Actorhood of the European Court of Auditors: a visual analysis

Purpose – This paper explores how public audit institutions establish themselves as distinct actors on the public stagethrough communication practices. Byfocussingon the journeyof the European Court of Auditors (ECA), this paper addresses the following research question: how does a transnational audit institution construct its actorhood through visual communication practices? Design/methodology/approach – Using the theoretical frameworkof actorhoodtheoryand inspired by the visual accounting methodology, this study explores the ECA actorhood journey through the visual analysis of front pages of its official journal (ECA Journal) from its inception in 2009 up to 2019. The visual analysis is conducted through content analysis and a two-step cluster analysis. Findings – By showing how combinations of different visual artefacts have evolved over time, this study highlights the ways transnational public audit institutions, such as the ECA, construct their actorhood and position themselves on the public stage. It further reveals the underlying legitimacy mechanisms through which organisations such as the ECA position themselves in the public eye. Originality/value – Thisstudyshedslightonthedepictionofindividualsandtheircontextsininteractionwith each other and how this interaction reveals the development of the actorhood journey of the ECA over time.


Introduction
The terms "audit explosion" and "audit society", first used by Michael Power (1994, 1997, have become abbreviations for a set of beliefs about the ambiguity of an audit regime that has had a perverse impact on public sector culture and its image in the public eye. The increasing demand for public sector auditing is associated with New Public Management (NPM) and the inherent Given gaps in the literature on the visual analysis of communication practices of public sector audit institutions and the limited understanding of transnational audit institutions, this study explores the ECA's journey of establishing itself as an actor on the public stage. Specifically, this study asks the following research question: how does a transnational audit institution construct its actorhood through visual communication practices? To address this question, this explorative study borrows from the field of organisation studies and is inspired by Michael Power's (2017) editorial "Accounting and Actorhood" in the journal Accounting Organizations and Society in which he urges the application of actorhood theory in the accounting and auditing context.
Relying on the visual accounting methodology (e.g. Baldvinsdottir et al., 2009;Duff, 2011;Kyriacou, 2016;Parker and Warren, 2017;Warren and Parker, 2009), this study explores the ECA actorhood journey through the visual analysis of the front pages of its official journal from its inception in 2009 up to 2019. The visual analysis is conducted by means of content analysis and a two-step cluster analysis. The ECA is particularly relevant for performing such an analysis given both its importance for the functioning of EU institutions and, by extension, the well-being of its citizens and the increasing visibility it has gained through various communication efforts (Tid a, 2021). Our analysis reveals patterns in this "journey" showing, for example, how the ECA communicates the importance of social capital in performing its work, its modernisation through the intertwining of personal, social, political and technological visual attributes, and its move toward de-personification. In highlighting these and other findings, this paper thus contributes to the public sector auditing literature in showing how transnational audit institutions, such as the ECA, construct actorhood over time and position themselves on the public stage.
The remainder of this paper is organised as follows. First, it presents a brief review of the academic literature on the evolution of public sector auditing. This is followed by a description of the theoretical framework for the visualisation of the actorhood of public sector audit institutions. The empirical research context of the ECA is then presented, followed by an explanation of the methods used to conduct the visual analysis. The findings are then presented along with an analysis of the results. The paper concludes with an overarching discussion, which includes limitations of the study and suggestions for future research.
The national public sector audit institutions also known as SAIs play a prominent role in public sector auditing because they are responsible for compliance and performance audits of governments and various public sector agencies (Stapenhurst and Titsworth, 2001;Bonollo, 2019;Cordery and Hay, 2021a). A comparison of various SAIs shows that their structures differ, because their approaches and the type of audits they carryout are not homogeneous (Cordery and Hay, 2021b). At the same time, in recent years the use of audit standards and the framework published by the International Organization of Supreme Audit Institutions (INTOSAI) has grown (Hay and Cordery, 2021).
Most of the studies in reference to SAIs have focused on the national government level. Primary concerns were to gain an understanding of their roles as "watchdogs of public value" at the national level (Moore and Gates, 1987) as well as the ways they create public value . The SAIs are seen as performing a strategic role in acting as an "agent of change" to expand public sector accountability boundaries (Gendron et al., 2001;Jacobs, 1998;Pallot, 2003). The SAIs can also reinforce the parliament's powers in scrutinising the government (Bunn et al., 2018;Ferry and Midgley, 2021).
