In the Line of Fire: European Financial Regulators before, during, and after the Crisis

A key argument in recent theorizing on the drivers of bureaucratic behaviour is that agencies seek to establish and maintain a unique reputation. While recent years have witnessed substantial empirical support for this claim, the field lacks comparative examinations of the dynamics of reputation and its management throughout crisis periods. This article draws on a systematic media content analysis to explore the exposure and communication responses of the German, Belgian and Danish financial regulators to reputational threats before, during, and af- ter the financial crisis. Our results point at the dynamic and context-sensitive nature of reputation management. syste- matischen Medienanalyse, welchen reputationsbe-zogenen Bedrohungen die Finanzmarktbehörden in Deutschland, Belgien und Dänemark vor, während und nach der Finanzkrise ausgesetzt waren und wie diese hierauf mit Hilfe öffentlicher Kommunikati-on reagiert haben. Die Ergebnisse zeigen, dass Re- putationsmanagement zeitlichen Schwankungen unterliegt und kontextabhängig ist.


Introduction
This article contributes to a growing body of literature that has established reputation as an important motivation for the behaviour of government agencies and their stakeholders (Busuioc & Lodge, 2017;Carpenter & Krause, 2012). Whereas institutional political science has long focused on the authority of political principals to control agencies, reputation-based approaches shed light on reputation as a source of bureaucratic power (Maor, 2015). In a context of information overload, individuals make choices based on impressions and experiences, or in other words: on reputation (Canel & Luoma-aho, 2019). Therefore, favourable and unique reputations serve as valuable assets for agencies to forge power and autonomy (Carpenter, 2001). Not surprisingly, then, agencies have been found to be concerned with how they are externally perceived and to react strategically to threats to their reputation.
This article addresses several shortcomings in the literature on bureaucratic reputation theory. First, bureaucratic reputation theory predicts that reputation-seeking agencies will react differently to public allegations targeted at different areas of agency activity. The literature shows that agencies calibrate their responses to the reputational costs of public allegations (Gilad, Maor, & Bloom, 2015;Maor, Gilad, & Bloom, 2013). The empirical literature testing this core claim in bureaucratic reputation theory is, however, limited in scope and in particular lacks validation across various national contexts. The theory's assumptions have mainly been tested for agencies in the United States (Carpenter, 2010a(Carpenter, , 2010b, Israel (Gilad et al., 2015;, and Australia (Maor & Sulitzeanu-Kenan, 2016). The article addresses this limitation by studying reputational threats and agencies' responses for agencies with similar tasks in three European countries.
Second, the concept of reputation is multi-dimensional. For instance, agencies scoring high on performative reputation (the agency's ability to achieve its mission) may not necessarily score high on procedural reputation (the agency's respect of norms and procedures of appropriate decision-making) (Carpenter & Krause, 2012). However, despite the intuitively powerful distinction between different reputational dimensions, this idea has received limited empirical attention so far (but see Christensen & Lodge, 2018). Hence, we do not know whether agencies are systematically exposed to some types of allegations rather than others, and whether they respond differently towards allegations addressing different dimensions of bureaucratic reputation. The article contributes to the literature by systematically analysing agencies' exposure and reactions to threats directed at performative reputation.
Third, although several authors have studied agencies' decision-making and communication in times of crisis (Carpenter, 2010a;Moschella & Pinto, 2018), we do not know whether agencies respond differently to reputational threats under crisis conditions as opposed to 'business-as-usual' conditions. Although reputation threats may be linked to an institutional crisis exposing significant errors of omission or commission in an organization, they may also take the form of criticism of single-case decisions or organizational practices well below the (malleable) threshold of a fullblown crisis. The financial crisis is a unique possibility for studying regulatory agencies being exposed to similar types of reputational threats across different national contexts.
In order to investigate how agencies react to reputational threats under various contextual conditions, we explore the exposure and reaction of three European financial regulators to reputational threats before, during, and after the financial crisis using systematic media content analysis. The article's comparative research design primarily serves as a means to test the robustness of bureaucratic reputation theory across national and situational contexts. We use systematic media content analysis and apply a novel content coding of major newspapers' reporting on financial regulators' reputations in Belgium, Denmark and Germany from 2003-2015. 1 2 Bureaucratic Reputation Theory The study of bureaucratic reputation and its implications is a nascent sub-field in the study of executive politics (Maor, 2015). A key assumption in bureaucratic reputation theory is that government agencies are motivated by a desire to demonstrate "reputation uniqueness", which refers to their ability to create solutions and provide services found nowhere else in the polity (Carpenter, 2001, p. 5). For government agencies, reputations are valuable political assets that are used to generate public support, to achieve autonomy, to protect the agency from attacks, or to attract and retain employees (Carpenter, 2002). Bureaucratic reputation scholars are interested in how agencies identify and respond to threats to their reputation, which is defined as "a set of symbolic beliefs about the unique or separable capacities, roles, and obligations of an organization, where these beliefs are embedded in audience networks" (Carpenter, 2010b, p. 45). Government agencies are seen as rational and politically conscious actors with incentives to avoid reputational damages (Maor, 2015).
