Images, Politicians, and Social Media: Methodological Challenges of Investigating Visual Political Communication on Social Media
Research on politicians’ visual communication strategies on social media has flourished in the past few years (Veneti et al. 2019). This is not surprising, considering that compared with written or spoken texts, images are not only easier to recall, but can convey more specific messages as well (Grabe and Bucy, 2009). Thus, as it is argued in the blog post of Marc Jungblut & Mario Haim, and Uta Russmann as well, visuals are an essential part of political communication both in general, and specifically on social media. On social media in particular, the role of visuals is even more prominent, and they are an important element in the strategic toolkit of political actors.
However, methodological challenges of studying visuals often lead researchers to consider them as illustrations of the textual main messages, and the visual content frequently remains understudied. In our recently published study, we aimed to fill four main research gaps in the current state of literature. We did this by creating a visual coding scheme that is suitable to 1) investigate images as objects of interest on their own; 2) explore political actors’ visual personalization strategies; 3) compare visual strategies on different social media platforms; 4) investigate the effects of different images on user engagement.
Personalization is a central focus of our research, and the notion is understood here as the process in which individual politicians, rather than their parties, move to the forefront of politics. Scholars identified two dimensions of personalization (see, Van Aelst et al., 2012) While individualization refers to the fact that individual politicians appear as central actors in the political arena, the privatization dimension of personalization is understood as the larger emphasis on politicians’ personal characteristics and everyday lives.
Our results show that visual political communication is strongly personalized both on Facebook and Instagram, but the level of personalization is higher on Instagram. Further, while the individualization dimension of personalization has a strong presence on both platforms, it is more dominant on Facebook, where images are used to provide insights into candidates’ political work. On Instagram images are rather used to privatization and to highlight the “human” sides of the politicians. Considering the findings on user engagement, it can be said that personalized images are more liked on both platforms, although formal elements are more favored on Facebook. Although based on the more personal nature of Instagram we expected that privatized images will be more likely to be liked there, interestingly, they are not more popular on the visual communication-based platform.
The methodological challenges
The main challenge in studying visual content was to create categories that help investigate as many aspects of the image as possible, but at the same time, ensure a portable and adaptable coding scheme for a quantitative content analysis. Moreover, although the depiction of a candidate in an image seems to be an obvious sing of personalization, we aimed to move beyond and identify more nuanced and informative visual features.
Thus, first of all, we conducted an inductive qualitative visual content analysis in order to identify the most meaningful and conceptually-fitting elements. Then we removed some of the aspects which were present in some images and seemed relevant from a theoretical aspect, but didn’t appear frequently enough. Since every small details of the images can carry visual messages, it was a challenge to leave out some aspects in order to ensure a rigorous coding scheme.
As a next step, the visual elements had to be connected to the literature that is mostly concerned with the verbal expression of personalization, both in the case of individualization and privatization. Even though our brain processes visual messages quickly, “reading” and identifying these messages was also challenging. Consequently, we grouped the visual features into formal and informal dimensions that we considered further on as visual expression of individualization and privatization. We argued that formal visual elements such as official context, campaign scenes or official clothing relate to the individualization dimension focusing on the political work of the particular candidates. On the other hand, informal visual elements such as selfies, popular cultural objects, private context or casual clothing are associated with the privatization dimension concentrating on the personal aspects of politicians.
In addition to the characteristics of personalization, important indicators of visual political communication, such as the type and the sentiment of the images, were also part of our investigation. Although these categories cannot be connected neither to individualization nor privatization, they tell us how the candidates intended to address the public. Thus, the general visual characteristics were used as control variables in the models.
To ensure that even the smallest details are counted, we decided to manually code the images. Hence, the next demanding task was the deductive quantitative visual content analysis, and training undergraduate coders. Compared to textual data, visuals may allow more space for subjective judgments. Consequently, in order to ensure reliability, each post was coded by three independent coders, and we accept only those codes that were recorded by at least two coders.
To sum up, there is no easy way to investigate visuals by quantitative content analysis. As we have shown, however, it is not impossible. Considering the prominent role of visual content in social media communication, it is worth tackling these challenges.