THE IMPACT OF POLITICAL LEADERS ON VOTING IN ROMANIA

Transcription

1 THE IMPACT OF POLITICAL LEADERS ON VOTING IN ROMANIA Andrei GheorghiŃă Department of Soiology Luian Blaga University of Siiu, Romania Paper prepared for the European Leaders and Demorati Eletions Workshop, ECPR Joint Sessions, Lison, April 2009 (Work in progress, please do not ite without the author s permission)

2 THE IMPACT OF POLITICAL LEADERS ON VOTING IN ROMANIA 1. POLITICAL PARTIES AND THE PERSONALISATION OF ELECTORAL POLITICS Classi models have generally plaed politial parties at the entre of any explanation of voting ehaviour during parliamentary eletions. Aording to suh a perspetive, eletions should e regarded as ompetitions among politial parties (hene group ators) legitimized through a speifi plaement in relation to the main leavages a soiety has to deal with at a partiular moment of time. As a onsequene, personal harateristis of a andidate for puli offie are not expeted to ear great influene upon the eletoral deision, espeially in soieties where latent onflits among large soial groups tend to set the general framework for the eletoral ompetition. A signifiant transformation in the relevane for the eletoral deision of suh frozen politial alignments is antiipated y Otto Kirheimer in a famous essay (1966), disussing a new speies of parties, ath-all, seriously hallenging the old-fashioned lass-mass parties and denominational mass parties. Cath-all parties are defined in terms of: (a) drasti redution of the ideologial aggage; () strengthening of top leadership groups; () downgrading the role of the individual party memer; (d) de-emphasis of the speifi soial-lass or denominational lientele; (e) seuring aess to a variety of interest groups (Kirheimer, 1966: 190). The hange antiipated y Kirheimer finally ourred, involving the generalization of the new party speies. Consequently, from the eletoral perspetive, politial leaders have egun to play an inreasingly important role. This role of politial leaders in the game of eletions enompasses at least two dimensions, namely presentation and impat (Mughan, 2000: 11). The first one relates to the way politial parties hoose to present themselves to the puli, setting their leaders as the ore element or the puli fae of the party. The seond one relates to the presumed effets of leader s personality and ehaviour on the voting patterns of the itizens. This paper is ased on work arried out during a visit (AG0108) to the European Centre for Analysis in the Soial Sienes (ECASS) at the Institute for Soial and Eonomi Researh, University of Essex, supported y the Aess to Researh Infrastrutures ation under the EU Improving Human Potential Programme. I wish to thank Cristina Stănuş, HoraŃiu Rusu, and Mirea Comşa for their valuale omments and help in the various stages of writing this paper. 2

3 The two dimensions define the phenomenon referred in the literature as the personalization of eletoral politis or the presidentialization of voting 1. It involves a growing tendeny to assoiate the eletoral deision in favour of a party or another with the image of its leader, ut also to identify the party itself with the leader. As a onsequene, the personal harateristis of politial leaders are expeted to have a growing impat on the party vote during parliamentary eletions. Although still deated in the literature, eause of rather inonsistent empirial findings, the personalization thesis is largely taken for granted y ampaign strategists all over the world. Thus, eletoral ampaigns are regarded as inreasingly personalized (Shmitt and Ohr, 2000), presidentialized (Crewe and King, 1994; Mughan, 2000; Poguntke and We, 2005) or andidate-entered (Wattenerg, 1991). Explanations for suh a hange in the voting patterns in parliamentary eletions, ut also in the way suh ampaigns are fought, are not linked exlusively to the sustantive transformations faed y politial parties, ut also to other relevant fators that are to e outlined in the next setion of this paper. 2. THE CAUSES OF PRESIDENTIALIZATION Change in the patterns of mass ommuniation, with a growing role of eletroni media (espeially television), is the phenomenon most widely referred to in the literature in relationship to the presidentialization of voting (Bean and Mughan, 1989; MAllister, 1996, 2007; King, 2000; Shmitt and Ohr, 2000; Poguntke and We, 2005). Television fouses on image for roadasting politial information and, for this purpose, is in a desperate need for faes. Thus, it is just natural for television to use politiians as a vehile for disseminating information, while disregarding astrat ideas, institutions or politial douments. Additionally, television has the apaity to reveal, more diretly and more authentially than any other mean of mass ommuniation, a multitude of nonveral attriutes of politiians: faial appearane, style, lothing, and ody language (Shmitt and Ohr, 2000: 6). On the other side, voters might find it easier to manage politial issues y referene to leaders and their media representations. Hene, the information osts of the individual voter are signifiantly redued: while, efore the age of the television, developing an 1 Presidentialization of voting is frequently regarded in the literature as a onstituent of a larger proess of presidentialization of parliamentary systems. As a proess, the presidentialization involves a transformation in the funtional logi of parliamentarianism along three diretions: an inrease in the power resoures of the prime ministers within the exeutive, an inrease in the autonomy vis-à-vis his/her own party, and the personalization of the eletoral proess (Poguntke and We, 2005: 4-5). 3