Individual auditors, as well as institutions performing an audit function on different levels (local, national and transnational), face the challenge of positioning themselves vis-a-vis their stakeholders, engaging in a constant effort to stay relevant, useful and legitimate . These efforts appear to affect the complexity of the interaction between the auditors and the institutions they represent. Furthermore, as with accountants (Rocher et al., 2021), auditors can also be considered characters "in action" because they interact in society with other actors and they are embedded within an organisation. Since auditing is executed in the public interest, auditors must maintain the confidence of the public for it to fulfil its primary function, something that can be achieved by building the image and reputation of auditors (Monroe and Woodliff, 1994;Rogers et al., 2005;Tid a, 2021). However, confidence in public sector auditing, defined as "the mutual understanding as to the appropriate responsibilities of auditors and a belief by the public that such responsibilities are being fulfilled" (AICPA, 1987), seems to have decreased (Holm and Zaman, 2012;Humphrey et al., 2007).
While research on auditing in local and national organisations highlights both the function and the outcomes of these organisations, studies on national audit institutions such as SAIs and transnational audit institutions such as the ECA highlight the position of these organisations, and their reputational work and image, in complex stakeholder contexts (e.g. Stephenson, 2020;Tid a, 2021;Justesen and Skaerbaek, 2010). Therefore, SAIs' response to stakeholder expectations is an area of interest (Yamamoto and Kim, 2019).
It is increasingly important for auditing institutions to enhance their legitimacy by building their own reputations and ensuring that stakeholders trust them (Johnsen, 2019;Funnell, 2015;Guthrie and Parker, 1999;Jacobs, 1998;Lonsdale, 2008). The perception of public sector auditing institutions should be that they can maintain solid social values (Funnell, 2015), prevent mismanagement and corruption , and avoid the so-called "credibility gap" (Bringselius, 2018). Therefore, public sector audit institutions not only need to establish communication practices showing that they remain anchored to regulatory compliance issues, but they also need to engage in social construction processes as vehicles for the activation, among others, of legitimation (Stephenson, 2017;Deephouse and Carter, 2005).

Visualising the actorhood of public sector audit institutions
The increasing complexity of stakeholder interactions challenges national and transnational public audit institutions. How these organisations deal with those challenges can be described in terms of actorhood theory (Bromley and Sharkey, 2017;Halgin et al., 2018). Actorhood theory suggests that public sector audit institutions are increasingly constructed as actors in society (Halgin et al., 2018), as people are, and act in a socially meaningful way (Meyer and Vaara, 2020).
Both individuals and organisations can be considered actors, which means that they have bounded autonomy, clarity of purpose, decision-making ability, technical action capability and effective self-control. They also possess substantial information about themselves and the environment (Meyer and Bromley, 2013), with individuals inhabiting institutions and institutions personified by people (Voronov and Weber, 2020). In striving to gain legitimacy with their stakeholders, organisations are increasingly depicting themselves as entities with a mission, strategy, identity and purpose (Glynn and Abzug, 2002), seeking to legitimise their actor role vis-a-vis or on behalf of the greater principle that empowers their agency (Meyer, 2010).
Adopting the actor stance, organisations strive to anthropomorphise themselves, developing personhood aspects that emphasise their "legitimate collective authority" (Meyer, 2010, p. 2) and corporate person status. The actor status also permits an organisation to transit from a closed to an open model, whereby depicting themselves as human-like entities allows them to present themselves as social actors engaged in social intercourse and possessing all the rights and responsibilities of a single individual (Whetten and Mackey, 2002).
