Paper prepared for the 7th ECPR General Conference, Bordeaux, September 5-7, 2013, Panel Variations of individual candidate campaigns

Transcription

1 Individual campaign strategies in a new democracy: The interplay between a single-member district proportional electoral system, party type and party popularity Mihail Chiru, Marina Popescu Department of Political Science Central European University chiru_mihail@ceu-budapest.edu Paper prepared for the 7th ECPR General Conference, Bordeaux, September 5-7, 2013, Panel Variations of individual candidate campaigns Abstract Taking as vantage point the separation of individualization and localization of candidate campaign strategies, this paper adds a further conceptual differentiation between role definitions, message content and campaign practices. Unlike previous scholarship, we argue that these strategies should be understood mainly through the interplay between inter- and intra-party competition and relevant candidate characteristics. To test this claim we use original data from the most recent Romanian parliamentary elections, which took place in a peculiar political context (huge unpopularity of the incumbent) and under an electoral system that incentives personalization and differentiation from fellow party candidates. Our multivariate models emphasize the importance for the three dimensions of campaign personalization of objective assessments of the likelihood to win a (direct or a redistributed) mandate, local candidacy as well as the parties organizational strength and level of popularity. 1

2 Introduction Contemporary political competition remains fundamentally party competition but the nature of parties and party-voter linkages have changed over time. There are significant decreases in party membership, a dwindling encapsulation capacity of intermediate partisan organizations (like churches and trade unions), as well as more distrust and disaffection from political parties and mainstream politics (Dalton and Wattenberg 2000). It is in this context that there is talk of a personalization of politics (Karvonen 2010; Adam and Maier 2010) which implies an increased relevance of party leaders in all matters of government (Poguntke and Webb 2007) as well as of leaders evaluations in vote choice (van Holsteyn and Andeweg 2010), and similarly of candidate traits in parliamentary elections (Glasgow and Alvarez 2000). At the same time, there is a discussion on the personalization of electoral systems that suggests that reforms have brought about systems with more candidate-centric features (Pilet and Renwick 2015) and that voters do make use of existing possibilities to differentiate between candidates and influence candidates chances to win a seat (Popescu and Toka 2012). Last but not least, there is evidence of different degrees of personalization of parliamentary election campaigns in the media across Europe (Karvonen 2010; Kriesi 2012). Therefore, it is expected that the role of individual candidates also increased and one way in which these roles are manifested refer to campaigning and communication foci. The subject was previously explored in the USA and more recently in a series of analyses about West European countries (Zittel and Gschwend 2008; Marsh 2009; De Winter 2009; Lutz and Selb 2010) and also East European ones (Chiru 2012) making use of candidate survey data, specifically of the Comparative Candidate Survey (CCS) 1 that was or will be conducted in some 30 consolidated and young democracies. In the same vein, we are interested in individual campaign strategies (i.e. choices of individualization and localization) seen to be determined by party competition factors and candidate characteristics and the puzzle is to try to disentangle them, which is however not an easy task. In this paper we try to make a few steps towards a clearer and more systematic understanding of the role definitions and campaigning methods used by individual candidates at the Romanian parliamentary elections in 2012 based primarily on the Romanian Candidate Study (Popescu and Chiru 2013). Our paper adds to this literature in three ways. First, more conceptual clarity in terms of personalization is necessary in order to be able to disentangle its various forms and their actual meaning and implications in terms of individual candidate roles in elections. We build upon the distinction between individualization and localization, recently examined by Karlsen and Skogerbø (2013) and add a separation between role definitions, message content and practical campaign practices. Second, one way to better gauge individual candidate roles is to understand them as delimited by the nature of inter- and intra-party competition and relevant candidate characteristics rather than just the latter and/or idiosyncratic campaign marketing. In doing so, we build upon the work of Zittel and Gschwend (2008). Third, the 2012 Romanian parliamentary elections present an interesting set of puzzles related to electoral systems and party systems that go beyond the pure case study interest of candidate campaigning in a new (post-communist) democracy. Romanian political parties are deeply distrusted, electoral volatility is high yet the same main parties have dominated the political arena; a new mixed proportional electoral system based on single member districts yet without intra-party candidate preferences was introduced in 2008 following an electoral reform debate in which personalization of political choices was a central issue. 1http:// 2

