The use of hashtags as a political branding strategy

Political brands are using Twitter intensively during campaigns. The objective of this study is twofold: to build a theoretical framework to pose that the use of hashtags can be a brand strategy; and to analyze how political brands are effectively using hashtags. First, drawing on the social identity approach, this article argues that brands are becoming social identities enacted through horizontal user-to-user relationships. Second, a sample of a million tweets from the last two Spanish elections in 2015 and 2016 is collected to analyze the type of hashtags posted by politicians and the networks of interaction. Results show that political brands mainly used three types of hashtags: party, candidate and branded hashtags; and mainly interacted with their own hashtags. This can be considered as a traditional brand identity strategy. However, a brand community strategy would be more aligned with the horizontal and top-down nature of social networking sites as well as with today’s allencompassing political debate. Political brands should foster a co-creative brand identity process to build more effectively the organization’s meaningful values.


INTRODUCTION
In the course of the past decade, Twitter has rapidly become one of the major digital campaigning tools used by politicians in many countries (Frame & Brachotte, 2015). Twitter offers a great capacity to target the different publics and the topics of their interest. In the digital arena, political parties are building their candidates as political brands (Guzmán, Paswan, & Steenburg, 2015;Winther Nielsen, 2015). Political brands can make an influence on public opinion by making decisions in two fronts of the social networking sites (SNS) arena: who to interact with and the topic of interaction. For instance, posting a hashtag with your campaign slogan will certainly encourage interaction with the candidate's followers (namely the like-minded), while making the decision to interact with some other user's hashtag will expand brand relationships. Similarly, posting a hashtag with a brand's candidate name is less interactive than posting a hashtag with a statement about a hot topic of campaign. Despite being closely related, the way brands act in these two fronts imply different branding strategies: the former will be building on the existing brand community by reinforcing the intended -illocutionary-brand identity while the latter will be expanding brand community by enacting brand identification (Kornum et al., 2017;Wallpach, Hemetsberger, & Espersen, 2017).
The prevalent concept of brand co-creation implies that the identity of the brand is determined by the dynamic interactions of users in multiple, overlapped brand communities (Brodie et al., 2013;Michel, 2017;Orth & Rose, 2017), thus the brand itself must engage in interactions with users and encourage interactions between users regarding multiple themes so that the brand identity is socially enacted while maintaining control over the intended brand identity (Habibi, Laroche, & Richard, 2014). This approach to branding is in line with Grunig's two-way symmetrical model for PR (Grunig & Dozier, 2002;Xifra, 2010). Although several recent researches advocate for a dialogical approach to social media from PR (Kent, 2013;Saffer, Taylor, & Yang, 2013;Seltzer & Zhang, 2011), not much theory from PR has been applied to political marketing cases (Strömbäck, Mitrook, & Kiousis, 2010). This study aims to fill this gap. This is particularly important in virtual brand communities (Habibi et al., 2014;Kornum et al., 2017;Wallpach et al., 2017), especially referring to ideological debates during political campaigns (Black & Veloutsou, 2017;Brodie, 2017).
The dialogical perspective clashes with current state of the art on the use of Twitter by politicians and citizens. Many studies have reported highly homophilic networks (Maruyama, Suthers, & Robertson, 2014;Guerrero-Solé, 2017;Guerrero-Solé, Corominas-Murtra & López-González, 2014) in which users rarely interact (retweet or mention) with the non-likeminded, but cluster around similar brand topics (Smith et al., 2014). The objective of this study is twofold: to build a theoretical framework to pose that the use of hashtags can be a brand strategy; and to analyze how political brands are effectively using hashtags-the type of hashtags posted by each political brand and the social networks that result from interacting with these or other hashtags. First, we argue that the use of hashtags is a branding strategy. Second, we define a classification of hashtags used by political brands.
Finally, we illustrate this approach with the analysis of two large samples of tweets from the two last Spanish campaign periods.

