Organizational Reputation, the Content of Public Allegations and Regulatory Communication ABSTRACT INTRODUCTION

Transcription

1 Organizational Reputation, the Content of Public Allegations and Regulatory Communication Sharon Gilad, Moshe Maor and Pazit Ben-Nun Bloom The Hebrew University of Jerusalem ABSTRACT How does the content of public allegations impact upon regulatory communication strategies? Which type of allegations pose a higher threat to agency reputation, and how do agencies manage these threats via communication? Based on a comprehensive dataset and quantitative content analysis of the Israeli banking regulator s responses to public expressions of opinion between 1998 and mid-2009, we demonstrate this agency s greater propensity to acknowledge problems when faced with claims that regulation is overly lenient, as opposed to claims of over regulation, and regarding functional areas with regards to which its reputation is weaker. These findings, while based on one institution, are important because they demonstrate how an agency s assessment of the relative threat to its reputation is implicated in distinct communicative patterns across functional areas. They also suggest that external audiences can shape agency attention and response by carefully framing their claims in light of their understandings of agencies distinct reputational vulnerabilities. INTRODUCTION How does the content of public allegations impact upon regulatory communication strategies? Which type of allegations pose a higher threat to agency reputation, and how do agencies manage these threats via communication? Recent accounts look at regulatory agencies strategic crafting and management of publics external perceptions of their roles, capabilities and performance, namely: their organizational reputation. The scholarship focuses on the way in which regulators concerns over reputation shape their decision making processes (e.g. time to drug authorization) and outputs (e.g. the accuracy of macroeconomic predictions). In his generalized portrayal of regulators as constructors and protectors of a unique organizational reputation, Carpenter (2010) has summed up the essence of this reputation-oriented perspective in the following way: The lesson of this scholarship is that, when trying to account for a regulator's behavior, look at the audience, and look at the threats (Carpenter 2010b, 832; italics in original). This article seeks to contribute to this growing body of scholarship in the following important ways. First, it focuses on regulators communication strategies, a terrain that has been neglected by previous studies of bureaucratic reputation. Second, it analyzes how regulatory communication 1

2 strategies are shaped by regulators simultaneous management of different types of threats that are imposed upon them by their multiple audiences. The article builds on our recent contribution to the field of bureaucratic reputation, in which we examine how an agency s communication patterns vary given its multidimensional reputation across functional areas (Maor, Gilad and Ben-Nun Bloom, working paper, 2012). We argue there that an agency is likely to keep silent when facing allegations over issues regarding which it generally enjoys an established positive reputation, and to respond to opinions about core functional areas with regards to which its reputation is poorer or still emerging. These findings are important because they indicate that agencies sensitivity to and management of - external opinions is shaped by their understanding of their distinct and multidimensional reputations. They also demonstrate that words are action, and, occasionally, so is regulatory silence. That study, however, analyzes regulatory response as a binary choice between silence and talk while avoiding more elaborated analysis of the content of regulatory response. This article delves into the nuances of regulatory response to audiences assessments of their performance. We contend that the content of external opinions, that is, the type of allegations, should have a profound effect on whether the regulatory agency will keep silent, deny a problem altogether or admit there is a problem and either shift blame to others or assume responsibility. In particular, we focus on regulatory differential reaction to audiences claims of under- and over regulation. 1 By over-regulation we mean audiences claims that regulatory standards, processes, or behavior impose an excessive burden on the economic, professional or societal field under regulatory jurisdiction. By contrast, allegations of under-regulation proclaim that regulatory standards, processes or behaviors are overly lenient towards the regulated sector, and fail to adequately protect the public interest. Such allegations may focus on the weakness of existing standards, on failure of the regulator to enforce existing regulation as well as allegations of its capture in the hands of business to the detriment of the general public. Based on quantitative media-content analysis of the Israeli banking regulator s responses to opinions and criticism from 1998 to mid-2009, we find that this agency s 1 Of course, not all audiences claims embody suggestion of over or under regulation. Nonetheless, criticisms that target a regulator s efficiency, effectiveness, expertise, competence, operation and so on often involve a claim that regulation is either excessive or overly lenient. 2

