LABOUR UNIONS AND WAGE INEQUALITY AMONG AFRICAN MEN IN SOUTH AFRICA

LABOUR UNIONS AND WAGE INEQUALITY AMONG AFRICAN MEN IN SOUTH AFRICA MIRACLE NTULI PRUDENCE KWENDA DPRU WORKING PAPER 13/159 DECEMBER 2013

LABOUR UNIONS AND WAGE INEQUALITY AMONG AFRICAN MEN IN SOUTH AFRICA DEVELOPMENT POLICY RESEARCH UNIT MIRACLE NTULI School of Economc and Busness Scences Unversty of the Wtwatersrand mraclenb@yahoo.com PRUDENCE KWENDA School of Economc and Busness Scences Unversty of the Wtwatersrand prudence.kwenda@wts.ac.za Workng Paper 13/159 ISBN 978-1-920633-00-4 December 2013 DPRU, Unversty of Cape Town 2013 Ths work s lcenced under the Creatve Commons Attrbuton-Non-Commercal-Share Alke 2.5 South Afrca Lcense. To vew a copy of ths lcence, vst http://creatvecommons.org/lcenses/by-nc-sa/2.5/za/ or send a letter to Creatve Commons, 171 Second Street, Sute 300, San Francsco, Calforna 94105, USA.

ABSTRACT One Achlles heel of post-aparthed South Afrca s the growng ntra-racal ncome nequalty, partcularly among Afrcans. Ths paper examnes the role of labour unons n explanng ths phenomenon among Afrcan men gven that labour markets are at the core of ncome nequalty n South Afrca. Usng cross sectonal data drawn from labour force surveys for 2001-10, we fnd a monotoncally declnng unon wage premum. Further, our results ndcate that unons have both compressonary and dsequalsng effects on wages. The dsequalsng effect domnates the compressonary effect suggestng that unons have a net effect of ncreasng wage nequalty among Afrcan men n South Afrca. Ths fndng mples that there s scope for unons to reduce nequalty through ntatves that promote wage compresson. JEL Classfcaton Codes: D31; J31; J51 Keywords: Income Inequalty; Wage Dstrbuton, Labour Unons, Inequalty Decomposton, South Afrca Acknowledgements The authors of ths paper acknowledge helpful comments receved from partcpants of the 2013 MASA and CSAE (Economc development n Afrca) conferences, and Statstcs South Afrca for provdng data used for the study. Workng Papers can be downloaded n PDF (Adobe Acrobat) format from www.dpru.uct.ac.za. A lmted number of prnted copes are avalable from the Communcatons Manager: DPRU, Unversty of Cape Town, Prvate Bag X3, Rondebosch, Cape Town, 7700, South Afrca. Tel: +27 (0)21 650 5701, emal: sarah.marrott@uct.ac.za Correspondng author Mracle Ntul, Assocate Professor School of Economc and Busness Scences, Unversty of the Wtwatersrand, Johannesburg, 1 Jan Smuts Avenue, P. Bag 3, Wts 2050, Braamfonten, Johannesburg, South Afrca Tel: +27 (0)11 717 8123, emal: mraclenb@yahoo.com Recommended ctaton Ntul, M. and Kwenda, P. (2013). Labour unons and wage nequalty among Afrcan men n South Afrca. Development Polcy Research Unt Workng Paper 13/159. DPRU, Unversty of Cape Town. Dsclamer The Workng Paper seres s ntended to catalyse polcy debate. They express the vews of ther respectve authors and not necessarly those of the Development Polcy Research Unt (DPRU).

CONTENTS 1. INTRODUCTION 2 2. BACKGROUND ON INEQUALITY AND LABOUR UNIONS AMONG AFRICANS 3 3. LITERATURE REVIEW 4 4. 4.1 4.2 METHODOLOGY AND DATA ANALYSIS Methodology Data 5 5 8 5. 5.1 5.2 5.3 5.4 5.5 RESULTS Heckprobt Models of Employment and Labour-force Partcpaton Unon Membershp Models Wage models Unons and Wage nequalty Robustness checks 11 11 13 15 16 18 6. DISCUSSION AND CONCLUSION 20 REFERENCES 22 APPENDIX 24

DPRU WORKING PAPER 13/159 1. INTRODUCTION That an unequal dsperson of labour market earnngs les at the heart of an extraordnarly large and ncreasng problem of ncome nequalty n post-aparthed South Afrca cannot be overemphassed. More recently, ths has manfested through growng ntra-racal ncome nequalty where the ncrease s relatvely hgher for Afrcans than other races. Hence conclusons abound that ncreasng ntra-afrcan earnngs nequalty dctates the evoluton of aggregate ncome nequalty n the country (Lebbrandt et al., 2012). Followng South Afrcan hstory and nternatonal lterature, some studes attrbute the ncrease n earnngs nequalty to the dstrbuton of, and returns to, educaton among Afrcans (Mwabu & Schultz, 1996, 2000; Branson et al., 2012). Whle human captal partly explans the de facto wage dstrbuton, the segmentaton of the labour market nto unon and non-unon sectors cannot be gnored, as unons play an mportant part n wage determnaton. Followng the post-aparthed Labour Relatons Act (1995) whch promotes collectve barganng, unonsaton ncreased sgnfcantly and fortfed unons. Ther strength also stems from an nconspcuous lne between ther efforts and poltcal partes n resstng the Aparthed regme (Schultz & Mwabu, 1998; Butcher and Rouse, 2001). As such, labour unons n South Afrca are labelled as beng far too powerful for the country s level of ncome (Banerjee et al., 2008). The de-racalsaton and strengthenng of labour unons n post-aparthed South Afrca motvated research on unons nter- and ntra-racal effects on wages (see for nstance, Moll, 1993; Azam & Rospabé, 2007). A common fndng from the lterature s that unon members receve hgh wage premums. Based on the presence of a unon premum there are perceptons that unons are nstrumental to wage nequalty n South Afrca, yet there s lttle emprcal evdence on ths noton (Hofmeyr & Lucas, 2001). Internatonal lterature shows that the socal effect of unons s two-pronged,.e. unons have a compressonary (wthn group) effect on ther members wages and a dsequalsng (between group) effect on wages of unon and non-unon members va the wage premum (c.f. Card et al., 2004). Thus, focussng on the premum leaves a dearth of knowledge on the net effect of unons on wages. In lght of ths, the contrbutons of ths paper are two-fold. Frst, snce unons dsequalse wages through the premum, we examne the evoluton of the wage premum to assess whether the dstortonary effect was growng over tme. In addton, ths wll update avalable studes whch used data for 1985-2004.e. Hofmeyr & Lucas (2001) and Banerjee et al. (2008). Results from ths wll also enable us to verfy whether labour unons have successfully preserved the benefts of ther members over the 2001-10 decade. Ths s topcal gven contemporary worker protests whch are lnked to loss n confdence towards unons nter ala. Second, the paper wll assess the net effect of unons on wage nequalty among Afrcan men over the same perod. If the compressonary (dsequalsng) effect domnates the dsequalsng (compressonary) effect t mples that unons n South Afrca reduce (ncrease) wage nequalty. The fndngs could sgnal whether labour unons are among enttes drvng ncome nequalty n South Afrca. The study wll use ndvdual data drawn from South Afrcan Labour Force surveys. The rest of the paper s organzed as follows. In secton 2 we present a bref background on nequalty and labour unons n South Afrca. We revew relevant lterature on nequalty and labour unons n secton 3. The methodology and data are dscussed n secton 4. Secton 5 dscusses the results and secton 6 concludes the paper. 2

