Teachers ' Strikes in Argentina : Partisan Alignments and Public-sector Labor Relations *

In a context of increasing teachers' militancy in Argentina, this article provides the first empirical analysis of teachers' strikes in all twenty-four Argentine provinces during the 1990s. Using a cross-provincial statistical analysis, it explains the wide variation across provinces and across time of Argentine teachers' strikes. It demonstrates that political alignments between provincial governors and teachers' unions explain these patterns better than organizational and institutional variables, which strongly shape public-sector labor relations in other countries. We emphasize the discretion of provincial governors, for both the application of labor regulations and budgetary appropriations in the politicization of labor relations in the education sector. Our findings confirm the politicization of provincial public-sector labor relations in Argentina, especially after the decentralization of education resulted in the provincialization of teachers' protests.


78
Studies in Conrparative In{ernational l)erelopment / Spring 2{}04 tine teachers' nlilitanc)' is u'ide: while 3E5 class da1's were lost in all trventy-four provinces in 1989. onl)'forty-srx days were lost in 1992.l\4oreover, whereas each province lost on average I06 da1,s ofclass between 1984 and 2000, the province of La Parnpa lost onll' a single day of class during this period while the province of Tucumdn lost 3.13 da1,'s.
The increase ir-r Argentine teachers' rnilitancy is related to their declining work conditions as a result of flscal adjustment. rvhereas its growth relative to other economic sectols is associated with the diminishing militancy' of other sectors of the econoitt\' (N4cGuire 1996). Policl' reforms. such as decentralization, also increase teachers' militancl, a-tainst reforms that challenge their income security (Murillo 1999). Hou'ever. tirese reasons are insufficient to explain the u'ide I'ariatiol across provinces and time in Argentine teachers' strike activity. which is the puzzle that u'e address in tiris arlrcle.
By focusng on the puzzle of u'hat explains the various patterns of Argentine teachers'urilitancr'. tliis article contributes to the iiterature on public-sector labor relations in Latin Amedca and 1o our understanding about provincial protests in Argentina. Because teachels are public service pror-iders. their militancy afI-ects rhe daily' Iives of farnilies u'hose children cannot altend school. rather than the profit of emplo;'ers (althougir these families vote forteachers 'employers, i.e., govemors). Iu the United States. the public character of teachers' employment real-finns the imporiance oi Iabor regulations on influencitrg teachers' behavior (Freeman 19861. Hoq'er,er. u,e demonstrate here that in Argentina the patterns of teachers' rnilitancl,' are iretier erpllinrd hv political alignments betil'een public ernployers and teachels' uniclns than hr, labor instituticins. lfr/e suggest that executive discretion in labor relatiolts -senerates uncertaintl, about the working of fbrmal institu-1ions, uhereas rhL-srrong social links between Argenline political parties and iabor unions make poliiical alignmenfs-bsiu:gsn governors and provincial teachers' unions-intluential on explainin-s pattems of teachers' strike activity. l\{ore broadlr'. teachers'strikes reflect a larger but uneven pattern of pror-incialization c1'protests. In the case of tcachers. this process resulted fiom the decentralization of educaiion. u,hich was part of the social refbnn agenda of the 1990s inArgentina. Due to dec'entralization of social services. since the 1990s labcr protests have concenlrated on provincial pubiic seclors and included not only strikes. but also urban riots and road-blockades.r 'l'he cornbination of decentralization and pr-rblic-sector ad.lustntent caused Argentine teacirers to suffer fiom the contrast betu eeil their^ ntrddle-class status and the proletarization of their u'ork conditiotts uhile experienc'ing ihe accuniulation of salarl' arrears by provincial employers. lnciced. a teacltcfs' stiike causeci b1' unpaid salaries staned the so-called "Santiagazo" of 1993. uftich became the first ur:ban riot of Ar:gentina in the 1990s.' This ar-ticle is orgaitized into four sectjcrns. The first section revieu,s clifl'erent iabor theodes tr,i exprlsili lrne l'ns ot i uriatitrn irr,{r.tentine teacher-q' militanci'. Thc seconrl secti(]n pi.senls tire organization of labor relations in the Argentine educalit-rn sector. Tite thiri -cecti()i.t te!t.! i)Llr trsluilent usinc a data set of teachers'strikes \Iurillo and Ronconi 11) in ali Ar-sentine provinces Lretueen 199tr rrnd l{)()0. I-he finlj se,;tion crinch:cies w'ith the irnplicrttions of this studr'.