4 individual position related to various issues was a rather diffiult (perhaps even prohiitive) task, the use of individual politiians as referene frames for politial topis has signifiantly simplified the hoie of the voter. The voters might empathize or not with the politiian s goals, might put themselves or not in their shoes. However, it s likely that the voters shall gain a etter understanding of the politiian s views and understandings (MAllister, 2007: 579). Thus, for television, politial leaders represent a visual shortut in the effort of apturing and retaining the viewers attention on politial issues (MAllister, 2007: 579), while for the viewers they at as an information shortut for forming eletoral deisions in favour or against parties (Popkin, 1994: 61). Later prevalene of ommerial private television has further amplified suh hanges (Poguntke and We, 2005: 15). In order to gain larger shares of viewers, private television frequently omined information and entertainment in a mixed infotainment (Shmitt and Ohr, 2000: 6), leaving no plae to anything else than the human element of politis. Thus, politial leaders turned fast into TV superstars (Shwartzenerg, 1995: 12-14, Gidengil et al., 2000: 3). Politial parties adjusted their strategies as a reation to transformations in the media system. Leaders turned into the puli faes of politial parties and their disourse adapted to the logi and format of television y fousing extensively on symolism instead of sustane (Poguntke and We, 2005: 15). This is mainly a onsious hoie taken y politiians in order to exploit the visual media s potential for simplifiation and symolism for their own purposes (Poguntke and We, 2005: 15). Erosion of traditional eletoral alignments is another ause of presidentialization widely mentioned in the literature (Wattenerg, 1991; MAllister, 1996, 2007; Shmitt and Ohr, 2000). The traditional ties etween parties and soial groups, developed along politial leavage lines, have faed a ontinuous deline during the last deades, in spite of onsiderale efforts done y politial parties to revitalize them. Suh dealignments refleted in a ontinuous deline in oth party memership and turnout (MAllister, 2007: ). While traditional leavage politis had previously assumed a oherent politial program addressing the aligned eletorate is the key to eletoral suess, dealignment has largely suverted the very sustane of suh a strategy. Politial parties had to fae more heterogeneous eletorates, oth soially and ideologially, less interested in the ideology and the loyalties of the soial group they elong to (Poguntke and We, 2005: 15), and onsequently more volatile (Wattenerg, 1991: 2). Thus, group identifiation has turned into a rather lousy preditor of voting ehaviour, with a limited and dereasing moilization potential. One the ath-all strategy has een adopted y a large majority of 4

5 politial parties, the signals transmitted towards soial groups eame more amiguous, as politial manifestos were targeting rather heterogeneous ategories of voters. As a onsequene, large ategories of voters eame availale (Poguntke and We, 2005: 16) on the eletoral market, politially dealigned and ideologially onfuse, potentially easy to ath and, eventually, moilize. Parties reated with a strategi inrease in the fous given to politial leaders and their attriutes, onsidered to e the appropriate tools for re-moilizing the eletorate. Thus, personal harateristis of the andidates eame elements largely exploited in the ampaign strategies, in a delierate attempt to extrapolate the positive image of andidates towards their parties as a whole. A general shift in voters awareness from loal eletions to national eletions might e regarded as a simultaneous result of the deline in the ties etween parties and soial groups and of the hanges in the patterns of mass ommuniation aout politis. National politis is y its nature more spetaular and hene more ale to draw itizens attention. Moreover, resoures invested y the politial ators in national ampaigns are signifiantly higher than in loal ones, meaning a more fasinating representation for the puli. If in ontext of print (or even audio) media national politis was hardly aessile to most of the voters, television has radially transformed things. National politis is nowadays easily aessile to everyone having a TV and it eame a muh simpler usiness than ever efore, thanks to parties efforts of adaptation to the television formats and standards. Furthermore, the growth in the role of the state in ontemporary soieties, after the seond world war, has supplementary nourished this orientation of the voters towards the national ontext of politis instead of the loal one. Sine national politis mainly uses politial leaders as agents of eletoral moilization (rather than loal networks, ased on group loyalties, as it was the ase in loal politis), this shift should e regarded as ringing up additional leadership-enteredness in eletoral politis. The internationalization of politis has largely redued the relevane of olletive deisionmaking odies, at least at national level. A growing share of the politial prolems governments have to fae an only e dealt with via international ooperation, involving inter-governmental ooperation (Poguntke and We, 2005: 13). The gloalization of politial deisions has shifted the power to high rank exeutive ators (prime ministers, presidents), while delierative odies (parliaments, even ainets) find themselves in the unomfortale position of simply ratifying deisions taken elsewhere (Poguntke and We, 2005: 13-14). Thus, individual ators (leaders) onentrate more deisional resoures and, onsequently, more visiility, to the detriment of group ators (parliaments, parties). 5

6 The inreasing omplexity of politial issues in the last deades has additionally nourished the voters severe need for a simplifiation of their politial representations. Thus, they might find easier to assoiate politial power and authority with a readily identifiale politial personality than with an astrat institution or politial ideal (MAllister, 1996: 287). For the ordinary itizen, it s easier to hold individuals aountale for politial ations than an institution or an ideology. Suh a mehanism of uilding politial representations will hold diret onsequenes upon how ordinary people design their representations on politial parties and eletions. 3. ANALYTICAL STRATEGIES Analysing the effets of politial leaders on party voting is a diffiult task. There are rather few authors that ontest the very existene of leader effets. Usually, the dispute is foused upon the strength of leader effets in the eletoral deision-making proess. At the origins of this dispute lie the diffiulty of disentangling the effets of leaders personalities and other personal attriutes orne on individuals vote deisions from the influenes of all the other fators, suh as group loyalties, partisan likes or dislikes, personal values, ideologial preferenes, previous performane of politial parties (King, 2002: 14-15). In other words, it is extremely diffiult, if not impossile, to isolate pure leadership effets without eing ontaminated with influenes from other variales. Broadly speaking, there are three analytial strategies mentioned in the literature for measuring leadership effets: the experimental strategy, the improved-predition strategy, and the ounterfatual strategy (Crewe and King, 1994: ; King, 2002: 15-21). Following, we shall riefly disuss eah of the three strategies, in an attempt to review the speifi advantages and disadvantages involved y suh approahes. A pure experimental design would imply running a onsiderale numer of national eletions with a strit ontrol of alternative explanatory variales, allowing exlusively variations of stimuli related to the personalities and other personal harateristis of the parties leaders or andidates. Oviously, in this pure format, the experimental strategy is impossile to exeute. Additionally, suh a design would fail to onsider the indiret leader effets on voting (the influene exerted via leaders ativity inside the party organization or the government). Experimental studies fousing on the personalization of eletoral politis attempt to simulate an eletoral ontext in whih stimuli are represented y fititious andidates that 6