3 choice between silence, problem denial and problem admission is shaped by the content of the allegations that it faces. When audiences claim that regulation is overly lenient (namely: under-regulation), the regulator is relatively inclined to admit a problem rather than either keep silent or deny a problem altogether. In addition, extending our previous work (Maor et al., 2012), we show that the regulator is more inclined to admit a problem when facing allegations regarding a functional area wherein its reputation is poor (consumer protection) and to either keep silent or deny a problem regarding an area wherein it enjoys a positive reputation (prudential regulation). Finally, we further find that the regulator s choice between silence, problem denial and problem admission is shaped by the context of allegations, namely their salience and the objective severity of the underlying incidents that they refer to. We conclude that regulatory choice of response is shaped by the content of allegations and the relative threat that they pose to the agency s distinct reputation. We argue that certain types of claims are inherently more threatening, and therefore more likely to induce an acknowledgment of a problem. More specifically, we argue that claims of under-regulation pose a greater threat to regulatory reputation because they imply agency failure to accomplish its distinct core mandate for protecting a specific public value (e.g. protecting the public from unintended side effects of dangerous drugs). By comparison, claims of over regulation refer to the negative externalities of regulation (e.g. firms loss of profits that could have been made had a drug been authorized faster; or patients continued suffering, which could have been mitigated by a new drug pending authorization). An alternative explanation, which we cannot exclude on the basis of our case study, is that the relative threat that claims of under vs. over-regulation pose to agency reputation is context specific. Namely, given the public s mood and media discourse in a specific time and place, some type of allegations can be effectively ignored or denied, whereas others have to be acknowledged. These findings offer several new contributions to the study of bureaucratic reputation and communication. The first is the importance of appreciating that regulators are more/ less sensitive to certain audience allegations given their content. This finding challenges the assumption that agencies are equally responsive to all external signals (Noll, 1985) and that they simply pursue an equilibrium wherein positive feedback is maximized and adverse feedback is minimized. It also means that the way in which audiences choose to frame their allegations e.g. as a problem of 3

4 over vs. under regulation - matters for the content of regulatory response. The second and related implication is that an agency s established, and distinct, reputation whether good or bad mediates its sensitivity and reaction to public allegations. Third, and more particular, the findings presented here suggest that regulatory communication strategies are crafted with a close attention to the agency s distinct reputation and to the consequent relative threat that allegation content and context poses to agency reputation. These latter findings contribute not only to the study of bureaucratic reputation, but further challenge the assumptions of the economic literature on central banks communication, which overlooks the political, agencyregarding, as opposed to task-related, aspects of this phenomenon. The article proceeds in four sections. The first briefly discusses recent advances in the study of bureaucratic reputation management. The second provides a more detailed discussion of the research questions and hypotheses. The third section describes the selected case and the research design. A fourth section presents the results, and the final section concludes. BUREAUCRATIC REPUTATION AND DIFFERENTIAL RESPONSE TO EXTERNAL THREATS The scholarship on bureaucratic reputation and analyses that build on this perspective (Joskow 1974; Kettl 1986; Carpenter, 2001, 2002, 2004, 2010; Carpenter and Krause 2012; Carpenter, Zucker and Avron 2008; Carpenter, Chattopadhyay, Moffitt and Nall 2012; Douglas and Krause, 2005; Gersen and O Connell 2008; Maor, 2007, 2010, 2011; Maor and Sulizeanu-Kenan, forthcoming; Moffitt, 2010; Whitford 2005; Yackee and Yackee 2010) emerged as a challenge to the assumption of institutional-choice theory that in the absence of exogenous political control bureaucrats will exploit their office in pursuit of their narrow self interests (Horn, 1995; Huber and Shipan, 2002; McCubbins and Schwartz, 1984; Wood and Waterman, 1991). The reputation-centered challenge to institutional-choice theory builds upon an earlier stream of research that proposes that bureaucracies carefully calibrate their response to multiple external signals from their environment in pursuit of maximizing overall support and minimizing negative feedback (Noll, 1985; Olson, 1995, 1996, 1997). This earlier approach suggests that holding budget constrains constant, public agencies are equally responsive to all external signals irrespective of their content and context. In contrast, a reputation-centered perspective suggests that 4

