How mnortes fare under referendums. A cross natonal study * Danel Bochsler and Smon Hug CIS and IPZ, Unverstät Zürch Département de scence poltque, Unversté de Genève Paper prepared for presentaton at the ECPR General Conference Potsdam 10 12 September 2009 Frst verson: August 2008, ths verson: August 8, 2009 Abstract The way n whch mnortes and cvl rghts fare when referendums can be used has preoccupped scholars for a long tme. Emprcal studes, so far, have tred to deal wth ths research queston at the subnatonal level by comparng ether referendum or polcy outcomes across subnatonal unts. These unts are, however, often constraned by the natonal level of government. Hence, to understand the full effect of referendums on mnorty polces and cvl rghts, cross natonal comparsons are requred. Relyng on game theoretc * An earler verson of ths paper was presented at the Annual Meetng of the Swss Poltcal Scence Assocaton (Unversty of St. Gallen, January 7 8, 2009). We express our grattude to the partcpants at ths event and especally Isabelle Stadelmann Steffen for helpful comments and the Swss Natonal Scence Foundaton (Grant No. 100012 108179) for generous fundng. Center for Comparatve and Internatonal Studes, Unversty of Zurch, Poltcal scence departement, Selergraben 53, 8001 Zürch, Swtzerland, phone ++41 44 634 50 98; emal: bochsler@pz.uzh.ch Département de scence poltque, Faculté des scences économques et socales; Unversté de Genève; 40 Bd du Pont d Arve; 1211 Genève 4; Swtzerland; phone ++41 22 379 89 47; emal: smon.hug@unge.ch 1
models we thus test the proposton that also the area of mnorty polces the effect of referendums depends on the voters preferences. We test ths proposton wth natonal level nformaton on preferences and nsttutons as well as polcy outcomes n the area of mnorty polces. The set of countres used for the emprcal analyss spans the whole globe. 1. Introducton How mnortes fare n representatve democraces that allow for referendums on partcular polces has occuped scholars and poltcans for some consderable tme. Researchers have attempted to address ths ssue both from a theoretcal perspectve and also n emprcal research. Conclusve results, both at the theoretcal an emprcal level, are, however, stll elusve. Most emprcal work focuses on the subnatonal level, mostly n the Unted States and Swtzerland, or deals wth a sngle country. Comparatve work on ths ssue s, however, largely absent. The present paper wshes to address hs research gap. Drawng on the nsghts from postve theoretcal work, we propose hypotheses that we test on the bass of data dealng wth varous mnorty related polces. Whle we fnd for several polces the expected effect, namely that nsttutonal provsons bas polces towards the voters preferences, ths result s not paramount. Further research needs to address the ssue under what crcumstances the expected effects appear, and when they fal to appear. The paper proceeds as follows. In the next secton we dscuss brefly the theoretcal debate, both from a normatve and postve vewpont, concernng mnortes under referendums. As the postve theoretcal work suggests partcular problems n assessng the effects of referendums on mnortes specfcally and any polcy outcome n general, we dscuss these emprcal problems n secton three. In secton four, we present the emprcal strateges that we deploy to assess the effect of referendums on mnortes cross natonally. Secton fve, after ntroducng our data, presents our emprcal results, whle secton sx concludes by sketchng out a research agenda. 2
2. Mnortes and referendums n theory Whether mnortes mght fare worse n representatve democraces when referendums are possble has not only preoccuped emprcally orented researchers but also scholars workng from a theoretcal perspectve. 1 Theorsts workng from both a normatve and a postve perspectve have attempted to shed lght on ths mportant queston of democratc theory. At least snce Barber (1984) has touted partcpatory democracy as a normatve deal 2 nterest n delberatve democracy related to referendums has ncreased (see for an overvew Chambers 2003). In several studes Bruno Frey wth several co authors (e.g. Frey and Krchgässner 1993; Bohnet and Frey 1994; Frey 1996) argues that campagns on referendums create condtons approachng those of a dscursve deal. Baurmann & Klemt (1993) crtcze specfcally Frey & Krchgässner (1993) by argung that especally also for Habermas (1992) real delberaton s most lkely only possble n parlaments wth repeated nteractons among the protagonsts. Even more crtcal s Chambers (2001) who argues that delberaton, also n the realm of referendum campagns, s undermned f at the end of the process a majorty decson looms (see also Sanders 1997). Based on these normatve elements t appears dffcult to assess how referendums mght affect the poston of mnortes n socetes. If the delberatve element really were to strengthen n drect democratc decsons, then mnortes should be better protected through drect democratc nstruments than through the representatve process. But gven that already the premse of ths clam s heavly dsputed, normatve poltcal theory seems to be of lttle help for the queston how mnortes n contexts wth drect democratc nstruments mght be treated. In the postve poltcal theory, the problem of mnorty protecton under drect democracy goes back at least to the Federalst Papers by Hamlton, Jay und Madson (1787). They have argued that drect democratc decsons by majorty rule mght have a problematc effect on the rghts of mnortes, resultng n a majorty decdng n ts 1 Ths part draws heavly on Hug (2009). 2 In ths context authors often also refer approvngly to Pateman s (1970) work, wthout notng that ths author has a very nuanced, at some nstances crtcal, poston regardng partcpatory democracy. 3