\\'hereas economic conditions can varv br,'prcvince. their eftect on pubiic-sector workers is iimiteil. Labor strength, by'conlrast. can !ary across provinces and shoLrld affect the problbilitl of strike success and. therefbre. pattems of militanci,. Other thrngs being equal. empiol'ers r.vho knolv about labor stlength should preemptivelv eive in to strong unions to avoid sultering the losses generated by the threatened strike.b That is. labor strength should be ne-eativelv associated with militancy because stron_s unions do not need to exercise their mnscle to obtain concessions from ernployers. Due to the requisite of union registration and automatic deduction of union clues tiom salaries in Argentina. emplolers' information about labor strength-wrth its negative effect on militancy-can be assumed tbr publicsector employees.
Labor stren_eth is aff'ected bv organizational variables that facilitate collective action. such as unionization rates, nniolr monopolies. and centralized labor organization. \&'hereas unionization increases the representativeness of labor unions and the number of u'orkers they can involve in the strike (Francozi 1995). union monopoly reduces coordination problems and faciiitates bar-gaining by redticing interunion competition (Golden 1993). Centralization. on the other hand, is a national-levei characteristic related to coilective bargaining (Cameron l9B4; Calnifors and Driffil 19981. and thus of little use tor explainin-q variation in teachers' miiitancv across Drovlnces. In addition. Iabor re-quiatiotls affect union behavior in the public sector by changing the costs of a strike or bv in.lirrovin-t rnechanisms for non-conflictive bargaining. In discussilg public-sector unionislr in the Llnited States. Freeman (1986) argues that unionizatiou gro\r's u,here state labor regulations are more favorable, whereas altematives to strike. such as art-ritration. as well as state-level regulations on strike actiiitl, allbct labor milttancl'. Because public-sector emplo1'ment is less exposed to busine ss c1.,cles and trade competition. labor institutions should have a stronger effect on the hehur ior of public-:cctor lhan prir ate-secltrr unions'' Beyond organizational and institutional effects. political exchange theories assume thar political alliances betu'een national-level unions and labor-allied political parties change the strate-eic behavior of labor unions. Based on their trust of labor aliies and expectations about long-temr benefits, unions increase their shortterm tolerance for less satisfactory u,ork conditions. Trust is sr-rstained by previous erperiencesantl lcld*toloriermilitancr uhettallie:areinpoucrrndaccesstothe state can be used for obtaining concessious. Pizzotno (1978), Korpi (1978), and Hibbs (1978) har,e atlvanced this argument and demonstrate its empirical standing for \Vestern European colrntries. in addition to this effect of trust in extending time horizons. parrl' links also prolide informal channels of communications and generate interactions in multiple arenas. rvhich increases the benefits of keeping credibility for labor aliies (N{urillo 2001).