This view stands in contrast to political economy approaches which understand the power of agencies as defined by the powers delegated by elected politicians in a principal-agent relation. The bureaucratic reputation approach, similar to other bureaucratic politics approaches, puts the protection of agency reputation and the carving of a distinct turf by public organizations at the centre stage of the analysis (Bach & Wegrich, 2019). A theory of bureaucratic reputation is not merely interested in how audiences perceive agencies, but assumes that reputation-sensitive agencies deliberately engage in managing their reputation among those audiences (Maor, 2015).
Those studies show that regulators operate in an environment comprising diverse stakeholders with varying expectations, including elected officials, clientele groups, the media, policy experts, and citizens (Carpenter & Krause, 2012). Audiences' expectations can be directed at several dimensions of an agency's reputation. Carpenter (2010b, pp. 45-46) proposes a framework of four reputational dimensions: (1) performative, i.e. the ability to execute its tasks effectively, both with respect to outcome and output; (2) moral, i.e. the ability to meet normative expectations, such as protecting citizens and ensuring transparency; (3) technical, comprising the "expertise" and professional qualifications of the organization; and (4) procedural, which refers to the organizations' conformity to procedures and legislation. How agencies perceive, process and respond to multiple audiences' expectations is a core interest of reputation theorists (Christensen & Lodge, 2018). Agencies face a major challenge that reputational dimensions "neither stay nor move in harmony" (Carpenter & Krause, 2012, p. 27): maximizing one dimension might come at the expense of another. Audiences have different (and potentially conflicting) expectations about the way the agency should exercise its tasks, which ultimately means that agencies have to prioritize whether and how to respond to reputational threats.
While the core assumptions of reputation theory have seen considerable empirical validation, several questions remain unanswered. First, empirical studies are limited in scope and in particular lack cross-national validation. Most studies have focused on the reputation-seeking behaviour of single agencies in a limited number of policy domains. This leads to the question whether existing results can be generalized across country contexts and policy domains. Are all agencies equally reputation-sensitive, and if not, how can we explain these differences (see also Maor, 2015)? This article adds to this research agenda by expanding the geographical scope of existing research on financial regulators' reputation management.
Second, although Daniel P. Carpenter's (2010b) theoretical conception of reputations as multidimensional constructs is widely cited, there are hardly any empirical studies that have analysed reputational threats using his four-dimensional scheme. An exception is a recent article by Tom Christensen and Martin Lodge (2018) who study how agencies manage their reputation through the use of symbols on their websites. However, this study does not address agency responses to reputational threats, but aims at teasing out country-and domain-specific variation in agencies' exercise of social accountability. Among others, these authors find task-specific differences in reputation management. Accordingly, it seems advisable to control for tasks when studying agency responses to reputational threats in different contexts.
Third, reputation scholars have been silent on the question whether regulators' sensitivity to reputational threats, and their subsequent responses, are different in crises compared to business-as-usual conditions. Most studies focus on reputation management of agencies during normal times, yet differences might be expected between day-to-day reputation management and reputation management during crisis times (see Baekkeskov, 2017 for a study of reputation management under crisis conditions). Christensen and Lodge (2018) speculate that the procedural dimension might be more prominent during crises as agencies seek to absolve themselves from blame by denying responsibility or by stressing conformity to procedural rules. Moreover, a key question is whether crises permanently damage a regulator's reputation, or whether everything turns back to the previous state after the smoke has settled. More attention is needed to these dynamics of agency reputation before and after a crisis. This article will tackle these questions. We explore regulatory agencies' exposure and communication responses to reputational threats before, during and after the global financial crisis (2003 -2015) in Germany, Denmark, and Belgium. As we elaborate below, financial regulators in these countries are largely comparable in terms of formal autonomy and task portfolio. Quantitative analyses will shed light on the evolution in the volume of reputational threats and the communication responses of each regulator to these threats. Furthermore, this article is not only interested in the volume of reputational threats, but also in their content. We distinguish between threats towards the performative dimension of each regulator's reputation and threats to other reputational dimensions.