3 The paper will first define and discuss personalization individualization localization and present how we see their determinants. In the second section we develop specific hypotheses on the determinants of candidate campaign individualization and localization. Then, following the main dependent variables, we look at what is actually going on in the Romanian parliamentary elections through descriptive methods followed by multivariate models. Personalization individualization localization: some clarifications Personalization of politics is used frequently in a range of contexts and its meaning is often ambiguous. We refer to personalization very broadly (and loosely) to mean simply any focus on specific individual candidates. The relevant clarification is that we do not assume that this focus is necessarily automatically opposed to a party-centric campaign message or aimed to cultivate a personal vote in the sense of a separate reputation from the party (Carey and Shugart 1995). These are separate and important issues that we do not deal with in this paper. We see individualization and localization as distinct but related dimensions of (individual) candidate campaigning, each of them referring to candidates campaign communication efforts. Both may or may not include messages that distinguish the candidates from the party or just highlight party campaign issues or topics that are relevant for the candidate and the local context. Individualization of campaigning does not necessarily need to be a form of cultivating a personal vote but rather either a desire or a need to pursue a separate individual campaign in terms of means and/or message for instance given the lackluster or limited central party campaign. Individualization of campaign means is related to resources but is not necessarily equated with personal financial resources, but can also be related to resource allocations within the party. To what extent personal funds are used may have different relations with the position within the party depending on the strength of the party its centralization but more importantly to the usual structure of funds within the party, in turn shown to be related to party finance rules and norms. Individualization may be gauged at the level of role definitions how the candidates see their campaign needs. Depending on the electoral systems candidates may need to differentiate themselves not only from candidates of other parties but from candidates within their party, such as in the case of preferential voting on open/flexible lists. Therefore what candidates beliefs as to what their appropriate campaign focus means depends on the electoral system. Within an electoral system with single-member districts and no preference voting but county and national pooling of votes for the party, candidates may believe to have to focus on themselves more rather than the party yet try to campaign to get out the vote for the party as a whole. The way this desire is put in practice may take various forms - it may mean a separate campaign in terms of message or in terms of means of communication. Yet, these two forms of individualization may not necessarily (a) happen together, (b) be related to a vision or role definition but rather to a decentralized or even unfocused party campaign and (c) the individualized message content may differ both in terms of its link with the overall party message and in terms of candidate focus, from candidate personality features to policy achievements. Thus, one can distinguish three inter-related but not fully overlapping levels of individualization: belief/role definition, campaign message and campaign means. One way to personalize a campaign message may be through a focus on local issues, i.e. localization, which can encompass different messages related to the diverse concerns to be found within the district or found to be of appeal specifically to voters in the district. 3

4 Therefore, localization can be one form of individualization but it does not need to be. Localization or constituency focus can be for instance about the good deeds of a government or of government policies for the particular local constituency and be a part of the centralized party campaign strategy. It can refer also to the good deeds of the candidate or his specific relevant skills. The decision to focus on local issues or not may not belong to the candidate but can be part of the party strategy, therefore the candidate may not perceive it as part of a candidate centric campaign norm/ role definition, and even if it is related to candidates role definition it may not necessarily take the form of a separate campaign operation on the ground. Therefore, before exploring all the options about this mix of individualization and localization at the levels of role definition, message content and campaign organization, it is necessary to disentangle the determinants of the various components. HYPOTHESES Our hypotheses focus primarily on the determinants of candidate campaign individualization and localization that we posited above to be distinct but related dimensions of (individual) candidate campaigning, each of them referring to candidates view on campaign communication foci that may take different shape in practice depending on factors related to inter and intra party competition patterns and candidate characteristics. We also only look at individualization and localization in terms of stated focus and independent campaigning. Following this logic, both at the level of party competition patterns and candidate characteristics, there are a series of common determinants that render individualization and localization more likely and a number that are related only to one or the other, as well as that define the specific shape individualization and localization may take. Although these can be considered just further refinements and clarifications of the general model of individualization and localization of candidate campaigning, we believe that they are valuable in their own right as they tap into the puzzle of how actual candidate roles come about beyond the idiosyncrasies of campaign marketing. COMMON DETERMINANTS: Party/party competition factors Previous scholarship has shown that both objective (Eder et al 2010; Chiru 2012) and subjective (Zittel and Gschwend 2008) evaluations of the likelihood of winning the mandate influence heavily the candidates decisions regarding campaign personalization. A positive effect was documented: hopeful candidates engaged much more in individualized campaigning (De Winter 2012; Zittel and Gschwend 2008; Chiru 2012) than their colleagues who had no chance of winning. Beyond the rational personal incentives which are likely to make a difference between the levels of campaign individualization of the two types of candidates, an additional argument was advanced concerning the quality of the candidates. Thus, open-contest SMDs would attract more competitive candidates (Zittel and Gschwend 2008; Chiru 2012), whereas in districts where a party s candidate is doomed to fail one would mostly see residual candidacies, and little effort put into promoting anything else than the official party manifesto. Last but not least, Karlsen and Skogerbø (2013) have shown that hopeful candidates run both more independent and more localized campaigns. H.1: The higher the likelihood of winning the mandate the more will the candidate personalize her campaign. 4