Follou ing this literatut'e. \\,e ar-que that political aligntnents should have a stronger ef-fect on public-tltan on private-sector labor relations because public-sector emplol,ers aie nt()re likel1, to be concemed about political sustainability than profit naximizaiion. Indeed. Hibbs (1976) argues that golernment intervention in labor nrarkets causes the politicizatjon of strikes in ltalv. Hence. u'e should expect the public character of the eurplover to fufiher politicize labor relations (Pencavel 1997Cox Edu'arcls 1997 and therebf increase the influence of political alignments in definrn-e atiitudes betu'eeu emplo1'ei:s and employees. W-e expcct political alignnents to explain the patterns of teachers' militancy in Argentina becanse. rn adclition to tlreir be ing pubiic entplo1,ees, a context of executir.e discretion-in the application o1'l:rbor regulations and the allocation of bud-set3rv expenditure5-l1s1si1sgs, the value of informal channels of cotnlnunication. irust basecl on prei iitus erperiences. and credibility based on muitiple interactions. and therefore of panisan links. N{oieover. dern()cratic governments in general, and Argentinc piiti'incial giilernors in particuiar. are elected as party candidates. thus increasing the use of partisan identities as a tool to generate expectations about their behar,ior. In Ar.sentrna. political parties have a strong tradition of linkages u ith socje6' iil cenerrai. and ri ith unions in particular (James 19881 Levitsliy 2003) reinforcing the inrpcrtance o1'political alignments based on partisan affiliation.t Althor"r-gir a najori$, of unions are aifiliated u'ith the Peronist party in Argentina. education is a sector characterized bi' irs political pluralism. and the ieaders of provincial leachers' ur-rior.rs hr.r'e clifferent part)' identities (Ydzquez and Balduzzi l0r.lij 1." ln -slrm. or_e anizaticrnal. institrltionai. ol politicai exchange theories could be used i,, c:,plll;;: irc iri;' i-' .r; " u:j;il.r;l in 1,:re;l.i:lc ttrciteis'mililancl acros\ pror inces and time. rVe expect political alignntents tc provide a better explanation becaltse Ar-uerline -so\ernr)fs enjov hirh discretii.rn in the appiication of institutions that \Iurillo and i{onconi increases the r,'aiue oi politrcal a1i_snlients due to the stiong fink; esiablishcrl between politicel pailics and labor unions in Argentir"ra. In lhe next seciion. we riescribe the Arge ntine eCucatic:n and labor-reiatron:, s),stem to iho..'; that the pro..'ince is the appropriate unit of analr'sis to evlluritc our hvpothcres. *hicli rre icsted in the tollor.r, ing secti0n.
In labor relations. provincial _rrovernors hirve the po\,ver to decide on the le-eality of strikes, to frx the arnount ot'pay strikers lose. and to call for compulsory arbitration to helt conf-iicts. Hor,vever, thev also enjoy high levels of discretion in their application of labor re-rulations. The application of provincial decree 2202 of 1992. which established attendance bonuses in the province of Buenos Aires. ilh.rminates this discretion because it was suspended by'a _sovernor in l99.1and reactivated by another in 2002. In Tierra del Fuego province. the process of uniol re-eistraticn. which should be linked officially only to the fr-rlfillment of specific requirements. provides another erample. In 2000. the governor tried to deny resistration to the lar,eest union and _sive it to a triendlier one.
Provincial goverxors also enjol'' discretion over educational expenditures. whereas provincial budgets are relativell.' isolated fiom the cycles of their ou.n economies. Provincial budgets clepend on a revenue-sharing system whose fonnula results tiom the historical bargainin.r betr.vee'n the provinces and the federal govemment (Eaton 2001. Cetriingolo. and Jim6nez 1996: Jones. Sanguinetti. and Tommasi 2000. Revenue sharing pays over 7 1 percent of provincial expenditures but has little relation to locally collected taxes-which in most provinces are less than 2-5 percent of revenue-or population shares. and -eenerates snch fiscal imbalances that provincial fiscal deficits averaged 5.9 percent of their expenditures between 1997 and 2000 (although there was provincial variation).rr Because the governor has discretion over 90 percent of prol'incial resources derived from revenue-sharing, publicsector salaries are more aff'ected by fiscal manasement than by the perfornrance of the local econor.n\,. Indeed. in many provinces. teachers' salaries also include provincjal atteudancc bonuses, uhich are established by provincial executi\/e decrees, and affect strike costs because thev are contingent on not missing days of class.lz Traditionally'. nrost primarv teacher nnions har,e been organized at the provincial level. aithough a majority cf theil are members of a national confederation called CTERA (Contederation of Education Wbrkers of the Argentine Republic).r3 Furthennore. as public service providers. teachers suffer principal-agent tensions betu'een hufiing families rvhose children miss school. and pressin-u public employers, u,ho are roted into oflice b1'these farnihes. Because the government rvill only internalize the costs of strikes if it expects that tlie alienation of families will have negative electcral consequences. teachers should consider public suppor-t in defin-in_s their miiitancy. In turn. public suppol't should be related to pubiic perceptions about the provincial iiscal managerrent.
A Cross-Provincial Test on Teachers' l{ilitancl,' In this seclion. u'e provide the first (to our knowled-ee) empirical study of subnational r'ariation in Argentine teachers' miiitancy. \Ve test the alternatir,e theories explarning the puzzle of variation in Argentine teachers'rnilitancl' across provinces and across time after the decentralization of education in 1993. ril/e used data collected tbr the flrst time on the variables of interests. although lirnitations in the availabiiitv of information about some explanatory variables reduced the scope of our studl'to the 1996-1000 period. rfiie flrst erplain the variables we used to test the alternatile theode s discussed abtrle and then provide the tests and their interpretation.