Financial Regulators and Reputational Risks
Most studies using a reputation theory framework investigate regulatory agencies, and several studies have focused on the behaviour of financial regulators and central banks facing reputational threats (Carpenter, 2010a;Gilad, 2015;Gilad et al., 2015;Moschella & Pinto, 2018). Financial regulators are particularly interesting for the study of bureaucratic reputation. First, they are part of the growing population of independent regulatory agencies that supervise important economic sectors (Gilardi, 2008). The financial and sovereign debt crisis of the late 2000s has made clear how much modern societies depend on a well-functioning financial sector, which includes banking, securities, and insurance. Arguably, cultivating and maintaining a favourable reputation among those industries is a crucial element of effective financial regulation.
Second, regulatory agencies are usually granted wide-ranging powers in both policy implementation and formulation (Gilardi, 2008). Importantly, bureaucratic reputation theory assumes that organizational behaviour is not primarily driven by formal design. The cultivation and maintenance of a distinct reputation rather results from autonomous actions of agencies themselves. That said, we assume that formal autonomy provides a fertile ground for developing a distinct organizational identity that becomes accepted among relevant stakeholders. Moreover, following Anthony M. Bertelli, J. Andrew Sinclair and Haram Lee (2015) who stress the importance of organizations' evaluability and identifiability for holding them accountable, we suggest that formal autonomy makes regulators more identifiable and their work more evaluable compared to less autonomous bodies (Boon, Salomonsen, & Verhoest, 2018).
Finally, financial regulators are facing distinct reputational risks, many of which have come to the fore in the wake of the financial and sovereign debt crisis. These include allegations of lenient enforcement vis-à-vis regulated industries and cosy relationships with regulatees (regulatory capture). Another reputational risk consists of quite the opposite, including criticism about over-regulation, which is said to affect the competitiveness of financial industries. Moreover, financial regulators may face allegations of their staff lagging behind the financial industry in terms of expertise. While all of these claims have been present during the financial crisis, the claim of a lack of effective regulatory standards and their enforcement was arguably most prominent (Lodge & Wegrich, 2011). This criticism clearly taps into the performative dimension of agency reputation or the ability to achieve its mission ("to get the job done").
What this implies for the reputational threats of financial regulators is the following: if these regulators have been granted substantial independence and the setup or reform has been justified with a need for a depoliticized, technocratic regulation of the growing financial sector, then a diagnosis of regulatory failure ("not getting the job done") might go to the core of the legitimacy of these institutions. This article hence explores if these regulators really have been in the line of fire during the financial crisis, the degree to which the performative dimension has been targeted, and how these regulators responded to reputational threats. Before moving to the data and methods section, we highlight key characteristics of these regulators within the respective governmental systems.

Context: Financial Regulators in Three European Countries
In the following, we provide an overview of financial regulators in three countries (Germany, Belgium and Denmark). This contextual knowledge is necessary for understanding the dynamics of reputational threats and praises in different contexts. Financial regulators might differ with regard to their formal powers and formal autonomy from political oversight, which potentially impacts on how much leeway they have to react to reputational threats. Moreover, they are part of complex regulatory architectures, where various organizations perform different types of regulatory activities. Finally, financial regulators have regularly been subject to reform activities, both before and after the financial and sovereign debt crisis.

The German Financial Regulator: BaFin
The German financial regulator, BaFin (Bundesanstalt für Finanzdienstleistungsaufsicht), was established in 2002 as a result of a merger of three sectoral regulators for banking, insurances and financial services. The main argument was that changes in the financial sector demanded an integrated regulator able to address regulatory issues across these three branches. While the underlying argument was one of strengthening oversight by solving coordination problems through organizational integration, this argument was also situated in a discourse of regime competition between 'financial market places'. Given that the wider policy context in Germany was one of financial market liberalization, limiting regulatory burden was an important underlying motive.
The predecessor organizations of BaFin were semi-autonomous agencies (type 1), whereas BaFin is a legally independent agency (type 2) following the classification by Sandra van Thiel (2012). This classification will be used throughout this section, as it allows us to locate and compare agencies in different countries on a continuum of formal autonomy from political oversight. Legally independent agencies (type 2) operate at a greater distance from political executives and are delegated more managerial autonomy (and often also policy autonomy in substantial decisions) compared to semi-autonomous agencies (type 1).