THEORETICAL FRAME: THE SOCIAL BRAND
The social identity approach (Tajfel & Turner, 1979) explains how groups can form rather arbitrarily and how the feeling of belongingness makes them social by making salient personal beliefs, attitudes and behaviors shared with other members (Hornsey, 2008;Schau & Muniz, 2002). Since the cognitive and self-categorization processes are considered as important mediators of the communication interplay, networks of interaction can be seen as legitimizers of the group existence, that is, interaction is a vehicle to make salient these shared norms (Hogg & Rei, 2006: 14). This approach has emerged as a suitable framework for marketing and branding studies in social media (Kenski, Filer, & Conway-Silva, 2017;Kornum et al., 2017;Orth & Rose, 2017), particularly to explain how brand communities take shape and evolve through identification processes with the group (Popp & Woratschek, 2017;Zhou et al., 2012).
In line with this theoretical framework, branding can be conceived as a relational process in which consumers allocate social values and meanings to brands in a dynamic interactive way (Brodie, 2017). The idea of brands as social entities matches the current definition of brand identity as a dynamic, fluid and plastic concept (Michel, 2017;Schau & Muniz, 2002;Wallpach et al., 2017) (Brodie et al., 2013).
No wonder the acceptance of this rather new conception of branding runs parallel to the establishment of social media as one of the main touchpoints between brands and consumers (Brodie et al., 2013;Gao & Feng, 2016;Habibi et al., 2014;Schau & Muniz, 2002;Yan, 2011;Wallpach et al., 2017). The interactive nature of Social Network Sites (SNS) can be used to accomplish several branding goals: build a sense of belonging to the brand, accept brand values, interact with the brand, associate the brand with other social actors, raise awareness and influence attitudes, and so on (Yan, 2011). During the last decade SNS have become one of the most popular tools to spread and share brand-related topics and build brand communities (Schau & Muniz, 2002). This perspective clearly challenges the applicability of the previously well-established conceptualization of brand identity and brand community based on a stable description of the brand`s attributes (Kapferer, 2012) to the SNS field.
In the SNS arena users engage in a great variety of interactions to perform a great variety of nested identities, thus the very concept of stakeholder as a homogenous entity falters (Wallpach et al., 2017). Brands are expected to evolve as these conglomerates of identities shape continuously into many of the brand`s communities. Thus brands` identities are cocreated by the members of the community when interacting, including the brand as a member (McAlexander, Schouten, & Koenig, 2002). As forcible as it may sound, networks create the value and identity of brands (Fyrberg & Jüriado, 2009). Therefore, brands are very interested in social interaction as a way to reinforce what the brand is and control what the brand is becoming every moment. Kornum et al. (2017) also distinguish between the intended brand identity and the enacted brand identity. The former refers to the brand communication with its internal and external stakeholders, while the latter refers to the contribution of brand communities to the brand identity (Michel, 2017;Wallpach et al., 2017). A distinction between brand identity and brand identification stems from this perspective: brand identity is how the brand sees itself and brand identification is how the brand is seen by the brand community. However, the compatibility between the a-priori brand identity defined by marketers and the identity enacted by the subsequent brand relationships does not seem straightforward (Brodie, 2017;Wallpach et al., 2017). Although the brand is naturally interested in matching those two approaches (Orth & Rose, 2017), the brand identity strategy is not fully compatible with a brand relationships strategy as each pursues a different goal in the brand community building process: reinforce an existing community or expand it.

HASHTAGS AS A WAY TO REINFORCE A STATIC BRAND IDENTITY OR EXPAND BRAND COMMUNITY
Not much research has focused on the use of hashtags by political parties as a branding strategy. There are two strategic dimensions when it comes to use hashtags in Twitter during campaigns: the content of the hashtags and the users of interaction in political discussions (Enli & Skogerbo, 2013). Regarding the former political brands can generate content to state the intended brand identity. On Twitter, such a strategy would encompass the definition of hashtags with some of the brand`s key attributes and core ideascandidate, political party, slogan (reinforce brand identity); or the definition of interactive public-debate-based hashtags in order to expand the brand community on these topics (Gao & Feng, 2016).