Labour Unons and Wage Inequalty Among Afrcan Men n South Afrca 2. BACKGROUND ON INEQUALITY AND LABOUR UNIONS AMONG AFRICANS Income nequalty n South Afrca has ts roots n the system of poltcal Aparthed whch was nsttutonalsed n 1948. Aparthed was geared to advance the economc welfare of whte people at the expense of Afrcans, coloureds and Indans (Lebbrandt et al., 2001; van der Berg & Bhorat, 1999). Ths saw the creaton of separate economes for the dfferent racal groups wth whtes beng more advanced and servced by other racal groups. In the process whtes had relatvely hgher access to qualty educaton, superor jobs, wages and workng condtons, and ncome generatng opportuntes than non-whtes n all these Indans fared better than coloureds who n turn were better than Afrcans (Lebbrandt et al., 2001; van der Berg & Bhorat, 1999). Furthermore, poor whtes had the prvlege of state grants whle these were ratoned for non-whtes. Ths undoubtedly skewed the dstrbuton of wages, ncome and access to opportuntes n the economy. More mportantly, the dsadvantaged poston of Afrcan workers was nsttutonalsed n labour relatons, the Bantu Labour Act (1953) and the Industral Conclaton Act (1956) prohbted Afrcans to form or jon regstered labour unons (van der Berg & Bhorat, 1999:7; Bendx, 2001:67). Ths strategcally left whte unons wth power to set wages for all workers n the varous sectors of the economy. Ths arrangement enabled whte workers to ensure that ther wages remaned hgher than others. Afrcan workers ressted ths oppresson whch culmnated nto legslaton of ther labour unons n 1980 leadng to the expanson of unon membershp (Bendx, 2001:77). Most of the new unons represented the nterests of unsklled and sem-sklled workers (manly non-whtes) who were mltant n the face of entrenched unfar labour practces (Bendx, 2001:78). Despte havng a voce n the ndustral relatons systems, Afrcan workers were stll poltcally oppressed. The oppresson led them nto formng allances wth poltcal partes n fghtng for democracy (Fnnemore & van der Merwe, 1996:33; Bendx, 2001:81). Snce the attanment of democracy n 1994, the post-aparthed government has made major changes to the South Afrcan consttuton and effected new polces amed at reducng ncome nequalty. As mentoned earler, the wage determnaton process n the new dspensaton s underpnned by the Labour Relatons Act (1995), whch gves workers freedom to jon labour unons and provdes a platform for collectve barganng. The collectve barganng process occurs at two echelons; centralsed and plant levels. Centralsed barganng occurs when one or more regstered labour unons bargan wth one or more regstered employer organsatons for wages and workng condtons n a partcular ndustry or sector (Bendx, 2001:271; Butcher & Rouse, 2001; Bhorat et al., 2012). Ths happens when labour unons and employer organsatons are representatve of the majorty of ther members, and they have been granted mnsteral approval (Bendx, 2001:271; Butcher & Rouse, 2001). The barganng councl agreements can be extended to non-unon workers n covered sectors. Non-members can, however, seek mnsteral exempton from the condtons (Bhorat et al., 2012). The second level of collectve barganng occurs at ndvdual plants: unons for workers at a plant can bargan wth ther employers for plant specfc wage adjustments and condtons of employment. Ths level of barganng can supplement agreements from barganng councls - unon workers covered by barganng councls receve the same unon premums as those outsde the barganng councl (Butcher & Rouse, 2001). Whle wages n the unon sector are set as dscussed above, those n the compettve sector are set by employers presumably on the bass of ther proft maxmsng decsons. In concluson, the wage settng process for the unon sector has been well documented by earler studes hence here we only presented a summary. See for nstance Moll (1993), Butcher & Rouse (2001) and Bhorat et al. (2012) for more detals. 3

DPRU WORKING PAPER 13/159 3. LITERATURE REVIEW Ths secton presents a bref revew of lterature on ncome nequalty n South Afrca, as well as extant lterature on the mpact of labour unons on wages. There are numerous studes n South Afrca whch document the extent and evoluton of ncome nequalty at household level and n the labour market. Accordng to Hoogeveen & Ozler (2005), the overall Gn of household per capta ncome ncreased from 0.57 n 1995 to 0.58 n 2000. Concomtantly, the contrbuton of between populaton group nequalty to overall nequalty decreased from 38-33%. Ths mples that an ncrease n ntra-racal nequalty accounted for the overall ncrease n nequalty between 1995 and 2000. The ncrease was mostly attrbuted to a rse n ntra-afrcan nequalty as the assocated Gn ncreased from 0.47 to 0.50 over the same perod. Ths result s also supported by Lebbrandt et al. (2012) who further dstngushed the labour market as the man source of aggregate nequalty n South Afrca. Labour market ncome contrbuted between 85% and 88% to the overall Gn coeffcent. Ths fndng concurs well wth Lete et al. (2006) and Tregenna &Tsela (2012). These authors also found that there has been a slght decrease n earnngs nequalty from 2000-2007, when all races are consdered. It s noteworthy that although the lterature has smlar qualtatve conclusons on nequalty trends, t lacks consensus on the magntude of nequalty due to use of dfferent data sets, dfferent ways of constructng ncomes from nterval data, and dfferent samples. Gven that labour markets are at the core of earnngs nequalty n South Afrca, concerted efforts have been nvested towards understandng the role of labour market nsttutons, partcularly unons, n shapng ncome nequalty. Interest n unons was partly roused by the strengthenng of Afrcan labour unons post ther legslaton n 1980, a need to verfy whether labour unons also compress the wage dstrbuton as posted by nternatonal lterature, and a need to understand whether unons lead to wage nflexblty n an era of hgh unemployment (Card et al., 2004; Moll, 1993; Butcher & Rouse, 2001; Schultz & Mwabu, 1998; Azam & Rospabé, 2007; Banerjee et al., 2008; Hofmeyr & Lucas, 2001). A general result from these studes s that unon members receve a wage premum relatve to non-unon members. Although, these studes provde rrefutable evdence for the exstence of a unon wage premum, they do not present a consstent measure of the premum. Some report values as low as 6% (Bhorat et al., 2012), others around 24% (Moll, 1993; Barnajee et al., 2008) and some n excess of 100% (Azam & Rospabé, 2007; Mchaud & Vencatachellum, 2001). Ths large varaton n the premum s largely drven by use of dfferent estmaton samples, methodologes and data sets. An addtonal fndng pertans to the evoluton of the unon wage premum. For nstance, Hofmeyr & Lucas (2001) explored the trend controllng for selecton nto employment and unons. The results showed that the premum ncreased from 1985 to 1993, although the magntudes were not robust across models. In a related study, Barnajee et al. (2008) found a premum whch ncreased from 14% n 1995 to 27% n 2000 and then declned to 23% n 2004. The decrease n recent years rases curosty was t fortutous or t marked the begnnng of a declnng trend n the premum. Ths rases a queston whether the mpact of unons on wage nequalty has weakened concomtantly. Other studes provde evdence suggestng that Afrcan labour unons compress the wage dstrbuton of ther members (Schultz & Mwabu, 1998 and Butcher and Rouse, 2001). Usng quantle regresson, Schultz & Mwabu (1998) found a unon wage premum of 145% at the bottom of the wage dstrbuton and 19% at the top. Ths huge varaton suggests that unons reduce wage nequalty among ther members. In synopss, ths dscusson shows that unons have two relatve wage effects. Frst they reduce wage dsperson n the unon sector. Second, 4

Labour Unons and Wage Inequalty Among Afrcan Men n South Afrca they dsequalse wages between unon and non-unon sectors. Whle ths nformaton s mportant, t does not allow us to deduce the overall wage effect of unons, yet ths s crtcal for understandng whether unons ncrease or reduce wage nequalty n South Afrca. 4. METHODOLOGY AND DATA ANALYSIS 4.1 Methodology The standard treatment effects model s wdely used to estmate the unon non-unon wage dfferental. In ths model, a unon membershp dummy s ncluded n the wage functon to capture the unon premum or penalty. One drawback of ths approach s that t assumes that the coeffcents of personal characterstcs such as educaton and age are the same across sectors (Lee, 1978; Hetmueller, 2006; Azam & Rospabé, 2007). Ths assumpton mples that the wage settng mechansms are the same n unon and non-unon sectors whch mght not be the case. In lght of ths nherent lmtaton of the standard treatment effect model, we estmate the unon wage premum usng an endogenous swtchng regresson model appled by Lee (1978). Ths model allows us to estmate separate wage functons for unon and nonunon members thereby yeldng rcher results wth regards to structural dfferences n wage determnaton between the two sectors. In the endogenous swtchng regresson model, a swtchng equaton sorts ndvduals over two states (Lokshn & Sajaa, 2004) n ths case unon and non-unon sectors. The state n whch an ndvdual s selected nto determnes the wage regme faced by that ndvdual. Formally, the model s specfed as follows: ln w = X β + u U N U ln w = X β + u N U N (1) (2) I = δ w ln w ) + Z γ ε (ln (3) U N Where: I s a latent varable that determnes the sector n whch ndvdual s workng. If I > 0, s selected nto regme U (unon sector) otherwse regme N (non-unon sector) s the outcome; w j the wage of ndvdual n state j = U, N; X Z a vector of ndvdual and job characterstcs that nfluence wages, that s, age, educaton, occupaton, economc sector and locaton. a vector of observed characterstcs that nfluence the sector nto whch ndvdual s selected n ths case t conssts of age, educaton, economc sector, locaton and a dummy varable ndcatng whether an ndvdual lves wth other unon members. An nverse mlls rato to control for selecton nto labour force partcpaton and subsequent employment s also ncluded followng the correcton procedure suggested by Bhorat et al. (2001); δ, γ, β are vectors of parameters to be estmated; u j, ε dsturbance terms (j = U, N). 5

DPRU WORKING PAPER 13/159 Equaton (3) s the swtchng functon whch determnes the sector n whch an ndvdual s employed. Equatons (1) and (2) are the wage functons for a specfc sector. It s assumed that the error terms have a trvarate normal dstrbuton, wth mean vector zero and the followng covarance matrx: 2 σu Ω = σ N1 σ εu σ σ σ UN 2 N εn σ σ Uε Nε 1 Followng standard practce, the varance of the error term of the reduced form swtchng equaton s set to 1 n order to dentfy the parameters δ and γ (Harthog & Oosterbeek, 1991). Gven the assumptons placed on the dstrbuton of the error terms, the logarthmc functon for the wage equatons and swtchng functon s (c.f. Lee, 1979; Lokshn & Sajaa, 2004): ln L = { I ω [ln( F( ηu ) + ln( f ( uu σ U ) σ U ) + (1 I ) ω[(ln(1 F( η N )) + ln( f (( u = 1 N σ ) σ )]} N N (4) 2 Where η j = [ γz ρ j µ j σ j ] 1 ρ j ; F s the cumulatve normal dstrbuton and f s a normal densty dstrbuton functon, ω s an optmal weght for observaton and ρ j the correlaton between μ j and μ. Snce both w U and w N cannot be observed smultaneously for ndvdual, σ does not appear n the lkelhood functon as ths parameter s undentfed. UN Condtonal on unon status, the unon wage equaton s gven by: E(ln w U I = 1) = f ( γz ) X + βu σ Uε (5) F( γz ) Condtonal on non-unon status, the non-unon wage equaton s gven by: E(ln w N I = 0) = X N σ N f ( γz ) 1 F( γz ) β ε (6) The estmated wage equatons allow comparson of wage dfferences between unon and nonunon sectors. To measure the average percentage ncrement of the wage rate for unonmembershp we predct wages usng the estmates: yˆ = ln w = U U X βˆ U (7) yˆ N = ln w N = X βˆ N (8) 6