5 At the same time, the degree of candidate selection centralization was shown to influence the politicians decision whether or not to build a personal reputation and to cultivate a personal vote (Carey and Shugart 1995; Hazan 2002; Shomer 2009). When the recruitment is centralized what matters for gate-keepers is the candidates loyalty to the party and to them as leaders, and their ability to promote and explain to voters the collectivistic policies the party advocates for.conversely, decentralized selectorates will tend to favour candidates that can campaign convincingly on local issues (Hennl and Zittel 2011), which might not appear evident or salient enough to national campaign coordinators. H2: The closer the nomination to the district level the more will the candidates personalize their campaigns. COMMON DETERMINANTS: Candidate characteristics Local politicians are more likely to have the knowledge (Shugart et al 2005: 441; Tavits 2009: 797) necessary to campaign on issues that are salient for the communities where they live and run. Moreover, because of their probable notoriety and access to local, nonpartisan social networks (Tavits 2009: 797), they are also in a better position to organize an independent campaign and have a personal-oriented communicative focus. H3: Candidates who live in the districts where they run are more likely to personalize more their campaigns. DETERMINANTS OF INDIVIDUALIZATION: PERCEPTIONS & INDIVIDUAL CAMPAIGNING Strong party organizations can provide their candidates with resources for campaigning and usually they also set up a national coordination system for planning and ensuring the consistency of campaign messages sent to the electorate (Denver et al. 2003). Moreover, organizational cultures such as that associated with the mass party might be less tolerant of individualized campaigning (Karlsen and Skogerbø 2013: 4). Candidates from strong party organizations will (a) be less likely to perceive as necessary (or acceptable from a party point of view) to emphasize their own features rather than the party and (b) to be encouraged to develop their own individual campaigns. Therefore: H.4: Weaker parties from an organizational point of view are more likely to be associated with a higher propensity of campaign individualization both in terms of communication foci and independent means of campaigning NOTE: Not relevant for localization since strong or weak party organizations alike may decide that it is in the best interest of the party to focus on constituency issues The way in which the candidate s nomination took place is another aspect that has the potential to influence both dimensions of individualization. How contested the nomination was, is likely to influence the quality of the candidature, and in turn more competitive candidates will inherently believe in a need to focus on themselves when campaigning. Similarly, a competitive nomination may also suggest an increased disposition to invest resources, including personal ones, and to go solo, i.e. run independently a campaign in order to maximise the chances of winning the seat. However, following a contested nomination a candidate may not just be willing but actually need to use his own means given a possibly less favourable allocation of resources within the party for a host of reasons, including the actual lack of support within some of the party branch factions. H.5: Candidates whose nomination was challenged will personalize more their campaigns. 5

6 SPECIFIC DETERMINANTS OF INDIVIDUAL COMMUNICATION FOCUS In the context of short term party competition factors, individualization may be part of a strategy to divert attention from the reasons of their party s unpopularity, i.e. some specific policy choice or overall government performance. Therefore, candidates from an unpopular incumbent party are likely to believe that individual campaign strategies are the better choice and to try any means of campaigning that can focus on potential individual positives. H. 6: Candidates of parties associated with negative government performance evaluations are more likely to believe in the need to individualize their campaign focus In the specific Romanian context this means ARD candidates are more likely to individualize their campaigns. Being closer to the party, ideologically or in an organizational way - more embedded in party structures, long term party soldiers have been shown to be associated with a lower inclination to personalize one s campaign strategy and communication focus (Zittel and Gschwend 2008). They are however unlikely to be systematically related to the actual capacity to develop an individual campaign organization. Thus, candidates who have been for longer periods in the party may or may not have better access to partisan resources and donations from party sponsors depending on their overall capacity to influence party decisions. Moreover, localization of campaigns by staunch party soldiers is likely to vary depending on contextual circumstances. H7. Longer party experience is likely to be associated with less individual focus in terms of campaign norm H8. Ideological proximity is likely to be associated with less individual focus in terms of campaign norm DETERMINANTS OF LOCALIZATION: Candidate characteristics Experience in local government or regional legislative bodies is known to encourage a particular conception of the representation role (Norris 1997: 33-34) i.e., defender of local interests, which can spill-over into the campaign strategy. Moreover, such experience should favour localized campaigning since these candidates are more acquainted with the region s specific problems and can advertise their already existing linkages with local authorities and voters (Freeman and Richardson 1996). Also, it is likely that they are more visible and have already developed strong ties with the local media. H.9: Longer local politics experienceis likely to be associated with more localization. Individualization and localization: Party choices at the 2012 parliamentary elections in Romania After five electoral cycles under closed list PR, Romania has switched since 2008 to a mixed proportional electoral system in which all candidates are elected in SMDs: some directly by obtaining an absolute majority of votes in the SMD and others through the allocation of party votes at county and national level. This particular electoral system reform was specifically promoted as a way to vote for people not parties in order to increase accountability and the quality of MPs. Although certainly there was no discussion on what 6