In all this, communication is of considerable importance (Voronov and Weber, 2020;Bitektine et al., 2020;Meyer and Vaara, 2020) as it is the nodal point of the social construction of social reality (Berger and Luckmann, 1967), able to represent the reputation of professional organisations (Maass and Arcuri, 1996;Stangor and Schaller, 1996;Rocher et al., 2021). Public sector audit institutions use reports, media announcements and the Internet to communicate with their stakeholders and, in turn, gain visibility from both their stakeholders and the wider public (Gonz alez et al., 2008). Using these tools as part of their communication strategy allows these organisations to build their reputation and gain recognition as well as legitimacy by completing the cycle of SAIs' accountability (Gonz alez et al., 2008). With respect to the ECA, Stephenson (2020) suggests that the ECA's special reports provide insight into the role this institution plays in the EU context. Tid a (2021) argues instead that an understanding of the ECA's role can be achieved through the exploration of their media presence and coverage.
Although previous studies acknowledge the importance of communication in relation to public audit institutions in general, and the ECA in particular (e.g. Waeraas, 2010;Gonz alez-D ıaz et al., 2012), they fall short in not recognising that communication aimed at the visibility of public sector audit institutions can be achieved through means other than just textual representation. To capture communication strategies, visual analysis might be an important alternative to textual analysis by avoiding the complexity of textual representations (cf. Spoehr and Lehmkuhle, 1982). Visual analysis is considered one of the best tools to capture organisational actorhood via communication practices, given its ability to grasp the animation and complexities of actors (Halgin et al., 2018;Guthey and Jackson, 2005). Images can depict multiple meanings and evoke a variety of complex and abstract feelings and interpretations (Langer, 1953). An organisation's presentation of themselves as actors can be best achieved through images, given that visual images represent a transformation from something invisible and unobservable into something that has a form and is visible (H€ ollerer et al., 2013). It is the specific "performativity" of visual artefacts and visual discourse (quite different from other styles of communication) that holds great potential (Meyer et al., 2013, p. 489). A focus on visual elements can be useful to illuminate the ways in which organisational, professional and personal identities are formed and communicated.
Despite a growing interest in visual analysis, Davison (2015) acknowledges that its use is still rather limited in accounting and auditing research, with most visualisation studies directing their attention towards financial and non-financial reporting rather than exploring communication strategies of accounting and auditing organisations. Given the relative paucity of studies aimed at understanding public sector audit institutions' communication strategy, visual analysis might represent a way forward in exploring the patterns of the ECA's communication practices and understanding the actorhood of transnational public audit institutions.
To summarise, and in line with  and Halgin et al. (2018), this study states that organisations engage in the construction of organisational actorhood through the images they communicate to external audiences, and it is through the analysis of these images that this actorhood journey can be explored.

Research context and methodology
The paper performs a visual analysis of the images on the covers of the ECA's official journal, the ECA Journal, from its inception in 2009 up to 2019. The analysis entails a mixed method consisting of both content analysis and a two-step cluster analysis. A short overview of the ECA's history precedes descriptions of the data sources, the data collection method and the analysis.  (3)).
The mission of this institution is to contribute to improving EU financial management, guiding European policy makers and legislators on how to manage EU policies and programmes. Moreover, it promotes accountability and transparency, informing European citizens on how their resources are being spent (European Court of Auditors, 2019). Each year the ECA provides an annual statement of assurance to the European Parliament and the Council, but it may also at any time communicate observations on specific issue or deliver opinions.
The ECA carries out the audit task in liaison with SAIs at the national level (Stephenson, 2017). As stated earlier, the ECA delivers accountability and oversight. The ECA shows its work programme on its website (European Court of Auditors, 2019), and uses the ECA journal to provide its stakeholders with insight into the organisation's work and role. In addition, the journal provides a platform for court auditors and employees, EU politicians and academics to express opinions on matters related to ECA work. The journal also publishes interviews with members of the court as well as experts in areas of interest. Prior to 2017 the journal published general issues with no specific themes; as of 2018 the journal has a new format and now publishes ten issues per year, each issue dedicated to a specific theme, for example, climate change and audit, big data and digital audit and EU pre-accession aid.

Methodology
Images are an important tool for reflecting reality and shaping it by communicating meaning (Kyriacou, 2016). As such, images are able to construct the identity of professionals (Parker and Warren, 2017;Warren and Parker, 2009). The front page of the ECA journal was chosen as the source of data for analysis because the journal is the primary communication tool of the ECA and it is through this tool that "construction of reality" in a wider sense takes place (Hines, 1988). Between January 2009 (the date of the first issue) and March 2019 (when data collection ended), 106 issues of the journal were published. The 325 distinct images from these covers were analysed.