Labour Unons and Wage Inequalty Among Afrcan Men n South Afrca Followng Lee (1978), the unon-membershp premum s computed as follows: P = 100 n (ey U e y N )/e y N =1 Where n s the sample sze. (9) Usng the estmates from the wage functons, we then examne the mpact of unons on earnngs nequalty (see appendx for formulas). Frst, nequalty measures are computed usng predcted earnngs from equatons (5) and (6). Second, whle mantanng the predcted earnngs for non-unon members usng equaton (6), we replace predcted earnngs of unonmembers by the followng counterfactual earnngs dstrbuton: f ( γz ) E(ln w = 0) = U I X β N σ (10) Uε 1 F( γz ) Equaton (10) provdes the earnngs of unon members under a non-unon regme. We then recompute nequalty measures and assess the extent to whch they dffer. Ths exercse wll enable us to assess the extent to whch the premum contrbutes to nequalty. The second approach used n assessng the mpact of unons on earnngs nequalty s N adopted from Card et al. (2004). Usng a two sector model, let W be the log wage of U ndvdual when employed n the non-unon sector and W be the log wage of the same ndvdual when employed n the unon sector. Assume that: N N W = W + e N (11) U U W = W + e U (12) N U Where e, e are random error terms wth condtonal means of zero. Assumng that n the absence of unonzaton, current unon workers wll receve the same average wage as nonunon members, then the unon non-unon wage gap s gven by: w U = W W N (13) Let Var = U U ( e ) V and Var = N N ( e ) V denote the varance of potental wage outcomes of workers n the unon and non-unon sectors respectvely; then the varance gap s gven by: = V V U N v (14) Let α be the unonzaton rate, then overall varance of log wages s gven by (c.f. Freeman, 1980; Card et al., 2004) V N = V + α + α) v 2 ( 1 w (15) The effect of unons on the varance of wages relatve to what would preval f all workers were pad accordng to the non-unon age regme s: 7

DPRU WORKING PAPER 13/159 V V N = α v + 2 ( 1 α) w (16) Equaton (16) shows that the presence of unons exerts two effects on the dsperson of wages relatve to the counterfactual V N. Frst, the wthn sector effect that s assocated wth the fact that wage dsperson s dfferent n unon and non-unon sectors s captured by α v. Second, the between-sector effect reflectng the wage dsequalzng effect caused by unons s 2 captured by the term ( 1 α ) w. Although ths analyss does not ncorporate dfferences n the rate of unonzaton across dfferent types of skll groups, ths smple framework provdes basc nsghts nto the mpact of unons on earnngs nequalty. 4.2 Data The data used for the study were obtaned from Statstcs South Afrca s September LFSs (Labour Force Surveys) for 2001, 2004, 2007 and the 3rd Quarter of the 2010 QLFS (Quarterly Labour Force Survey). The surveys are natonally representatve contanng socoeconomc nformaton for ndvduals lvng n 30 000 households across the country. The LFSs were reweghted to establsh comparablty wth the QLFS on the bass of the 2001 populaton census. We restrct our sample to Afrcan men aged 15-64 years, provded they had nformaton on our key varables defned n Table A.1, n the appendx. The full samples for 2001, 2004, 2007 and 2010 had 50 485, 51 384, 50 352 and 88 998 ndvduals, respectvely. Table 1 presents economc postons of these ndvduals; that s, strct and broad LFP (labour force partcpaton) rates, employment and unon status. The broad labour force ncludes employed people, actve job seekers and dscouraged workers, whle the strct labour force excludes dscouraged workers. In the ntal sample, the broad LFP rates range from 59 percent to 68 percent whle the strct LFP rates are relatvely lower rangng from 51 percent to 54 percent. Among the partcpants, approxmately 55 percent were employed n 2001 and 2004; ths ncreased to 60 percent n 2007 and 2010. 8

Labour Unons and Wage Inequalty Among Afrcan Men n South Afrca Table 1: Economc Status by perod 2001 2004 2007 2010 mean std. dev mean std. dev mean std. dev mean std. dev Panel A: labour-force partcpaton status Broad labour-force partcpaton 0.667 (0.471) 0.672 (0.470) 0.676 (0.468) 0.587 (0.492) Strct labour-force partcpaton 0.539 (0.498) 0.511 (0.500) 0.537 (0.499) 0.510 (0.500) No. observatons 50 485 51 384 50 352 82 998 Panel B: Employment status Employed 0.559 (0.497) 0.558 (0.497) 0.604 (0.489) 0.608 (0.488) Strct unemployment 0.250 (0.433) 0.203 (0.402) 0.191 (0.393) 0.260 (0.439) Broad unemployment 0.441 (0.497) 0.442 (0.497) 0.396 (0.489) 0.392 (0.488) No. observatons 32 890 32 890 31 643 45 832 Panel C: Unon status Unon member 0.537 (0.499) 0.417 (0.493) 0.406 (0.491) 0.412 (0.492) No. of observatons 4 511 4 332 5 340 8 822 The samples n Panels A and B are used for estmatng sample selecton correcton terms (for LFP and subsequent employment) to be ncluded n our unon membershp functons. To derve the fnal sample used n wage functons, we restrct the sample of employed workers to full-tme employees n the formal non-agrcultural sector. The excluded workers fall outsde the doman of South Afrcan labour unons. Ths restrcton leaves a fnal sample composed of 4511, 4332, 5340 and 8822 Afrcan men n 2001, 2004, 2007 and 2010, respectvely. The correspondng unon membershp rates for these sub-samples are 54 percent, 42 percent, 41 percent and 41 percent. Our statstcs suggest there was some de-unonzaton among Afrcan men between 2001 and 2004, but ths stablzed n later years. Based on the fnal sample depcted n Panel C, Table 2 presents some basc descrptve statstcs of ndvdual characterstcs by unon status and perod. We fnd that log hourly wages ncreased over the perod, and they are on average hgher for unon than non-unon workers. The hgher unon wages unveled here are consstent wth prevous studes n South Afrca and other countres. As ndcated n Table 2, there are notable dfferences across unon and nonunon workers unon members have hgher marrage rates and are generally older than nonunon workers. In addton, the proporton of ndvduals wth educaton above matrc s hgher among unon-members relatve to non-unon members. For nstance, n 2001 the proporton of workers wth educaton above matrc s 11 percent n non-unon compared to 19 percent n unon sector-. Ths pattern s consstent across all years and more evdent n 2007 where the proporton of workers wth educaton above matrc s 10 percent n non-unon sector compared to 25 percent n unon sector. In contrast, workers wth educaton levels below matrc are somewhat concentrated n the non-unon sector. 9