7 express themselves y the means of pitures/flyers (Rosenerg et al., 1986) or video lips (Jenssen and Aalerg, 2006). The experimental group (university students or ustomers of a loal store) is introdued to pitures/video lips of the andidates, ut also to additional information aout their party affiliation, issue positions adopted and their akground (eduation, oupation, politial experiene) (Rosenerg et al., 1986: ). In the Norwegian experiment, video lips reprodue disourses of real politiians, y the means of fititious andidates played y professional ators (Jenssen and Aalerg, 2006: ). Hene, suh studies attempt to reprodue an eletoral ontext, under whih they manipulate different stimuli related to personal harateristis of the andidates, with a strit ontrol of alternative explanatory fators for the eletoral deision. However, there is very little to do in order to avoid the artifiiality of suh an eletoral ontext, although the sujets appeared to have taken the task of voting quite seriously in oth ases (Rosenerg et al., 1986: 117; King, 2002: 16; Jenssen and Aalerg, 2006: 261). Both experimental studies onfirm the hypothesis of a signifiant impat that personal harateristis of the andidates ear on the vote deision 2. The improved-predition strategy hooses a different approah for dealing with the need to measure pure leadership effets. It attempts a statistial measurement of the additional explanatory power rought in a model of party vote y the personal judgments on politial leaders. In other words, it assumes the explanatory variales of vote oming from alternative models (voters values, ideologial affiliations, party preferenes, group loyalties, retrospetive pereptions of reent performane in offie) that are entered to multiple regressions as independent variales, in order to statistially ontrol their effets. The goal of suh a design is a preise measurement of the additional explanatory power (improvement in predition) rought to suh multiple regression models y the evaluations of politial leaders. The improved-predition strategy is extensively used into analyses of the personalization of the eletoral politis. It is employed in a famous series of studies y Warren E. Miller and J. Merrill Shanks dediated to the U.S. presidential eletions of the 1980s and 1990s (Miller and Shanks, 1982, 1996; Shanks and Miller, 1990, 1991). Miller and Shanks explain the hoies of individual voters as eing a umulative effet of several fators temporally ordered: (1) partisan and ideologial predispositions, (2) poliy preferenes, (3) performane evaluations, (4) issue distanes etween respondent and the andidates, and (5) other (non-politial) andidate evaluations (Miller and Shanks, 1982: 2 Oviously, the two experimental studies are not omparale, neither in terms of design, nor of results. There is a signifiant numer of variales that should also e taken into onsideration, like the politial system, the eletoral system, the party system, the type of eletions simulated et. 7

8 341). Thus, eah of the five fators has a net ontriution (and theoretially measurale) in the formation of the vote deision. Their ordering is not at all aritrary, as the sequene starts with more stale fators (far from the eletoral deision) and ontinues with less stale and more ontextual ones (nearer in time to the individual s final vote deision) 3. Perhaps the most expliit use of the improved-predition strategy in the analysis of the personalization of the vote deision omes from Shmitt and Ohr (2000). They employ the so-alled R² strategy for analysing the relevane of leader effets on voting in Germany for the period etween 1961 and Hene, they measure the additional variane in the party vote (R²) explained y leaders evaluations (net effets), after ontrolling for the effets of soial-strutural fators and party/ideologial identifiation (Shmitt and Ohr, 2000: 16). The ounterfatual strategy is somehow linked to the experimental strategy and relies on the tehnique of thought-experiments, emphasizing the expliit answers to What if? questions. What if John Edwards were the Demorat hallenger of George W. Bush for the Amerian presideny instead of John Kerry? An infinite numer of suh ounterfatual questions an e formulated, at least theoretially, aout a single eletoral ompetition, and onsequently an infinite numer of thought-experiments are needed. In pratie, as Mughan (2002: 19) notied, the few authors adopting the ounterfatual strategy have hosen to run their thought-experiments on ounterfatuals for whih real-world data are availale. Bean and Mughan employ this strategy in the final setion of a highly influential artile on leadership effets in parliamentary eletions in Australia and Britain (1989). Their ounterfatual question refers to the possile eletoral effet of a reversal in the personal profiles of the two main party leaders in eletions of 1987 in Australia and 1983 in Britain (the Laour leader John Hawke and the Conservative John Howard in the first ase, respetively Conservative Margaret Thather and Laour leader Mihael Foot in the seond ase). Based on British Eletion Study (1983) and Australian Eletion Study (1987) data, Bean and Mughan develop a projetion of the eletion results in the hypothetial ase of the losers (Howard and Foot respetively) would have een pereived y the eletorate similarly to the winners (Hawke and Thather respetively) in terms of personal harateristis 4. Results suggest a signifiant influene of leaders personal harateristis 3 In their view, earlier more stale fators will simultaneously have a diret influene on the vote deision, ut also an indiret one, via influening later less stale fators (King, 2002: 18). 4 Respondents were offered a limited set of attriutes (aring, determined, shrewd, likale as a person, tough, listens to reason, deisive, stiks to priniples, effetive) and asked aout the degree to whih the main two 8