!\re also test the eftbct of organizational variables aftecting labor strength. We use uttioniacttion to measure the ratio of r,rnionize d teachers to the number of teachers rr,,ithin a provincc and thereby labcrr stre ngth. vvhich we expect to be ne-uatively relatecl to militancl'.re To assess the etTect of union monopoly or fragmentation over collective action lve use the variable rtnionJrugntentstion. which measures the number of teachers' unions per provinct'holdin-e more than l0 percent of total union members.ro \\'e use 10 percent as a cut-otT point under the assumption that r.'ery small unions are unable to generate coordination problems if a large number of workers are members of one irirge union that bears the' costs of collective action. Thus, we expect uniott ft'ttgmetfiution to be positil'ely related to militancy.
W-e test the eff'ect of institutional variables by using Legal recognirlon (LR) as a dummy vanable. where 1 denotes that the principal provrnciai teachers' union has a legal monopoly of representation and 0 otherrvise.rr Followin-e institutional theories, we expect that unions rvith legal reco_tnition should be more successful at noncont-lictive bar-uaining and less likely to incur in militancy. The Table I summarizes our expectations re_sardin-s the effects oi each r,'ariable on the dependent variables.
Although w,c har.'c not ibuncl adequate data to measure popular support tbr primary teachers' demirnds in blaming the _eovernor for iiscal mismanagement, the eftect of strikes on families cannot be understated. We use as proxies both the number of days lost in strikes by other public-sector workers and tl-re provincial fiscal deficit as a percenta,ge of revenues in order to captLrre perceptions about the -qovernor. but neither of these is prorirnate enough and we find no public opinion or arrears diita. We ran our models u,ith these two pro-ries ancl the results were robust to their inclusion although we are not presenting thern here.I The descriptive sta- \\/e also add a r,ector of time effects and lag the observations on unionization and real \l'ase 'u'ariation b1' one )rear to account for the temporal sequence of causaiity.
Year dummies are included to account for variatiol'rs in econorlic conditions o\/er the course of our iime period. Whereas I996 and 1997 were years of relative growth. recession characterized the subsequent tluee 1'ears. There are three :;aljent aspe.ts of the data u'e studied. First, the r':riues of our dependent variables-both the nurnber of class da-vs lost due to teacher strikes and the nurnber of stlilies-are n()uilegative inregers or coutlts. For some proriinces, the value o1'thrs var:rabie varies irom zero to manr'. The "zero" r'aiue is the usual outt:otre Llecause strihes al'e rir'a evenls. alinost J-5 percent of the obscr\ations \\cre zero.23 SeconC. u,e ha-,,e repeated obsenations {five 1'ears) for the same tu'ent,vfour pror rnces. That isr. our iiata fornr a combrned time-series cross-section ponel.  Third. the data. however, span a short period of time. ancl the values of some of the independent variables in our model did not change over that period (i.e. unionfragme-ntation, legal recognition and attend(nce bonus). or they did so very sporacli_ cally. such as political ctlignment, because only six provinces ihanged theii governor in the studied period. Thus. u'e Lrelieve that the fixed ef'fects (u:ithin) provinces model is not an appropriate test nrethod to test our h),potheses. The preponderance oi zeros and the clearly discrete nature of the dependent varialrle su-csest that least squares and the linear model are not the appropriate specrfications. Therefbre \\.e use a Poisson model.} as suggested by Carneron and Tlir.edi (198(r. i998). The Poisson nrodel assumes that the conditiorial mean of the outcome is equal to the conditional r ariance . This assumption l-rolds for the dependent variable ruunber (of strikes). but ue find over-dispersion (i.e. the conditional variance excc-eds the conditional mean) for tlrrrutiotr (number of class days lost).15 Theretbre. !\'e use a Negative Binonrial nrodel for this dependent variable (Table  2 ).r6 Our results shoq' tlrat pttliricol oligrurturt has the expected negative eflect on militanct'-rneasured both as duratittu v11fl 11un1bgy-and that this relation is statrsticalit' si{Inificant. B1'coutrast. the control variable lagi.slcttit'e support" ruhich measures execlltive capacit-v at lhe provincial let'el. does not have a significant effect on either de'peudent variable.rr The alternative explanatory variables afTectin-q labor st.rength are not as good as pttlitic'ul ttligrmtcnt in elplainin-t militanct'. Llrtiorti:utiort. urirm.fi'ctgntentation. and legol t'e cctgnition are not statisticaliv si-cnificant and iu sonte cases they do not have the expected si-gns. Yet. the control variables. oltettdonce boutrs and real v'age t'ltcrrtge. have the erpected signs and are statisticalll, significant. urhereas unernplotrrrt:ttt rule is sisnitlcant ri'ith a ne-cative sign that is discussed belorv.
