (1984) | Defends Simkins' series as cyclically sensitive and indicating a genuine structural upward trend; argues absolute levels depend on arbitrary definitions and matter less than movements over time. |
| Bell and Padayachee (1984) | Rejects it as the main cause: low interest rates also have a scale effect, and manufacturing data show employment slowed mainly because output and capital stock growth fell, not because capital intensity rose. |
| Bell and Padayachee (1984) | Holds that long-term involuntary structural unemployment is real and explicable by global and domestic growth slowdown. |
| Bell and Padayachee (1984) | Sceptical of privatisation, restrictive monetary and fiscal policy, and reliance on the informal sector. Gives regional policy only limited value, and sees foreign exchange and trade policy as crucial. |
| Bromberger (1978) | Favours the Simkins/Knight 'full-time job equivalent' approach for capturing labour underutilisation. He notes that Sadie and van der Merwe resist it, and that the CPS definition is narrow and excludes discouraged workers and migrants resting in the Reserves. |
| Bromberger (1978) | Treats the level as unresolved, but argues that all studies agree on a rising trend in the 1970s and on levels likely above 10%. He judges Sadie's census-anchored estimate 'fatally open' to the objection that the unemployed were undercounted in 1970. |
| Bromberger (1978) | Argues that the possibility deserves serious attention, as choices made within constraints (influx control, low pay, deferred pay, labour allocation). He rejects the view that Africans simply work as little as possible and says the bulk of unemployment is involuntary. |
| Bromberger (1978) | Rejects low capital accumulation as the cause. Treats capital deepening, imported technology, high African wages, subsidies, the colour bar and political incentives to mechanise as plausible but under-researched explanations. |
| Fedderke (2006) | Argues that real wages have outrun productivity in mining and elsewhere, that wage elasticities are large and negative, and that wage moderation is the key employment lever. |
| Fedderke (2006) | Finds output is insignificant for gold employment while real labour cost is significant; labour cost is the extra source of job loss, alongside falling demand. |
| Fedderke (2006) | Finds no evidence that liberalisation caused manufacturing job losses; Stolper-Samuelson effects raised demand for labour, while technology was labour-saving. |
| Fedderke and Mariotti (2002a) | Argues that real wages outpacing productivity and supply-side labour market features, along with capital-augmenting technical change, explain weak job creation. Recent macro or trade policy is not blamed; trade liberalisation is not a cogent explanation. |
| Fedderke and Mariotti (2002a) | Favours flexibility, defined as letting wages adjust to productivity. It also calls for human capital investment. |
| Fedderke and Mariotti (2002a) | Finds 1990 plausibly a structural break in sectoral employment growth, but sees the break as general across factor markets and not solely caused by labour policy. |
| Fedderke and Mariotti (2002b) | Argues that real wages relative to productivity are an important determinant of employment, and that labour market policy raising wage and non-wage costs reduced incentives to hire. It also accepts that technological change was capital-augmenting and that structural problems predate the 1990s. |
| Fedderke and Mariotti (2002b) | Rejects this. Poor job creation is a long-standing structural feature since the 1970s, and trade liberalisation demand factors do not explain it, citing Fedderke, Shin and Vaze (2000). |
| Fedderke and Mariotti (2002b) | Finds an output elasticity of labour demand of about unity, so manufacturing growth is not jobless. |
| Fedderke and Simkins (2004) | Blames real wages rising faster than productivity and labour-market rigidity; reports strongly negative wage elasticities and argues that trade liberalisation did not cause labour shedding. |
| Greenberg (1987) | Notes that van der Merwe, Loots and Simkins disagree on size but agree that unemployment rose through the 1970s. Greenberg favours the higher estimates (Simkins 22.4% in 1977) and sees a structural surplus. |
| Harvey (1974) | Argues the effect is more complex than claimed. Higher wages can bring productivity gains, the wage bill share is small, and demand shifts can create jobs, so a strong case remains for higher minimum wages. |
| Heintz (2000) | Treats South African unemployment as largely classical (insufficient labour-absorbing capital accumulation), with some cyclical Keynesian elements. |
| Hindson (1978) | Structural and chronic since at least the Second World War and probably most of the century. Cyclical unemployment is only the visible tip. |