DPRU WORKING PAPER 13/159 Table 2: Descrptve statstcs for unon and non-unon sub-samples by perod 2001 2004 2007 2010 non-unon unon non-unon unon non-unon unon non-unon unon Log hourly wage 2.259 (0.655) 2.510 (0.610) 2.445 (0.706) 2.932 (0.670) 2.751 (0.784) 3.229 (0.762) 3.114 (0.720) 3.723 (0.801) Demographcs: Age 36.002 (9.935) 39.190 (9.014) 35.190 (9.667) 39.761 (9.799) 34.322 (10.081) 39.280 (9.999) 34.468 (9.622) 39.641 (9.954) Marred 0.595 (0.491) 0.741 (0.438) 0.593 (0.491) 0.729 (0.445) 0.493 (0.500) 0.681 (0.466) 0.472 (0.499) 0.635 (0.482) Educaton: No Schoolng 0.052 (0.222) 0.052 (0.221) 0.038 (0.191) 0.050 (0.218) 0.031 (0.173) 0.030 (0.170) 0.022 (0.146) 0.018 (0.131) Prmary 0.210 (0.407) 0.268 (0.443) 0.193 (0.395) 0.215 (0.411) 0.174 (0.379) 0.164 (0.370) 0.135 (0.342) 0.113 (0.317) Incomplete secondary 0.354 (0.478) 0.279 (0.449) 0.365 (0.482) 0.299 (0.458) 0.404 (0.491) 0.320 (0.466) 0.415 (0.493) 0.266 (0.442) Matrc 0.271 (0.445) 0.217 (0.412) 0.295 (0.456) 0.251 (0.433) 0.273 (0.446) 0.283 (0.450) 0.329 (0.470) 0.350 (0.477) Certfcate 0.017 (0.130) 0.020 (0.140) 0.018 (0.133) 0.016 (0.127) 0.015 (0.121) 0.022 (0.147) 0.011 (0.103) 0.022 (0.146) Dploma 0.052 (0.223) 0.117 (0.322) 0.056 (0.230) 0.111 (0.314) 0.067 (0.249) 0.130 (0.336) 0.068 (0.252) 0.166 (0.373) Degree 0.044 (0.204) 0.047 (0.212) 0.034 (0.182) 0.057 (0.233) 0.036 (0.187) 0.052 (0.222) 0.020 (0.141) 0.066 (0.248) Employment sector / ndustry: Publc sector 0.161 (0.367) 0.392 (0.488) 0.119 (0.323) 0.360 (0.480) 0.090 (0.286) 0.335 (0.472) 0.056 (0.229) 0.386 (0.487) Mnng and quarryng 0.044 (0.205) 0.234 (0.424) 0.029 (0.168) 0.185 (0.388) 0.027 (0.161) 0.148 (0.355) 0.025 (0.156) 0.120 (0.325) Manufacturng 0.247 (0.431) 0.197 (0.398) 0.202 (0.402) 0.206 (0.405) 0.203 (0.402) 0.202 (0.402) 0.191 (0.393) 0.205 (0.404) Electrcty, gas & water supply 0.018 (0.133) 0.018 (0.132) 0.009 (0.093) 0.020 (0.139) 0.007 (0.084) 0.013 (0.112) 0.005 (0.072) 0.020 (0.140) Constructon 0.108 (0.310) 0.038 (0.192) 0.146 (0.353) 0.035 (0.184) 0.179 (0.383) 0.035 (0.183) 0.176 (0.381) 0.046 (0.209) Wholesale & retal trade 0.216 (0.412) 0.091 (0.288) 0.230 (0.421) 0.092 (0.289) 0.250 (0.433) 0.133 (0.340) 0.266 (0.442) 0.078 (0.268) Transport, storage & communcaton 0.099 (0.299) 0.069 (0.254) 0.077 (0.267) 0.069 (0.254) 0.085 (0.279) 0.062 (0.241) 0.070 (0.256) 0.080 (0.271) Fnancal servces 0.131 (0.337) 0.060 (0.238) 0.176 (0.381) 0.085 (0.279) 0.148 (0.355) 0.104 (0.305) 0.180 (0.384) 0.105 (0.307) Communty, socal & personal servces 0.137 (0.344) 0.292 (0.455) 0.130 (0.337) 0.308 (0.462) 0.102 (0.302) 0.304 (0.460) 0.086 (0.281) 0.347 (0.476) Occupaton & locaton dummes yes yes yes yes yes yes yes yes No. of observatons 1 819 2 692 2 266 2 066 2 887 2 453 4 856 3 966 Standard devatons n parentheses. 10

Labour Unons and Wage Inequalty Among Afrcan Men n South Afrca We also fnd that most ndvduals are employed n the prvate sector; 60-67 percent of unon workers and 80-94 percent of non-unon workers; compared to the publc sector. A relatvely large proporton of non-unon workers were employed n trade ndustres and fnancal servces, whle those n unon jobs were concentrated n mnng, manufacturng and servces ndustres. For the manufacturng sector, the workers seem to be equally concentrated across unon and non-unon sectors. Most workers are employed as sales workers, artsans, operators and n elementary occupatons regardless of unon status. Also about 80 percent of non-unon workers were n urban areas whereas the same apples to 68-78 percent of unon workers. In general, the majorty of workers were employed n Gauteng and KwaZulu Natal provnces. 5. RESULTS Gven that our model estmates a choce model for unon membershp and wage functons, our dscusson of results s structured as follows: frst, we dscuss the Heckprobt models of employment and labour-force partcpaton, then the correlates of unon membershp. Ths s followed by a dscusson of wage functons. Thereafter we dscuss the lnk between unons and wage nequalty and robustness checks. 5.1 Heckprobt Models of Employment and Labour-force Partcpaton The fndngs for the Heckprobt models of labour force partcpaton and employment are presented n panels A and B of Table 3 respectvely. They show that there s an nverted U age-labour force partcpaton profle for Afrcan men, and smlarly for employment. The probabltes of partcpatng, and employment, are also shown to be hgher for marred men relatve to ther non-marred counterparts. Ths suggests that famly responsbltes whch could be fnancal may drve Afrcan men not only to partcpate, but to be actually employed.. In lne wth human captal theory, the prospects of partcpatng and fndng a job are postvely correlated to educaton relatve to no schoolng (.e. havng matrc educaton and above for employment). The outcomes for the excluson restrctons for the labour force partcpaton model.e. lvng wth a penson-elgble woman (aged above 59 years) or man (aged above 64 years) n a household proxes for non-labour ncome show that havng an elderly woman n a household has non-labour ncome effects on partcpaton, whle the presence of an elderly man s statstcally nsgnfcant. The fndng for Afrcan women s n lne wth Bertrand et al (2003). Panel B also shows that the prospects that an ndvdual wll be employed are postvely correlated wth lvng n a household comprsed of a large proporton of other adults who are employed. In lne wth South Afrcan lterature on socal networks and job status, ths result suggests that other employed household members act as a proxy for the respondent s socal networks on employment opportuntes (e.g. Schoer and Lebbrandt, 2006). Nonetheless, ths fndng should be nterpreted wth cauton, as t s dffcult for us to establsh the tmelnes of the employment events for the respondent and other household members, gven that we are usng cross secton data. It s also noteworthy that ths varable could be overestmated, as Burns et al. (2010) mantan that t does not fully capture all network related effects. The fndngs for the athrho are postve and statstcally sgnfcant, suggestng that the error terms for the labour force partcpaton and employment equatons are postvely correlated; hence, t s sutable to estmate the employment equaton controllng for sample selecton bas. 11

DPRU WORKING PAPER 13/159 Table 3: Labour force partcpaton and Employment Probt models 2001 2004 2007 2010 coeff. std. err coeff. std. err coeff. std. err coeff. std. err Panel A: Labour-force partcpaton Age 0.349 (0.005) *** 0.338 (0.004) *** 0.716 (0.006) *** 0.347 (0.003) *** Age-squared -0.004 (0.000) *** -0.004 (0.000) *** -0.004 (0.000) *** -0.004 (0.000) *** Marred 0.496 (0.029) *** 0.385 (0.042) *** 0.412 (0.040) *** 0.526 (0.028) *** Prmary 0.299 (0.049) *** 0.286 (0.045) *** 0.282 (0.039) *** 0.191 (0.040) *** Incomplete secondary 0.124 (0.047) *** 0.119 (0.047) ** 0.117 (0.034) *** 0.281 (0.042) *** Matrc 0.720 (0.056) *** 0.906 (0.057) *** 0.831 (0.050) *** 0.767 (0.045) *** Certfcate 0.367 (0.110) *** 0.390 (0.091) *** 0.301 (0.098) *** 0.652 (0.105) *** Dploma 0.609 (0.097) *** 1.139 (0.123) *** 0.888 (0.099) *** 1.114 (0.052) *** Degree 0.658 (0.128) *** 0.829 (0.212) *** 0.962 (0.224) *** 1.039 (0.090) *** Presence of elderly men n household (age>64) -0.033 (0.038) -0.073 (0.050) -0.048 (0.040) -0.080 (0.033) ** Presence of elderly women n household (age>59) -0.254 (0.026) *** -0.250 (0.021) *** -0.273 (0.024) *** -0.241 (0.018) *** Constant -5.391 (0.101) *** -5.249 (0.090) *** -5.341 (0.104) *** -5.812 (0.052) *** Panle B: Employment Age 0.152 (0.007) *** 0.156 (0.008) *** 0.164 (0.009) *** 0.155 (0.008) *** Age-squared -0.002 (0.000) *** -0.002 (0.000) *** -0.002 (0.000) *** -0.002 (0.000) *** Marred 0.911 (0.032) *** 0.937 (0.031) *** 0.936 (0.030) *** 0.864 (0.029) *** Prmary -0.002 (0.046) 0.038 (0.053) 0.095 (0.047) ** -0.187 (0.076) ** Incomplete secondary 0.032 (0.057) -0.075 (0.054) 0.054 (0.055) -0.226 (0.079) *** Matrc 0.113 (0.059) * 0.026 (0.062) 0.152 (0.073) ** -0.055 (0.098) Certfcate 0.258 (0.148) * 0.287 (0.132) ** 0.293 (0.147) ** 0.181 (0.086) ** Dploma 0.357 (0.102) *** 0.488 (0.127) *** 0.555 (0.082) *** 0.205 (0.108) * Degree 0.518 (0.128) *** 0.797 (0.195) *** 0.590 (0.137) *** 0.538 (0.124) *** Prop. of other adults n a household who are employed 5.165 (0.109) *** 5.366 (0.102) *** 5.199 (0.082) *** 4.932 (0.071) *** Constant -4.826 (0.141) *** -4.993 (0.164) *** -4.877 (0.151) *** -4.726 (0.169) *** athrho 0.472 (0.035) *** 0.468 (0.029) *** 0.451 (0.027) *** 0.367 (0.039) *** Locaton dummes are also ncluded n the labour force partcpaton and emploment equatons. Reference group: noschoolng. Sgnfcance level: ***=1%, **=5%, *=10%. Standard errors n parentheses. 12