In addition, to gain a deeper insight into the ECA journal's publishing and editorial processes, two explorative interviews were conducted with the current editor in chief and the deputy editor. The interviews each lasted 1.5 h and focused on the development of the ECA journal over the years as well as the formation of its visual and textual content. Interview data were used to corroborate the results of the visual analysis.

Content analysis.
The data capture and coding were inspired by pioneering studies on the visual images of accountancy (e.g. Benschop and Meihuizen, 2002;Davison, 2010, Kyriacou, 2016Parker and Warren, 2017;Warren and Parker, 2009). In particular, the studies by Benschop and Meihuizen (2002) and Kyriacou (2016) provided strategies for data capture and coding of texts and visual artefacts, including photographs, pictures, paintings, drawings, sketches and other material forms of visuality, and also more abstract manifestations such as colour, typography and charts (Meyer et al., 2013, p. 489). The visual artefacts were coded directly from the covers, without any transcription of the content (Fors and B€ ackstr€ om, 2015). A content analysis (Bell and Davison, 2013) was employed to explore and scrutinise the 325 images and to capture the frequency of certain visual elements.
Data were coded by two authors independently of each other, after which the codes were compared and consensus-based final coding completed. This approach aimed to minimise risks related to inter-coder variability (Bell and Davison, 2013). The codes attributed to the 325 images available on the covers of the ECA journal were recorded in Excel. The two coding authors paid particular attention to how the images could be perceived by stakeholders and the possibility that the authors' interpretation could differ from stakeholder perceptions. Notes the authors had made during code comparison on their interpretations of the images were later discussed with all the authors engaged in this study to secure coding credibility.
Data coding used both realistic (factual) and interpretive codes. The primary purpose was to understand how individuals personifying the actorhood of the ECA were depicted and the context in which that took place. Regarding individuals, five aspects were coded, namely their gender, facial expressions, ethnicity, approximate age and clothing. Regarding context it included both the context provided by the entire cover of each issue and the context of each single image.
The cover context was coded in terms of colour schemas. The single image context was more complex, covering both situational aspects and focus. For example, each image was coded in terms of the occurrence of situational dimensions or gestalts which were divided into the following ten categories: close-up one person, shaking hands, interviews, individuals addressing a group, group posing, conference, roundtable discussions, individual in a chair, buildings and other. The primary focus could differ from that of the situation; for example, while a person could be the main focus of the picture, the situational dimension could be conference gatherings. Thus, the following eight focus categories were used: portrait (when the face or head and shoulders of individuals were depicted), group, close-up object, building, industry, symbols, office work and other.
A more interpretative coding was performed on the domains the images represented, with a specific focus on artefacts that symbolised the domain, where a domain is understood to be an aggregate conceptualisation of the context (cf. Umans et al., 2019). For example, when coding an industrial domain, we considered industrial objects such as railway tracks, heavy machinery, or items used in a laboratory or in construction. An industrial domain image can be found on the March 2018 cover (https://www.eca.europa.eu/lists/ecadocuments/journal18_ 03/journal18_03.pdf). Another example is the coding of the social domain. To be assigned a social domain code, the image had to include interaction or conversation between two or more people. One such image can be found on the November 2016 cover, where two men are engaged in conversation with each other and gesticulation confirms their interaction (https:// www.eca.europa.eu/lists/ecadocuments/journal16_11/journal16_11_en.pdf). The 13 domains and their features are presented in Table 1.
Finally, the images were also coded in terms of location specificities, such as inside/ outside, in the workplace or not and in terms of more specific details such as staircases and the interior and exterior of prestigious buildings.