9 upon the party vote 5. Thus, Bean and Mughan s strategy provides an estimation of the vote gains rought y having the right party leader (Bean and Mughan, 1989: 1175). 4. DEFINING EXPECTATIONS: LEADERS IN THE ROMANIAN SOCIAL, POLITICAL, AND CULTURAL CONTEXT History gives leaders a entral part in Romanian politial life, e them voievod, prines, kings or Communist ditators. With the notale exeption of kings, in eah of these ases we an identify leaders that went far eyond aeptale arriers for the exerise of power in a modern onstitutional regime. The last of the series, the Niolae Ceauşesu regime, ontriuted through its extreme degree of personalization and aritrary deisionmaking to defining the ontemporary onept of sultanism (Linz and Stepan, 1996: 44-45, 51-54; Pasquino, 2002: ). Nevertheless, the periods when politial life was dominated y single ators alternated with periods when olletive ators suessfully ounteralaned the autorati tendenies of leaders. Among these olletive ators we should mention assemlies of aristorats, the Diet of Transylvania (oth during the medieval age), and the proto-politial parties and the parliament (in the modern age). Given these aspets, it is almost impossile to define some lear expetations regarding the role and relevane of politial leaders for the eletoral ompetition, ased on analyzing historial events or traditions. Despite the numerous examples of politial leaders exerising power in a disretionary manner, we an hardly say that these examples are harateristi of a given historial period, no matter what that period would e. Nevertheless, olletive memory seems to e dominated y the images of suh leaders, skethed in highly positive shades, whih determines many to see nostalgia of powerful leaders at population level. Fresh memories of the sultanisti regime determined the 1990 Constituent Assemly to opt for semipresidentialism, whih would limit the potential authoritarian tendenies of the president y giving him limited prerogatives 6. Regardless of this aspet, the Constituent party leaders possess suh attriutes. Additionally, they were asked to offer a sore of the pereived effetiveness of alternative leaders (Bean and Mughan, 1989: 1168). 5 Regression models employed y Bean and Mughan in the British ase suggest that if Foot had had Thather s personality profile, other things eing equal, Laour s vote in 1983 would have een 6-7 perentage points higher. In the Australian ase, the ontrast in personality profiles would have translated into 4 perentage points (Bean and Mughan, 1989: 1175). 6 The prerogatives of the president of Romania an e separated into three ategories: prerogatives of representation (he or she is representative eause he or she is legitimated via diret expression of the will of the people, ut also he or she represents the state, inside and outside Romania), prerogatives of safeguarding (the state and the onstitution) and prerogatives of mediation (etween different puli authorities or 9

10 Assemly also opts for a diretly eleted president, solution whih is often seen as a orretive mehanism in the event of the emergene of a highly fragmented party system (Siaroff, 2003: 308). Consequently there is a ontradition etween the position of a president, well aove the typial president in a parliamentary regime, and his real prerogatives, that ontriutes to diminishing his role within the state. The results was often haraterized as parliamentarism with a presidential orretive (Siaroff, 2003: 287, 308), with unlear impliations for the personalization of voting. On the one hand the eletoral ompetition for the presidential offie will stimulate the appearane of strong personalities, apale of launhing multiple messages aimed at the voters; while on the other hand, the prerogatives suggest a seondary position one the eletoral proess is finished. Despite these unlear onstitutional impliations, our expetations onverge toward strong leader effets, reated y the joint organizing of presidential and legislative eletions. At the same time, it is our expetation that these leader effets will diminish in time, as a result of the separation of presidential and legislative eletions following the hanges rought to the onstitution y law no. 429/ Another regulation regarding eletoral ompetition with potential impat upon the intensity of voting personalization is law no. 68/1992 regarding the eletion of the Chamer of Deputies and the Senate, whih introdues a proportional representation system with losed lists. This eletoral formula gives the image and the reputation of the andidates the lowest possile influene on voting (Carey and Shugart, 1995: 424), transferring to the party lael the essential part in the formation of the voting deision. It is to e expeted that suh an institutional framework will suessfully inhiit the personalization of eletoral politis at onstitueny level, ut it will not have a similar effet on the transfer of popularity from the leaders of the national party, espeially sine legislative and presidential eletions are simultaneous. A different set of expetations is outlined y the hange of eletoral formula introdued y law no. 35/ from PR with losed lists to a mixed system (majority voting in single seat onstituenies, with redistriution mehanisms). It is expeted that this hange would give extra meaning to the image and personality of loal andidates to the expense of the leaders of the national party and party etiquette in general. But suh onsequenes an only e antiipated and, sine the effets of this hange will eome ovious at a moment of time outside the reah of this paper, I will not address them. etween soiety and puli authorities) (Deleanu, 1998: ). 7 A hange to the Romanian Constitution y Law no. 429 /2003 expands the presidential term from 4 to 5 years, while the term of the memers of parliament remains 4 years. Thus, the presidential and legislative eletions are separated and an only e organized simultaneously every 20 years. 10