In line with the traditional concept of brand identity (Aaker, 1996), political hashtags tend to be conceived as a top-down vertical tool of communication to spread slogans of the campaign (Small, 2011) and to set the agenda of topics during the campaigns (Vargo et al., 2014). Small (2011) found that Canadian politicians use three types of hashtags: partisanrelated to their political party, political and non-political. First, partisan hashtags are those created by a political brand following the brand's strategy; second, political hashtags are those that elicit political non-partisan debate (cannot be aligned with certain brand's strategy); and third, non-political hashtags deal with different topics. Also, Kenski et al.
(2017) categorized tweets as candidate, party and ideological labels during the US 2016 presidential preprimary and primary campaigns. These authors found that candidates use self-advocacy strategies over social comparison as a way to foster their social identity.
In this paper we classify hashtags used by political brands during campaigns as party Besides the content of hashtags, brands can also make decisions on whom to interact with, which would lead to either participate in self-defined hashtags or in other brands' and users' hashtags. Previous research on political debate in Twitter has showed highly homophilic networks in which users only interact with the like-minded (Maruyama et al., 2014;Guerrero-Solé, 2017) and brand clusters networks in which users interact with the brand but not between themselves (Smith et al., 2014): RQ2: What is the degree of homophily of the networks resulting of political brands' interaction with own or others' hashtags during campaigns?
Both the definition of different types of hashtags and the interaction with own or others' hashtags are indicators of the strategy followed by the political brands (Smith et al., 2014).
Defining party and candidate hashtags and mainly participating in these (own) hashtags would add on a brand identity strategy while defining interbrand or public debate hashtags and mainly participating in other political brands' hashtags would be an indicator of a brand community expansion (Kenski et al., 2017). where politicians, media, journalists and users participate and interrelate on equal terms (Cohen & Ruths, 2013). For each tweet and retweet, we gathered its identifier, the username, the content of the message (including links, mentions and hashtags), and the user's self-description. We then created a register for each user that contained the number of messages posted by her/him, the number of times the messages were retweeted and the usernames of the users retweeting these messages. We then classified the top 5000 most retweeted users for both samples. In the #20D, 39% were classified as politicians, 8.4% as media outlets and journalists and 53% as citizens or others; whereas in the #26J, distribution was 44%, 48% and 8% respectively. Overall, data on the political brands' participation in hashtags #20D and #26J show that each brand published between 3000 and 7500 tweets, and over 31000 retweets, in each hashtag. The classification of hashtags was defined following an open coding process. The open coding is an iterative process in which the items of analysis are selected according to the focus of the research, namely political categories (Neuman, 2003). Firstly, the 20 most used hashtags for 20D and 26J were qualitatively coded as partisan or non-partisan hashtags (Small, 2011), this is, brand or non-brand hashtags. In doing so, each item was examined regarding its potential contribution to one of the four political brand's identity strategy). In case the hashtag could be attributed to any of the brands, it was coded as a brand hashtag. If no strategy could be identified, it was coded as non-brand hashtag. Secondly, both the partisan and non-partisan hashtags were broken down into subcategories through an iterative cumulative process: the different political  Branded content (BH): keywords, slogans, mottos, headlines or references to those or to any topic from the brand's campaign.

METHOD AND SAMPLE
And two types of non-brand hashtags on the other:  Interbrands (IB): any hashtag with nonpartisan or media content, this is, any topic that cannot specifically identified as belonging to a single political brand's agenda. Data on the use of these hashtags by the four political brands are provided as percentages of use in the two elections. A proportion of use of own and others` hashtags is also provided as well as two visual networks of the political brands` interactions.

