Exploring the Social Media Impact of Voting Advice Apps: A Case Study on the Representation of the Wahl-O-Mat on Twitter

1 Introduction

A central aspect of democracies is that every citizen has the right to vote, however, not every eligible voter actually participates in the elections. In Germany, voter participation decreased from 91.1% in 1972 to 72.4% in 2013 [1]. In consequence, the elected parliament does not represent the entire society but only a subpart, i.e., those who participated in the elections. Especially the group of young adolescents shows a relatively low political participation. For instance, during the last elections to the German Bundestag in 2013, only 64.2% of people younger than 21 years voted which is more than 8 percentage points below the average voter turnout [1]. Given that empirical evidence suggests that young people who feel knowledgeable have a higher voting intention [2], an obvious question is how to support the political information acquisition process. One increasingly popular way is represented by web-based Voting Advice Applications. The next section describes in greater detail the effects of Voting Advice Applications on political information and opinion formation processes. Subsequently, we outline the method and results of the current research and discuss practical implications as well as design recommendations for future product improvement.

2 Theoretical Background: Voting Advice Applications

Voting Advice Applications (VAAs) are web-based tools designed to help voters to find a political party that matches their political views. The procedure is simple: users fill in a questionnaire on a wide range of political issues. The answers are then compared to the programmatic stands of several political parties (in the current version users can select up to eight parties for comparison to be listed in the results list). The result is a rank-ordered list of political parties that are the closest to the user’s preferences regarding the surveyed political issues. The major benefit of VAAs is to provide condensed information in a vivid, interactive and easy-to-handle manner. In the past decades, VAAs have been developed for almost every European country [3]. The corresponding VAA that is used in Germany is called ‘Wahl-O-Mat’. It was developed by the Bundeszentrale für politische Bildung and first used for the elections to the Bundestag in 2002 [4].

Users of the Wahl-O-Mat are mainly young, with over 40% being younger than 30 years of age. Surveys showed that 50% to 60% of Wahl-O-Mat users declared to be motivated to search for further political information [4], [5]. This is, after all, good news, as we know that people who feel well informed report to have a higher voting intention [2]. And in fact, VAAs have already been shown to enhance the voting intention of young people [6], [7]. Therefore, VAAs can be considered as a suitable tool for increasing the voter participation of young and first-time voters.

3 Study Focus: The Interplay of VAAs and Social Media

Given that nowadays social media such as Twitter play an increasingly important role for political discussion and potentially opinion formation [8], [9], an interesting question is which role VAAs play within this, and to what degree VAAs serve as a trigger for further discussion in social media. Hence, the present research explores the representation of Voting Advice Apps in social media with a special regard to user intentions, emotions and evaluations of the VAA results as well as the product itself.

4 Method

We used an exploratory research approach to look at the opinions and statements related to VAAs in social media. In this study, we focused (a) on the German version of VAAs, the Wahl-O-Mat, (b) on the social network Twitter, and (c) on the federal state election in North Rhine Westphalia, Germany, in 2017. The observed period was from the launch of the Wahl-O-Mat on the 24^th of April 2017 to the Election Day on the 14^th of May 2017.

First, we extracted all tweets with hashtag or keyword Wahl-O-Mat and Wahlomat from Twitter using the Data Miner extension for Google Chrome. The data set consisted of n=2019 tweets. We applied a quality check to exclude tweets that were not understandable or unrelated to the Wahl-O-Mat. Second, based on a random selection of 300 tweets, we developed a categorization system (see table 1) along several dimensions by applying an inductive research approach. Additionally, we utilized existing theoretical foundations or frameworks in a deductive manner where suited (e.g., the theory of basic emotion [10]). The development of the categorization system was oriented on existing procedures of qualitative content analysis [11].

Then, we coded n=500 randomly selected tweets according to the categorization system. The following tweets were translated by the authors to give an overview of the coding procedure (see table 2).

Tweet 1: ‘I truly love the #wahlomat. Everybody should consult it before elections.’

Tweet 2: ‘The #wahlomat makes me angry.’

In order to obtain more insights into our data, we created a composite called “Social Impact”, which clustered the sum of retweets and likes that each Tweet obtained. The final number was then categorized as Low (0), Medium (1) and High (>1) and was matched with the content categories. In the following section, we discuss the main descriptive and correlational results across the different dimensions of analysis.