5 agencies pay uneven attention to external signals given their assessment of the relative threat to the agency s distinct reputation (Gilad, 2012). Before elaborating how reputation shapes agencies differential response to external signals and demands, we need a workable definition of this concept. In the broadest sense, reputation regards audiences association between an organization and distinctive traits (Carpenter, 2010; Fomburn, 1996, forthcoming). That is, reputations, like perceived personal characteristics, are unique, and thereby differentiate an organization from other similar entities. Organizational reputations are formed in interaction between an agency s claims regarding its roles, capabilities, performance and so forth, and its multiple audiences judgments of these claims (Carpenter, 2010; Gilad, 2008). This linkage between an organization and its perceived traits, once established, is relatively stable (cf. Albert and Whetten, 1985), although, needless to say, good reputations can be tarnished and bad reputations can be repaired. Finally, and significant for regulatory agencies that are the subject matter of our study, organizational reputations are multi-dimensional (Carpenter, 2010: 35; Maor 2011). One aspect of this multidimensionality is that audiences judgments of an agency may simultaneously regard its output performance, the expertise of its staff, its values and/or the legality of its processes (Carpenter, 2010). Another aspect of multidimensionality is that agencies reputations may vary across functional areas (Maor et al., 2012). And finally, as we emphasize in this article, audiences may differ in their normative assessments of a regulatory agency as either under or over regulating the industry under its jurisdiction. So how does reputation shape agencies behavior, and their uneven attention and responsiveness to external signals and demands? Current studies show that agencies management of risks to their reputation is embedded in their decisionmaking processes (Carpenter, 2002, 2004; Gilad, 2009; Gilad and Yogev, forthcoming; Maor and Sulizeanu-Kenan, forthcoming) and outputs (Douglas and Krause, 2005; Maor, 2011). Few studies further investigate how reputational concerns shape agencies endogenous structure of their jurisdiction or domain (Gilad, 2008; Maor, 2010). The contrasting findings of two studies, both regarding the American Food and Drug Administration s (FDA) decision-making processes, are particularly noteworthy. Carpenter s (2002) empirical analysis of FDA processing of businesses applications for drug approval demonstrates this agency s tendency to delay its decisions, and to require multiple testing for potential drug adverse effects, so as to 5

6 safeguard its reputation as guardian of the public s safety. It is only when the salience of relevant diseases is high, and when the agency faces intense interest-group demand to expedite its decision making, that the FDA is inclined to shorten its time-toauthorization. Conversely, in the context of enforcement, Maor and Sulizeanu-Kenan (forthcoming) find that the FDA s normal approach is to quickly issue formal warnings to firms following an investigation. It is only when media valence of the agency is positive and agency profile salient that the FDA allows itself more time to consider a case before issuing a formal warning. Hence, taken together these studies demonstrate, but do not acknowledge, the FDA s differential management of external signals across tasks. Namely: its relative disregard for businesses and patients demand for quick drug approval, alongside its hyper responsiveness to public demands for swift enforcement. Put differently, the findings of these studies suggest that agencies manage simultaneous threats to their multidimensional reputation by means of distinctive procedural patterns. Building on the above studies, the focus, and intended contribution, of this article is on agencies simultaneous management of multiple threats to their multidimensional reputations via communication. We are interested in agencies complex management of different allegation content, while accounting for the changing context of allegations and the agency s low or high reputation reserves across functional areas. But why focus on communication? First, it seems patent that communication is a key terrain wherein agencies manage threats to their reputation, alongside their adaptation of processes, outputs and claims of jurisdiction. To the best of our knowledge, Maor et al. (2012) is the first systematic investigation of how agencies reputational concerns shape their choice between silence and response to external allegations. In this article we build on that work to further investigate the logic underlying agencies content of response; i.e. their choice between silence, problem denial and problem admission. By focusing on communication we are able to analyze agencies simultaneous management of multiple audiences claims of under and over-regulation, across regulatory functional areas, while controlling for various contextual factors. Second, our dual focus on reputation and communication further contributes to the scholarly debates in economics on central bank communication (for a survey, see Blinder et al. 2008). To date, this literature focuses almost exclusively on central bankers use of communication as an integral means of the regulatory task, while 6

7 overlooking the agency-regarding aspects of communication. Key empirical questions and findings are derived from the desire to identify the optimal communication strategy in terms of its contribution to the effectiveness of monetary policy (e.g., how central bank talk increases the predictability of its actions, hence reducing volatility in financial markets). The very few contributions in this area that acknowledge the political value of central banks communication strategies revolve around the concept of central bank transparency (Faust and Svensson 2001), assuming that the more independent a central bank is, the more accountable, and hence, transparent it should be (Blinder et al. 2008). This assumption is based on the claim that independent agencies have a compelling need to account for their actions (Blinder et al. 2008, 8). A key limitation of this research dealing with optimal levels of transparency (e.g., Cruijsen and Eijffinger 2007; Cukierman and Meltzer 1986; Geraats 2002) is its focus on central bankers talk with financial markets, to the exclusion of the general public, and on the variance in communication strategies either across central banks or across time, rather than across functional areas (e.g., monetary policy, fiscal policy and banking supervision). Our article contributes to existing studies of central bank communication by emphasizing that communication has an independent role to play in government agencies attempts to actively construct the way in which their multiple audiences judge them. The argument advanced in this article offers a reputationcentered explanation for variation in central bank responses to multiple audiences and their diverse allegations. In the following section we draw concrete hypotheses regarding agencies communicative response to external signals, in the form of allegations in the public domain, given their relative threat to agency reputation. MANAGING MULTIPLE THREATS TO REPUTATION VIA COMMUNICATION Our analytical framework focuses on the shaping of communication as a reputationmanagement strategy. It sets out from an assumption that regulators are generally rational and politically conscious organizations that cautiously construct and protect their unique reputations (Carpenter 2001, 2004, 2010a, 2010b; Gilad 2009, forthcoming; Gilad and Yogev forthcoming; Heimann 1997; Krause and Douglas 2005; Maor 2007, 2010, 2011; Maor and Sulitzeanu-Kenan forthcoming; Quirk 1980; Whitford 2002; Rourke 1984; 1989; Wilson 1989). Organizational reputations are valuable political assets they can be used to generate public support, to achieve 7