7 kind of personalization was meant to be promoted and how accountability could be achieved, the space for individual candidate campaigning has definitely increased given the change in the electoral system. This does not necessarily mean that candidate characteristics are likely to matter more in vote choices especially given political polarization. Indeed only 14.7% (18.5% if we exclude the missing cases) of the respondents of a post-electoral survey (RES 2012) said that who the candidate was mattered for their vote decision. But, it is exactly for the questions of this paper that these elections are an appropriate testing ground since one can expect both types of individualization, and localization, as well as variation depending on the party competition factors hypothesized. Specifically, Romanian party competition is polarized and party centric but it is not programmatic (Pop-Eleches 2008) and its local dimension is constructed more around the so-called local barons (presidents of county councils) that dominate the area through patronage and clientelistic networks (Gallagher 2009) than on specific local differences in citizens values and interests. Parties also run in alliances that may make candidates more relevant actors. Moreover, there are significant differences between the parties in terms of strength and centralization, with the PSD being the dominant party, fairly stable in terms of vote numbers even when it lost elections, with stronger party organizations as well as higher membership. To these long term differences is to be added the short term contextual party competition outlook, specifically the likely distribution of votes according to most opinion polls. First, the Social Liberal Union (USL) were expected to obtain a large share of vote and to win a high number of seats with a majority of votes in the SMDs. This increased the probability that few of their candidates would be allocated a seat in the redistribution stages, which created additional incentives for USL candidates to run a personal campaign to make sure they would not fall behind the 50% threshold. Second, the main opposition party, PDL, which had governed Romania for most of the legislative term (from December 2008 until April 2012) was hugely unpopular. This provided an incentive for their candidates to tone down party-centered messages as well as to differentiate from and do better than other candidates of the party with whom they were likely to compete for a seat in the redistribution stage but also to campaign through whatever means, including individualized campaigning and use of personal resources in order to ensure their own and the party success. In order to grasp what was going on in terms of campaign strategies chosen by the candidates of the main parties at the 2012 elections, we use the Romanian Candidate Study (Popescu and Chiru 2013) that includes a number of 407 cases from a total of 1802 candidates, i.e. all the candidates of the parties that passed the electoral threshold. Overall, the response rate was 22.6%, a figure in line with the completed samples of other elite surveys and well above the usual reported for Romania. Most responses were received via the online survey (393) and only a few (14) via self-administered paper questionnaires. Approximately one in five respondents was an incumbent (81) and 234 were unelected candidates (57.5% of all). Figure 1 below illustrates the high degree of representativeness of our sample for the entire universe of elected and unelected candidates with respect to their partisan affiliations. Thus, most of the points are extremely close to the line that indicates a very good match between the sample and the population. 7

8 Figure 1: Survey party representativeness DV Individualization measurements Individualization roles or the campaign norm. We measure the overall campaign role, i.e. what was called campaign norm (Zittel and Gschwend 2008) or communicative focus (Karlsen and Skogerbø 2013) through the standard question asking the candidates to rate the goal of their campaign on a 0-10 scale, where 0 meant attracting attention to their party and 10 attracting attention to themselves. Figure 2 below brings evidence in favour of the party unpopularity individualization strategy hypotheses, as the ARD candidates assumed on average the most personalized campaign focus. At, the other extreme, we find, as expected, the UDMR, a party with a consistent identity based programme that has candidatures in all constituencies (for electoral administration reasons primarily) even though their chances are confined to a relatively clear set of constituencies and do not expect to win votes outside the ethnic appeal. Figure 2: Campaign individualization - perceptions and actual practice 8

9 Individualization can take in practice a large range of shapes and forms. Independent campaigning does not necessarily imply a different campaign message than the party but surely has been associated with the set up of an independent campaign organization that designs and uses its own means and mechanisms of communication. We develop an index of campaign materials and methods that the respondents claimed to have used independently from their party based on seven items asking about the organization of social events and consulting hours, and the design of own placards, flyers, website, newspaper ads and TV or radio commercials. When examined with factor analysis, all the seven items loaded highly on one factor, which had an Eigen value of 2.922, while the next factor retained had a value of only.220. The resulting scale had 8 points, with 0 indicating that the candidate has not used independently any of these means and 1 that she has used all of them. The construct proved particularly reliable displaying an overall Cronbach Alpha coefficient of.828. Figure 2 above reveals that the candidates of the most unpopular alliance, ARD, also used the most independent means of campaigning. Another interesting aspect is the fact that there seems to be no difference in terms of independent campaigning between the UDMR and PPDD candidates, although the latter had declared more personalized campaign norms. Campaign spending is usually perceived as a valuable indicator of campaign intensity (Giebler and Wüst 2011; Lutz and Selb 2010). Whether spending is largely subsidized by the party, by private donors or it comes from personal funds is of key importance for understanding campaign individualization. We chose to look at the share of own campaign funds because this is a direct measure of independent campaign organization, while at the same time funding of individual campaigns by parties seems less relevant in Romania (see table 2 below). Figure 3 below shows that the candidates of the parties with weaker organizations had to provide, in average, for more than 60% of their campaign budgets. Figure 3: Campaign individualization: personal spending and independent means 9