Although BaFin was granted more formal powers compared to its predecessors, especially in terms of managerial autonomy and self-financing through fees collected from regulatees, the parent ministry's formal powers to instruct the agency, including single-case decisions, remained untouched. BaFin can thus be characterized as a hybrid organization, combining high levels of managerial autonomy (a type 2 agency) with low levels of policy autonomy, at least in formal-legal terms (Bach & Jann, 2010). However, the parent ministry struggled with keeping the newly created agency under control, which could leverage its international activities and its technical expertise to increase its actual autonomy (Ruffing, 2015).
BaFin is part of a complex regulatory architecture. The agency formally reports to the Ministry of Finance, which is the key policymaker in financial regulation. In terms of policy implementation, BaFin alone is in charge of supervising the securities and insurance sectors. For banking regulation, BaFin shares responsibilities with the Bundesbank (the central bank), which is in charge of on-site inspections. Although reform plans for concentrating banking supervision within the Bundesbank were tabled in 2009, this reallocation of responsibility never materialized (Handke & Zimmermann, 2012). This reform mainly had a symbolic character aiming to discipline BaFin, which was considered as acting too autonomously in international negotiations, and which had a rather assertive chief executive at that time (Jochen Sanio). The reform was therefore not a 'rational' response to inter-organizational coordination problems, which according to experts never played an important role during the crisis. FSMA is a legally independent agency (type 2) according to van Thiel (2012). FSMA (and CBFA before it) independently carries out the tasks in the general interest entrusted to it by parliament. FSMA mainly reports on its activities and decisions by means of its annual report and its website. The Chairman of the Management Committee, or the Management Committee as a whole, may be heard by the competent committees of the Belgian federal parliaments. Individual decisions taken by FSMA may be appealed before the administrative or judicial courts. FSMA's budget is paid for by regulatees, within the limits of and according to the specific rules determined by Royal Decree.
In 2011, the Belgian legislature opted to move from an integrated supervisory model to a bipartite supervisory model ("Twin Peaks"). In this model, the micro-and macro-prudential supervision of financial institutions is grouped at the central bank (National Bank), while supervision of the rules of conduct of financial institutions is entrusted to another supervisory authority (FSMA). Before, CBFA was responsible for the prudential supervision of financial institutions. FSMA acquired new competences related to the supervision of financial institutions' compliance with rules of conduct, supervision of distribution of financial products, and financial education of the public. Because of the restructuring, about half of CBFA's personnel moved to the National Bank, where they continued to fulfil the prudential supervision roles that were now performed by the National Bank.

The Danish Financial Regulator: The Danish Financial Supervisory
Authority (DFSA) DFSA was established in January 1988. It has around 297 full-time equivalents and a budget of approximately 40 million EUR (Finanstilsynet, 2015). DFSA is a semiautonomous agency (type 1) "without any legal independence but with some managerial autonomy" (van Thiel, 2012, p. 20, italics in original). It operates under the Ministry of Industry, Business and Financial Affairs. However, the minister has no hierarchical authority in relation to the agency's performance of its regulatory functions (which are regulated by law). There is an ordinary hierarchical relation bebetween the minister and the agency only in relation to the provision of information on the situation of the financial industry (Ministry of Industry‚ Business and Financial Affairs, 2015, p. 164). The tasks of DFSA consist of three main parts: (1) supervision of financial companies (banks, mortgage-credit institutions, pension and insurance companies) and markets through inspection and based on continuous reporting from the companies; (2) regulatory tasks by developing rules for regulating the financial area; and (3) information tasks by providing insights on the development within the financial sector in Denmark through statistics and information (Finanstilsynet, 2015, p. 4).
In the wake of criticism regarding the agency's lenient regulatory approach during the financial crisis, an independent governing board with seven members appointed by the minister was established in 2014. The governing board is tasked with strategic decision-making, for instance regarding strategic targets for supervisory activities and fundamental supervisory principles. Moreover, it adopts rules and guidelines in areas of delegated rulemaking and DFSA's annual report (Finanstilsynet, 2016). Hence, the board holds rather substantial formal powers in relation to the agency's management, which can be interpreted as a reduction in the agency's autonomy.
DFSA collaborates with the Danish central bank, Danmarks Nationalbank, which operates independently from the political system in Denmark (independence is regulated by law). This collaboration is formalized by two collaboration agreements (memoranda of understanding). The key purpose of the collaboration is to ensure financial stability. The collaboration is performed through various coordination committees. The memoranda are described as confirmations of the "pragmatic and informal" collaboration among the parties, explicating that the collaboration is performed in respect for the areas of responsibility in view of ensuring financial stability and does not "overrule" the involved authorities' areas of competence and responsibility (Danmarks Nationalbank, 2005, 2014.