4.2.2 Cluster analysis. Data were further analysed by means of a two-step cluster analysis, which allowed natural groupings (clusters) to be identified within a data set that would otherwise not be apparent. Two-step cluster analysis was chosen because of the size of the data set and because it contained both scale and ordinal variables (Chiu et al., 2001). This method used a log-likelihood distance measure and determined the number of clusters automatically based on changes in distance (Chiu et al., 2001). The automatic selection in twostep clustering provides objectivity and avoids arbitrariness in pattern emergence (Stanley et al., 2017), allowing the ECA's actorhood journey to be unveiled and understood. Two-step cluster analysis involved two steps. In the first step, the original cases were grouped into pre-clusters through construction of cluster features tree (Rundle-Thiele et al., 2015). In the second step, the pre-clusters were clustered through a standard hierarchical clustering algorithm (Norusis, 2011). In this second step the analysis produces a range of solutions which is then reduced to the best number of clusters on the basis of the Schwarz's Bayesian information criterion (BIC). BIC suggested number of clusters is considered to be one of the most objective selection criteria given that it avoids the selection arbitrariness (Rundle-Thiele et al., 2015). Once the cluster solution was formed, three validation measures were used. First, the silhouette measures of cohesion and separation were checked. These should be above the required level of 0.0 to suggest that the within-cluster distance and between-cluster distance is valid (Norusis, 2011). Three cluster analyses performed for this study had silhouette measures above 0.0 and ranged between 0.4 and 0.5 suggesting fair to good validity. Second, ANOVA and Scheffe's test were performed on categorical and continuous variables respectively to determine the importance of individual variables within each cluster as well as to indicate significant differences between clusters (Rundle-Thiele et al., 2015). Both tests performed for validation of the clusters for the three cluster analyses indicated significant differences between the clusters as well as that the clusters varied significantly across the segmented variables. Third, we halved the clusters to determine whether the two-halves of the cluster were similar in terms of size, number and characteristics of the clusters (Norusis, 2011). Our results indicated that for all three cluster analyses performed, the halves were similar to each other on all the aforementioned dimensions. images) (see Table 2) and primarily opted for grey background colours, which were found in 92% of all issues, with a combination of blue, green and orange also being common. The image colour schemas diverged from that of the background, with blue and green being found in 13 and 15% of the cases, respectively. 5.1.2 Gestalts. The majority of the images depicted a close-up of one individual, an individual addressing a group, or a group posing. In 25% of the images the gestalt was difficult to identify and a category of Other was devised; it was the largest category (See Table 3).

Results and analysis
According to the journal editors, the shift from the depiction of individuals and towards other gestalts was motivated by the journal's aim of depicting happenings, that is, events or activities and the dynamics of what happens within the ECA, and its relationship to societal challenges.
5.1.3 Focus. More than 70% of the images focused on an individual or a portrait (represented by the category Portrait). A focus on groups (represented by the category Group) was depicted in 38% of the images. A focus on close-up objects (represented by the category Close-up object) was depicted in 25% of the images (See Table 4 Table 3. Gestalts and Technological (See Table 5). The journal editors reported that it was important for the journal at its inception to introduce the people who worked for and represented the ECA. Later, showing the interactive nature of the relationships the ECA had with its stakeholders became important to the journal. 5.1.5 Visible attributes. The data showed that there were on average 3.1 individuals per image, each containing on average 1.7 men and only 0.4 women. On average each image contained 1 neutral face and 0.8 smiling faces. Caucasians were overrepresented in the images with 2.3 Caucasian individuals per image. On average there were 1.9 individuals in formal clothing, and on average 1.7 individuals in the images were between 45 and 60 years of age (See Table 6).
5.2 Two-step cluster analysis 5.2.1 Focus. In the two-step cluster analysis related to focus in the images, six clusters emerged as an optimal solution based on the Schwarz criterion. The silhouette coefficients (0.5) suggested good validity for the within-and between-cluster distances. The results of the ANOVA test indicated significant differences between the clusters, and Scheffe's test confirmed that the clusters varied significantly across the segmented variables (See Figure 1).