11 Literature emphasizes the erosion of traditional alignments among the fators that ontriuted to the inrease in personalization of eletoral politis. This erosion left ehind an ideologially onfused, de-aligned and volatile eletorate (Poguntke and We, 2005: 16), that needed to e aptured y appeal to the personal qualities of the andidates. Suh an explanation loses all relevane when we look at new demoraies in Central and Eastern Europe, where we do not have stale party systems or long term party and ideologial identifiation materialized in onsolidated eletoral alignments. Nevertheless, we an not neglet some similarities etween the erosion of traditional alignments in Western Europe and the formation and onsolidation of Central and Eastern European party systems that manifest at the voter level: low ideologial and politial party involvement; weak party identifiation; generalized dissatisfation, lak of trust and alienation regarding politis; and high eletoral volatility. One we delineate this similarity etween the two historial periods in party system evolution, there seem to e enough premises to onsider the party system development and onsolidation stage a failitating fator in the personalization of politis in Central and Eastern Europe, in general, and Romania, in partiular. Moreover, individual personalities remain an element of staility in the ontext of a young, fragmented and unstale, party system (for a similar argument see Kitshelt, 1995; Ştefuriu, 2003). TVR 1 TVR 2 ProTV Antena 1 PrimaTV AasăTV Others ,2% 5% 35,1% 17,4% 4,8% 4,1% 11,3% ,5% 3,1% 33,7% 22,5% 4,9% 6,9% 10,2% ,6% 3,7% 25,4% 21,5% 5,7% 8,4% 16,3% ,4% 4,3% 21,8% 19,9% 9,7% 9,1% 18,6% ,5% 3% 21% 17% 7% 8% Tale 1: Market shares for the main TV stations, uran areas (ities) aording to AGB Data Researh. Soure: Coman, 2003: 111 The aelerated development of the Romanian media system, espeially of roadasting, is proaly the most important failitating fator for the personalization of eletoral politis. Dominantly uninformed, politially or ideologially not aligned, and laking a onsolidated politial ulture, voters found an unexpeted simplifiation of politial life in the mass media, espeially in television programs, given y the projetion of the human element into politis. It is expeted that the personalization of politis found a atalyst in the inreased market share of private ommerial television stations, at the expense of the puli servie, espeially after 1998 (Coman, 2003: ). For a while this domination only materialized 11

12 in uran areas (see Tale 1) eause private television stations laked the tehnial means to over rural areas as well. In time this domination generalized, ommerial roadasting ompanies suh as MediaPro (Pro TV) and Intat (Antena 1) otained dominant positions not only with the generalist TV stations ut also with the speialized ones: sports, news, et. (Coman, 2003: 111). Commerial television programming inluded more and more shallow politial information, impregnated with sensationalist or entertaining elements, while popularizing politial haraters in detriment of politial ideas, issues and parties. A similar phenomenon is identified in the print media, where taloidization via reorientation on image in detriment of text and sensational in detriment of usual is the norm. Given the fat that taloidization is a top-down phenomenon motivated y ommerial onstraints (Hallin, 2000: 233) appliale to all mass media used y voters in extrating politial information, it is very unlikely that media did not exerised signifiant influene over segments of the eletoral ody, segments that finish up pereiving politis in terms of haraters rather than issues and symoli rather than sustantial terms. In the end of this disussion we should analyze the impat on vote deision of three other ategories of fators, frequently mentioned in the literature: the internationalization of politis, the inreased omplexity of politial issues and the transformations in the role of the state within the soiety. Approahing these fators as a whole is justified, in our opinion, y the fat that we elieve they exerised a low degree of influene, at least for the period of time analyzed in this paper. The internationalization of politis is on an asending trajetory in olletive pereptions during the post-ommunist transition, starting with sporadi ontats with international finanial institutions (IMF, World Bank) and ulminating with visile influene exerised y international organizations to whih Romania proposed and managed to adhere (NATO, EU). Even if, for most of the population, the plae and part of these institutions in Romania s demoratization remained unlear, it is expeted that the multipliation of ontats with these institutions rought into the spotlight a series of individual politial ators (presidents, prime ministers, foreign affairs ministers, finane ministers, negotiators, et.) in detriment of olletive ators (parliament, government). Aompanied y an inreased omplexity of politial issues, it is very likely that this phenomenon generated an inrease in the prestige and visiility of individual politial ators, materialized in a growing relevane for the formation of voting deisions. Thus, we expet that oth phenomena stimulated the personalization of voting, without onsidering their role as essential, sine they oth marginally affet most of the eletoral ody. 12

13 Institutional fators: - Diret eletion of the president. - Simultaneous presidential and legislative eletions. Historial and ultural fators: - Colletive memory dominated y images of authoritarian leaders. Stimulate - Anteedent sultanisti regime, followed y a diffiult transition. the personalization Strutural fators: of eletoral politis - Unstale and unonsolidated party system. - Lak of onsolidated eletoral alignments. + Contextual fators: - Domination of ommerial TV stations. - Taloidization. - Internationalization of politis. - Inreased omplexity of politial issues. Institutional fators: - Semipresidentialism with limited presidential prerogatives. - Eletoral formula ( ) PR, losed party lists, one round. - Separation of presidential and legislative eletions (starting 2008). Historial and ultural fators: Inhiit - Powerful rememrane of the personality ult of the sultanisti regime. the personalization Strutural fators: of eletoral politis. - Party system onsolidation. Contextual fators: - - Derease in the role of state in soiety during the post-ommunist period. Figure 1: Defining expetations regarding the personalization of voting deisions in Romania. The need for deentralization speifi to the transition from ommunism to demoray and free market naturally refleted itself in a step y step diminishing of the role played y the state in soiety. Suh a retreat of the state resulted in a natural derease of the resoures availale to the national top leadership, as well as in a lear delineation of the leverages for intervention in eonomy and soiety. Still, these transformations refleted very late at the level of soial pereptions of these offies and their holders. If we orrelate these hanges with party alternation in these offies, and also with the inreased ompetition etween national and loal leaders, it is very likely that a depersonalization in the pereption of puli offies eomes manifest, aompanied y a similar effet in terms of ompetition for these offies. Summarizing, we may say that Romania s ase generates unlear expetations aout the personalization of politis. Even though there is a signifiant numer of institutional mehanisms meant to inhiit this phenomenon, there is also a signifiant numer of 13