RESULTS
The four political brands participated heavily in the two neutral hashtags. Figures on participation show that each brand published between 3000 and 7500 tweets in the 20D and 26J, and over 31000 retweets, in each one of the two elections.
In response to research question 1 (RQ1), the four political brands were consistently defining brand hashtags. As can be seen in to 87%, with the exception of Cs in the 20D that tagged half of its tweets with candidate hashtags (CN, 49%) and more than a third of tweets with party hashtags (PH, 37%). This is highly aligned with the strategy of this political brand, determined by the charisma of their leader, Albert Rivera.
Brands were also tagging conversations with party hashtags (PH), although to a lesser extent than branded hashtags (see Table 1).. In the 26J conversations, remarkably, all the four parties coincided in using branded hashtags, while the candidate hashtags (CN) were not considered by any of them. The party name hashtag (PH) was still employed by Podemos (34%) and Cs (41%). RQ2 inquired about the target of interactions in the networks, this is, the types of hashtags political brands were interacting with. In response to RQ2, political brands tended to participate in own hashtags, specifically brand hashtags (PH, CN and BH).
First, as can be seen in Table 2, there was a direct correlation between the definition of each type of brand hashtags and the use of it in conversations. The participation of brands in others' PH, CN and BH hashtags oscillated between 0,1% and 1%. In other words, more than 90% of all the tweets published with a brand hashtag of a given political brand were published by users of the community of the political brand. Second, the four political brands participated in very few neutral (non-partisan) hashtags during 20D (see Table 3). For the 26J, this unbalanced distribution was more skewed. Clearly, the PP was an exception during the 20D, as they mostly used interbrand hashtags.
These figures complement with the content of these hashtags. Firstly, for the 20D, the interbrand hashtags most used by parties were #Spain and #indecisos20D (undecided20D).
For the 26J, #Spain and #Vota (vote). We may note that in 26J #Spain was mostly used by PP to position its brand, while #Vota was mostly used by the right wing (PP and Cs). There is a clear tendency to convert interbrand and media debate hashtags into branded hashtags according to the strategy of each brand, which, coherently with the data reported, would hinder real interbrand conversations and the participation of all political brands in media debates. Therefore, the resulting retweet networks for 20D and 26J are characterized by homophily and polarization (Figure 1). Brand users cluster around the brand with no interaction with other brands whatsoever across the two campaigns. If anything, homophily and polarization increase in the second campaign.

DISCUSSION AND CONCLUSIONS
Results show that the four political brands mainly used party, candidate and branded hashtags, and that they mainly interacted with own defined hashtags, resulting in highly homophilic networks. In particular, each political brand tends to exclusively interact with conversations tagged with its own party name, candidate name or campaign messages hashtags, and ignore the other brands` party name, candidate name or campaign messages hashtags. Further, the four political brands did not engage in conversations tagged with interbrand hashtags or media debates hashtags. Thus the four political brands were using hashtags to exclusively transmit their brand identity with no horizontal interactive power and no user-to-user relationships beyond each brand-voter vertical interaction. These results are consistent through the two samples of analysis.
Only in one case we found a political brand that massively used what could be considered, at first, other party's brand hashtag: PSOE used #Rajoy (PP's candidate name). Clearly, PSOE`s brand identity strategy focused on the value of its candidate as opposed to the PP's candidate -certainly, the polls showed a low assessment of Rajoy and a high assessment of Sanchez (PSOE`s candidate). Conversely, PP did not use this hashtag (#Rajoy, their own candidate) in any of its tweets.
In sum, each brand`s network is highly homophilic indicating that each brand`s users were exclusively interacting with the brand vertically. Hashtags defined with own party or candidate names are not very interactive, thus no participation of other brands in these hashtags and no horizontal interaction between each brand`s users were expected.