own nterest, whch mght nclude the volaton of the rghts of a mnorty. More precsely, the volonté générale does not always need to be equal to the opnon of the majorty. Gven that the protecton of cvl rghts and mnorty rghts s an mportant element of any democratc state, they can be better protected by a system of checks and balances than through drect legslaton through the ctzens. 3. Emprcal problems of the study of drect democracy The fears of such a tyranny of a majorty have been reflected n a vast emprcal lterature dealng wth the effect of drect democratc nsttutons on mnorty rghts and cvl rghts. Mnorty ssues have taken an mportant place n the study of the usage of drect democratc nsttutons and of ther effect. The mpact of drect democracy has been analysed namely for the rghts of same sex couples (Hader Markel and Meer 1996; Gerber and Hug 2002; Donovan et al. 2008; Bowler et al. 2006; Donovan and Bowler 1998, 1997; Gamble 1997; Hader Markel and Lndaman 2007; Ncholson Crotty 2006; Matsusaka 2007), aborton rghts (Bowler and Donovan 2004; Matsusaka 2007) or parental notfcaton requrements for teenage abortons (Gerber 1996, 1999; Matsusaka 2007), Englsh only laws, prescrbng Englsh as the only language to be used for state acton n US states (Gamble 1997; Gerber and Hug 2002; Matsusaka 2007) or Englsh as only language n schools (Bowler et al. 2006), affrmatve acton for racal or ethnc mnortes (Gerber and Hug 2002; Bowler et al. 2006), recognton of relgous mnortes (Chrstmann 2008), ctzenshp rghts (Helblng and Kres 2004; Bollger 2004), death penalty (Gerber 1999; Hug 2004), ant dscrmnaton laws for job dscrmnaton (Matsusaka 2007), housng and accommodaton, school desegregaton polcy, AIDS polces (Gamble 1997), or same sex marrage (Matsusaka 2007). Hajnal et al. (2002) and Frey and Goette (1998) look at a broad set of mnorty relevant referendums. So far, the comparatve emprcal research on the effects of drect democracy on mnorty rghts has concentrated manly on two countres that practce referendums partcularly frequently: In Swtzerland, drect democracy s practced at all levels of government (see for nstance Helblng and Kres (2004) for a comparson at the muncpal 4
level, Frey and Goette (1998) for a study of natonal and muncpal referendums, or Chrstmann (2008) for the cantonal level). In the US, the effect of drect democracy has been nvestgated comparng drect democratc and non drect democratc states (see Gamble 1997; Donovan and Bowler 1998; Hajnal et al. 2002, and many others). The debate n the US has been fuelled by Gamble s (1997) study, whch reported that ballot propostons (referendums and ntatves) that would ncrease cvl rghts of mnortes are most frequently rejected at the polls. Investgatng drect democratc votes on ssues that are senstve to dfferent mnorty groups, she fnds that n 78% of the cases, the result s lttle favourable of the majorty. But her concluson that drect democracy most often results n a tyranny of the majorty has soon been nuanced and contested. Donovan and Bowler (1998) show that a tyranny of a majorty s most lkely n small communtes. Namely, drect legslaton that would strengthen the rghts of homosexual mnortes s most often beng rejected n referendum votes n states wth small populatons. In large states, however, gay rghts seem not to suffer strongly under drect democracy. These results are contested, however, by Hader Markel and Lndaman (2007), who show that after ncludng addtonal varables, the hypothess about lberal outcomes n large states can be dsmssed. Hajnal and colleagues (2002) do not drectly contradct ths result. Indeed, analysng the ndvdual voters' preferences and vote outcomes n Calforna, they report that members of ethnc or racal mnortes mostly belong to the losers of referendums that drectly target mnorty rghts (Hajnal et al. 2002). However, n many of the votes on mnorty rghts, parts of the mnorty voters as well report to have voted for postons that are classfed as antmnorty. Further, Hajnal and hs co authors show that the same mnortes belong very often also to the wnners of drect democracy n other ssues, ncludng ssues that they consder as very mportant. Even f members of ethnc and racal mnortes do not belong to the wnners of referendums as often as members of the domnant ethnc group, the tyranny of the majorty seems thus to be reduced only to a few questons when the mnorty majorty relatons become drectly an ssue. 5
Whle these results focus on the US, the same can not be sad for Swtzerland, where a look at 64 mnorty relevant referendums at the natonal level and the cty level reveals that outcomes of drect democracy are not partcularly hostle towards mnortes (Frey and Goette 1998). In a replcaton, Vatter and Danac (2008) fnd that the effect needs to be more nuanced by polcy feld, but stll, the Swss voters seem to decde more often n favour of mnortes than voters n the US. These studes have all manly nvestgated the outcome of drect democratc decsons, and looked f they were n favour of mnorty groups. Ths, certanly, captures a subjectve apprecaton of the drect democratc effect by concerned mnortes: gatherng a majorty of supporters behnd mnorty rghts n a referendum mght have the functon of approval for a mnorty wthn a socety, whle the contrary mght lead to a deteroraton of the subjectve ntegraton. However, ths s not necessarly always related to the real consequences n terms of publc polces, because ths vew neglects two mportant aspects. On the one hand, drect democratc nsttutons have an ndrect mpact on the decson makng process n legslatve nsttutons, and on the other hand, not all drect democratc verdcts are fully mplemented. Consderng the ndrect effect Whle the results show that many voters who belong to a mnorty often are among the wnners n referendums, so that not all mnortes need to feel always or systematcally as losers after the holdng of referendums, ths does not teach us very much about the full effect of drect democracy on mnorty and cvl rghts. All these studes neglect the ndrect effect of drect democracy on legslaton. Namely, f the people can veto the decsons of ts representatves, there s not only a drect effect that s manfested n the drect democratc votes, but lkewse, already the threat of a referendum mght affect the polcy. Because legslators are aware that ther laws mght be subject to a referendum, they antcpate a possble negatve verdct of the people, so that the decson process s ndrectly affected by drect democratc nsttutons, even f they are not used. Ths s why drect democratc 6