Figure 1

Expressed emotions within Wahl-O-Mat tweets.

5 Results and Discussion

How much communication on Twitter is initiated by the Wahl-O-Mat?

The majority of tweets about the Wahl-O-Mat were posted within three days after the launch (first day: 989 tweets, second day: 223 tweets, third day: 105 tweets) and on the election day (161 tweets). On all the others days of the observed 21 days period, the number of tweets ranged from only 13 to 57. This shows that only very salient events like the launch of the Wahl-O-Mat or the Election Day triggered political discussion related to the VAA. However, potential adaptations of the marketing campaign as discussed in the implications section might help to create more impact of the Wahl-O-Mat on Twitter.

Is the Wahl-O-Mat rather positively or negatively associated?

In sum, 55% of the coded tweets expressed an overall appraisal. Among those tweets, 62% contained a positive judgement, whereas 38% expressed a negative judgement about the application.

Which emotions are expressed in tweets about the Wahl-O-Mat?

Overall, 35% of tweets expressed an emotion in line with the theory of basic emotion [10]. Among those tweets that implied a basic emotion, the most frequent emotions were surprise (37%), joy (33%) and contempt (21%). Anger (7%), sadness (2%), fear (1%) and disgust (0%) were not or only barely shown. In order to assess the social impact of these tweets, we looked at the sum of retweets and likes for each tweet. As described in the methods section, this resulted in three categories of social impact: Low (0), Medium (1) and High (>1). We then analyzed potential differences in the social impact of tweets for all basic emotions that had a frequency of at least 5% (see figure 1). Especially for surprise, we found a high ratio of high impact tweets. Compared to this, other emotions such as contempt or anger seemed not connected to particular social impact.

Which intentions do tweets about the Wahl-O-Mat express?

Clear intentions were revealed by 81% of all tweets. Interestingly, the most frequent intentions were to show one’s own result (i.e., self-portrayal, mostly through a screenshot of the Wahl-O-Mat result, 43%) or to recommend the Wahl-O-Mat to others (37%). Compared to this, tweets that criticized the Wahl-O-Mat as a product (10%), made jokes about it (5%) or criticized (3%) or emphasized a political party (2%) were rare. We then obtained the social impact elicited by each intention for all categories with an overall frequency of at least 5%. Surprisingly, most of the tweets that recommended the Wahl-O-Mat had almost no social impact (see figure 2), contrary to tweets that displayed the users own result, which elicited the most retweets and likes. These findings enable to detect potential for product improvements which are discussed in the implications section of this paper.

Figure 2

Intention of tweets among social impact categories.

How was the result of the Wahl-O-Mat evaluated on Twitter?

Overall, 28% of all tweets expressed an evaluation of the users’ result, commenting on their matches with the different political parties provided by the Wahl-O-Mat. The most represented categories within the expressed opinions about the results were agreement in the first place and confusion in the second (see table 3 for frequencies). This is in line with previous research that showed that surprise emerges if the preferred party is not displayed as the best match. If this irritation is perceived intensely, this may lead to a change of one’s own voting intention against the originally preferred party [12].

How was the Wahl-O-Mat as a product evaluated on Twitter?

In 36% of tweets, an explicit evaluation of the product took place which formed five broad categories (see table 4 for frequencies). Among those tweets, more than half rated the product as informative, compared to only 11% as useful and 6% as fun.

Does the representation of political parties on Twitter predict the election result?

In sum, 23% of the analyzed tweets mentioned a political party. Among those tweets that mentioned a political party, the most often mentioned parties were Die Partei (22%) and AfD (19%, see figure 3). Both parties are considered as controversial parties: The AfD is a right-wing populist and Eurosceptic political party. Die Partei is a satirical party. However, the election results of these two parties were lower than their representation on Twitter (AfD: 7%; Die Partei: <1%). On the other hand, the parties that achieved the highest election results (CDU: 33%; SPD: 31%) were less frequently mentioned on Twitter (CDU: 12%, SPD: 5%). The overall correlation between representation on Twitter and the actual election result was r=−0.38. This result indicates that frequent mentioning of a party on Twitter was associated with a lower election result. When interpreting this finding, it has to be taken into account that the voters of both, CDU and SPD, might be underrepresented on Twitter due to their higher age-range compared to other parties.

Figure 3

Representation of different political parties within tweets and election results.