Cluster 1 represents a joint focus on groups and portraits, aspects that were popular at the journal's inception but that have since declined (see the 2009 December issue at https://www. eca.europa.eu/lists/ecadocuments/journal09_12/journal09_12_en.pdf). Cluster 3 represents a focus on close-up objects and portraits, aspects that were increasing from 2010 until 2012 but have been in steady decline since (see the 2012 May issue at https://www.eca.europa.eu/lists/ ecadocuments/journal12_05/journal12_05_en.pdf). Cluster 6 is dominated by symbolic  03.pdf). The focus area of Other is also included in Cluster 4. Cluster 5 is a mixed cluster and, with some slight dominance of portraits, it is similar to Cluster 2, which is dominated by a focus on office work (see the 2018 January issue at https://www.eca.europa.eu/lists/ ecadocuments/journal18_01/journal18_01.pdf). Both clusters 5 and 2 had very low representation over the years but were always present. According to the editors, the journal attempts to present the ECA in the different contexts in which its work is being performed, which explains the recent emergence of a focus on industrial and building aspects (along with aspects in the Other category).

Domains and Gender.
In the two-step cluster analysis related to domains and gender, four clusters emerged as an optimal solution based on the Schwarz criterion. The silhouette coefficients (0.4) suggested fair validity for the within-and between-cluster distances. The results of the ANOVA test indicated significant differences between the clusters, and Scheffe's test confirmed that the clusters varied significantly across the segmented variables.
Cluster 4 is primarily represented by the Other and Political domains, where both men and women appear jointly. This cluster is also characterised by limited representation of the Social and Technological domains, as well as the lack of men in the images. This cluster was particularly strong between 2012 and 2016 (see the December 2016 issue at https://www.eca. europa.eu/lists/ecadocuments/journal16_12/journal-dec-2016-final.pdf). Cluster 1 is represented by the Personal domain, with only men appearing in the images. This cluster is also characterised by the absence of the Social and Other domains, as well as the appearance of both men and women in the same image. This cluster peaked in 2012, with a steep and steady decline thereafter (see the July-August 2012 issue at https://www.eca. europa.eu/lists/ecadocuments/journal12_07/journal12_07_en.pdf). Cluster 3 is primarily represented by the Social domain, with some Other and Technological domain influences. Men appearing alone are depicted within this cluster and fewer images have men and woman appearing simultaneously. This domain was prominent at the journal's inception, with some renewed interest in 2016 (see the November 2009 issue at https://www.eca.europa.eu/lists/ ecadocuments/journal09_11/journal09_11_en.pdf). Cluster 5 is also dominated by the Social domain, but with some presence of the Technological. This cluster includes both men and women. While present over the years, this cluster had no peaks and appears to be on the decline (see the April 2012 issue at https://www.eca.europa.eu/lists/ecadocuments/journal12_ 04/journal12_04_en.pdf). Finally, Cluster 2 is dominated by the Environmental and Life domains, followed by the Economic, Professional, Industrial, Cultural and Historic domains. This cluster has only women on the covers. This cluster is seeing the greatest growth, with a substantial increase starting in 2016 (see the July-August 2016 issue where a woman appears alone on the cover at https://www.eca.europa.eu/lists/ecadocuments/journal16_07_08/ journal16_07_08_en.pdf).
The journal editors acknowledged that the journal covers have been shifting from representation of individuals to the representation of current events and societal challenges that the ECA's work is related to. The shift was associated in part with a reconsideration of Figure 1. Cluster analysis -Focus the journal's stakeholders. Editors explained that at the journal's inception it aimed to provide information for EU institutions it was working with. Since then, it has been widening its stakeholder focus and now provides information to a larger audience, which also includes the general public (see Figure 2). 5.2.3 Domains and diversity. In the two-step cluster analysis related to domains and diversity three clusters emerged as an optimal solution based on the Schwarz criterion. The silhouette coefficients (0.4) suggested fair validity for the within-and between-cluster distances. The results of the ANOVA test indicated significant differences between the clusters, and Scheffe's test confirmed that the clusters varied significantly across the segmented variables (See Figure 3).