10 Campaign Localization is derived from the question: During your campaign did you raise any issue specific to your district, not promoted by the party at the regional or national level?. The variable is a self-assessment and a dummy, which limits its capacity to capture the extent of variation in focus on local issues and is also open to interpretation in terms of what different candidates understand by constituency issues. Specifically this is a question about the UDMR whose candidates may not define problems of the Hungarian minority in terms of local schools or infrastructure as constituency issues, while those running in majority Romanian constituencies are unlikely to emphasize anything else than their ethnic minority stances. In future research we intend to include an objective measure of local campaigning agenda or at least to test the correlation between self-assessments and campaign messages. Figure 4 illustrates a clear difference between the Hungarian party and the rest of the competitors. Thus, only around one third of the UDMR candidates declared they promoted a constituency agenda whereas for the other parties the corresponding proportion is larger than three quarters. One has to note also the propensity of the PPDD candidates to campaign on local issues, especially since they were among the least likely to use independent campaign means. An explanation for this situation is provided when discussing the results of the multivariate analyses. Figure 4: Localized and independent campaigning LINKS BETWEEN THE 3 DIMENSIONS Table 1 indicates that the various facets of campaign personalization can also be empirically distinguished in the Romanian case. Campaign localization is mildly correlated with all the 3 indicators of individualization. Nevertheless, the limited nature of this overlap corroborates the theoretical intuition that these dimensions ought to be separated. The overall communicative focus and the means of campaigning are moderately positively correlated whereas, surprisingly, there seems to be no relation between the two and the 10

11 spending that the candidates do from their own pocket. Possible explanations are related primarily to the variance in resources that parties were able to allocate to their candidates. Table 1: Bivariate correlations between DVs Variables Campaign norm Ind. camp. Own funds Campaign norm 1 Independent campaigning.208*** 1 Own funds Constituency agenda.276***.231***.134** Variables Operationalization Party competition factors Likelihood to win In the Romanian mixed member proportional system all candidates compete in an SMD but only those that win 50%+1 of the votes win the mandate. The other candidates have to wait for the additional rounds of mandate redistribution based on the party vote at county and national level and the hierarchy of the absolute number of votes won by the party s candidates. For these reasons we need to disentangle the likelihood to win the two types of mandate. The direct mandate likelihood is a 0-1 scale derived from the number of extra votes the candidate would have needed to win directly the SMD (i.e. obtain the absolute majority). 1 implies the highest likelihood of winning the mandate, while 0 is assigned for the largest difference of votes, i.e. smallest chance. As already explained, there is a need to include separately a redistributed mandate likelihood variable.the variable is derived from the difference between the highest number of votes obtained by one of the party s candidates in the county and the respondent s share of votes. The difference was then rescaled and reverted so that 1 means the respondent had the highest score among her party candidates in the county and 0 that she registered the poorest score. For the UDMR respondents running in counties where they stood no chance of winning a redistributed mandate, the variable takes the value of the largest difference. Otherwise the variable would be useless for these candidates, e.g.: with 130 votes you are the best UDMR candidate in county such as Giurgiu, the difference would be 0 but nevertheless there is no mandate to be redistributed to UDMR. The party organizational strength variable assigns parties in three categories based on membership density, i.e. the ratio between their members and the population. The Social Democratic Party (PSD) with ratios of 2.1% and Democratic Alliance of Hungarians in Romania (UDMR) with 5.9% 2 were included in the largest organizational strength category. The medium category is comprised of PNL: 0.6% and PDL: 0.58%. All the other parties, whose membership is smaller than 0.2%, are part of the weak organizational strength category. Candidate characteristics The ideological distance scores were computed by using the absolute difference between the candidate self-placement on a 0-10 left-right scale and the party s perceived ideological 2For UDMR the ratio is computed relative to the size of the Hungarian community in Romania. 11

12 position on the same scale. The latter is represented by the mean of the party placement scores assigned by all the respondents of the respective party. Party socialization records the number of years spent by the candidate in her current party, whereas local politics experience registers years in offices such as mayor, local or county councillor and prefect. In addition to the variables for which we formulated hypotheses we also control for incumbency, type of residence and university education. The operationalization of the rest of the variables is straightforward and can be retrieved from the codebook in Appendix 1. In order to make easier the interpretation of the effects all the independent variables were standardized on a 0-1 scale. Table 2 below shows the descriptive statistics for both the dependent and independent variables. Table 2: Descriptive Statistics Variable Mean SD Min Max N Campaign Norm Constituency Agenda Independent Campaigning % Own Funds % Party Funds Direct mandate likelihood Redistributed mandate likelihood Nomination contest Level of nomination Party organizational strength Party socialization Local candidate Incumbent Local politics experience Ideological proximity Type of residence University educated Individualized campaigning: Candidates perceptions of a need to cultivate a personal vote as a chosen communicative focus The models predicting the campaign norm (table 3) have a moderate fit, explaining between 15 and 25% of the variance. Two of the three variables related to the organizational profile of the party make a difference in the first model. To begin with, the difference in the communicative focus between candidates from the parties with the highest membership density and those from the lowest party organizational strength category amounts to 1.3 points. As expected, candidates from stronger organizations individualize to a lesser extent their campaigns. Surprisingly enough, it appears that nomination at a higher level is associated with increased levels of campaign individualization. Thus, candidates nominated at national, central level place themselves 1.3 to 1.5 points higher on the scale of campaign personalization compared to those nominated at county level. Most probably, these candidates campaigns focused on personal qualities rather than on party policies because the candidates needed to compensate their lack of local notoriety. Being the only candidate or having one or more challengers for nomination does not seem to change the degree of personalization. 12