Labour Unons and Wage Inequalty Among Afrcan Men n South Afrca 5.2 Unon Membershp Models Table 4: Margnal Effects for Unon Membershp Probt models, by perod 2001 2004 2007 2010 Reduced form probt Structural form probt Reduced form probt Structural form probt Reduced form probt Structural form probt Reduced form probt Structural form probt Age 0.030 *** 0.030 *** 0.017 *** 0.019 *** 0.030 *** 0.026 *** 0.031 *** 0.033 *** Age-squared 0.000 *** 0.000 *** 0.000 *** 0.000 *** 0.000 *** 0.000 *** 0.000 *** 0.000 *** Marred 0.015 *** 0.024 *** -0.038 *** -0.035 *** 0.053 *** 0.069 *** -0.024 *** -0.024 *** Prmary 0.055 *** 0.073 *** -0.051 *** -0.037 *** 0.102 *** 0.076 *** 0.088 *** 0.077 *** Incomplete secondary 0.037 *** 0.057 *** -0.004 ** 0.012 *** 0.174 *** 0.142 *** 0.132 *** 0.126 *** M atrc 0.044 *** 0.054 *** 0.049 *** 0.059 *** 0.221 *** 0.148 *** 0.223 *** 0.221 *** Certfcate 0.161 *** 0.191 *** 0.039 *** 0.022 *** 0.261 *** 0.184 *** 0.320 *** 0.311 *** Dploma 0.153 *** 0.182 *** 0.098 *** 0.115 *** 0.239 *** 0.220 *** 0.300 *** 0.288 *** Degree 0.043 *** 0.087 *** 0.051 *** 0.123 *** 0.129 *** 0.216 *** 0.213 *** 0.229 *** Mnng and quarryng 0.266 *** 0.266 *** 0.248 *** 0.236 *** 0.385 *** 0.376 *** 0.364 *** 0.370 *** Electrcty, gas & water supply -0.139 *** -0.151 *** -0.082 *** -0.070 *** 0.023 *** 0.084 *** 0.119 *** 0.150 *** Constructon -0.243 *** -0.261 *** -0.217 *** -0.192 *** -0.244 *** -0.219 *** -0.215 *** -0.209 *** Wholesale & retal trade -0.119 *** -0.126 *** -0.147 *** -0.109 *** -0.092 *** -0.026 *** -0.189 *** -0.182 *** Transport, storage & communcaton -0.085 *** -0.093 *** -0.076 *** -0.088 *** -0.063 *** -0.068 *** -0.018 *** -0.021 *** Fnancal servces -0.081 *** -0.087 *** -0.118 *** -0.083 *** -0.011 *** 0.015 *** -0.088 *** -0.071 *** Communty, socal & personal servces -0.043 *** -0.054 *** -0.070 *** -0.076 *** 0.042 *** 0.056 *** -0.049 *** -0.058 *** Large frms (50/more employees) 0.144 *** 0.043 *** 0.157 *** 0.042 *** 0.147 *** 0.037 *** 0.209 *** 0.039 *** Job tenure 0.041 *** 0.043 *** 0.043 *** 0.042 *** 0.041 *** 0.037 *** 0.038 *** 0.039 *** Job tenure squared -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** Selecton term -0.019 *** -0.019 *** -0.012 *** -0.011 *** -0.014 *** -0.013 *** -0.034 *** -0.034 *** Other unon members 0.248 *** 0.148 *** 0.314 *** 0.150 *** 0.334 *** 0.150 *** 0.179 *** 0.208 *** Wage dfferental 0.117 *** 0.153 *** 0.234 *** 0.072 *** Other varables ncluded n all the models are constant term; urban, provncal and publc sector dummes. Reference groups: no schoolng, manufacturng, Gauteng provnce. Results for age-squared are negatve but appear as postve due to roundng off. Sgnfcance level: ***=1%, **=5%, *=10%. 13

DPRU WORKING PAPER 13/159 Table 4 shows margnal effects for unon membershp probt models for 2001, 2004, 2007 and 2010 n reduced and structural forms. The results for reduced form models show an nverted U shaped age-unon membershp profle. Ths suggests that unons attract mddleaged workers relatve to younger and older counterparts. Ths result also apples to job tenure. Also, resdng wth other unon members s postvely correlated wth the prospects of jonng a unon. Accordng to Moll (1993), ths could be reflectng household tastes for unonsm or frms recrutment strateges. Apart from the 2004 sample, the fndngs for educaton show that havng some form of educaton ncreases the probablty of jonng a unon relatve to those wthout schoolng. The prospects were, n many cases, hghest for those wth certfcates and dplomas and the effects ncreased over tme. These fndngs suggest that workers of all educaton levels could be seekng unon protecton aganst possble employer explotaton gven hgh unemployment rates n the country. Workng n large frms ncreases the chances of jonng a unon relatve to small frms because t s easer for unons to organse workers n large frms. The fndngs for 2004-2010 also show that beng n an urban area ncreases opportuntes to jon a unon relatve to a rural area; the converse apples to 2001. Furthermore, workng n the mnng sector avals hgher opportuntes to jon a unon relatve to the base category (manufacturng). The reverse s the case for other ndustres except electrcty, where the effect s not robust. Ths s expected snce mnng and manufacturng ndustres are tradtonally unon hubs. Beng n the publc sector also ncreases chances of unonsm relatve to the prvate sector. Provncal varatons n the probablty of unonsm were also detected. Ths could be due to regonal dosyncratc factors that affect atttudes towards unonsm (Moll 1993). Sample selecton correcton terms are statstcally sgnfcant, mplyng that unobservable characterstcs whch nfluence the employment decson also affect the probablty of jonng a unon once employed. These were computed from the fndngs of Heckprobt models of employment as unon and non-unon wages are only observed for employed people who are a non-random sample of the labour force. Therefore, we controlled for selecton nto employment followng Bhorat & Lebbrandt (2001). Fndngs for the structural models are qualtatvely smlar to those of the reduced form models. They also show that the unon wage dfferental s postvely assocated wth the chances of jonng a unon. 14

Labour unons and wage nequalty among Afrcan men n South Afrca 5.3 Wage models Table 5: Wage Equatons by Unon status 2001 2004 2007 2010 unon non-unon unon non-unon unon non-unon unon non-unon Age 0.015 * 0.019 * 0.015 * 0.025 *** 0.010-0.008-0.007 0.008 Age-squared 0.000 0.000 0.000 0.000 ** 0.000 0.000 0.000 0.000 Marred 0.023 0.096 *** 0.039 0.056 ** 0.012 0.084 *** -0.003 0.002 Prmary -0.023 0.136 ** 0.010 0.100 * 0.110 0.003 0.185 ** -0.073 Incomplete secondary 0.116 *** 0.290 *** 0.088 * 0.195 *** 0.223 *** 0.088 0.271 *** 0.067 Matrc 0.338 *** 0.420 *** 0.332 *** 0.381 *** 0.535 *** 0.215 *** 0.412 *** 0.226 *** Certfcate 0.376 *** 0.666 *** 0.476 *** 0.789 *** 0.721 *** 0.361 *** 0.386 *** 0.049 Dploma 0.493 *** 0.743 *** 0.749 *** 0.696 *** 0.890 *** 0.695 *** 0.889 *** 0.427 *** Degree 0.807 *** 0.973 *** 0.854 *** 1.194 *** 1.214 *** 1.347 *** 0.996 *** 1.042 *** Urban 0.093 *** -0.007 0.112 *** 0.065 * 0.142 *** 0.141 *** 0.180 *** 0.192 *** Mnng and quarryng 0.266 *** 0.266 *** 0.089 ** 0.006 0.068 0.024 0.117 ** 0.027 Electrcty, gas and water supply -0.139 *** -0.151 *** -0.062 0.008 0.006 0.281 ** -0.273 *** 0.122 Constructon -0.243 *** -0.261 *** -0.160 ** 0.060-0.052 0.077 ** -0.129 * -0.031 Wholesale and retal trade -0.119 *** -0.126 *** -0.319 *** 0.012-0.311 *** -0.004-0.260 *** -0.063 Transport, storage and communcaton -0.085 *** -0.093 *** 0.044-0.062-0.087-0.115 ** -0.027-0.124 *** Fnancal servces -0.081 *** -0.087 *** -0.282 *** 0.058-0.215 *** -0.062-0.413 *** -0.044 Communty, socal and personal servces -0.043 *** -0.054 *** 0.002-0.006 0.014 0.056 0.024-0.099 * Large frms (50/more employees) 0.012 0.056 * 0.077 *** 0.034 0.073 *** 0.086 *** 0.010-0.029 Job tenure 0.022 *** 0.014 ** 0.012 ** 0.002 0.023 *** 0.007 0.020 *** 0.015 ** Job tenure squared 0.000 * 0.000 * 0.000 0.000 0.000 ** 0.000 0.000 *** -0.001 *** Selecton term -0.398 *** -0.189 ** -0.128 ** -0.387 *** -0.094 * -0.330 *** -0.103-0.218 ** Chow test (F-statstc) (7.240) 6.170 (6.820) P-value (0.000) (0.000) (0.000) Other varables ncluded n all the models are constant term; urban, provncal, publc sector and occupaton dummes. Reference groups: no schoolng, manufacturng, Gauteng provnce and machne & plant operators. Sgnfcance level: ***=1%, **=5%, *=10% 15