Financial Regulators Compared
What are the main similarities and differences between these financial regulators? First, in terms of regulators' formal autonomy, we find that regulators in Belgium and Denmark are formally protected from political interference in their regulatory activities, which is not the case in Germany, despite BaFin's independent legal status. Second, in all three countries, we find complex regulatory architectures (see on the complexity of regulatory architectures Mathieu, Verhoest, & Matthys, 2017). However, all three agencies perform largely similar tasks regarding prudential supervision of financial institutions, at least until the Belgian regulator was fundamentally restructured in the wake of the financial crisis, losing almost half of its staff to the central bank. The Danish regulator was re-organized through the establishment of a board with considerable powers vis-à-vis the agency's management. Finally, the German regulator 'survived' the crisis unscathed.
What are the implications of those differences and similarities for the quantitative analysis of financial regulators and their communicative responses? Until now, reputation theory has been silent on the relationship between formal autonomy and reputation-seeking behaviour. Arguably, more formally autonomous regulators potenpotentially have more discretion in responding to reputational threats, and hence the German and Belgian regulators might be more likely to react than the Danish regulator. At the same time, the Danish regulator is formally protected from political oversight and therefore might have a similar leeway in responding to threats compared to its German and Belgian counterparts. All in all, we have no reason to believe that our results are driven by differences in the regulators' formal autonomy. This reasoning also squarely fits research showing that there is not necessarily a linear relationship between agencies' formal autonomy and actual autonomy (Bach, 2016). Likewise, all three agencies have largely similar tasks, which leads us to expect similar patterns of reputation management (Christensen & Lodge, 2018). We might however expect reputational threats and response behaviour in Belgium to be different after the structural reform, as the Belgian post-reform regulator has a different set of tasks, which may attract other types of reputational concerns.
Overall, the comparison of regulators and regulatory architectures indicates a high degree of comparability of the three cases. Hence, according to bureaucratic reputation theory, we would expect fairly similar patterns of how financial regulators respond to threats targeting their performative reputation.

Data and Methods
The data consist of coded news articles in three newspapers: Berlingske in Denmark, De Standaard in Belgium and Frankfurter Allgemeine Zeitung in Germany. All three are centre-right, quality newspapers published daily and distributed nationwide, and all three cover political and business news. To identify articles where the financial regulators' reputation was potentially under threat, we searched for the name of the regulator in databases of all articles published in the period from 1 st of January 2003 to 31 st of December 2015. After excluding letters, the total sample consists of 7,981 articles in the three countries.
We coded whether articles contained a threat to the financial regulator's reputation. The coding was based on a codebook developed by the Danish and Belgian researchers as part of a larger project on bureaucratic reputation (the Rep Gov project, see Boon et al. (2018)). In total, 984 articles (12.3% of all articles) contained an explicit threat to the respective agency's reputation. If articles had an explicit reputational threat, we further classified whether the threat was targeted towards the performative dimension of reputation or not. The performative dimension addresses judgments on the quality, efficiency and/or effectiveness of the services that are considered an agency's outputs and outcomes. These services can be a set of activities, policy instruments, but also an initiative, a program, or a report that is the final manifestation of the agency's core task delivered to society, to politicians or to other public actors (Carpenter, 2010b;Carpenter & Krause, 2012). 2 In addition, we coded agency responses to the articles where a threat was evident.
While this does not capture all the reputation management activities performed by the agencies in relation to their reputation in general and their media reputation in particular, it is a well-established way of studying reputation-seeking behaviour of regulators (Gilad et al., 2015;. How regulators are held accountable in the media, and how they respond to those allegations, will provide relevant insights into the drivers of bureaucratic behaviour, due to the relative importance of an organization's media reputation vis-à-vis other reputations. Moreover, the media serves as an important accountability forum, which will pick up agency relations with key audiences such as political superiors and stakeholders (Jacobs & Schillemans, 2016). In addition, generic reputation research suggests that an organization's media reputation is related to the reputation held by the general public, both in terms of salience and valence of specific organizational attributes (Carroll & McCombs, 2003;Deephouse, 2000).
To assess intercoder-reliability, three coders from Denmark, Belgium and Germany coded 40 articles on the British financial regulator (Financial Services Authority) in The Daily Telegraph. The intercoder-reliability coefficient (Krippendorff's alpha) for negative threats and performative threats were 0.80 and 0.74, respectively, and hence at an acceptable level.