13 Table 3: Predicting Candidates Campaign norm (OLS models) 3 Variables Model 1 M 2 M 3 Party organizational strength *** (.512) ** (.562) (.859) Nomination contest.525 (.371).345 (.392).060 (.396) Level of nomination 1.593*** (.565) 1.511*** (.565) 1.346** (.552) Party socialization (.585) (.578) -.993* (.586) Ideological distance (.898) (.907).393 (.919) Local candidate.832*** (.312).738** (.330).641** (.316) Incumbent 1.324*** (.370) 1.008** (.391).430 (.405) Type of Residence 1.083** (.452).907** (.446).741* (.440) University educated.242 (.955) (.932).041 (.823) Direct mandate likelihood 1.622* (1.048).924 (1.272) Redistributed mandate likelihood.787 (.642).568 (.634) USL 1.164* (.577) ARD 2.339*** (.740) PPDD (1.037) Constant 3.660*** (1.046) 2.574** (1.080) 2.519* (1.348) R N *Significance at * p<.1; ** p<.05; *** p<.01 ** Robust standard errors in parentheses. A longer period of party socialization goes hand in hand with a party oriented campaign communicative focus, even if the effect reaches conventional levels of statistical significance only in the last model. It appears that candidates who live in the SMD were they run have a significantly more personalized campaign norm compared to the rest - the difference amounts to almost one point in the first model. Incumbency produces an effect of larger magnitude that goes in the same direction. The ideological distance vis-à-vis the party mainstream, and the level of educational attainment do not influence the overall campaigning strategy. Instead the type of residence seems consequential. Thus, candidates who live in large cities appear much more inclined to personalize their campaigns than politicians from rural areas. 3 All models were re-run with observations weighted by party. The results were not significantly different. 13

14 The second model indicates that politicians who were virtually certain of obtaining the absolute majority of votes in their SMD chose a more personal campaign focus compared with candidates which had no chance to win a direct mandate. Thus, the former are situated in average 1.6 points higher on the campaign individualization scale. Having a good chance to obtain a redistributed mandate seems also to foster an individual communicative focus, but the effect did not reach statistical significance. All the other effects behave virtually identical to the first model both in terms of their magnitude and direction. Last but not least, model 3 illustrates a number of interesting party effects. Thus, ARD candidates seem to have individualized the most their campaigns, fact which is not at all surprising given the huge unpopularity of the main party behind their electoral alliance. On the contrary, PPDD candidates are not significantly different from the UDMR politicians, the reference category, who were the least likely to individualize their campaigns given the virtual impossibility of obtaining a mandate in most SMDs. The decision of PPDD candidates to orient their campaigns towards their parties and not themselves can be explained if one takes into their sociological profiles. On the one hand, a large share of PPDD candidates was represented by people who entered politics only recently, having virtually no public notoriety. On the other hand, one third of their candidates (115 out of 446) were politicians that migrated from other parties, and almost half of them had had public offices under their older party label. As migration is largely regarded as dishonest by the general public, it is likely that these candidates also preferred to run party-centered campaigns. Individualized campaigning: the use of independent means of campaigning and personal resources The independent campaigning models presented in table 4 explain between 8 and 19% of the variance. The party organizational strength is a significant predictor in all the 4 models. For each of these models, candidates from the parties with the strongest organizational capacity appear to use significantly less independent means to campaign compared to their counterparts from the parties with the weakest membership density. Candidates who had to compete with more politicians for nomination are situated half a point higher on the independent campaigning scale compared to those that had an uncontested nomination. The effect loses statistical significance after controlling for the likelihood of winning a direct or redistributed mandate. The other variable related to nomination, which registers the level at which the decision was taken does not make a difference. The same is true for incumbency. Party socialization has a positive effect on the use of independent campaign methods in 3 of the 4 models. Similarly, candidates who live in the SMD where they run engaged significantly more in independent campaigning compared to the rest, the magnitude of the effect being very similar to the nomination contest variable. Politicians having a high chance of winning directly a mandate engage much more in independent campaigning than their counterparts who had no such chance. The magnitude of the effect varies between 1.5 and 3.3 points on the 8 points index. Similarly, the politicians who were favorites to win a party mandate in the redistribution stage also used more independent campaign means than the rest, although the effect is somewhat smaller, compared to that of the previous variable. Unexpectedly, ideological distance decreases engagement in independent campaigning in all models. Thus, almost perfect ideological congruence amounts on average to an increase of 1.5 points on the scale of independent campaigning. This might well be an artifact resulting 14