DPRU WORKING PAPER 13/159 Some of the fndngs for age and tenure exhbt an nverted U age/tenure-earnngs profle n both the unon and non-unon sectors. Ths suggests that regardless of unon status, mddleaged workers receve more wage benefts than other workers. The results for educaton demonstrate that earnngs ncrease wth educaton levels n both sectors. Whle ths outcome accords well wth the theory of human captal, t could also be reflectng the sklls-basedness of the South Afrcan labour market. Whle unonsed workers have relatvely hgher wages n 2001 and 2004, the opposte apples n 2007 and 2010 except for workers wth a degree. Overall, unon wages are hgher (than non-unon wages) n urban areas than n rural areas, and ths urban wage effect has ncreased over tme. The sample selecton terms were also negatve and statstcally sgnfcant, whch mples that the samples of unon and non-unon workers were selected, whch justfes our correcton for sector choce. Usng Equaton 9 we computed unon wage prema for 2001-2010 and the fndngs are presented n Table 6. The unon premum exhbts a declnng trend over the perod 2001-2010. In 2001, unon workers earned 57 percent more than non-unon workers. The premum decreased to 54 percent, 48 percent and 42 percent n 2004, 2007 and 2010 respectvely. Dfferences n these prema are statstcally sgnfcant at the 5 percent level. Ths shows that the wage gan of unon workers decreased by 15 percent ponts over the perod 2001-2010. Ths trend harmonzes wth that for Banerjee et al. (2008) for the perod 2000-2004. It suggests a weakenng n the ablty of unons to preserve the benefts of ther members. Table 6: Unon Wage Premum Year Unon premum Std. error Dfference 95% confdence nterval 2001 0.567 (0.003) [ 0.572 ; 0.562 ] 2004 0.541 (0.004) 0.025 [ 0.549 ; 0.534 ] 2007 0.476 (0.004) 0.066 [ 0.483 ; 0.469 ] 2010 0.424 (0.002) 0.052 [ 0.428 ; 0.420 ] 5.4 Unons and Wage nequalty We further examne the mpact of unons on wage nequalty usng two procedures. The frst procedure uses predcted wages for unon and non-unon workers, computed usng Equatons 5 and 6, to compute measures of nequalty. These results are presented n Table 7, Panel A, labelled wth premum. We then re-compute nequalty measures usng a counterfactual dstrbuton for unon members (computed usng Equaton 10) wages that workers would receve under the non-unon wage regme, mantanng predcted wages for non-unon members. The nequalty measures derved from ths dstrbuton are labelled wthout premum n Table 7, Panel A. The second procedure s based on Card et al. s (2004) model dscussed earler, and the varance decomposton (based on Equaton 16) results are presented n Table 7, Panel B. 16

Labour unons and wage nequalty among Afrcan men n South Afrca Table 7: Impact of Unons on Wage Inequalty Panel A: Inequalty measures 2001 2004 2007 2010 Wth premum Gn 0.297 0.349 0.367 0.370 Thel ndex 0.147 0.210 0.243 0.249 Atknson (e = 0.5) 0.069 0.096 0.108 0.110 Atknson(e = 1) 0.129 0.175 0.191 0.196 Atknson(e = 2) 0.225 0.289 0.304 0.311 Wthout premum Gn 0.267 0.286 0.323 0.320 Thel ndex 0.137 0.163 0.237 0.215 Atknson (e = 0.5) 0.061 0.071 0.098 0.091 Atknson(e = 1) 0.110 0.126 0.163 0.156 Atknson(e = 2) 0.179 0.200 0.240 0.240 Panel B: Effect of unons on wage structure Mean predcted log wages unon workers 2.725 2.833 3.099 3.639 non-unon workers 2.159 2.299 2.622 3.208 unon gap 0.566 0.534 0.477 0.431 Varance decomposton Overall varance 0.296 0.379 0.410 0.516 Varance n non-unon sector 0.226 0.306 0.355 0.477 Wthn sector effect -0.014-0.013-0.017-0.019 Between sector effect 0.084 0.086 0.072 0.058 Total effect 0.070 0.073 0.056 0.039 In accordance wth the evoluton of South Afrcan ncome nequalty, the Gn coeffcent wth unon wage effects ncreased over the perod 2001-2010. In 2001, the Gn was 0.3 and ncreased by fve ponts n 2004. Though small, a further ncrease n nequalty s regstered n 2007 and 2010. When consderng the dstrbuton wthout premum where all workers are pad under the non-unon wage regme we fnd that the Gn slghtly decreased for all years. In 2001, the Gn decreases from 0.3 (wth premum) to 0.27 (wthout premum), whle n 2004 t decreases from 0.35 to 0.29. Ths change n the Gn suggests that unons have a modest effect on wage nequalty. Ths fndng s consstent across the years, and s supported by the Thel ndex and Atknson s measures of nequalty. It s noteworthy that the magntude of the Gn reported here s much smaller than n prevous studes whch report Gn coeffcents n excess of 0.5 (see Tregenna & Tsela (2012), for nstance). Ths dfference stems largely from the sample used n ths study whch s restrcted to Afrcan men employed n formal sector non-agrculture occupatons, whle other studes ncluded all workers rrespectve of gender and race. Ths restrcton s, however, relaxed n the robustness check whch s based on an all-nclusve sample. The above fndngs on nequalty trends are corroborated by the overall varance of earnngs. We fnd that the varance ncreased from 0.3 n 2001 to 0.52 n 2010. Wth regards to the decomposton of varance, the negatve sgn of the wthn-sector effect ndcates that unons have a compressonary effect on wages wthn the unon sector. Ths result s consstent throughout the 2001-2010 perod. The between-sector effect s postve, suggestng that unons also have a dsequalsng effect on wages between unon and non-unon sectors. Ths dsequalsng outcome s consstent wth the common fndng of a unon wage premum. 17

DPRU WORKING PAPER 13/159 Overall, unons n South Afrca seem to ncrease wage dsperson snce the between-sector effect outweghs the wthn-sector effect over the perod. Relatve to the overall varance, the compressonary effect ranges from 4-5 percent, whle the between-sector effect ranges from 12-29 percent. From ths result t s clear that unons have a weak compressonary effect relatve to the dsequalsng effect. 5.5 Robustness checks To ascertan that the results presented n ths paper are robust ths secton presents a seres of robustness checks. The frst check s concerned wth the senstvty of the results to the selected sample, whle the second set of checks examnes the senstvty of the results to compostonal changes of unon and non-unon sectors over tme. Selected sample: The results dscussed earler are based on a sample whch excludes ndvduals workng n the nformal sector. To check f the results are not senstve to ths, we re-estmate the model on a broader sample whch ncludes both nformal and formal sector workers. The results from ths exercse are presented n Table 8, along wth the baselne results. The fndngs ndcate that when usng a broader sample, we fnd a larger unon wage premum relatve the baselne result (whch excludes nformal sector workers). Ths result s expected gven that nformal sector workers often engage n precarous actvtes whch offer relatvely low wages (see Perry, 2006). Consstent wth the baselne results, we also fnd a declnng unon wage premum from 86 percent n 2001 to 50 percent n 2010. Wth regards to varance decomposton, the results suggest that the net effect of unons s to ncrease wage dsperson, snce the dsequalzng effect domnates the compressonary effect. The qualtatve smlarty of the results s reassurng, suggestng that our fndngs are not too senstve to the selected sample. 18

Labour unons and wage nequalty among Afrcan men n South Afrca Table 8: Robustness checks Baselne: excl nformal sector workers Incl. nformal sector workers Workers wth age > 25 yrs Workers wth educ. <= matrc 2001 2004 2007 2010 2001 2004 2007 2010 2001 2004 2007 2010 2001 2004 2007 2010 Mean predcted log wages unon workers 2.725 2.833 3.099 3.639 2.442 2.740 2.968 3.551 2.719 2.826 3.185 3.607 2.590 2.634 2.903 3.373 non-unon workers 2.159 2.299 2.622 3.208 1.580 2.050 2.434 3.055 2.127 2.247 2.640 3.108 2.009 2.079 2.419 2.971 unon gap 0.566 0.534 0.477 0.431 0.863 0.690 0.534 0.496 0.592 0.579 0.544 0.499 0.581 0.556 0.483 0.402 Varance decomposton Overall varance 0.296 0.379 0.410 0.516 0.646 0.520 0.531 0.553 0.320 0.426 0.421 0.503 0.298 0.309 0.336 0.470 Varance n non-unon sector 0.226 0.306 0.355 0.477 0.475 0.423 0.474 0.504 0.239 0.337 0.358 0.454 0.219 0.241 0.283 0.436 Wthn sector effect -0.014-0.013-0.017-0.019-0.040-0.025-0.016-0.020-0.010-0.016-0.027-0.027-0.010-0.015-0.008-0.018 Between sector effect 0.084 0.086 0.072 0.058 0.211 0.123 0.077 0.069 0.091 0.105 0.090 0.076 0.089 0.083 0.061 0.053 Total effect 0.070 0.073 0.056 0.039 0.171 0.097 0.060 0.049 0.081 0.089 0.063 0.049 0.079 0.068 0.054 0.035 Excl. and ncl are abbrevatons for excludng and ncludng respectvely. 19