| Hindson (1978) | Hindson criticises academic advocates (van der Merwe, Knight), saying they reinforce an ideology that treats African employment and residence in 'white' areas as a withdrawable privilege. |
| Hofmeyr (1984) | Inconclusive. The evidence is mostly consistent with both models, but the author favours a dual labour-market model. In it search or queuing for high-wage jobs explains much unemployment, while simple job inadequacy cannot be ruled out. |
| Hofmeyr (1984) | Sides with the segmentation and supply-side view for this peri-urban sample during an upswing or early recession, while stressing that the finding is specific to time and place. |
| Hofmeyr (1984) | Adopts a recognition-based definition after Sen. A person must want work at the going wage and be actively searching. Low-income 'poverty' underemployment is excluded as a separate question. |
| Kaplinsky (1995) | Rejects the World Bank (Fallon), SACOB and Normative Economic Model view that high black wages cause capital intensity and unemployment. Factor prices are a relatively unimportant factor, and the explanation is mainly political. |
| Kingdon and Knight (1999) | Finds a negative wage curve in South Africa, with a convex or U-shape whose upward part appears only at very high unemployment rates. It treats Harris-Todaro as a long-run relationship compatible with a short-run wage curve. |
| Kingdon and Knight (1999) | Supports the broad definition: broad unemployment significantly affects wages while narrow unemployment does not, so discouraged workers act as genuine labour supply. |
| Kingdon and Knight (1999) | Sub-group results are consistent with efficiency-wage behaviour and local bargaining. Union and public sector wages are flat because non-local bargaining, notably through Industrial Councils, insulates them. |
| Kingdon and Knight (1999) | The homeland/non-homeland division is the clearest fault-line; the rural-urban division does not capture segmentation well. |
| Kingdon and Knight (1999) | Reconciles a positive wage-unemployment relation over time with a negative one across space as an outward shift of the wage curve, probably from increased bargaining power. |
| Knight (1976a) | Sceptical of a general scarcity. Argues that residual labour and underemployment have grown in absolute numbers, that shortages are specific (physically fit young men, particular regions), and that the data cannot settle the question. |
| Knight (1976b) | Argues the evidence points to a large and rising surplus of unskilled labour. The mining industry's view of a drying-up pool is reconciled with this only by a shortage of physically fit young men. |
| Moll (1996a) | Moll sides with the second view: ergo omnes creates a 'wage cliff', protects insiders and large firms, depresses small-firm entry and growth, and raises measured unemployment. He grants that the evidence is suggestive, not conclusive. |
| Moll (1996a) | He treats institutional wage floors, not only apartheid-era racial barriers, as a source of the covered/uncovered divide that persists after apartheid, shifting the divide from white vs African to insider vs outsider. |
| Nattrass (1977) | Argues that in the short run Black employment is not very wage-elastic and that substitution of cheaper Black for dearer White labour can make rising Black wages compatible with rising Black employment. |
| Nattrass and Seekings (2010b) | Argues unemployment rose from the 1970s as semi-skilled capital-intensive production displaced unskilled labour, reinforced by union and industrial-council wage floors. |
| Nattrass and Seekings (2012a) | Nattrass and Seekings argue that councils and extensions destroy low-wage, labour-intensive jobs and that aggregate studies understate this. They reject Godfrey et al. (negligible employment effect), Altman and Valodia, and Roskam. |
| Nattrass and Seekings (2012a) | They criticise Pollin et al. for dismissing the link, arguing that reform would create new low-wage jobs rather than cut existing wages. They also hold that high wages for less skilled workers exacerbate poverty and inequality. |
| Nattrass and Seekings (2012a) | They see continuity: the industrial council system of 'civilised wages' was reincarnated by COSATU unions as 'living wage' and 'decent work' bargaining. |
| Sitas et al. (1984) | Rejects the view that unions and the state together block capital's job creation. Argues that organised labour's class interest lies in full employment, while acknowledging real tensions between the employed and the unemployed. |
| Spandau (1980b) | Argues that South Africa has no coherent policy and that the state has done little to remove factors it created that favour capital-intensive choices, in contrast to Zimbabwe. |