To differentiate between before, during and after the financial crisis, we use country-specific dates. The dates are taken from an international comparison of the financial crisis in the European countries (European Systemic Risk Board, 2017). Hence, we operationalize the crisis period independently of our measurement of reputational threats. Table 1 lists the defining dates. In the figures below, we counted the threats towards the regulators on a monthly basis and calculated the relative share of articles in which the agencies face a reputational threat in the total sample of articles. Hence, our data do not merely reflect a higher salience of financial regulation during the crisis, but specifically measure (negative) attention targeted at the regulators. To better illustrate the development of threats across time, we inserted a smoothed, bold line that takes a running mean of the present month, three months before and three months after. In the tables, we compare the three time periods within the three countries, and even though we collected all articles mentioning the regulator, and hence have population data from 2003 -2015, we perform statistical significance testing as data might be prone to stochastic variation and/or measurement error, and to make predictions on further cases. In the appendix, we included country-specific figures on the threats over time and the logistic regression analysis the tables are based upon.

Media analysis of Reputational threats and agency responses
We start by briefly presenting descriptive findings on reputational threats over time in the three countries. Figure 1 provides an overview of the total and relative number of reputational threats per agency (see appendix for country-specific figures including absolute and relative numbers of threats). These are aggregated figures that do not distinguish between different reputational dimensions under threat. First, we find that in all three countries, the number of threats peaks at the onset of the crisis. In Denmark and Germany, the peak corresponds with the start of the crisis, while there is a delay in the Belgian case. Likewise, at the end of the crisis, the number of threats decreases in all countries. This decrease in the number of reputational threats is temporary, however, as we observe increases in the number of reputational threats in all three countries after the end of the crisis, albeit at overall lower levels.
That said, we also find that a relatively high level of (relative) threats persists in Belgium after the end of the acute crisis management phase. Those threats are to a substantial degree related to a controversy surrounding the reappointment of the FSMA's president who was widely perceived as a main responsible for the mishandling of the financial crisis. In contrast, the relative share of threats falls below crisis levels in Denmark and Germany. Yet these two countries' regulators, in particular the German BaFin, had a much more strained reputation at the outset of the crisis. After its establishment, BaFin quickly established a reputation as a 'harter Hund' ('tough dog'). Its regulatory style was perceived by the industry not so much as lighttouch, but rather strict, particularly by smaller public and cooperative banks. Hence, public criticism prior to the crisis primarily was about the agency's performative and procedural reputation. In contrast, the critique of DFSA prior to the crisis reflected allegations of a too reluctant regulatory role.
Another observation is that the crisis lasted for about the same time in all countries (71 months in Denmark and Germany, 61 months in Belgium). What varies, though, is the salience during the crisis (the number of articles mentioning the agency) (see Table  2): In Denmark, 60% of all articles identified and coded were published during the crisis; in Belgium (47.6%) and Germany (47.1%), these shares are considerably lower. 3 This suggests that the Danish regulator was under higher pressure during the financial crisis relative to the other periods and compared to the Belgian and German ones, which might also be reflected in how the Danish regulator responded to reputational threats under different contextual conditions (see below). Turning to the figures for the individual countries (see Figures A1-A3 in the appendix), we see for Denmark (A1) that the absolute number of threats peaks at the start of the crisis (one month with around 25 threats) and stays relatively high during the crisis period. Just before the crisis ends, the absolute number of threats peaks again and then decreases as the crisis is ending. At the beginning of the crisis the agency was, as already suggested, accused of taking a too reluctant regulatory approach towards the banking sector. In particular, criticism was raised in relation to the bankruptcy of two banks in Denmark. During the crisis, there was a change of agency head. At the end of the crisis, the agency head was described as 'a sheriff' who was accused of being too powerful and performing over-regulation.
In Belgium (A2), the number of threats peaks around one year after the crisis started -in 2008, the Belgian banks Fortis and Dexia were running into serious problems -and then stays at a relatively high level until halfway through the crisis when the number of threats is back at pre-crisis levels. This indicates a rather intensive threat level to FSMA that lasted for about two years.
As to the German regulator, BaFin (A3), the number of threats also peaks around the start of the crisis like in the Danish case. The criticism revolved around lenient supervision practices, a lack of highly qualified staff, and unilateral regulatory action, among others. The increase in threats towards the peak for BaFin is nevertheless not as steep as for DFSA in Denmark, which can be attributed to pertinent criticism by public and cooperative banks prior to the crisis, as well as a scandal regarding corrupt procurement practices. The reputational threats to BaFin were thus far more spread out in time compared to its counterparts in Denmark and Belgium.
Moving to the comparative analysis of reputational threats and agency reactions, we look at differences in reputational threats before, during, and after the crisis. Are regulators more exposed to criticism during the crisis compared to before/after the crisis? Moreover, do levels of reputational threats go 'back to normal' after the crisis? Table 2 compares threat levels between these three different periods, using three pairwise comparisons (before vs. during, before vs. after, during vs. after). In the tables, statistically significant differences are present if margins do not share the same letter.