DPRU WORKING PAPER 13/159 Sector composton: One potental ptfall of the nter-temporal analyss n ths study s that t mght be smply reflectng changes n the composton of unon and non-unon workers. Indeed, there are a number of trends durng the 2001-2010 perods that are lkely to affect the composton of workers n unon and non-unon sectors. For nstance, the fnancal crss mght have affected the employment probabltes of low sklled workers whch mght nfluence the results. In addton, the mplementaton of the employment equty polcy n ths decade s lkely to have nfluenced the wages of sklled Afrcan men, as employers competed to employ sklled Afrcan men. Gven that sklled workers are less lkely to jon unons, the returns n the nonunon sector mght affect the estmated average unon wage premum. To address these potental ptfalls; frst, we re-estmate the model on a sub-sample whch excludes young workers (aged 15-25 years). The excluson of young workers enables us to obtan results based on a sample purged of workers who were lkely to become unemployed durng the fnancal crss. Second, we re-estmate the model on a sub-sample composed of low-sklled workers (.e. ndvduals wth an educaton level whch s less than or equal to matrc), to examne the senstvty of the results to the mplementaton of the employment equty polcy. The results presented n Table 8 ndcate that when excludng workers aged 15-25 years, and low sklled workers, we fnd a unon wage premum whch s above the baselne result. The hgher premum s n lne wth the noton that unons aval hgher premums at the bottom of the wage dstrbuton (Schultz & Mwabu, 1998; Casale & Posel, 2012). We also fnd a declnng unon wage premum despte the possblty of sector compostonal changes. Ths suggests that our results are farly robust. Smlar to the baselne results, we also fnd that the dsequalzng effect domnates the compresson effect, ndcatng that the net effect of unons s to ncrease wage nequalty for both sub-samples. 6. DISCUSSION AND CONCLUSION Understandng sources of ntra-afrcan wage nequalty n South Afrca s mportant for ntatves workng to reduce rsng ncome nequalty n the country. Whle there are varous drvers of wage nequalty such as human captal, hstorcal factors and labour market nsttutons, ths paper focuses on the latter by consderng labour unons. The paper has two nterestng fndngs. Frst, we fnd that the unon wage premum among full tme, wage employed Afrcan men has been decreasng over the 2001-2010 decade. Ths fndng s contrary to the results for the 1990s whch show an upward trend for Afrcan men (Hofmeyr & Lucas, 2001). Although the latter studed a dfferent sample (.e. Afrcan, urban men) from the one heren, t stll provdes nsghts on the evoluton of the unon wage premum. The ntal ncrease n the wage premum as uncovered n Hofmeyr & Lucas (2001) was expected gven that South Afrca was emergng from a hghly dscrmnatory system characterzed by low wages for Afrcans. Consequently, Afrcans unons made concerted efforts to secure szeable premums for ther members. In recent years, the declnng trend uncovered here suggests that unons have been weakenng. Possble reasons for ths could be the sklls-basedness (towards hghly sklled workers) of the labour market, whch s juxtaposed to hgh unemployment. In ths context, t mght be dffcult for unons to bargan for hgher premums as most of ther members are lowly educated and vulnerable to unemployment. Second, we fnd that unons n South Afrca have both compressonary and dsequalsng effects on wages among full tme, wage employed Afrcan men. The compressonary effect s however, domnated by the dsequalsng effect for ths partcular sample, suggestng that the net effect of unons s to ncrease wage nequalty. On the bass of the above fndngs, t s clear that unons partly contrbute to wage nequalty among Afrcan men. When placed n a global context, our fndngs dffer from what has been found elsewhere. For nstance n the Unted Kngdom, Unted States and Canada, Card et al. (2004) found that 20

Labour unons and wage nequalty among Afrcan men n South Afrca unons reduced overall wage nequalty snce the compressonary effect outweghed the dsequalsng effect. Ths could be due to the dfferences n human captal, hstorcal factors, labour market nsttutons and sample of analyss. To some extent ths mples that South Afrcan unons can contrbute to the overarchng goal of nequalty reducton through more compresson of the wage dstrbuton. We conclude wth a dscusson of the lmtatons of ths study. Frst, the Card et al. s (2004) model used the assumpton that non-unon members do not beneft from provsons of barganng councl agreements. In South Afrca, some non-unon workers beneft from barganng councl wage agreements; therefore t s mportant to verfy the extent of the overlap. Unfortunately, ths s mpossble to check usng our data. Nonetheless, Bhorat et al. (2012) analysed the barganng councl wage prema and found that non-unon workers covered by barganng councls receve wage premums of 9-10 percent. Ths premum s small n the South Afrcan context where premums can be n excess of 100 percent, suggestng that the spll over effect can be deemed as trval. Also, Butcher & Rouse (2001) reported that only 16 percent of non-unon workers were covered by barganng councls, whch further ndcates that the spll over effect s not substantal. The second lmtaton of ths study s that t does not account for unobservable characterstcs whch mght affect sector choce and wages. Future studes can address ths usng panel data. Thrd, although we have corrected for sample selecton bas n our endogenous swtchng regresson model, our results could potentally be based. Casale & Posel, (2012) pont out that unon wage equatons are senstve when controllng for selecton through dfferent technques, and there s no consensus on a superor technque. 21

DPRU WORKING PAPER 13/159 REFERENCES Atknson, A.B. (1970). On the Measurement of Inequalty. Journal of Economc Theory 2, pp. 244-263. Azam, J. & Rospabé, S. (2007). Trade unons vs. statstcal dscrmnaton: Theory and applcaton to post-aparthed South Afrca. Journal of Development Economcs 84(1), pp. 417-444. Banerjee, A., Galan, S., Levnsohn, J., Mclaren, Z. & Woolard, I. (2008). Why has unemployment rsen n the New South Afrca? Economcs of Transton 16(4), pp. 715-740. Bendx, S. (2001). Industral Relatons n South Afrca. Juta and Co. Ltd, Cape Town. Bhorat, H. & Lebbrandt, M. (2001). Modellng vulnerablty and low earnngs n the South Afrcan labour market. In Bhorat et al. (Eds.), Fghtng Poverty: Labour Markets and Inequalty n South Afrca, pp.107-129. UCT Press, Cape Town. Bhorat, H., Goga, S. & van der Westhuzen, C. (2012). Insttutonal wage effects: revstng unon and barganng councl wage prema n South Afrca. South Afrcan Journal of Economcs 80(3), pp. 400-414. Branson, N., Garlck, J., Lam, D. & Lebbrandt, M. (2012). Educaton and Inequalty: The South Afrcan Case. A Southern Afrca Labour and Development Research Unt Workng Paper Number 75. SALDRU, Unversty of Cape Town. Butcher, K. & Rouse, C. (2001). Wage effects of unons and ndustral councls n South Afrca. Industral and Labour Relatons Revew 54(2), pp. 349-374. Burns, J., Godlonton, S. & Keswell, M. (2010). Socal networks, employment and worker dscouragement: Evdence from South Afrca. Labour Economcs, 17, pp. 336-344. Card, D., Lemeux, T. & Rdell, W.C. (2004). Unons and Inequalty. Journal of Labour Research 25(4), pp. 519-562. Casale, D. & Posel, D. (2012). Unons and the gender wage gap n South Afrca. Journal of Afrcan Economcs, 20(1), pp. 27-59. Freeman, R.B. (1980). Unonsm and the dsperson of wages. Industral and Labour Relatons Revew 34, pp. 3-23. Fnnemore, M. & van der Merwe, R. (1996). Introducton to labour relatons n South Afrca. Butterworth publshers (Pty) Ltd. Gn, C. (1912). Varablta e Mutablta. C. Cuppn, Bologna. Harthog, J. & Oosterbeek, H. (1991). Publc and prvate sector wages n the Netherlands. European Economc Revew 27, pp. 97-114. Hetmueller, A. (2006). Publc- prvate sector pay dfferentals n a devolved Scotland. Journal of Appled Economcs 9, pp. 295-323. Hofmeyr, J.F. & Lucas, R. (2001). The rse n unon wage premums n South Afrca. Labour 15(4), pp. 685-719. 22