14 historial, ultural, strutural and institutional fators, as well as some institutional inonsistenies, that diminish the effets of the institutional mehanisms mentioned. Nevertheless, taking into aount oth the failitating and inhiiting fators (Figure 1), it is our expetation that the manifestation of this phenomenon is well aove the average Western European onsolidated demoray. 5. DATA SOURCES AND DIRECTIONS OF ANALYSIS The hypothesis of personalization is oviously a dynami one. The nature of suh a hypothesis demands a speifi strategy for analysis longitudinal, on an extensive period of time. The attempt to apply suh an analyti strategy to the phenomenon of personalization of voting in post-ommunist Romania is loked y the limited amount of data availale, espeially for the first deade after the fall of ommunism. Basially, the numer of independent studies dediated to voting ehaviour etween 1990 and 1999 is limited 8, most of the times atteries of questions referring to this phenomenon eing inluded in roader researh endeavours. In this ategory, in terms of data onsisteny, we mention the Puli Opinion Barometers (BOP), a series of surveys funded y the Open Soiety Foundation etween 1994 and , whih onstantly inluded a attery of questions foused on voting intentions and onfidene in puli personalities. The last eight years were marked y a more oherent presene of studies foused on voting ehaviour 10, ut results are still far from the plae and signifiane given to this phenomenon in other demorati ountries. In the attempt to analyze the personalization of eletoral politis from a longitudinal perspetive we used the attery of politial questions from the BOPs from 1995 to 2006, data olleted during 12 years and 4 eletoral yles. Given the harateristis of the data availale, we foused on indiators of ehavioural intentionality and not on indiators of 8 Among these we shoud mention: the May 1990 exit poll y IRSOP and INFAS Bonn (oordinated y Petre Datulesu and Klaus Liepelt); two Septermer 1992 IRSOP surveys done with three weeks, respetively four days, efore eletions; the Septemer 1992 exit poll y IRSOP and INFAS Bonn; the June-July 1993 survey y Siene Center Berlin and Informatix România as part of the Eletion Studies from Eastern Europe projet (oordinated y Georgeta Muntean); the Romanian Politial Culture survey from Novermer- Deemer 1996, whih inluded a attery of questions for the Comparative Study of Eletoral Systems (oordinated y Cătălin Baa, Gariel Bădesu and Paul Sum). 9 Data olletion y CURS, Metro Media Transilvania, Gallup România and LUAS. 10 Among these we should mention: the Politial Culture and Civi Engagement in Romania study, Novemer-Deemer 2004, inluding the Comparative Study of Eletoral Systems attery of questions (oordinated y Gariel Bădesu); a study y the Romanian Soiety for Politial Siene (SRSP) - Parties, Organizations, and Individual Ators: Behavior, Influene, and Politial Reruitment (oordinated y LaurenŃiu Ştefan, Aurelian Muntean, Răzvan Greu and Andrei GheorghiŃă); as well as two dediated Puli Opinion Barometers, from Otoer 2004 and Novemer

15 atual politial ehaviour. More exatly, we foused on survey data olleted regularly (every three or six months), ontaining at least the following variales: delared voting intentions for legislative and presidential eletions, delared onfidene in the politial leaders, soio-demographial variales. For the purposes of this analysis some variales were sujet to suessive reoding. First, the dependent variale voting intention in the legislative eletions - was transformed, for eah politial party inluded in the analysis, in a dummy variale, with two response ategories: voting for the party X and voting for any other party, with non-responses eing eliminated. Only voting intention for politial parties deemed relevant 11 was inluded in the analysis, taking into aount politial relevane (represented in the parliament during the time interval analyzed) and eletoral relevane (umulative, they over over 95 perent of the voting intentions). Seondly, the independent variale onfidene in politial leaders 12 was transformed in a onfidene in politial leaders sore as follows: a lot of onfidene answers were assoiated with 2 points, some onfidene with 1 points, little onfidene with -1 point, very little onfidene with -2 points. Non-responses were exluded, as well as I do not know him or her answers, whih we interpreted as an indiator of lak of politial visiility. Based on individual sores, we uilt an index of onfidene in a ertain politial leader, alulated as an average etween sores given to the respetive individual y the entire sample at a ertain moment in time. The index otained varies in the [-3; +2] 13 interval, with positive values refleting a favourale level of onfidene in the respetive politial leader and negative values refleting in an average situation of soial mistrust 14. From the longitudinal perspetive, our analysis is strutured on two levels. On the first level, dominantly desriptive, we look at the evolution of the onfidene in politial leaders indexes for over a deade, in onnetion with main politial events. The seond level is dediated to the analysis of 11 From the relevant polital parties we exluded the Demorat Union of Hungarians in Romania (Uniunea Demorată a Maghiarilor din România, UDMR), for reasons related to its exlusivelly ethni eletoral akground. 12 Measured in the Puli Opinion Barometers via a losed question, How muh trust do you have in the following politial personalities: (a) a lot; () some; () little; (d) very little; (e) I do not know him or her; (f) I don t know/not responding?. 13 The apparent disequilirium of the index s variation interval is due to a disequilirium in the struture of the ategories of answers provided to respondents in the POBs starting May There is a supplementary negative response ategory ompared to older POBs, namely not at all to whih no positive ategory orresponds. We assoiated -3 points to this ategory whih, in ase of 100 perent answers plaed in this ategory, would lead to the index eing equal with Traditional analyses of onfidene in politial personalities resoure, for presentation purposes, to summing up the positive (onfidene) and negative (no onfidene) ategories and omparing the totals as follows:: x% of the population displays high and very high onfidene in Z, while (100-x)%of the population displays low and very low onfidene in the same personality. The index used in this paper in the formula presented aove has the ovious advantage of introduing a qualitative riterium into the final evaluation. Moreover, y otaining single sores for eah andidate ar a ertain moment in time, the idenx allows for easier and nuaned omparison etween the onfidene sores of the politial leaders. 15