Cluster 1 is dominated by the domains Personal, Life and Environmental, with diversity of individuals not emergent. The cluster is also characterised by no appearance of the Social, Other or Technological domains. The cluster is characterised by a decreasing number of simultaneous domains over time. The cluster appears to be increasing, and while less dominant at the beginning is more so recently (see the April 2019 issue at https://www.eca. europa.eu/lists/ecadocuments/journal19_04/journal19_04.pdf). Cluster 3 is dominated by the Other domain but is also influenced to some extent by the Political domain. In this cluster, diversity in terms of gender, ethnicity and age appears to emerge. The cluster was increasing in 2014 with a steep decline in the later years (see the October 2014 issue at https://www.eca. europa.eu/lists/ecadocuments/journal14_10/journal_102014.pdf). Cluster 2 is dominated by the Social, Political, Technological and, to some extent, Other domains. In this particular cluster, gender diversity (both men and women appearing in the image), age diversity and ethnic diversity appear more than in the other clusters. This cluster was particularly dominant at the journal's inception but has been consistently declining over the years since (see the May 2009 issue at https://www.eca.europa.eu/lists/ecadocuments/journal09_05/ journal09_05_en.pdf).
The interviews with the journal editors revealed that the journal covers first reflected the diversity of individuals working at the ECA. Recently the focus has shifted to the diversity of assignments performed by the ECA employees.

Discussion
This paper attempts to understand how ECA actorhood has been constructed, and by means of cluster analysis to identify patterns in this journey. Seeking inspiration in the actorhood theory (Meyer, 2010;Meyer and Bromley, 2013) and using the visual accounting method (Davison, 2010) of gathering the data from the front pages of the ECA's official journal, this paper has mapped the patterns of manifestation of organisational actorhood. The study has focused on different actors and the context within which these actors were depicted in. The primary focus of the exploration was to reveal how the ECA depicts and positions individuals in contexts that serve the purpose of communicating ECA actorhood to its audience (i.e. the stakeholders) (Bromley and Sharkey, 2017;Halgin et al., 2018).
The content analysis suggests that the ECA primarily uses photos to establish its actorhood position, with subtle background colours that serve as a context for brighter spots of blue and green. This can be interpreted as an effort to ensure that attention remains focused on the happenings and actions in the image. To establish their actorhood, the ECA opts to depict individuals in different constellations, thereby showing the importance both of individuals in context and also indicating the interactive joint nature of the work performed by the ECA. Portraits of one or several individuals are most dominant, which serves to establish that human capital and its value is what characterises the ECA positioning. In the depiction of the contextual domains, the ECA brings out the categories of Personal, Social, Political and Technological. An emphasis on the Personal and Social contexts serves to depict an organisation as a live actor on the social stage. Paying homage to its status as a "public servant" within the political landscape, the ECA makes reference to the political context it has to respect as it plays its role. Finally, to signal its modernity as well as possible relevance, the Technology domain provides important contextual aspect. Regarding characteristics of individuals, the covers primarily show middle aged white men with neutral faces and in formal clothes. Thus, the modernity of the context intertwines with the formality of appearance, which could be a way for the ECA to communicate the stability of professional values and its identity in the modernising world . According to the editors, the journal has been gradually moving from depicting social actors in the professional environment to depicting social actors in modern settings that might be outsides of the ECA or EU agency contexts. Also mindful of the gender debate in society and possibly unwilling to take a stance, the journal has been gradually de-gendering its front page and moving towards context-specific attributes rather than individuals of one or another gender. Similarly, while the ECA journal was previously highlighting ethnic and age diversity in their images, these practices have subsided, with rare appearances of individuals or groups that could signal these types of diversity.
To sum up, the ECA appears to position itself as a human and social actor that would like to become well embedded in both the political and technological spheres. It also appears to opt for a social actorhood rather than a professional actorhood. The latter is somewhat surprising given the professional nature of ECA work. Yet as our interviews reveal, the journal attempts to present ECA as an institution that appeals to the general public and tackles issues of relevance for this public. Finally, while the contexts depicted in the images signal the dynamic nature of the ECA, the individuals positioned within the contexts appear to be homogeneous rather than diverse in demographic characteristics, in contrast to earlier practice.
Exploration of the patterns over time through cluster analysis showed that the focus of the images on the covers has changed over the years. While in the formative years of the journal, it focused on groups and portraits, it has gradually changed to a focus on close-up objects and symbols (a sign of de-personification) and has further developed into a focus on undetermined, usually abstract, other aspects as well as industry and buildings. This development might suggest a shift towards an abstract rather than a real actorhood stance. While content analysis suggested an attempt at personification, cluster analysis suggests that while this effort characterised the ECA in its formative years, in recent years the emphasis has been on making the ECA a symbolic and abstract artefact rather than an actor. This shift has been confirmed by the editors of the journal, who believe that depicting pressing societal issues on the cover rather than highlighting the individuals dealing with these issues is a good strategy.