Looking at all kinds of threats (without differentiating between reputational dimensions), we find the following statistically significant patterns ( Table 2): In Germany, the level of threats after the crisis is below the crisis and pre-crisis period, hence there is no increase in threat levels in connection to the crisis. In Denmark, we find a similar pattern, namely a drop of the level of reputational threats after the crisis both compared to the situation before and during the crisis. In Belgium, the pattern is different: Here, we observe an increase in threat levels with the onset of the crisis, yet other than that, we cannot find any significant differences. This confirms our preliminary observation of persistently high levels of reputational threats in Belgium after the crisis.
An important take home message is that major crises do not necessarily translate into reputational threats to sectoral regulators. But what can plausibly explain that reputational threats ebbed after the crisis in Denmark and Germany whereas they remained at crisis levels in Belgium (in terms of the share of negative articles, though not in terms of overall attention)? The reform of the Belgian regulator was completed before the end of the crisis, yet a major controversy was sparked surrounding the agency's chief executive, who was reappointed in 2013 despite being blamed for the poor handling of the financial crisis. Hence, even though the reform of the regulator can be seen as blame avoidance by politicians through "defensive reorganization" (Hood, 2011, p. 70), which blurs present day organizations' responsibility for past events, several stakeholders continued to hold the agency head accountable for mishandling the crisis. Note: Margins sharing the same letter are not significantly different at the 5% level. Margins without any letter are significantly different at the 5% level. This only applies to diachronic comparisons within the same country. Margins calculated from Table A1.
Source: own analysis.
We now look more specifically at the performative dimension of reputational threats ("did the organization do its job?") (first results column in Table 3). In all countries, the relative share of articles with performative threats relative to other kinds of reputational threats increases throughout the period. This increase is statistically significant in all countries when comparing before and during the crisis. In Denmark, the share of performative threats increases from 73.0% before to 84.6% after, in Belgium from 64.4% to 84.0% and in Germany from 55.9% to 70.4%. Moreover, we find that performative threats clearly dominate among all reputational threats throughout the entire observation period, especially in Denmark and Belgium. However, despite an overall increase, we find no statistically significant differences between crisis and post-crisis levels of performative threats in any of our three countries. This means that even after the crisis ends, performative threats persist. In Denmark and Germany, an important difference of post-crisis performative threats relative to pre-crisis performative threats is that a smaller percentage of articles also include threats to other dimensions after the crisis (second results column in Table 3). Put differently, in Denmark and Germany, reputational threats focus more exclusively on agency performance, whereas the Belgian regulator increasingly faces other types of threats after the crisis. In Belgium, causes of malfunctioning were often linked to (a) the regulatory architecture (which was changed in 2011); and (b) the politicized character of top officials in institutions within the regulatory architecture (both FSMA and National Bank). Finally, moving to the core of reputation theory, we analyse regulators' reactions to reputational threats (Table 4). We coded regulators' reactions if they were included in the same article containing a reputational threat. There are different types of possible reactions, including explicitly not commenting (despite being asked to do so by a journalist), explanatory statements, or blame shifting to others. We combined all of these categories into one common response category for the present analysis. Although we would have preferred to distinguish between agency reactions towards reputational versus other kinds of threats in our analysis, we are facing problems of low numbers of observations for agency responses. Therefore, we opted for an aggregated analysis of agency responses. We hence compare whether agencies remain silent when facing criticism as opposed to reacting to this criticism . The analysis indicates no statistically different change in reactions in Denmark over the three periods. This stability of responses is striking, given a clear increase in threats during the crisis. For Belgium, the regulator's responses both during and after the crisis are statistically different from before. During the crisis, the Belgian regulator reacted much less to reputational threats by using media communication than before the crisis.
Remarkably, this more reluctant attitude to communicate in media in response to reputational threats remained in place after the crisis. For Germany, we find that BaFin was less likely to respond to a reputational threat by means of communication during the crisis compared to before the crisis. After the crisis we observe again more reactions compared to during the crisis, but these are not statistically significant due to few observations after the crisis (n=12). Both before and after the crisis, we see the same pattern of reactions in Germany. Those findings raise interesting questions regarding the generalizability of bureaucratic reputation theory. Most importantly, we find that financial regulators in different contexts and under changing conditions (before, during, and after the crisis) react differently to reputational threats. Whereas regulators responded less to reputational threats during the crisis in Belgium and Germany, we do not observe such a pattern in Denmark. As argued above, the three agencies do not differ substantially in their formal autonomy nor their tasks. A possible explanation might be that the peak of the crisis simply did not allow using many resources for reputation management, yet this does not explain cross-country differences.