nsttutons restrct thus the representatve to respect the wll of the majorty of voters (Gerber 1996; Gerber and Hug 2001; Hug 2004). For ths reason, some studes have been nterested n evaluatng both the drect effect of drect democracy and the ndrect effect on decson makng n the representatve nsttutons. Ncholson Crotty (2006), studyng the passage of same sex marrage amendments n the US states, apples a dfferent methodology, lookng drectly at the resultng polces both n referendum states as n states wth purely representatve nsttutons. He fnds that purely representatve systems are leadng to more mnortyfrendly polces, and the dfference s even more pronounced n partcularly dverse socetes. Chrstmann (2008) focuses on the recognton of small confessonal communtes n Swss cantons, namely Jewsh or Muslm communtes. Whle there s only mnor nsttutonal varance among Swss cantons regardng drect democratc rghts, so that there are no fully negatve cases wth no drect democracy, n some cantons, lberal solutons for the recognton of relgous communtes have passed n total revsons of cantonal consttutons. Such packages of varous reforms allow the lawmakers to combne dfferent dsputed ssues n one decson, and to create artfcal majortes. 3 Chrstmann s (2008) result show that all fully lberal rules for the recognton of relgous mnortes have only passed n general consttutonal reforms, where they were only mnor aspects among many more relevant changes to the consttutons. All other full lberalsatons of the recognton rules have ether faled already n the parlamentary debate, or the degree of lberalsaton was substantally reduced, and n both cases, probably antcpatng a possble negatve verdct of referendums. Whle the study shows a negatve ndrect effect of drect democracy nsttutons on one partcular type of mnorty groups, manly drven by the ssue of offcal recognton of Muslm communtes, t s questonable though f the results can be generalsed. Other cvl and mnorty rghts mght not have such a hard stand as Muslms n 3 Furthermore, packages can be used to create a problem of nformaton and to nhbt publc dscusson on certan ssues. The referendum campagn wll possbly focus only on the major aspects of the consttutonal reform, leavng mnor changes out of sght. 7
referendums, and the study desgn does not nclude the potental drect and ndrect effect of popular ntatves, whch can be used to affect cvl and mnorty rghts through drect democratc nstruments. The problem of mplementaton of drect democratc verdcts The full effect of drect democratc nsttutons can stll not be estmated, f consderng the drect effect and the ndrect effect on decson makng. A further problem that has been dentfed by the lterature s the frequent lack of mplementaton of drect democratc verdcts. Even f decsons of a majorty of voters would be legally bndng, they are not always mplemented. There s stll an nsttutonal flter after the voters verdct, snce ntatves that were accepted by a majorty of voters mght ether not be mplemented by governments, or challenged by courts (Gerber et al. 2001; Gerber et al. 2004). 4 Especally, mnorty and cvl rghts are partcularly protected by declaratons of human rghts that are bndng to executves and to courts, so that the effect of referendums s medated by these nsttutons that can affect how the people's decsons get mplemented. For both reasons, t appears wshful to study not only the way how referendums on mnorty ssues are scheduled and decded, but to study as well the full effect of referendums. Studyng the full effect of drect democratc nsttutons In order to capture all dfferent polcy effects of referendums, Gerber (1999) compares the polces of US states, based on the dstncton f they know drect democratc nstruments or not. She shows that drect democracy has rather a conservatve polcy effect. For nstance, n US states that use drect democracy, parental consent s more frequently requred for teenage abortons than n states wthout drect democracy (after controllng for the publc opnon), and captal punshment s more often allowed. 4 Ferraolo (2007) opposes the generalsaton of the stolen ntatve hypothess beyond Calforna, but her method of nvestgaton ntervews wth leadng members of parlamentary groups mght be based, gven that such ntervewees mght not necessarly admt f they do not fulfl the people s verdct. 8
Gerber and Hug (2002) argue that drect democracy needs not necessarly to have a postve or a negatve effect on mnorty rghts, but ths manly depends on the voters preferences on a certan ssue. In a comparatve analyss of mnorty polces n US states, they look at three polces the adopton of Englsh only laws (wth ant Latno undertones), the use of affrmatve acton n publc contracts, and gay and lesban protecton polces. They fnd that the effect of drect democracy on publc polces depends on the voters preferences. If a mnorty frendly polcy fnds strong support among the voters n a certan state, then easly accessble nstruments of drect democracy wll lead to a more mnorty frendly polcy, whle the contrary wll be the case f the voters are rather reluctant to accord a mnorty more rghts and protecton. Bowler and Donovan (2004) employ a smlar method for an nvestgaton of aborton rghts, and Matsusaka (2007) for seven mnorty rghts ssues n the US states. Whle Matsusaka s study s based on a dstncton of US states that use popular ntatves from non ntatve states, argung that the ntatve s the strongest nstrument of drect democracy, Bowler and Donovan nvestgate several operatonalsatons of drect democracy, ncludng measures of the dffculty to qualfy for a referendum/ntatve. Both studes fnd that the polcy outcomes s n stronger congruence wth the opnon of a majorty of voters n states wth (easy) access to drect democratc nstruments. The qute trval result that under drect democratc regmes polces would more closely correspond to the medan voter s preferences than n solely representatve democraces, s however all but obvous. The argument reles on the assumpton that the decson makers are nformed about the preferences of the medan voter. If ths would not be the case, then the argument would only hardly hold. Namely, f the people representatves do not know what the medan voter wants, then nterest groups can put pressure on the decson makers wth the threat of a referendum or a popular ntatve (Broder 2000; Matsusaka and McCarty 2001). Further, nterest groups mght fnance expensve campagns n order to nfluence the referendum outcomes (Garret and Gerber 2001; Gerber 1999). When voters 9