The combination of the domains and gender revealed a change in that initially only men were depicted in the Other (abstract) domain, followed by a period of showing both men and women in the Social domain, and recently switching to showing only women in various domains. By opting for only female individuals in multiple domains the ECA might be striving to signal the gendered characteristics of its organisation and to show the relevance of gender in various contextual domains. Finally, exploring diversity characteristics in relation to the contextual domains, the more recent lack of diversity in the Personal, Economic and Professional domains shows the neutral and often homogeneous nature of the ECA's actorhood. Age and ethnic and gender diversity were primarily "staged" at the journal's inception and although often associated with the Political domain, have become less relevant in recent years. It appears that the ECA is putting more emphasis on the context within which individuals are depicted, which gives stakeholders more they can relate to through the context rather than through individuals. However, this appears to lead to de-personification of the ECA and to showing the context in which actorhood takes place rather than revealing itself as an actor through individuals and their actions in the images. Interviews with the editors revealed that the journal seeks to address current trends in society and contribute to public debate on pressing social issues, which means de-personification becomes inevitable.

Conclusions
Our findings might reveal what Meyer and Vaara (2020) term co-construction of institutions and social actors through visual communication. By studying "communicative construction" (Luckmann, 2006) rather than opting for social construction of reality our study shows how communicative construction of institutional actors develops over time and how the ECA both shapes and is shaped by changes in institutional context. Moving beyond verbal or textual communication, our study highlights how the use of multiple sign systems (Berger and Luckmann, 1967) and the placement of individuals into specific "physical" spaces permits the ECA journal to provide a rich representation of the organisation itself. Finally, our findings reveal the journey over time of the ECA to establish legitimacy through visual communication. They also show how the use of both individuals and symbols in specific contexts aims at positioning the ECA on the public stage.
This paper makes several contributions to the public sector auditing literature. First, this work highlights how transnational public organisations such as the ECA construct their actorhood through the use of images. Second, the study introduces the actorhood theory to the field of public sector accounting and auditing, highlighting its instrumentality in revealing the underlying legitimacy mechanisms through which organisations position themselves in the public eye. Third, the paper shows the actorhood development of the ECA's journal, highlighting how individuals and context interact and develop over time. Fourth, it introduces a new technique to visual accounting methodology, namely cluster analysis, a tool that is primarily used in marketing but which is shown to be of value also for the public sector accounting and auditing literature. Finally, this paper makes an empirical contribution in showing the actorhood journey of the ECA.
This study has a number of limitations. The restricted time window within which the ECA actorhood journey has been analysed precludes a more thorough quantitative exploration. Second, the coding could be criticised for its arbitrary nature. Further studies could explore further ways of refining and expanding the coding schemas by drawing on research in semiotics, which has seen substantial development in recent years. Third, the study uses only one source of data (apart from the explorative interviews with the editors). It could be of benefit to examine representations of the ECA in other sources, in social and traditional media, for example, and going back in time also oral sources through interviews with key actors (such as the individuals depicted on the covers of the journal) that are associated with the journal and the ECA context. This paper has dealt with the actorhood of the ECA and its legitimacy, which can be connected to the organisational identity of the selected transnational audit institution. Future research could focus on the emotions, feelings and perceptions that auditors have when dealing with images in order to achieve further insights into auditors' professional identity construction. In addition, comparison of feelings and impressions of different kinds of stakeholders could also be studied. For example, experiments could be conducted with students to explore their impression of the public auditing profession. Future research could also focus on both images and related content in the entire ECA journal that could reveal ECA's journey of establishing itself on the public stage. Our interviews with the editors have revealed that in addition to the externally available ECA journal aimed at the general public, the ECA also has an internally available journal (mECAzine) aimed at ECA employees. Gaining access to mECAzine and comparing it with the ECA journal could provide a valuable research opportunity to study consistency of internal and external communications at the ECA.