6 Implications and Design Recommendations

Based on the results of our study several practical and design implications emerge.

How could the social media impact of the Wahl-O-Mat be increased?

It becomes clear that the extent to which the Wahl-O-Mat stimulated communication in social media is particularly low after a few days after the launch of the VAA. In order to enhance its popularity and usage, so called social influencers could help to activate its full potential. Social influencers are social media users who established credibility and have access to a large number of followers. Hence, social influencers have a high potential to persuade others to use the Wahl-O-Mat, too. From a psychological perspective, it is important that those social influencers are perceived as peers by the target group on Twitter and not as politicians or organizations.

How could the user experience of the Wahl-O-Mat be improved?

Only 6% of the analyzed tweets expressed that the usage of the Wahl-O-Mat was fun. In the light of the VAAs goal of fitting the user’s needs, it might be helpful to include more interactive and enjoyable features to the app, i.e., drawing on elements from gamification or making explicit use of the interactive potential of touch displays, for users who use the Wahl-O-Mat via smartphone. Single approaches in the field of VAAs already followed this idea. The new app WahlSwiper [13] for instance applied a technique comparable to Tinder: Swiping right for ‘yes’ and swiping left for ‘no’ makes the usage more fun and interactive.

With regard to the expressed emotions of the users, we found that 32% of the users were confused about their results and the emotion of surprise was more frequent in tweets of high (vs. low) social impact. Developers of the VAA or researchers should focus on further exploring the reasons for confusion and surprise to facilitate a voting decision.

For example, a basic improvement could be provided by not limiting the number of parties for comparison, but also allow users to see their matches with all parties. In the present form of the Wahl-O-Mat it might happen that the user does actually not learn which party best matches his or her view, just because he or she did not pick this party in the pre-selection. In a way, this is as if a dating app would ask you to pre-select ten potential partners before you even know which of the many makes the best match with you. In this respect, one could also critically question the currently realized default selection of the established bigger parties.

Another improvement could be provided by simplified links to additional information on particular parties. If, for example, the result of the VAA is not in line with the initial opinion of the user, one might wish to compare the initially preferred and the best matching party two parties directly to make a validated voting decision. Here, the Wahl-O-Mat could offer further information materials or links to additional tools. For example, the WahlSwiper provides neutral explanations for all political issues in short YouTube videos [13].

Which additional features should be integrated into the Wahl-O-Mat?

A further main finding of our study is that tweets which recommended the app as a product did not gain high social impact, whereas tweets with self-portrayal content (e.g., posting one’s own VAA result) elicited more engagement by other Twitter users. Hence, it is reasonable to include a button in the application that enables to share one’s own VAA result directly on social media to increase the social impact.

7 Limitations and Future Research

The present study provides profound insights into the critical elements of interactive VAAs such as the Wahl-O-Mat on social media. It enables the detection of potential product improvements which can be helpful to gain more social impact on social media like Twitter. It has to be noted though that our study comes with several limitations. First, we had no insight into actual voting behavior and did not assess to what degree the Wahl-O-Mat affected the users political attitude. Second, the study solely followed an exploratory approach. To test the emerging theories and hypotheses, more experimental research needs to be conducted. Third, it is important to mention that no representation of an entire population took place as the sample consisted of only 500 tweets and was limited to one social network and one election. Also, the demographics of Twitter users are unknown (e.g. age, political attitude, socio-economic status), which would have been another interesting issue for correlational analysis. Moreover, one has to take into account that the clarity of tweets might be restricted due to the lack of insights into voting behavior and concrete affections. There was no possibility to verify interpretations as no direct communication with users was conducted. Further research could fill this gap by using follow-up questionnaires for the users to clarify on their emotions, demographics as well as on their final voting decision. Besides, it is a critical point that emotions and judgements about Wahl-O-Mat might depend on the context of the election. For example, elections to the Bundestag might be perceived as more important and so may lead to stronger and different emotions regarding political topics or the usage of the Wahl-O-Mat. Therefore, further studies should make an investigation of the use and judgement of the Wahl-O-Mat on different social media platforms and elections to improve the accuracy and generalizability of the results.

In general, the present findings highlight the importance of integrating psychological research in the process of VAA development. We hope that the present insights will inspire further research on the issue and help to improve the design of VAA systems, to actually have the positive intended effect on democracy, in the sense of facilitating political information and opinion formation processes, and eventually increase voter participation.