are uncertan about a ballot proposton, they tend to reject t, so that costly campagns aganst a new bll can help to make voters doubt about a project and can lead to a negatve outcome (Bowler and Donovan 1998; Lupa and Matsusaka 2004). 4. Method Our methodology s amed at testng the theoretcal model, argung that the effect of drect democratc nstruments reles on the voters' preferences. Namely, we have argued that n countres that employ nstruments of drect democracy, polces (PO ) n countres should closer correspond to the medan voters' preference X m than n countres wth no drect democracy. X s a dummy varable that measures the exstence of drect democracy n a country. PO X m = f ( X ) (1) The medan voters preference s estmated usng comparatve survey data. We face, however, the problem that PO and X m can not, or only exceptonally, be measured on exactly the same scale. Matsusaka (2001) shows that, lackng nformaton about the comparablty of the scales, we can measure an effect of drect democracy, but we do not know f ths effect brngs the polcy outcomes closer to the voters preferences, or f t dstances them. Model for bnary and ordnal scaled dependent varables The problem of scale comparablty can however be resolved for models that are based on naturally dchotomous polces (Gerber and Hug 2002), or for polces that are measured on ordnal scales. prob P ) = f ( X, X, X X ) (2) ( m m We estmate ths model for each of our sx mnorty or cvl rghts. Our dependent dchotomous varable s coded 1 f the rght exsts n a country, and 0 elsewhere. For ordnal varables, ncreasng values ndcate ncreasng protecton of cvl or mnorty rghts. 10
We expect that the effect of drect democracy on polces wll rely on voters preferences. Ths s captured by the nteracton term of preferences and nsttutons, X m *X. We expect that at the exstence of drect democracy, the extent of mnorty and cvl rghts wll correspond more closely to the voters preferences. Ths means that nteracton term wll have a postve coeffcent, f the preference measure and the polcy measure are coded n the same drecton, or otherwse negatve. Model for metrc scaled dependent varables 5 For dependent varables whch are not bnary or ordnal scaled, the model that uses an nteracton varable can not be employed n order to estmate for the effect of drect democracy towards the voters preferences (Matsusaka 2001). 6 Based on Matsusaka's (2001) suggestons Hug (20010 forthcomng) proposes a swtchng regresson model whch allows for a drect test of whether under partcular nsttutonal arrangements polcy s based toward the voters' preferences. The basc mplcaton of the theoretcal models has been specfed n equaton 1: PO X = f X ) (1) m If the voters' preferences could be measured wthout error on the same scale as the polcy outcome PO, equaton 1 could be estmated drectly. 7 Whle ths s the case when the polcy outcome s bnary or ordnal scaled, n other stuatons the voters' preferences are measured, however, on the one hand wth error and second through proxes and thus not on the same scale as the polcy outcome PO. Hence, X m must be estmated as a functon of these proxy varables accordng to the followng equaton, where P are a set of proxes for the voter preferences: X m = g( P ) (3) ( 5 Ths part draws on Glland Lutz and Hug (2007). 6 Ths problem s closely related to the approprate measurement of the "representatveness" and "responsveness" of legslators (e.g. Achen 1977; Bartels 1991). 7 Agan, the lterature on representatveness and responsveness s llustratve. Representatveness would be assured f the slope estmate would be close to 1 and the ntercept 0. Responsveness requres, however, only a close correlaton between voter preferences and those of ther representatves. 11
If we assume that the functon f n equaton 1 3 s lnear, Hug (2010 forthcomng) demonstrates that all the parameters of nterest can be estmated n a swtchng regresson model wth endogenous swtchng. Gven that both n equaton 1 and 3 error terms are attached, three varance covarance terms must be estmated. 8 The swtchng regresson model derved from equatons 1 and 3 under the assumpton of lnear relatonshps looks as follows: f PO X m > 0 PO f PO X m 0 PO = P β + X γ + ε + θ = P β X γ + ε + θ (4) Thus, negatve estmated values for γ (the coeffcent for the referendum ndcator) would suggest that polces match more closely voter preferences n states wth referendums. We thus estmate ths second, more approprate, model whch allows the effect of referendums to depend on voter preferences. 5. Emprcal study and results In the emprcal part of our paper, we are nterested n estmatng the effect of drect democracy on mnorty ssues and cvl rghts n an nternatonal comparson. Namely, we analyse f n poltes wth drect democratc nstruments, the medan voters preference has a stronger mpact on mnorty rghts and on cvl rghts than n solely representatve democraces. Our study goes beyond the set of two dozens advanced ndustral democraces, whch are commonly used n comparatve poltcs. Namely, we nclude as well systematcally all post communst democraces n Central and Eastern Europe and all countres n Latn Amerca, f the needed measures of voter preferences, taken from the World Value Survey, are avalable. 8 Hug (2010 forthcomng) dscusses ths dervaton n much more detal, provdes evdence from Monte Carlo smulatons that ths estmator performs well and llustrates ns performance n a reanalyss of the data used by Lascher, Hagen and Rochln (1996) and fnds substantvely radcally dfferent results. 12
How to measure drect democratc nsttutons The nsttutonal varable measures whether there s any form of a referendum whch restrcts the power of the exstng veto players n a country. Our measure of drect democratc nsttutons reles on the typology by Hug and Tsebels (2002). We dstngush dfferent types of drect democracy nsttutons, accordng to f they can be used n order to veto a reform that was amended by parlament (referendums), or f t can ntate and decde a reform n a popular vote, possbly aganst the wll of the parlament (popular ntatves). Referendums can be further dstngushed by ther trgger: Mandatory referendums always take place when a certan reform s decded by parlament, but n all nvestgated cases, they ether apply only on changes of the consttuton, or they are lmted on a few pre defned ssues. Non mandatory referendums can be ether trggered by veto players (typcally a majorty of the parlament, or one of the parlamentary chamber, or the executve n presdental systems), or by non veto players (a certan number of voters, or a state nsttuton wth no veto power). We aggregate ths nformaton n a sngle ndcator that measures the exstence of drect democratc nstruments whch empower a non veto player to trgger a referendum that s bndng. The varable s coded postvely, f a country has ether any type of a (bndng) requred referendum, a non veto player referendum, or a popular ntatve for the polcy feld that we nvestgate. The varable s coded as zero f a country does not foresee any drect democratc nstruments, f they are trggered by exstng veto players, so that they can decde on ther own f they want to submt controversal legslaton to referenums, or f drect democratc nstruments are not applcable for the ssues under study. For the codng of our varables, we cross valdate the data wth several dfferent sources of data. The man source s C2D, 9 the orgnal countres consttutons whch are accessble through C2D and the Unversty of Bern. 10 Further, for European countres Councl of Europe, Vence Commsson (2005), for countres n Central and Eastern Europe Auer and 9 Centre for Research on Drect Democracy, Drect democracy n the world, avalable at http://www.c2d.ch [last accessed on 17 June 2008] 10 Unversty of Bern, Internatonal Consttutonal Law Project Informaton, avalable at http://www.servat.unbe.ch/cl/ [last accessed on 17 June 2008] 13