15 from the better capacity of senior, well-established politicians to assess the ideological position of their party and declare themselves close to it. Table 4: Independent campaigning (OLS models) Variables Model 1 M 2 M 3 M 4 Party organizational strength *** (.197) -.989** (.385) -.731* (.419) * (.703) Nomination contest.595* (.308).579* (.305).389 (.311).446 (.305) Level of nomination (.455).024 (.445) (.443).034 (.450) Party socialization 1.030** (.463).988** (.459).866* (.449).696 (.455) Ideological distance ** (.620) ** (.621) ** (.623) * (.611) Incumbent.490 (.334).482 (.332).134 (.343).023 (.336) Type of Residence.416 (.360).535 (.362).314 (.363).290 (.355) University educated.798 (.680).760 (.691).478 (.667).471 (.600) Local candidate.648** (.251).512** (.254).589** (.246) Direct mandate likelihood 1.463* (.769) 3.339*** (.927) Redistributed mandate likelihood 1.172** (.514) 1.109** (.465) USL *** (.465) ARD.140 (.535) PPDD * (.806) Constant 2.804*** (.799) 2.392*** (.821) (.841).903 (.964) R N *Significance at * p<.1; ** p<.05; *** p<.01 ** Robust standard errors in parentheses. Neither of the control variables appears to modify the engagement in this kind of activities. Party effects are interesting in themselves. It appears that both USL and PPDD candidates used less independent campaign means compared to the UDMR candidates that had a good chance to win the mandate. Table 5 below includes the models which predict the shares of personal contribution to the campaign budget. They explain between 13 and 19% of the variance. As expected, candidates from stronger party organizations could rely more on party funds and donations. Thus, the share of personal contribution of candidates from the parties with the strongest 15

16 organizational capacity was in average between 14% and 25% smaller than that of their counterparts from the parties with the weakest organizations. Virtual certainty with respect to direct mandate election seems to decrease the incentives of candidates to spend their own money in campaigns. Thus, for these candidates the share of the personal contribution to the campaign budget is 33 to 38% lower on average. Instead, candidates who hoped to receive a party mandate in the redistribution stage contributed much more to their campaign budget than the rest. This is, in fact, the only model in which the 2 variables related to the chance to win a seat produce effects that go in opposite directions. Table 5: Personal contribution to campaign budget (OLS models) Variables Model 1 M 2 M 3 M 4 Party organizational strength *** (8.132) *** (8.135) * (8.514) (13.729) Nomination contest (5.144) (5.119) (5.139) (5.283) Level of nomination (7.833) (7.724) (7.458) (7.490) Party socialization (9.103) (9.071) (8.851) (9.084) Ideological distance (12.262) (12.045) (11.883) (12.067) Incumbent (6.129) (6.154) (6.104) (6.563) Type of Residence (6.448) (6.476) (6.407) (6.468) University educated (15.266) (14.204) (14.830) (14.750) Local candidate 7.973* (4.677) (4.653) (4.797) Direct mandate likelihood *** (14.830) * (18.795) Redistributed mandate Likelihood *** (9.476) ** (10.017) USL (10.475) ARD (11.232) PPDD (16.312) Constant *** (17.632) *** (16.830) *** (20.095) *** (23.555) R N *Significance at * p<.1; ** p<.05; *** p<.01 ** Robust standard errors in parentheses. Almost all other independent and control variables do not influence significantly the share of personal contributions to campaign spending. The sole, partial exception concerns the 16

17 variable referring to local candidates. The latter appear to contribute with 6 up to 8% more to their campaign budgets. Localized campaigning Table 6 below illustrates that a number of effects in the constituency agenda models are quite consistent, irrespective of the changes in specification. To begin with, candidates who had the largest chances of winning the absolute majority of votes in their SMD are 4 to 5 times more likely to promote in their campaigns issues related to the constituency. Similarly, a high likelihood of winning a mandate through redistribution increases from 2.6 to 8 times the odds of campaigning on local issues. Table 6: Promoting constituency agenda (Logit models) Variables Model 1 M 2 M 3 M 4 Local candidate Local politics experience Direct mandate likelihood 4.652** 5.124** 5.103* Redistributed mandate Likelihood 8.015*** 4.134*** 3.270** 2.617* Incumbent Type of Residence University educated 5.242*** 6.978*** 7.376*** 8.201*** Party organizational strength.327***.256*** Nomination contest 2.347** 3.033*** 2.698*** Level of nomination Party socialization Ideological distance USL ARD 4.785** PPDD 6.165* Nagelkerke s R N *Significance at * p<.1; ** p<.05; *** p<.01 ** Cell entries are odds ratios Surprisingly enough, being a local candidate and having local politics experience do not significantly change the odds of talking about constituency related aspects in the campaign. Moreover, for local politics experience, even the direction of the effect is contradictory. A full move on the nomination contest scale: from no competition to having more countercandidates increases 2 to 3 times the likelihood of localized campaigning. On the contrary, the nomination level does not matter in any of the models. Party organizational strength is a significant predictor in 2 of the 3 models that include it. Thus, candidates from the parties with the highest membership density are 3 to 4 times less likely to run localized campaigns compared to their counterparts from the parties with the weakest organizational capacity. Incumbency, party socialization and ideological distance do not influence significantly the likelihood of promoting or not a constituency agenda. The same is true about the type of residence of the candidates. Instead, university educated politicians are 5 to 8 times more likely to run localized campaigns. 17