Another explanation is that this pattern was a more or less deliberate "keeping a low profile" blame management strategy during the crisis (Hood, 2011, pp. 58-62). This pattern was more pronounced in the Belgian case -and it apparently was not successful given the continuing exposure to threats in Belgium -where reputational threats remained at a higher level than in Denmark or Germany. Another plausible explanation is that both BaFin and FSMA were relatively new organizations (created in the early 2000s), which means that the period before the crisis was a time in which they attempted to position themselves and to carve out a distinct reputation. DFSA, in contrast, was created in 1988, and hence has had enough time to develop into a fullfledged institutionalized organization that has developed a distinct reputational profile guiding the agency's responses both in times of crises and under business-as-usual conditions.

Discussion and Conclusion
This article has drawn on media content analysis to explore the exposure and response of three European financial regulators to reputational threats. Our aim was to compare across these cases and over time, in particular to analyse the impact a crisis -the financial crisis -has on both the exposure to threats and regulators' responses. In doing so, we intended to bring a more dynamic perspective into debates about agency responses to reputational threats. The aim was also to shift attention to the dynamics of exposure to threats directed at different reputational dimensions, with a particular focus on performative reputation. Our analysis has confirmed some established claims of reputation theory, but also has revealed some unexpected dynamics.
The financial crisis has indeed put financial regulators into the line of fire as their exposure to threats peaked during the crisis, either early on (in Denmark and Germany) or with some delay (in Belgium). The significant increase of threats to the performative dimension of regulators' reputations (Table 3) points to the lack of regulatory effectiveness that the crisis exposed (see also Lodge & Wegrich, 2011). However, the post-crisis development in terms of performative threats shows some substantial differences between Denmark and Germany on the one hand and Belgium on the other hand. A possible interpretation is that regulators in Denmark and Germany have been able to use the crisis as an opportunity for defending and nurturing their reputation. The substantial post-crisis institutional reform in Belgium has evidently not contributed to a deflection of threats, which kept coming in at crisis level (in relative terms).
The overall observation regarding regulators' communicative behaviour is one of differential reactions to reputational threats to a core dimension of financial regulators' reputation. While we proposed an explanation at the organizational level above, with respect to the differences in responding during the crisis between Denmark on the one hand (no change in responses before, during and after the crisis) and Belgium and Germany on the other (less responses during the crisis compared to before), another tentative explanation is offered by the cross-national perspective, which relates to different national political-administrative cultures (Verhoest, van Thiel, Bouckaert, & Laegreid, 2012). Denmark can be characterized by a less legalistic Nordic tradition with a low power distance and low uncertainty avoidance. Denmark's political culture is rather consensual, having substantial experience with minority cabinets. Furthermore, Denmark is characterized by relatively trusting relations between political principals and agency heads and has a politically neutral bureaucracy. This stands in contrast to Germany and Belgium with strong legalistic traditions, comparatively high power distance and uncertainty avoidance, less minimum-winning coalitions, and relative distrust and high politicization (particularly in Belgium). Given the relatively hierarchical, low-trust and formal nature of political-administrative relations in Belgium and Germany, regulators in these countries are potentially more constrained in their external communication strategies. However, an empirical analysis of those assumptions is beyond the scope of this article.
Lastly, we should be careful with attributing too much weight to communication responses because (a) communication is not necessarily strategic (though one can expect that communication becomes more strategic during crisis times); and (b) regulators may display strategic behaviour other than by responding in the media. Those include the timing of decision-making (Carpenter, 2002;Maor & Sulitzeanu-Kenan, 2013) or the prioritization of tasks (Gilad, 2015). Moreover, reputational threats can also be overcome by non-strategic behaviour like a genuine effort to improve performance by adopting new internal working procedures, increasing technical expertise and enhancing cooperation with other regulators.
At the very least, our findings indicate that regulators respond differently to similar threats, which calls for a further refinement regarding the scope conditions of reputation-seeking behaviour. While we are not able to explore these configurations in detail, the implication for bureaucratic reputation theory is a call for methodological pluralism and contextual sensibility: it does not seem to make much sense to seek general statements about responses of regulators to threats independent from the national political and situational context.     Note: p: + < 0.10 * < 0.05 ** < 0.01 *** < 0.001. Standard error in parenthesis. Source: own analysis. Note: p: + < 0.10 * < 0.05 ** < 0.01 *** < 0.001. Standard error in parenthesis.