Bützer (2001). For Latn Amerca, addtonal sources for codng were Lssdn (2007), Zovatto et al. (2004), Gonzáles Rssotto (2007), Dons (2006). Among the 54 countres for whch we provde on preference measures, 23 use certan of these forms of drect democracy, whle 31 do not know any forms of drect democracy, or only such that are trggered by the exstng veto players. Dmensons of mnorty rghts and cvl rghts We have estmated the effect of drect democracy nsttutons on several dmensons of mnorty rghts and cvl rghts. We am at ncludng several aspects of cvl rghts and mnorty rghts, n order to get a broad vew on these polces, and n order to obtan results whch are wdely comparable to the fndngs of earler studes on the sub natonal level of poltcs. 11 Our data represent sx dfferent aspects of mnorty rghts or cvl rghts (see table 1). Each of these varables s coded such that ncreasng values ndcate more extended mnorty and cvl rghts, whle low values ndcate that the mnorty or cvl rghts are restrcted; detals for the bnary and ordnal scaled varables are lsted n the appendx, whle the ndcator for the rghts of sexual mnortes, that we constructed ourselves, s explaned below n the paper. Data on most dependent varables was taken from Cngranell and Rchards (2005), whle the aborton rghts were coded from the Harvard Annual Revew of Populaton Law (2008) and from the Center for Reproductve Rghts (2007), 12 and rghts of sexual mnortes were taken from Ottosson (2006). Varable name descrpton scale cr_wosoc05 Women's Socal Rghts, 2005 ordnal scale cr_wecon05 Women's Economc Rghts, 2005 ordnal scale cr_assn05 Freedom of Assembly and Assocaton, 2005 ordnal scale 11 Certan dmensons that were studed n earler research could not be nvestgated, because we are not aware of preference varables that would measure the polcy preferences (namely, death penalty), or because we lack a vald measure for the study of polcy outcomes n a cross natonal settng (rghts of racal or ethnc mnortes). 12 Source for Bulgara: Economc Poston of Women n Bulgara, Georga, Macedona, Poland, Serba and Montenegro and Slovena Summary nformaton sheet (49th Sesson of CSW, March 2005) 14
cr_speech05 Freedom of Speech, 2005 ordnal scale aborton Is aborton allowed, generally or wthn a tmeframe, 2007 bnary varable same_sex ndex of rghts of sexual mnortes metrc scale Table 1: Dependent varables, varable names and scale level Medan voter preferences We have used survey data from the World Value Survey n order to estmate the preferences of the medan voter on these polces. For some of the nvestgated polces, we were not able to select preference questons that would exactly ft the dependent varable. However, we are stll confdent that our preference questons are reasonably close to the measured polces. The followng lst ndcates fve preference measures on whch our analyss reles for sx polcy felds (table 2). 13 Gven that every wave of the World Value Survey ncludes only a selecton of countres, we have ncluded several waves, takng for each country the most recent survey avalable. Varable name wvs_v44m descrpton Labour market gender dscrmnaton ( When jobs are scarce, men should have more rght to a job than women, on a 3-pont-scale from 1, dsagree, to 1, agree) wvs_v60m Housework ( Beng a housewfe s just as fulfllng as workng for pay, on a 4- pont-scale from 1, strongly agree, to 4, strongly dsagree) wvs_v148m wvs_v202m wvs_v204m Strong leader ( Havng a strong leader who does not have to bother wth parlament and electons, on a 4-pont-scale, from 1, very good, to 4, very bad) homosexualty (on a 10-pont-scale from 1, never justfable, to 10, always justfable) aborton (on a 10-pont-scale from 1, never justfable, to 10, always justfable) Table 2: Preference varables Source: World Value Survey, several waves (1984 2007) 13 Theoretcally, veto player models rely on the preference of the medan voter. Emprcally, measures of voter preferences rely on surveys wth only a small number of categores, so that the medan preferences often vary only lttle across countres. If the preference dstrbuton s symmetrcal, the medan s however equal to the mean. For ths reason, we rely on the mean voters preferences. 15
Relevant control varables for mnorty rghts Furthermore, we nclude control varables that are often employed as statstcal controls n studes on mnorty polces. Typcally, studes have compared the mnorty rghts across the US states, and come up wth a set of control varables whch are measurable and vary among the US states. Such varables are for nstance household ncome, deology (conservatve lberal), the percentage of self reported Catholcs or fundamental protestants, the degree of professonalsaton of the legslature, populaton sze, percentage of whte populaton 14, educaton (college), share of non marred households, government deology (relyng on studes on rghts of sexual mnortes, on aborton, parent consent requrement for teenage abortons or death penalty, see for nstance Gerber (1996); Gerber (1999); Donovan/Bowler (1998); Ncholson Crotty (2006)). Certan studes nclude measures of the strength or actvsm of groups wth nterest n the polcy feld (Gerber 1996; Ncholson Crotty 2006), but such a varable mght be endogenous to our model: f the polcy proposed by the parlament strongly dffers from the polcy preference of the populaton, ths mght prepare the grounds for such actvty. We have found control varables whch am at measurng smlar aspects as the ones for the sub natonal comparson of mnorty rghts, but whch are (almost) systematcally avalable for a cross country study (lsted n table 3). GDP per capta measures the economc dmenson; the poltcal orentaton of the executve and the legslatve captures deology and the mpact of poltcal nsttutons; the share of students measures educaton; whle the degree of urbansaton captures the soco economc dmenson. For polcy felds where we expect that relgous values mght have an mpact, we further nclude the share of Catholcs and the share of Muslms. Earler research has shown that the confessonal structure of a country can have a strong mpact on values. 15 Varable name descrpton 14 Others employ the related ndcator of populaton dversty (Ncholson Crotty 2006). 15 In the model wth women s socal rghts (cr_wosoc05) as dependent varable, the share of Muslms n the populaton can not be ncluded as control varable, because otherwse the lkelhood functon becomes too flat to be estmated. 16