Labour unons and wage nequalty among Afrcan men n South Afrca Hoogeveen, J.G. & Ozler, B. (2005). Not separate, not equal: Poverty and nequalty n postaparthed South Afrca. Workng Paper 379, Wllam Davdson Insttute, Ann Arbor. Lee, L-F. (1978). Unonsm and wage rates: a smultaneous equatons model wth qualtatve and lmted dependent varables. Internatonal Economc Revew 19, pp. 415-433. Lebbrandt, M., Arden, F. & Woolard, I. (2012). Descrbng and decomposng post-aparthed ncome nequalty n South Afrca. Development Southern Afrca 29(1), pp. 19-34. Lebbrandt, M., Woolard, I. & Bhorat, H. (2001). Understandng Contemporary Household Inequalty n South Afrca. In Bhorat et al. (Eds.), Fghtng Poverty: Labour Markets and Inequalty n South Afrca, pp. 21-39. UCT Press, Cape Town. Lete, P.G., McKnley, T. & Osoro, R.G. (2006). The post-aparthed evoluton of earnngs nequalty n South Afrca, 1995 2004.Workng Paper No. 32, October, Internatonal Poverty Centre, Brasla. Ltchfeld, J.A. (1999). Inequalty: Methods and Tools. The World Bank. Lokshn, M. & Sajaa, Z. (2004). Maxmum lkelhood estmaton of endogenous swtchng regresson models. The Stata Journal 4(3), pp. 282 289. Mchaud, P-C. & Vencatachellum, D. (2001). The Unon Wage Premum for Blacks n South Afrca. Paper presented at the DPRU/FES Conference. Labour Markets and Poverty n South Afrca.15-16 November 2001, Johannesburg. Mwabu, G. & Schultz, T.P. (1996). Educaton Returns across Quantles of the Wage Functon: Alternatve Explanatons for Returns to Educaton by Race n South Afrca. Amercan Economc Revew 86(2), pp. 330 334. Schultz T.P. & Mwabu, G. (1998). Labour Unons and the Dstrbuton of Wages and Employment n South Afrca. Industral Labour Relaton Revew 51 (4), pp. 680-703. Schoer, V. & Lebbrandt, M. (2006). Determnants of job search strateges, evdence from the Khayeltsha/ Mtchell s plan survey. South Afrcan Journal of Economcs 74 (4), pp. 702-724. Mwabu, G. & Schultz, T.P. (2000). Wage premums for Educaton and Locaton of South Afrcan Workers, by Gender and Race. Economc Development and Cultural Change 48(2), pp. 307-334. Moll, P.G. (1993). Black South Afrcan unons: Relatve wage effects n nternatonal perspectve. Industral and Labor Relatons Revew 46(2), pp. 245-261. Perry, G., Maloney, W., Aras, O., Fajnzylber, P., Mason A., & Saavedra-Chanduv J. (2006) (eds). Informalty: Ext and Excluson, World Bank Latn Amerca and Carbbean Studes. Thel, H. (1967). Economcs and Informaton Theory. North Holland, Amsterdam. Tregenna, F. & Tsela, M. (2012). Inequalty n South Afrca: The dstrbuton of ncome, expendture and earnngs. Development Southern Afrca 29(1), pp. 35-61. Van der berg, S. & Bhorat, H. (1999). The Present as a Legacy of the Past: the Labour Market, Inequalty and Poverty n South Afrca. Development Polcy Research Unt: 01/29. Cape Town. 23

DPRU WORKING PAPER 13/159 APPENDIX A. Inequalty measures Ths appendx presents detals of the nequalty measures used n Secton 5 of the study. Gn Coeffcent (G) Gn coeffcent s the most commonly used measure of nequalty. It can be computed as follows (c.f. Gn 1912): 1 G = 2 2n y n n = 1 j= 1 y y j Where y s the ncome of ndvdual, y s mean ncome and n s the sample sze. When G = 1 t represents perfect nequalty, whle G = 0 represents zero nequalty. General Entropy (GE) measure The general Entropy measure s another class of nequalty measures defned as: n 1 1 GE( α ) = 2 α α n = 1 y y α 1 Measures of the GE are senstve to changes at the lower end of the dstrbuton for α close to zero and senstve to changes across the dstrbuton for α = 1. It s senstve to changes at the upper end of the dstrbuton for hgher values of α. When α = 1, ths yelds the Thel (1967) ndex. The values of GE ranges from 0 to, wth 0 representng an equal dstrbuton, and hgher values representng hgher levels of nequalty (Ltchfeld, 1999). Atknson measures of nequalty (Aε) Ths measure has a weghtng parameter ε whch measures averson to nequalty. The hgher the value of ε the more socety s concerned about nequalty (Atknson, 1970). The Atknson class of measures can be computed as follows: 1 ε = 1 n A n = 1 y y 1 ε 1 (1 ε ) Where y denotes mean ncome. Smlar to the Gn coeffcent, the Atknson class of measures range from 0 to 1, wth zero representng no nequalty and 1 representng complete nequalty. 24

Labour unons and wage nequalty among Afrcan men n South Afrca B. Defnton of varables used n the study Table A1: Varable defntons Varable Unon (dependent varable) Wage (dependent varable) Age Marred Educaton No educaton* Prmary Incomplete Secondary secondary Certfcate Dploma Degree Provnces Gauteng* Eastern Cape Free State KwaZulu Natal Mpumalanga North West Northern Cape Northern Provnce Western Cape Urban Industres Mnng Manufacturng* Electrcty Constructon Trade Transport Fnance Servces Frm-sze Job tenure Publc sector Occupatons Manager Professonal Techncan Clerk Servce Sklled Agrculture/Fshery Artsan Operator* Elementary LFP broad Employed Chldren<15 yearsᵠ Other hhld unon members Old menᵠ Old womenᵠ Prop_earn± Prop of other adults Descrpton Dummy varable: 1 f ndvdual s a unon member, 0 otherwse An ndvdual s log hourly wage deflated to 2001 values An ndvdual s age n completed years Dummy varable: 1 f ndvdual s marred cvlly or tradtonally, 0 otherwse Dummy varable: 1 f ndvdual has 0 years of schoolng or grade R, 0 otherwse Dummy varable: 1 f ndvdual s schoolng s n the range grade 1 to 7, 0 otherwse Dummy varable: 1 f ndvdual s schoolng s n the range grade 8 to 11, 0 otherwse Dummy varable: 1 f ndvdual s schoolng s grade 12, 0 otherwse Dummy varable: 1 f ndvdual s schoolng s vocatonal or techncal tranng, 0 otherwse Dummy varable: 1 f ndvdual s schoolng s dploma, 0 otherwse Dummy varable: 1 f ndvdual s schoolng s bachelor s degree and above, 0 otherwse Dummy varable: 1 f ndvdual resdes n Gauteng provnce, 0 otherwse Dummy varable: 1 f ndvdual resdes n Eastern Cape provnce, 0 otherwse Dummy varable: 1 f ndvdual resdes n Free State provnce, 0 otherwse Dummy varable: 1 f ndvdual resdes n KwaZulu Natal provnce, 0 otherwse Dummy varable: 1 f ndvdual resdes n Mpumalanga provnce, 0 otherwse Dummy varable: 1 f ndvdual resdes n North West provnce, 0 otherwse Dummy varable: 1 f ndvdual resdes n Northern Cape provnce, 0 otherwse Dummy varable: 1 f ndvdual resdes n Northern provnce, 0 otherwse Dummy varable: 1 f ndvdual resdes n Western Cape provnce, 0 otherwse Dummy varable: 1 f ndvdual resdes n an urban area, 0 otherwse Dummy varable: 1 f an ndvdual works n mnng sector, 0 otherwse Dummy varable: 1 f an ndvdual works n manufacturng sector, 0 otherwse Dummy varable: 1 f an ndvdual works n electrcty sector, 0 otherwse Dummy varable: 1 f an ndvdual works n constructon sector, 0 otherwse Dummy varable: 1 f an ndvdual works n trade sector, 0 otherwse Dummy varable: 1 f an ndvdual works n transport sector, 0 otherwse Dummy varable: 1 f an ndvdual works n fnance sector, 0 otherwse Dummy varable: 1 f an ndvdual works n servces sector, 0 otherwse Dummy varable for frms wth more than 50 employees number of years that an ndvdual has spent wth hs/her man employer Dummy varable: 1 f an ndvdual s employed n the publc sector, 0 otherwse Dummy varable: 1 f an ndvdual s a manager, 0 otherwse Dummy varable: 1 f an ndvdual s a professonal, 0 otherwse Dummy varable: 1 f an ndvdual s a techncan, 0 otherwse Dummy varable: 1 f an ndvdual s a clerk, 0 otherwse Dummy varable: 1 f an ndvdual s a servces worker, 0 otherwse Dummy varable: 1 f an ndvdual s a sklled worker, 0 otherwse Dummy varable: 1 f an ndvdual s a craftsman, 0 otherwse Dummy varable: 1 f an ndvdual s an operator, 0 otherwse Dummy varable: 1 f an ndvdual does elementary jobs, 0 otherwse Dummy varable: 1 f employed, actve job seeker or dscouraged worker Dummy varable: 1 f employed full tme n formal non-agrculture sector Dummy varable for the presence of chldren aged below 15 years n the household Dummy varable for co-resdence wth other unon members Dummy varable for co-resdence wth men aged 65 and above Dummy varable for co-resdence wth women aged 60 and above Proporton of earners n the household Proporton of other workng age people n a respondent s household who are employed Notes: 1. *Represents base category. 2. ᵠ and ± denote excluson restrctons n the broad labour force partcpaton and employment models respectvely. 3. denotes excluson restrcton n the unon membershp probts. The log hourly earnngs n the table were constructed from earnngs nformaton whch was reported n ether ponts or ntervals. Both categores were reflected n weekly, monthly or annual payment ntervals. We obtaned monthly earnngs from pont data through multplyng gross weekly wages by 4.3, dvdng yearly wages by 12 and retaned the gross monthly earnngs. To those who reported ntervals, we assgned mdponts of the bands as converted to monthly terms. Ths gave us a seres of monthly earnngs for 2001, 2004, 2007 and 2010. 25

DPRU WORKING PAPER 13/159 These were deflated to 2001 values usng deflators from Statstcs South Afrca. In turn the real monthly earnngs were converted to hourly wages as we dvded them by hours usually worked per month (.e. hours usually worked per week n man job multpled by 4.3). 26

Development Polcy Research Unt Unversty of Cape Town Prvate Bag Rondebosch 7701 Cape Town South Afrca Tel: +27 21 650 5705 Fax: +27 21 650 5711 www.dpru.uct.ac.za Lke us at www.facebook.com/developmentpolcyresearchunt