8 delegated autonomy and discretion from politicians, to protect the agency from political attack, and to recruit and retain valued employees (Carpenter 2002, 491). Reputation uniqueness according to Carpenter (2001, 5) refers to the demonstration by agencies that they can create solutions (e.g., expertise, efficiency) and provide services (e.g., moral protection) found nowhere else in the polity. Our categorization of regulators communication choices builds on the work of Christopher Hood and his colleagues (Hood, 2011; Hood et al. 2009), businessmanagement analyses of corporations impression management strategies (Elsbach, 1994) and on our own previous work (Maor et al. 2012). We differentiate between three forms communication that regulators may adopt: silence (i.e. non response), problem denial (i.e. rejecting the allegation that regulatory standards, actions or state of affairs are harmful to a legitimate public value or interest) and problem admission (i.e. agreeing that current standards, actions or state of affairs are problematic, and either shifting responsibility to others or accepting responsibility). 2 Following our analysis in the previous section, and the categorization of regulatory communication choices, the article focuses on the following concrete research questions: How do the content and context of public allegations impact upon agencies choice between silence, problem denial and problem admission? How does an agency s distinct, and multi-dimensional, reputation impact upon its choice between silence, problem denial and problem admission? The content of allegations The key innovation of this study is its investigation of whether the content of public allegations, in and of itself, impacts upon regulators perception of the threat to their reputation and their subsequent response. If we find that some allegations are inherently more threatening due to their content, while controlling for their context (i.e. salience, severity and so forth), this would have important implications for our understanding of audiences sources of influence in the public domain. It would mean that the way in which audiences frame their claims impacts on an agency s response, at least on the level of communications. 3 It would also substantiate the expectation 2 It should be noted that Hood et al. (2009) use this categorization in order to study ministers sequential responses to discrete scandals. We, by contrast, build on this categorization to study bureaucrats routine management of their reputation, as well as their response to scandals. 3 Of course, audiences framing of an issue is not wholly flexible. Nonetheless, we assume that at least some issues can be framed in different ways. 8

9 that agencies are not equally responsive to all external signals, but rather that they give greater weight to certain types of allegations. In particular, we examine, for the first time, how agencies react to claims of under versus over regulation. 4 For instance, pensioners low returns on their pensions can be framed as a consequence of regulators under-regulation of financial firms fee structures. Alternatively, it might be claimed that overburdening, and superfluous, regulatory requirements results in financial firms deduction of high fees from pensioners funds. We focus on this binary distinction between over and under regulation because it allows us to parsimoniously categorize a wide spectrum of allegations that are targeted against regulatory agencies. 5 We hypothesize that claims of under regulation are likely to be more threatening to regulators reputations, and to therefore result in their greater inclination to admit a problem rather than either keep silent or deny a problem altogether. This hypothesis rests on two arguments. First, allegations of over-regulation refer to the negative externalities of regulation. By comparison, arguments of under-regulation are directly targeted at the agency s fulfillment of is core function of protecting a specific public value. Namely, they directly threat the agency s claim for a distinct contribution to the public good (i.e., agency s unique reputation). Second, in certain domains, the harm of over-regulation is visible and salient only to concentrated groups (regulated business), whereas underregulation has greater resonance with the public at large. Consequently, in terms of reputation, claims of under-regulation pose a greater threat, although they may 4 Maor and Sulizeanu-Kenan (forthcoming) sought to investigate the implication of over vs. under regulation for FDA time-to-enforcement. Yet, in practice, they found that public allegations, in that context, exclusively point to the problem of under regulation. 5 The reasons claims of over- and underreagulation are employed here lies in the plasticity of these concepts and their multifaceted character. At the outset, under vs. over-regulation may be an objective matter (i.e. at any one time an agency excessively enforces some standards vis-à-vis some regualtees, while acting leniently towards others) but may also be an issue of norms and cognition (i.e. what one person perceives as under regulation another perceives as over regulation). Moreover, different audiences may strategically choose to frame an issue in terms of over vs. under-regulation in order to enhance their influence. In addition, claims of over- and unde-regulation may be simultaneously directed at a regulatory agency because an agency may objectively use some of its powers excessively and others leniently (Natapoff 2006, 1718). Once one splits a regulatory agency s audiences into distinct classes, overregulation may be aimed primarily at one class of regulatees while underregulation may be seen as harm by the other class of regulatees. For example, overenforcement aimed at firms that do not comply with regulatory standards may result in allegations of overenforcement, while underenforcement may be seen as harm suffered by the general public, hence the allegation directed at the agency of underenforcement. These audiences may be furthermore split into two because of the wide spectrum of both types of allegations. Either over- or underreaction may be perceived by some as positive and by others as negative. For example, regulatory restraint in the face of (minor) regulatory non-compliance may be perceived as positive by the regulatees and as negative by those who are directly affected. Overall, the broad spectrum of agency activity exposes the agency to multiple audiences whose response to agency actions is not easily predictable. 9