gle_gdp00 GDP per capta, 2000 dp_chleg75b Chef executve and one of the three largest partes n congress have left or center poltcal orentaton, 1975-2004 Source: Beck et al 2000; 2001; Keefer 2005, own calculaton van_urban98 urban populaton %, 1998 van_students98 students per 100,000 nhabtants, 1998 lp_muslm80 Muslms (populaton share), 1980 lp_catho80 Catholcs (populaton share), 1980 Table 3: control varables Source, where not ndcated dfferently: Teorell et al. (2008). Models wth bnary and ordnal dependent varables We employ a logstc and an ordnal logstc regresson model for the estmaton of the fve polces wth a bnary or an ordnal scaled dependent varable. The condtoned mpact of voter preferences s ncluded through an nteracton term of drect democratc nsttutons wth the medan voters preferences. Our models test the mpact of polcy preferences and drect democracy on fve dmensons of polcy outcomes. In a frst step, we have calculated for each polcy a model that ncludes drect democracy and the medan voter's preferences wthout an nteracton term (models 1a, 2a, 3a, 4a, and 5a). In three out of fve models, the preference varable has an effect n the expected drecton (see table 4). Namely, women s socal rghts are negatvely correlated wth a tradtonal vew on the women s role n socety. The more respondents reject the statement that beng a housewfe s just as fulfllng as workng for pay, the more women s socal rghts are developed (model 1a); women s economc rghts are postvely correlated wth the publc opnon that women should have the same rghts on the labour market as men (or negatvely wth the asked queston that advocates labour market dscrmnaton when jobs are scarce) (2a); n countres where voters have a rather lberal vew on aborton, aborton s more lkely to be allowed (wthn a certan tmeframe) than n other countres (5a). 17
In two models, explanng freedom of assembly and assocaton (3a) and freedom of speech (4a), our preference measure, measurng authortaran values (strong leadershp), correlaton wth the polcy outcome s extremely weak. In the same models, we have ncluded a smple, non nteractve measure for drect democratc nsttutons (ref3). In three models, explanng women s socal rghts (1a), women s economc rghts (2a), and freedom of speech (4a), drect democratc nstruments are related to lower levels of protecton of cvl and mnorty rghts. Only aborton rghts are slghtly stronger n countres wth strong drect democratc rghts (5a), whle n the case of freedom of assocaton, the correlaton s extremely weak. Overall, for the fve types of mnorty and cvl rghts that are nvestgated n these models, drect democracy seems rather to play a restrctve role, as long as the nteracton effects of drect democracy wth the preferences of the medan voters are not consdered n the model. Our estmatons are based on 48 to 52 cases, dependent on the avalablty of the relevant questons n the World Value Survey. The mpact of drect democracy looks more nuanced, once we consder t n nteracton wth our preference measures. For each of the fve polcy felds, we have estmated a second model (models 1b, 2b, 3b, 4b, 5b), where the preference measure s ntroduced n an nteracton effect wth the nsttutonal varable. We see now that the mpact of preferences on outcomes s moderated by an nsttutonal effect. In four out of fve models, there s a measurable nteracton effect of drect democratc nsttutons wth preferences, although t s not statstcally sgnfcant n any of the cases. And n four out of fve models, the correlaton fts wth our expectaton. In drect democratc countres, the preferences of the medan voters seem not only to be decsve for the degree of mnorty and cvl rghts; furthermore, ther mpact seems to be slghtly stronger than n solely representatve democraces. Ths relatonshp even holds for those polces for whch there was no correlaton n the non nteractve models: n drect democraces, less authortaran preferences of the voters lead to a hgher degree of freedom of assembly and assocaton 18
(model 3b) and of freedom of speech (4b). Only, wth regards to aborton (5b) no such effect could be found. Overall, n democraces wth drect democratc nsttutons, the polcy outcomes on cvl and mnorty rghts represent more closely the medan voter's preference than n solely representatve democraces. Among the control varables ncluded n the model, the GDP per capta seems to be related to stronger respect for cvl rghts and mnorty rghts, although the effect s not statstcally sgnfcant n all models. 19
Dependent cr_wosoc05 cr_wosoc05 cr_wecon05 cr_wecon05 cr_assn05 cr_assn05 cr_speech05 cr_speech05 aborton_gen aborton_gen varable (1a) (1b) (2a) (2b) (3a) (3b) (4a) (4b) (5a) (5b) Preference varable wvs_v60m wvs_v60m wvs_v44m wvs_v44m wvs_v148m wvs_v148m wvs_v148m wvs_v148m wvs_v204m wvs_v204m coeff. s.e. coeff. s.e. coeff. s.e. coeff. s.e. coeff. s.e. coeff. s.e. coeff. s.e. coeff. s.e. coeff. s.e. coeff. s.e. ref*preference 5.154 4.465-1.813 3.310 1.160 1.517 1.892 1.808-0.033 0.838 ref3-2.068 1.208-14.31 8 10.81 2-0.967 0.736-1.717 1.578-0.171 0.701-3.393 4.259-1.267 0.758-6.860 5.465 1.014 0.870 1.143 3.435 preference 4.126 2.476 0.729 1.988-2.083 1.732-1.361 2.166 0.197 0.897-0.280 1.081-0.150 0.996-0.712 1.113 1.381 0.632 1.398 0.764 lp_muslm80-0.020 0.066-0.011 0.067 0.000 0.000 0.000 0.000 lp_catho80 0.520 1.213 0.526 1.223 gle_gdp00 3.2-05 5.0-05 2.6-05 5.3-05 1.1-04 4.8-05 1.1-04 4.7-05 1.3-04 6.5-05 1.3-04 6.7-05 1.2-04 5.9-05 1.1-04 5.9-05 9.0-02 1.5-01 9.0-02 1.5-01 dp_chleg75b 1.648 1.748 1.453 1.785-0.355 1.109-0.219 