16 the intensity of assoiation etween the index of onfidene and voting intention for the politial parties represented y these leaders, as well as to variations in its intensity. It is ovious that, in order to support the hypothesis of personalization of voting for legislative eletions, we expet a signifiant and dynamially relevant level of assoiation. The longitudinal study of the personalization of eletoral politis is omplemented with a ross-setional perspetive on the phenomenon, foused on the Novemer-Deemer 2004 general eletions. We hose to fous on this eletion due to the extremely dynami harater of the eletoral rae, as well as due to availaility of data issues. We are referring to the Politial Culture and Civi Engagement in Romania (CPICR) 15 national two-wave (pre- and post-eletions) panel study. The ross-setional analysis tries to measure the raw and net effets of leader evaluations (independent variale, IV) on politial parties evaluations and voting intentions (dependent variales, DV), y introduing them into multiple regression models while ontrolling for alternative explanatory variales (soiodemographis, party identifiation, retrospetive evaluations). The proposed analysis is in line with a more general improved-predition strategy, eing uilt on a logi similar to the one proposed y Shmitt and Ohr (2000), and aims at explaining intentions to vote for the main ompetitors in the 2004 general eletions, namely the National Union PSD+PUR (Uniunea NaŃională PSD+PUR, PSD+PUR) 16 and the Justie and Truth PNL-PD Alliane (AlianŃa Dreptate şi Adevăr PNL-PD, ADA). We fous on two distint ategories of regression models: the first test the effets of leader evaluations 17 on the pereptions of the politial party, while the seond tests the impat of leader evaluations on the intention to vote for the politial party represented y the leader. In the first ategory of regression models the dependent variale (the evaluation of the politial party) takes the form of an evaluative 0-10 sore of the liking/disliking toward the respetive party 18 and we measure the raw impat of leader evaluations on party evaluation (R² of the model with one dependent and one independent variale) or net impat ( R² in multiple regression 15 Data olletion was finaned y joining several resoures omming from the CNCSIS A 382/2004 grant - Politial Culture and Civi Engagement in Romania (Gariel Bădesu), researh funds from Prineton University (Grigore Pop-Eleheş), Central European University (Marina Popesu) and OSI International Poliy Fellowship (Aurelian Muntean), and the CNCSIS BD 85/2003 researh sholarship (Andrei GheorghiŃă). 16 There are some differenes in CPICR questions when referring to the National Union PSD+PUR. The question regarding the pereption of politial parties referrs stritly to the PSD (with no referrene to the alliane with PUR), while the question regarding voting intentions referrs expliitly to PSD+PUR. 17 Measured in CPICR through a losed question: Please express your degree of liking for eah of the following politiians, on a sale from 0 to 4, where 0 means I do not like him at all and 4 is I like him a lot. 18 Measured in CPICR through a losed question: I would like to know what you think aout some of the politial parties in Romania. After I read the name of a party please give a mark etwen 0 and 10, where 0 means you do not like that party at all and 10 means that you like it a lot. 16

17 models). The net effet/impat is the inrease of explained variation we otain y introduing leader evaluation as an independent variale in the multiple regression model next to/together with other explanatory variales or, in other words, the inrease in R² otained y introduing IV (ompared to a similar model that only laks the leader evaluation IV). A similar strategy is ehind the seond ategory of regression models, where we try to measure the raw and net effets of leader evaluation on the intention to vote for a ertain politial party. In this ase, the dependent variales are otained y dummy-zation of the ategories of the losed question from the first CPICR wave 19. The hoie of supplementary explanatory variales ontrolled for in the regression models needs to e disussed separately. Naturally, there are two ategories of riteria that should guide suh a hoie: theoretial expetations (starting from alternative explanatory models of voting) on the one side, and availaility of data on the other side. A thorough taking into aount of oth riteria led to the hoie of three ategories of explanatory variales: (a) party identifiation variales (loseness to the politial party, ideologial distane, voting in the last loal eletions); () retrospetive evaluation of government ativity variales (evaluation of general government performane, evaluation of government performane in fighting orruption) and () soio-demographi variales (age, region, residene in uran or rural areas). A detailed explanation of the strategy for hoosing and uilding these explanatory variales an e found in Figure 2. Summarizing, the following setions fous on three diretions for analysis. First, we look at the evolution of the onfidene indexes in the politial leaders, for the period , in a dominantly desriptive manner. Seondly, we analyze the relationship etween the onfidene in the politial leaders and vote intentions for the party they represent, in an attempt to uild an image of the transfers of eletoral apital etween politial leaders and their parties. Thirdly, we try to measure the impat of leaders on the pereption of and vote intensions for politial parties. 19 Via uilding dihotomi variales with the ategories: vote for party X and vote for any other party. Nonresponses were eliminated. 17