10 actually imply lower threat to mobilized collective action against the agency. Thus, we hypothesize that - H 1a : when facing claims of under regulation a regulator will be relatively inclined to admit a problem (in comparison with either silence or problem denial). H 1b: when facing claims of excessive regulation a regulator will be relatively inclined to deny a problem (in comparison with either silence or problem admission). The context of allegations Although our focus in this article is on the consequences of allegation content, we expect the context in which audiences make their allegations to further shape regulatory communication strategies. In particular, we focus on two aspects of context: Salience: In line with previous studies, we expect visibility to render agencies more sensitive to external judgments of their performance (Carpenter, 2002; Maor and Sulizeanu-Kenan, forthcoming). Agencies have less of an opportunity to remain silent when media salience is higher. They get asked about an issue more frequently, and the absence of a response becomes an issue on its own with each additional choice to remain silent (Maor et al. 2012). Consequently, in general we expect an agency s inclination to admit a problem to increase with media salience (Cf. King, 2008). Thus, we expect that when media coverage is low an agency will tend to remain silent. When media coverage is intense, we expect an agency to experience a compelling need to admit a problem and to provide more substantive account as to who is to be blamed and/or what should be done. Finally, all else being equal, we expect that when media coverage is middling, the agency will be compelled to forgo silence, yet deny the problem altogether. We therefore hypothesize as follows: H 2a : under medium-level salience an agency will be inclined to deny a problem altogether rather than either remain silent or admit a problem (in comparison with either silence or problem admission). H 2b : under high-level salience an agency will be inclined to admit a problem (in comparison with either silence or problem denial). 10

11 Severity: The objective severity of an issue is also likely to shape the level of threat to reputation that an agency experiences. For example, Carpenter (2002) finds that the FDA rate of drug approval is faster when a new drug is intended for a more severe disease and/or in the absence of an alternative drug on the market. Of course, the severity of issues is, in part, a social construction. That is, just as much as more serious allegations tend to attract more media attention, the perceived severity of an issue is partly constructed by media-coverage intensity. Nonetheless, the catastrophic outcome of some events renders them objectively more severe. Consequently, our third hypothesis is the following: H 3 : when the event underlying public allegations regards an incident with catastrophic outcomes, an agency will be inclined to admit a problem rather than either keep silent or deny a problem altogether. Agency distinct reputation reserves Last, we expect that the relative threat that a regulatory agency experiences is either attenuated or augmented by its understanding of its reputation reserves, namely its established positive or negative reputation. This concept may be related to Maor and Sulitzeanu-Kenan s (forthcoming) distinction between reputationally-concerned and reputationally-relaxed agencies. 6 This expectation underlies Carpenter (2001, 2002, 2010) and others stress that a positive reputation is an asset that protects an agency from political attack on its autonomy. It also builds on experimental research, which shows that a company s poor reputation results in respondents higher attributions of blame for a crisis event whereas a positive reputation attenuates blame attributions (Coombs and Holladay, 2006; Klein and Dawar, 2004; Vanhamme and Grobben, 2009). In addition, few studies analyze the extent to which organizations reputation reserves actually shape their differential reaction to external pressures. King (2008) shows that companies that experience reputation decline in the years preceding a boycott are more likely to surrender to boycotters demands. Companies with a 6 Maor and Sulitzeanu-Kenan opens the possibility of studying the variance in baseline reputational sensitivity from which exogenous positive or negative shocks may shift outcomes. For example, according to their findings, enforcement decision duration is typically short for periods in which media coverage of the agency regulatory activity is low. This pattern suggests that FDA enforcement divisions are basically reputationally concerned, and thus, do not shorten their enforcement decision duration under reputational challenges, but rather extend the time for taking action when their reputation is visibly praised. The lack of reputational reserves over enforcement issues may attribute to precisely such a response. 11