Although the use of a flexible social brand identity strategy seems the most suitable for the SNS brand-users communication (Gao & Feng 2016), brands are still applying a traditional (static, controlled) brand identity strategy to political marketing campaigns in the SNS.
Arguably, political branding strategies are under the umbrella of pure political strategies, which nowadays mainly focuses on own voters. Only PP in the 26J was trying to expand their brand community by using media debate hashtags and building on brand associations with the interbrand hashtags.
The dominating political strategies have been illustrated by the brand clusters networks (Smith et al., 2014). These networks are characterized by users tweeting about brand topics with no interaction with their peers. The lack of user-to-user hinders the creation of brand communities (Zhou et al., 2012). Brands prioritized a one-way or two-way (but not interactive) top-down brand-user communication by defining non interactive hashtags and engaging in conversations tagged with own hashtags exclusively. Political brands made salient their respective brand`s attributes -the party and candidate names and the main slogans or messages from the campaign, to be recognizable to the brand supporters. ISSN: 2174-3681 Consequently, the resulting political networks in Twitter are branded networks supported by branded hashtags. The four political brands were using the same strategy when defining hashtags and the same strategy when engaging in conversations, thus brands failed at showing points of differentiation (Karlsson, Clerwall, & Buskqvist, 2013).
The classification of hashtags proposed here could be used by practitioners to brand-tag politically neutral networks strategically (Bode & Dalrymple, 2016). Specifically, hashtags can be used to design a social strategy or to mold the strategy as the political race progresses in election campaigns. The five types of hashtags defined here are relevant grounds to analyze networks of interaction as part of either the traditional brand identity approach or the social brand identity approach. This division between brand and non-brand hashtags mirrors Small's classification between partisan and non-partisan hashtags (2011).
In this regard, the Spanish case study is very informative. For the first time in the Spanish history, four political brands participated in the elections with chances to govern and the need to agree on political issues (Simón, 2017). Furthermore, Podemos and Cs parties define their identity as young modern political brands which leaders and supporters are very active in social media, not scared of debating about any issue of the public scene (Graham & Johnson Avery, 2013). Also, the second campaign required agreement between the four political brands, and yet no public debate ever happened in the form of brand relationships.
In this regard, in the 20D campaign, PP and Podemos based their strategy in strong campaign ideas (branded hashtags), whereas PSOE and Cs based their strategy on their candidates.
Half a year after the first elections, during the 26J campaign, the candidate became irrelevant for the four brands, and efforts were put on the party and the campaigning hashtags exclusively. In particular, almost all the hashtags used by the ruling party PP were branded, while the use of the name of the candidate and the party name was scarce.
Considering that the branded hashtags were stating core ideas and political measures from the campaign, how could the four political brands not interact with each other's branded hashtags? How could the four political brands not interact within the interbrand and media debates hashtags? One day the four candidates were engaged in an open interaction about these issues in the prime time live TV debates, next day their account in Twitter were ignoring these topics, being the later medium more participatory and engaging in political topics than the former.
This study is not without limitations. Firstly, the analysis of hashtags and the subsequent network analysis as a way to measure the implementation of branding strategies in Twitter is a theoretical contribution in itself, no effects or influence on beliefs or attitudes have been tested, hence all considerations with regard to branding strategies are exploratory and insightful. Secondly, regarding the network analysis, the use of neutral hashtags can be said to introduce a bias in the sample of analysis. However, in line with the objective of this research, these hashtags were defining ideologically unbiased conversations (Cohen & Ruths, 2013). And thirdly, this study focuses on Twitter as one of the main political marketing and PR tool in the political arena with no reference to other social media tools such as Facebook. This is a limitation for two reasons. On the one hand, Facebook can be a suitable tool for citizens' participation in local government (Haro-de-Rosario et al., 2018). On the other, the digital architectures of digital media may limit the validity of each social tool's data (Bossetta, 2018), especially taking into account that the penetration of Twitter is much smaller than Facebook. , D. A. (1996). Building strong brands.