18 Dimension Variales Values Constrution A. Party identifiation A1. Closeness to politial party Sore (0, 6) The CPICR questionnaire uses a filter question Please tell me whether there is a politial party you feel lose to?. This is followed y the Whih politial party? question, whih gives the respondent the possiility to name up to three politial parties. The loseness to a politial party sore is uilt as follows: 3 points if the politial party or alliane is the first option, 2 points if it is the seond option or 1 point if it is the last option. The loseness to a politial party sore varies from 0 (if the politial party or alliane is not mentioned at all) to 6 (if an alliane is mentioned as suh and two of the memer parties are also mentioned). For example, if a respondent lists first ADA, then PNL, and then PD, his loseness to ADA sore will e B. Retrospetiv e evaluation C. Soiodemographi harateristi s A2. Ideologial distane A3. Voting, last loal eletions B1. Pereption of government s general performan e B2. Pereption of government performan e in fighting orruption Distane (0, 10) Dihoto my 0, 1 Sore (1, 4) Sore (1, 4) C1. Age Years old. C2. Historial region Dihoto my 0, 1 C3. Residene Dihoto my 0, =6. The CPICR questionnaire asks respondents to plae themselves on a 0 to 10 left-right axis and to plae politial parties on the same axis. Ideologial distane is alulated as a differene (module, positive) etween the self-plaement sore and the sore orresponding to the ideologial plaement attriuted y the respondent to the respetive politial party. The smaller the ideologial distane (differene), the higher the proaility of voting for the respetive politial party. The answer to the losed question: To whih party elonged the andidate you voted for in the last loal eletions? was dihotomized as follows: 1=voting for party X and 0=voting for any other party. Measured y the losed question: Please think of the general government performane, how good or how ad do you think it was in the past 4 years?. The answer ategories were reoded as follows: 1=very ad, 2=ad, 3=good, 4=very good. Please note that this indiator measures the performane of the Adrian Năstase PSD government ( ), whih also inluded PUR ministers during the first three years in offie. Measured y the losed question: How do you think the urrent government performed in fighting orruption?. Answer ategories were reoded as follows: 1=very ad, 2=ad, 3=good, 4=very good. Please note that this indiator measures the performane of the Adrian Năstase PSD government ( ). Categories: Transilvania, Crişana-Maramureş, Banat, Oltenia, Muntenia, Dorogea, Moldova, Buureşti. For eah region we uilt dihotomial variales, 1 = the individual lives in the Y region and 0=the individual lives in any other region. Categories: large ity, over inhaitants; ity, inhaitants; town, inhaitants; very small town, under inhaitants; village, apital of a ommune (rural muniipality); village. For eah ategory we uilt dihotomial variales, 1= the individual lives in the Z area of residene and 0= the individual lives in any other area of residene. Figure 2: Explanatory variales ontrolled for in the regression models. Data soure: CPICR. 18

19 6. PARTY LEADERS PERCEPTIONS IN THE PUBLIC OPINION A longitudinal analysis of survey data should expliitly reflet the soial realities (and the hanges in soial realities) that shape the partiular moment of data gathering. Any signifiant historial event is expeted to ear some degree of influene upon the data. Thus, the first level of the analysis is dediated to a longitudinal perspetive on the evaluations of the main politial leaders in the Puli Opinion Barometer (POB) national surveys. An overview of the onfidene indexes in the main politial leaders alulated for the period (Appendix 1) suggests the ourrene of signifiant variations over time. Indexes vary etween a minimum of -1,81 (Valeriu Stoia, May 2001) and a maximum of +0,40 (Ion Iliesu, Marh 1995). Positive values are rare, only five politial leaders eing ale to sore positively over a period of twelve years: Ion Iliesu, Emil Constantinesu, Petre Roman, Teodor Meleşanu, and Traian Băsesu. For the eginning, it should e emphasized that, in the ase of three of them, positive sores are related to the eletion in/exerise of the presidential offie, assoiated with the highest eletoral legitimay. Furthermore, in the politial evolution of eah of the five politial leaders there is a unique positive peak over time. In other words, no matter how many or how large susequent variations might our, a su-zero desent from positive sores seems to e irreversile 20, none of the five leaders eing ale to sore positively over two different periods of time (separated in time y at least a negative value). Suh eginning remarks suggest the existene of a unique outurst of popularity/onfidene in the areer of a single high rank politiian, followed y a ontinuous erosive proess of that politial leader, whih ours faster or slower. The fat that Ion Iliesu has his last positive onfidene index in Otoer 1996 and is defeated in the presidential eletions only a few weeks later an not e ignored at this point. Four years later he wins the presidential eletions, although he does not sore positive in onfidene again. A hierarhy of politial leaders for the interval of analysis is dominated y Ion Iliesu, with ontinuous positive sores for two years ( ), during the seond half of his presidential term : +0,40 in Marh 1995, +0,21 in June 1995, +0,35 in Septemer 1995, +0,28 in Deemer 1995, +0,22 in July 1996, and +0,15 in Otoer Another leader soring remarkaly high on onfidene indexes is Teodor Meleşanu, 20 However, the evolution of onfidene in Traian Băsesu deserves a loser look in the future, in order to identify potential hanges in the pattern referred. For the remaining four leaders, a return to positive sores is highly unlikely, given their urrent middle rank politial ativity. 19