12 strong and stable reputation prior to the boycott were inclined to endure such pressure, even under intense media coverage of boycotters demands. Our previous study (Maor et al. 2012) similarly suggests that agency choice between silence and public response is a function of the strength of its reputation reserves across functional areas. Our analysis here further explores how an agency s differential reputation across tasks impacts its content of response. Following our previous study, we expect an agency to remain silent when facing allegations about functional areas regarding which it enjoys long-standing positive reputation, and to admit a problem and provide a substantial account when facing allegations about domains regarding which its reputation is already poor. We further expect to find that an agency will be inclined to deny a problem altogether, rather than either keep silent or admit a problem, when facing intense media coverage of allegations regarding an area wherein it enjoys a positive reputation. Thus, we hypothesize: H 4a : when facing allegations regarding a functional area for which its reputation is poor, a regulator will be inclined to admit a problem (in comparison with either silence or denial). H 4b : when facing intense allegation regarding a functional area for which the agency enjoys a positive reputation, a regulator will be inclined to deny a problem altogether (in comparison with either silence or admission). THE ISRAELI BANKING SUPERVISION DEPARTMENT AND ITS DISTINCT REPUTATION The Israeli regulator of banks the Banking Supervision Department (hereafter: the BSD) - is a statutory unit, located within the Bank of Israel, which is an independent central bank. The Supervisor of Banks, who heads the BSD, is appointed by the Governor of the Bank of Israel (Banking Ordinance, 1941, clause 5). Operating alongside the BSD are four other financial-markets regulators in Israel: the Supervisor of Capital Markets and Savings, the Supervisor of Insurance, the Securities Authority, and the Antitrust Authority (which is entrusted with the promotion of competition in the economy). The BSD s mandate is based on several laws. 7 Yet its main tasks have not been defined by law. Consequently, it has defined its role in its annual reports (Ben-Bassat 7 The Banking Ordinance, 1941; the Banking (Licensing) Law, 1981; the Banking (Services to Consumers) Law, 1981; the Bank of Israel Law, 1954, Cheques without Cover Law, 1981, and the Law 12

13 2007, 24). It perceives the maintenance of the banking system s financial stability (i.e. prudential regulation) as its primary goal, and consumer protection and competition as its secondary goals (Ben-Bassat 2007, 25). Regarding the sources of the BSD s reputation, of the aforementioned five regulators, the BSD enjoys a near perfect (external) independence vis-à-vis central government and its political executives. In terms of material resources, it is ranked second (after the Securities Authority) with respect to its relative wealth (Ben-Bassat 2007, 43). Being an integral part of the Bank of Israel is no less important. According to Maman and Rosenhek (2011, 2): Over the course of an incremental process that began in the mid-1980s, the Bank of Israel has succeeded in positioning itself at a pivotal position in the political-economic field and within the state s apparatus [ ] An important source of the Bank of Israel s power has been its success in positioning itself in the field as an apolitical agency with the authoritative voice of expertise in economic matters, an expertise which is based on consensual scientific knowledge whose veracity is taken for granted (p. 2). Being institutionally located at the local-global interface, the bank was able to instill and embed global logics (Maman and Rosenhek 2011, 141), especially with regard to the soundness of the Israeli financial system. The fact that the Israeli economy emerged relatively unscathed from the global financial crisis (International Monetary Fund, 2011) reinforced the strong reputation for maintaining banks stability that the Bank of Israel and its Banking Supervision Department already enjoyed. However, the BSD s reputation for maintaining the stability of banks has been attained at the expense of its reputation for the performance of other tasks. Some of the BSD s key audiences perceive its focus on prudential regulation as valuable per se, yet detrimental and rival to its performance in other domains. Leading Israeli economists criticize the BSD for failing to promote competition in the highly concentrated Israeli banking sector. They condemn it for encouraging mergers between small and large banks in order to increase the stability of the banking system at the expense of inter-bank competition (Ben-Bassat 2007, 114; Knot 1986; Rotenberg 2002; Yosha, Blei, and Yafeh 2007). However, this criticism arguably for Encouragement of Competition and Reduction of Conflicts of Interest in the Israeli Capital Markets,