18 PPDD candidates seem the most eager to talk about local issues, followed by the ARD politicians. Discussion and conclusion Taking as vantage point the separation of individualization and localization of candidate campaign strategies, this paper adds a further conceptual differentiation between role definitions, message content and campaign practices. Our bivariate tests confirmed that only a partial overlap exists between these dimensions. This theoretical contribution is matched by an empirical one: testing for the first time the effect that intra and inter-party competition variables alongside with candidate attributes have on the three dimensions of personalization. Table 7 below revisits the hypotheses in light of the multivariate analyses results. A correct assessment of winning chances appears indeed extremely consequential for campaign roles, contents and practices irrespective of the type of mandate the candidate hopes for. None of this is true for the level of nomination whereas having to compete with other candidates for nomination influences only the level of independent campaigning and the degree of localized campaigning. Table 7: Hypotheses and the results of the multivariate regressions Variables Campaign Norm Hy./Find. Independent Campaigning Hy./Find. Personal Spending Hy./Find. Localized campaigning Hy./Find. Direct mandate +/ + +/ + +/ - +/ + likelihood Redistributed mandate + / n.s. +/ + +/ + +/ + likelihood Level of nomination -/+ -/n.s. -/n.s. -/n.s. Local candidate +/ + +/ + +/ + +/ n.s. Party organizational -/- -/- -/- n.h./- strength Nomination contest +/n.s. +/+ +/n.s. n.h./+ Party unpopularity +/+ n.h./n.s. n.h./n.s. n.h./+ Party socialization -/- n.h./+ n.h./n.s. n.h./n.s. Ideological distance -/n.s. n.h./+ n.h./n.s. n.h./n.s. Local politics experience +/n.s. *Legend: + positive effect; - negative effect; n.s. not significant; n.h. no effect hypothesized. The third hypothesis is corroborated only with respect to individualization dimension. Thus, local candidates have more personalized campaign roles, engage to a larger extent in independent campaigning and seem to spend more personal resources than the rest. Of the other three variables dealing with candidate characteristics only the degree of party socialization makes a difference, whereas the level of ideological distance and record of local politics experience do not seem to matter at all. We find convincing support in favor of the party organizational strength hypothesis, a variable which has been largely neglected by previous studies dealing with individualized candidate campaigning. Last but not least, party unpopularity increases the personalization of campaign roles as hypothesized. 18

19 All these findings emphasize the importance of searching for the sources of campaign personalization at the interplay between inter- and intra-party competition and relevant candidate characteristics. The main limitations of the study come from the static nature of the data (capturing one case at one moment in time) and the self-reported character of the proxies used to measure individualized and localized campaigning. Both problems are shared by most studies in the relevant literature, and they ought to be addressed by the second generation of campaign personalization research. 19

20 Appendix 1: Variable Codebook Variable Operationalization Campaign norm 1 = The goal of my campaign was to attract the most attention to my party 11= Attract the most attention to my candidacy. Constituency agenda 0 = no issues related to electoral district, 1 = the campaign promoted constituency specific issues Independent campaigning 0 = no independent campaigning 1= the candidate organized independently from the party social events and consulting hours, designed her own placards, flyers, website, newspaper ads and TV/ radio commercials. Direct mandate 0-1 reverted scale based on the number of extra votes needed to win likelihood the SMD (difference between absolute majority of votes and number of votes received) Redistributed mandate 0-1 reverted scale based on the difference between the number of votes likelihood of the respondent and the highest number of votes received by a candidate of her party. Incumbency 0=not, 1 = Member of Parliament in the legislature Ideological proximity 0 = perfect congruence between the candidate s and the party ideological position Nomination contest 0 = no other candidate for nomination,.5= one counter-candidate, 1= more counter-candidates. Level of nomination 0 = nominated at local level,.33= nominated at county level,.67=nominated at regional level, 1=nominated at national level. Party socialization Years of membership in current party rescaled to 0-1 scale Party organizational 0= The candidate is member of PPDD, PNTCD, FC or PC;.5= The strength candidate is member of PNL or PDL; 1= The candidate is member of PSD or UDMR. Local candidate 1= the candidate lives in the SMD where she runs; 0= Not Local politics experience 0= no experience in local politics... 1= 22 years in offices such as mayor, local or county councillor and prefect. Type of residence 0= rural area,.33= small or medium town,.67= suburb of big city, 1= big city University educated 0= high school; 1= university degree 20