1.126-0.370 1.010-0.340 1.027-1.208 1.104-1.158 1.095-0.006 0.014-0.006 0.014 van_urban98 0.051 0.040 0.053 0.043 0.033 0.027 0.035 0.027-0.009 0.026-0.007 0.026 0.019 0.025 0.021 0.025-0.075 0.036-0.075 0.036 van_students98 1.5-03 6.7-04 1.5-03 6.8-04 4.7-04 3.7-04 5.0-04 3.8-04 -2.0-04 4.1-04 -2.0-04 4.1-04 -3.5-04 3.7-04 -3.3-04 3.7-04 1.2-04 4.5-04 1.2-04 4.5-04 constant -0.650 2.932-0.718 3.431 /cut1 9.964 6.208 5.671 7.280 3.822 2.139 3.802 2.100 1.225 2.930-0.583 3.387-3.884 3.237-5.938 3.761 /cut2 19.01 2 7.221 15.71 9 7.741 8.772 2.687 8.716 2.655 1.520 3.159-0.039 3.423 Number of obs 50 50 49 49 51 51 51 51 52 52 LR ch2(7) 30.37 31.44 36.6 36.91 12.63 13.24 15.1 17.81 27.23 27.23 Prob > ch2 0 0.000 0.000 0.000 0.049 0.067 0.020 0.013 0.001 0.001 Pseudo R2 0.472 0.489 0.389 0.392 0.185 0.194 0.198 0.234 0.393 0.393 Log lkelhood = -16.99-16.46-28.77-28.62-27.83-27.54-29.22-29.22-21.03-21.03 Table 4: Models wth a dependent bnary or ordnary scaled varable, logstc/ordnal logstc regressons, coeffcents and standard errors. 20
Model wth metrcally scaled dependent varable Further, we have ncluded a model that measures the mpact of drect democracy on the rghts of same sex couples, for whch we employ a metrc scaled ndcator. These polces have been addressed already n earler research on the effect of drect democratc nsttutons. Hader Markel and Meer (1996) nclude a whole range of aspects of same sex couples dscrmnaton, and nvestgate prohbtons aganst dscrmnaton n publc employment, publc accommodatons, prvate employment, educaton, housng, credt, and unon practces. In a cross natonal study desgn, a focus only on prohbtons of dscrmnaton of sexual mnortes would however narrow down the perspectve on only few aspects of polces towards sexual mnortes. Besdes the exstence of legal provsons aganst dscrmnaton, there s a number of ssues, rangng from same sex marrages, adopton through same sex couples, up to the possblty for transgender persons to get new documents. We rather want to construct an ndcator that depcts the general polcy aganst sexual mnortes, coverng dfferent aspects where mnortes mght experence dscrmnaton. For ths am, we construct an ndcator that covers eght dmensons, measurng f sexual mnortes have smlar rghts as heterosexual couples, and f they are protected aganst dscrmnaton (table 5). Comparable nformaton on all ssues has been compled by Ottosson (2006), coverng the legal stuaton of sexual mnortes n 2006. Dmenson Codng Mean (std. dev.) same age lmt for homosexual and heterosexual relatonshps (1) vs. hgher age lmt for homosexuals (0) Same-sex marrage allowed (1), same-sex cvl unons/partnershp laws (0.5), certan benefts for samesex couples (0.25), vs. others (0) Adopton rghts for same-sex couples (1) or step-chld adopton (0.5) vs. others Rght of artfcal nsemnaton for same-sex couples or sngle women Prohbton of dscrmnaton of homosexuals n employment 0, 1 0.963 (0.191) 0, 0.25, 0.5, 1 0.194 (0.306) 0, 0.5, 1 0.102 (0.281) 0, 1 0.407 (0.496) 0, 1 0.667 (0.479) General prohbton of dscrmnaton of homosexuals 0, 1 0.481 (0.504) Prohbton of homophobc propaganda 0, 1 0.333 (0.476) 21
Allowng new personal documents for transgender persons Overall addtve ndex Mn: 1, max: 8 0, 1 0.167 (0.376) 3.314 (2.057) N=54 Table 5: Index of rghts of sexual mnortes. Agan, we employ a varable on the preferences of the medan voters on homosexualty, coded from the World Value Survey, wvs_v202m, and the same set of control varables. Snce the dependent varable s contnuous a smple nteracton term between preferences and the presence of drect democratc nstruments wll not allow us to estmate the effect of these nstruments. Consequently, we employ the swtchng regresson model as dscussed above. OLS MLE (SR) coeff. s.e. coeff. s.e. (Intercept) -1.878 1.369-1.321 0.007 wvs_v202m 0.328 0.222 0.212 0.005 gle_gdp00 6.185 4.405 5.504 0.005 dp_chleg75b 0.818 0.733 1.840 0.011 van_urban98 3.684 1.913 3.229 0.003 van_students98-0.028 0.276 0.162 0.009 constant -0.761 0.212 ref3 0.093 0.511-1.904 0.653 varance 1 1.405 0.152 varance 2 1.921 0.152 varance 3 0.258 0.000 rse 1.650 llk 48.28 56.19 N 46 46 Table 6: Model for the estmaton of the effects on our metrcally scaled dependent varable. Legend: OLS stands for the lnear regresson model (ordnary least squares, whle MLE SR stands for the swtchng regresson model estmated wth maxmum lkelhood. Table 6 reports the results. In the frst column the results of a smple OLS regresson and are reported, whch suggest that the presence of drect democratc nstruments fals to affect the presence of polces related to same sex relatons. Our preference varable weakly affects 22
the extent of these polces, as does the degree of urbanszaton. When the same model s estmated wth the help of a swtchng regresson the substantve results change dramatcally. Whle the effects of the preference varable and control varables reman largely the same (though n terms of statsttcal sgnfcance change), we fnd a sold and statstcally sgnfcant effect for the presence of drect democratc nstruments. More precsely, f a country makes avalable such nstruments, polces regardng same sex relatons follow much more closely the preferences of the voters. 6. Conclusons The mpact of drect democracy on cvl rghts and mnorty rghts has been vvdly dscussed n the lterature on the effect of referendums. Whle the topc has manly been nvestgated on the sub natonal level, lookng at the poltcal effect of referendums n Swtzerland or n the US states, ths s the frst contrbuton that attempts to nvestgate ths queston n a cross natonal perspectve. Our database ncludes 54 democraces, gong beyond the typcal set of advanced ndustralsed democraces, and ncludng as well postcommunst democraces n Central and Eastern Europe and the countres of Latn Amerca. We rely on a model that takes nto account the preferences of the medan voter, polcy outcomes, and the nsttutonal settngs. Our model predcts that the poltcal effect of referendums and ntatves s not a genunely postve or negatve one, but rather, t depends on the preferences of the medan voter. Namely, the possblty of usng a referendum or an ntatve n order to veto parlamentary legslaton helps the medan voter to realse a polcy that s close to hs preferences. In sum, polcy outcomes n countres wth drect democratc rghts should be closer to the preferences of medan voters than n countres wth no referendums or ntatves. For our study, we have focussed on sx aspects of mnorty rghts and cvl rghts, rangng from the rghts of same sex couples over freedom of speech, freedom of assembly and assocaton, socal and economc dscrmnaton of women up to aborton rghts. After controllng for voters preferences and a set of control varables, we fnd that there s a 23