To assess the nrasnitude of the eif'ects prodLrced b1' our explanatorv variable s on tr'auhers' ntilitancl'. $'e present the discrete chanse in the erpected value of the nutnlrer ot'class cial'lost (tlurrttiortl and the number ol strikes (nurnber') for a unit of chan-te in ihe re-riessors coi.nputc-d u.ith ail r,ariables held at their means (because nciiher ti-re Pcrisscn nor the Negatir,e Binonrial recression models are linear) in Table  3a and b.18 As shor-v'n in Table -ia and b. pttliriccil uligrtntertt has the strongest substantive cifect on the nl'o dependeiit r ariables. altd as expected re.duces tlie propensitl, of teachers to strike. \Vhen Llte grrr crnor rrrd thc teachers'unions are politically aligned. the proi'ince loses l.:l feuer davs tn str"lkes per,vear than in pror,inces u'ithout pclittcai aligntlent. {leal v'ogc intproterncut ltnd qilendance bonuse,s as a percentage ol inconte also have the expected negative effects. In terms of rnagnitude. a ten pel'ccnl ittc;'.'a'e in ;'cul ',', a.ses dur ilg vcai i is corrciated u'ith I .1 fewer da\,s lost tct strikes during -\'e:1i'l+l u'hereas for everr,trn pcrecnt increase in attendance bonuses as a percelttiise of r.r a-ees. \r'e expect 1.2 f'erver dal's lost to strikes.
These iindings suppoll our iugui-r1erlt that. across all Argentine provinces, teacher's militancv is negatn'el-v re:lated to pofitical alignments between provincial governors and teachers' ulrions. er,en l'n'heu controlling for organizational and institutional I'ariables that also affect labor militanc_v-. Political alignments, therefore, influence teachers' mihtancf in a public sector u'ith considerable executive discretion or.er labor regLrlations and pulriic budgets. The public-sector character of teachers' emplol'ntent also explains that the pror,incial unemployment rate has a positive effect on teachers' inilitanc\'. controlling lor wa-re conditions and attendance bonuses.
Bv contrast. u,e clo not find strong sllpport frrr theories based on institutional or or_eanizational variables affecting labor capacity for collective action. We suspect that executir,e discretion dimiiiishes the effect of institutional aud organizational variabies by decreasin-g their value in fornring actors' expectations. Furthermore, u'e believe that b1' contributin-s to the politicization of labor relations in the education sector'. executive discretion explains the influence of political alignments as alternative rnechanisms to define expectatious about the use of executive discretion regarding fte allocation of education resources and application of labor regulatrons.