14 presents only a moderate threat to the BSD s reputation since promotion of competition is primarily entrusted with the Anti-Trust Authority (Maor et al. 2012). A greater threat to the BSD s reputation comes from its perceived deliberate under-regulation, and poor performance, in the domain of consumer protection. The Israeli Consumer Council, politicians and journalists all tend to be critical of the BSD s performance in this domain. For example, the recommendations of a working party, nominated by the Knesset s Economics Affairs Committee, published in 2004, suggested that, The Supervisor of Banks is responsible for goals that are not necessarily complementary. On the one hand, ensuring the financial stability of the banking system, which depends on banks profitability and prices. On the other hand, the Supervisor should protect and promote consumers interests Experience suggests that this duality of goals is detrimental to consumers (4). A subsequent parliamentary enquiry (2007) asserted: The enquiry has established that in the last few decades the Supervisor has given excessive weight to the [financial] stability interest, whereas the consumer interest has received partial attention (12). Both the working committee (2004) and the parliamentary enquiry (2007) recommended that the BSD s consumer protection function be reassigned, either to another independent unit within the Bank of Israel, or to a new regulatory agency. It is difficult to think of a stronger manifestation of the BSD s poor reputation for consumer protection, and the threat that this poses for its delegated autonomy, than these declarations by its overseeing Knesset committee. Consequently, if our aforementioned hypotheses are correct, we should find that the BSD is relatively inclined to admit (as opposed to keep silent or deny) problems concerning consumer protection regulation. DATA AND METHODS To test the above hypotheses, we constructed a database of opinions regarding the Banking Supervision Department. The aim of our database construction strategy was to produce a comprehensive list of opinions for the relevant years. To do so, we collated all the articles mentioning the BSD, either in the title or body of the article, that were published in Globes Israel s only dedicated business broadsheet between 1998 and mid-2009 (a total of 3,849 articles). This media outlet was a critical agendasetter for other media during the period under consideration. The first author and a research assistant read through the whole article population, and selected all articles involving opinions whether positive or negative about the BSD and its 14

15 performance. Opinions were either those that a source made during an interview, or reports by journalists of an actor s opinion made elsewhere (e.g., a Knesset committee, public report, conference, etc.). Our assessment of what makes an opinion vs. description intentionally favored inclusion over exclusion of articles. This process resulted in a database of 524 articles involving one or more opinions, rather than mere descriptions of facts or events. At a later stage, while coding each article, we created a separate row for each opinion source or opinion matter, so that our final database includes 634 opinions or rows. Each article was coded by a research assistant under close supervision of the second author (who herself coded 85% of the articles and 100% of those articles that the research assistant coded as involving BSD response). In unclear cases, the second author consulted with the first author and together they reached a decision. Finally, we asked a different research assistant, who had not been previously involved with the project to independently code a sample of the data (n=118). 8 Dependent variable The dependent variable is a categorical variable indicating the content of the BSD s response to opinions about its performance. We code the content of BSD response, as reported in the media, either directly to journalists queries or in response to audience opinions in a different venue (e.g. BSD s response to a Member of the Knesset during hearings at Knesset committees). Following Hood et al. (2009) we initially coded four types of responses: (i) no response inasmuch as there was no BSD response to the opinion/ allegation in the same article 9 (n=511, 81%); 8 The inter-coder reliability statistic (Krippendorf-Alpha) was above.7 for all relevant variables (ranging between.714 and.735), other than for our coding of the regulatory functional area (see below) where the reliability score was lower than warranted (α=.44). We think that the latter is not a reflection of the quality of our coding, but rather of the fact that the coding of this specific variable requires some expertise in financial regulation. Consequently, the second author, who specializes in financial regulation, verified 100% of the coding for this specific variable. 9 Our coding does not account for the possibility that the Supervisor response was reported in a different article or during the days following the original publication of an opinion. We have explored the frequency of such delayed response, during 3 days following an opinion, for a random sample (10%) of the Supervisor s non-responses to negative opinions. We find, for this 10% sample, that in 13.5% of cases the Supervisor has responded during the 3 days following the original article. While this rate of delayed responses is not insignificant, we have decided that it is not sufficiently large to justify exploring the possibility of delayed response for all of our cases. Moreover, it is noteworthy that rather than explicitly reporting a delayed response, journalists typically reiterate the original opinion and a response to it. Thus, delayed responses are not very frequent, and they are incorporated into our dataset as independent responses to subsequent reporting of an opinion. 15