general restrctng effect of drect democracy on mnorty rghts or cvl rghts. Only wth regards to aborton, drect democratc countres apply a more lberal polces. The pcture s gettng more shades, however, once we estmate the jont effect of drect democracy and voter preferences. Our results, whle remanng weak and below standard levels of statstcal sgnfcance, suggest for almost all nvestgated polces that referendum and ntatve move the results towards the preferences of the medan voter. Ths exploratory study opens new emprcal perspectves for the study of mnorty rghts and cvl rghts. In fact, many of these ssues are regulated at the natonal level of poltcs, and for ths reason, t s advsable to explore natonal regulatons, rather than concentratng on the sub natonal level, where only a small set of polces are defned. Our theoretcal model even f t s put up challenges for the estmaton remans rather smple, nvestgatng a genune effect of drect democracy, jontly wth voter preferences. Further research mght be done n order to dfferentate between dfferent nsttutons of drect democracy, or consderng further aspects that, for the sake of smplcty and lackng readly avalable comparable data, dd not consder, such as the dffculty to trgger a referendum or an ntatve, or the homogenety of voters preferences. 24
Appendx Descrpton of the polcy varables cr_wecon05 Women's Economc Rghts, 2005 In measurng women s economc rghts we are prmarly nterested n two thngs: one, the extensveness of flaws pertanng to women s economc rghts; and two, government practces towards women or how effectvely the government enforces the laws. Regardng the economc equalty of women: (0) There are no economc rghts for women under law and systematc dscrmnaton based on sex may be bult nto the law. The government tolerates a hgh level of dscrmnaton aganst women. (1) There are some economc rghts for women under law. However, n practce, the government DOES NOT enforce the laws effectvely or enforcement of laws s weak. The government tolerates a moderate level of dscrmnaton aganst women. (2) There are some economc rghts for women under law. In practce, the government DOES enforce these laws effectvely. However, the government stll tolerates a low level of dscrmnaton aganst women. (3) All or nearly all of women s economc rghts are guaranteed by law. In practce, the government fully and vgorously enforces these laws. The government tolerates none or almost no dscrmnaton aganst women. cr_wosoc05 Women's Socal Rghts, 2005 In measurng women s socal rghts we are prmarly nterested n two thngs: one, the extensveness of laws pertanng to women s socal rghts; and two, government practces towards women or how effectvely the government enforces the law. Regardng the socal equalty of women: (0) There are no socal rghts for women under law and systematc dscrmnaton based on sex may be bult nto the law. The government tolerates a hgh level of dscrmnaton aganst women. (1) There are some socal rghts for women under law. However, n practce, the government DOES NOT enforce the laws effectvely or enforcement of laws s weak. The government tolerates a moderate level of dscrmnaton aganst women. (2) There are some socal rghts for women under law. In practce, the government DOES enforce these laws effectvely. However, the government stll tolerates a low level of dscrmnaton aganst women. (3) All or nearly all of women s socal rghts are guaranteed by law. In practce, the government fully and vgorously enforces these laws. 25
The government tolerates none or almost no dscrmnaton aganst women. cr_worker05 Workers Rghts, 2005 Worker s rghts are: (0) Severely restrcted (1) Somewhat restrcted (2) Fully protected cr_assn05 Freedom of Assembly and Assocaton, 2005 Ctzens rghts to freedom of assembly and assocaton are: (0) Severely restrcted or dened completely to all ctzens (1) Lmted for all ctzens or severely restrcted or dened for selected groups (2) Vrtually unrestrcted and freely enjoyed by practcally all ctzens cr_speech05 Freedom of Speech, 2005 Government censorshp and/or ownershp of the meda (ncludng rado, TV, Internet, and domestc news agences) s: (0) Complete (1) Some (2) None aborton Aborton allowed, generally or wthn a certan tmeframe Not ncluded f aborton s only allowed as excepton for lfe danger or only to preserve mental health of the woman, n cases of rape, or n dffcult crcumstances 26
References Achen, Chrstopher H. 1977. "Measurng Representaton: Perls of the Correlaton Coeffcent." Amercan Journal of Poltcal Scence 21 (4):805 15. Auer, Andreas, and Mchael Bützer, eds. 2001. Drect Democracy: The Eastern and Central European Experence. Aldershot: Ashgate. Barber, Benjamn. 1984. Strong Democracy: Partcpatory Poltcs for a New Age. Berkeley: Unversty of Calforna Press. Bartels, Larry M. 1991. "Consttuency Opnon and Congressonal Polcy Makng: The Reagan Defense Buldup." Amercan Poltcal Scence Revew 85 (2):457 74. Baurmann, Mchael, and Hartmut Klemt. 1993. "Volksabstmmungen, Verhandlungen und der Schleer der Insgnfkanz." Analyse und Krtk 15:150 67. Bohnet, Irs, and Bruno S. Frey. 1994. "Drect Democratc Rules: The Role of Dscusson." Kyklos 47 (3):341 54. Bollger, Chrstan. 2004. "Spelt es ene Rolle, wer entschedet? Enbürgerungen n Gemenden mt Parlaments und Volksentsched m Verglech." In Paradoxen m Bürgerrecht. Sozalwssenschaftlche Studen zur Enbürgerungspraxs n Schwezer Gemenden, ed. P. Stener and H. R. Wcker. Zürch. Bowler, Shaun, and Todd Donovan. 1998. Demandng Choces: Opnon, Votng, and Drect Democracy. Ann Arbor: Unversty of Mchgan Press.. 2004. "Measurng the Effects of Drect Democracy on State Polcy: Not All Intatves Are Created Equal." State Poltcs and Polcy Quarterly 4 (3):345 63. Bowler, Shaun, Stephen P. Ncholson, and Gary M. Segura. 2006. "Earthquakes and Aftershocks: Race, Drect Democracy, and Partsan Change." Amercan Journal of Poltcal Scence 50 (1):146 59. Broder, Davd S. 2000. Democracy Deraled: Intatve Campagns and the Power of Money. New York (NY): Harcourt, Inc. Center for Reproductve Rghts. 2007. "The World's Aborton Laws." New York: Center for Reproductve Rghts. 27