Conclusion
This studl' contributes to the literature on public-sector labor telations by dernonstrating that political aligninent provides u b-tt.. explanation for the puzzle of I'ariation in Ar_centine teachers' niiitancr, than institutional or organizational variabies. In drrin_s so. it iiighii_ehts the pciiticization caused not only by public sector enplol'ment-protected frorn trade conlpetition and business cycles-but also by erecutive discretiou on labor reiations. \\/e suspect that it is executive discretion u'hat u'eakens the eff'ect of institutronal and organizational variables, especially because the requisite ol'union registration provides employ''ers with knou'ledge about labor strengih anci frag;ireutarion. Our stud,i' confinls cihcrs abcut ihc politicization of labor re la.,ions proCuced b1' gor,ernment interr,'ention in Latin American national polities (Zapata 1 986t Cotiier and Coilier i 991 : Murillo 2001 I and the public sector (Valenzuela 200'Ji. It also confirms the findings of othc-r studies of teachers' militancl in the regrcn. uhich have stressed political variables in defining their subnationalpatien:ls Lrf protests (Cook 1996: Foueraker 1993. The findings of this study ihus qualii'i' the applicaiion of institutionalist theories tc) c()ntexts where the uncenainil' generatc'd b'1, different appiications of the sar.ne rule weakens fonnal institutionaiization (O"Donneli 1 999). ts1' locusing on the sr:Lrnationai iei'el. this study clarifies the effect of explanatc;--..'r'ariat'ies. x'hir:: cln 5e ,::eC in ccnrplrrrltil'e analvses cf labor poiitics. Previous studies have shou'r-r the saiience of the national govemment party identit-v-in shapiirg laocr be h:ir i'or at the national le..'el. In demonstratin-t that this logic applies \Iurilio lnd 'lioncrni to the sub,national icr.'el. our stLlclv hi-:hi;shts which varitbics ailect ihls reiarionship. in part;cuiar. !!e sho\i'the \aiue of-i'ccrjsinc on nhc) is tnc ir-.ciis of iai;cr reiations tnd wage brrgrinine-tlie en-lpic.",er-as weii a-. lhe inilLience of the political process irnd par ti' itlentitie: on delining .-vho are the barsaining purtner-l-fhc empltt.,,s1 and the uniitn ielder. We rs:;ilme that the iiiscretion ot'iiilpioyers in dctining r,vork conditions reinibrces thr-epFligxlion oi thrs logic. For insiance. althttu-gh Nlexicln telchers hirre been iranslerer-l icr the srates in lhe-,-.arly 199Lrs. the i'ederal go\.elxment defines their saiar-,'aporcpriliion' r:lLrcin-q tire so.,'emors' Cistretion itn teacher'' salaries. Hence. w'e do not e{pect poiiilcal ali_gnments between the state govetxor ancl the ieadership of the local unicn seition (aithough there is variation in the pollticai atfiiiations of both pariies) io ha.,'e ihe same el'f'ecrs as in Arsentine provinces. Similarlv. althcu-gh eciucation rs decentralized. Chilean majors have little discretion to cietlne salaries. u'hich are linked to test resuirs and the number of students per school. \1'e thus dc not expect subnational par.tisan alignnlents to be as important as in the Ar-ecntine case. Instead. we wouid expect a similar logtc to appli, in Brazil. rvhere majors-elected on partv slates-re in charge of labor relations in the education sector ancl teachers' r-inions aiso have identities associated rvith poiitical parties.-r') More broadlv. the provincial decentillization of functions and resources had the unintended etfect of tosterin-q the concenu'ation of protests at the provincial ievei in Ar-uentina. Political ali-enments bet*'een Linions and govemors reduced conilictive interactions in the public education sector during the studied pericd. We su_ggest that these finciin-es are explained b1 the eficct of thc discretion of pror.,inciatr governors in pubiic-sector labor relations. r"'hen controliing tbr economic incentives (i.e.. incotne, bontrses) and labor strength (i.e.. unionization. union fragmentation.;. Tit fur-ther our understandin-e of decentralize d public-sect.rr labor relations, .,ve believe this first empirical studv on Argentine teachers' militancy should be complernented by others. to rissess in particular the eftect of provincial public suppoft on protest, which quaiitatir''e evidence si-unals is important (Ar.ryero 2003) although vu'e were unable to nlei.rsure it. Additionally. llrrther studies should fbcus on other patterns of popular discontent across Argentine provinces to build a more general theory about the politics of pror inciitl protesrs.
Frnally. we tbund support for the politicization of public-sector labor relations in Argentina. Because inissed class da,vs have negative policlimplications fbr the perfbrmance of the Argentine education system as a whole . we believe that policy reforms should strengthen labor institutions and constrain the discretion of provincial governors.'t1 The effect of -eovernors' budgetary' discretion on provincial fiscal mismanagement became ob''.ious during the 2001 Argentine crisis. when provincial currencies, debt defaults. and protests spread across the country.s2 However, the cost of pror'.incial erecutil'e discretion in terms of teachers'protests and decline in human development cannot be understated in a country where illiteracy has been growing in the last decade.