16 (ii) problem and responsibility denial 10 (n=79, 12.5%); (iii) admission of problem and shifting of blame 11 (n=34, 5%); and (iv) admission of problem and acceptance of blame 12 (n=10, 1.5%). As evident from this distribution silence is the BSD s predominant response. In addition, because there are very few cases of blame admission (n=10) we are compelled to integrate categories III and IV into one category of problem admission. Independent variables Our analysis focuses on the effects of three key factors: the content of allegations (the level of regulation), the context of allegations (salience and severity) and regulatory reputation reserves across functional areas. The level of regulation. Each opinion was coded to reflect the source s opinion that regulatory standards and/or their enforcement are overly lenient (n=178, 28%), adequate (n=152, 24%), excessive (n= 180, 28%) or mixed (n=51, 6%). In addition, in 83 cases (13%) we were unable to code the opinion as relating to the level of regulation. Media salience of the Banking Supervision Department. For each article, we computed the number of all other articles that were published about the BSD during the relevant month. These figures were drawn from our greater database of 3849 Globes articles mentioning the BSD. We test the impact of this variable below and over its median value. Severity of the incidents underlying the allegation. To test the impact of severity, we constructed a binary measure of all incidents where a bank collapsed ( 1 = opinions regarding incidents of banks collapse; and 0 = all other allegations). 13 Bank collapses are a good proxy for severity given their direct impact on shareholders, customers and their indirect impact on investors confidence in the 10 This category includes any of the following: full denial of the event or harm done, denial of the seriousness of the event or harm, claim that the action taken (or inaction) is fully justified given the public interest, professional considerations and/or international practice. 11 This category includes any case wherein the agency acknowledges that the public s interest has been harmed, yet suggesting that force majeure or someone else is to be blamed for this bad outcome. 12 This category involves cases wherein the agency acknowledges that the public s interest has been harmed and either assumes full responsibility (e.g. declaring the agency s head s resignation, the initiation of an internal enquiry or agency implementation of remedial action) or offers excuses and explanations to partially reduce the level of blame. 13 Two banks collapsed during the research period. One of them - the Commercial Bank collapsed as a consequence of fraudulent behavior by a bank employee. The second bank the Industrial Development bank collapsed due to a high rate of bad debt resulting from its imprudent creditallocation policy and consequent bank run. 16

17 banking system. Of the 634 articles in our dataset, 67 pertain to incidents of banks collapsing. Agency reputation reserves. As detailed in the previous section, based on secondary sources (academic literature and Knesset enquiry reports), the BSD is known to have a strong and positive reputation for prudential regulation and a weak reputation for carrying out consumer protection and competition regulation. In addition to these areas with regards to which the BSD has an established (good or bad) reputation, other type of issues/ functions came in our media content analysis. Ultimately, we coded each opinion as relating to one of the following functional areas: (1) prudential: opinions regarding the BSD s capital adequacy requirements and its supervision of banks financial stability (n=213); (2) consumer protection: opinions regarding the BSD s regulation of banks services and sales to retail customers and investors (n=161); (3) competition: opinions regarding the BSD s regulation of competition among banks and between the banking sector and other financial institutions (n=51); (4) fraud: opinions regarding the BSD s regulation of bank employees alleged fraudulent behavior (n=53); (5) internal governance and control: opinions regarding the BSD s regulation of banks internal controls and ownership structure (n=76); and (6) opinions regarding other, as well as indistinct, regulatory functions or actions (n=79). To further validate our assumptions regarding the variance in the BSD s reputation across functional areas, we coded for each article its overall tone positive, negative or neutral, accounting for both the source s opinion and the journalist s assessment of this opinion where relevant (n=211, 419, and 4, respectively). Table 1 presents the extent to which each regulatory functional area is associated with an overall positive or negative coverage. [Table 1 about here] Results from this cross-tabulation confirm that the BSD s reputation as reflected in media tone - varies across tasks (χ 2 =40.59, p<.001). To begin with, in congruence with the above secondary sources, our data confirms that, during the relevant research period, the BSD enjoyed an overall positive media assessment of its performance regarding prudential regulation (50% positive), in comparison to consumer protection (27%) and competition (28%). In addition, with regards to areas 17

18 regarding which the BSD lacked an established reputation, Table 1 demonstrates that the BSD attracted particularly high levels of negative media coverage in relation to fraud (9.6% positive). This was mainly due to the collapse of one bank as a result of an employee s fraudulent behavior. Control Variables In addition to the above key independent variables, we controlled for: (1) the direction of the source s judgment, differentiating between positive, negative or mixed opinions (n=157, 422 and 54 respectively); (2) the year in which the article was published (1998 to mid-2009); and (3) the venue of the opinion, allowing a test for the extent to which the content of the BSD s response is shaped by the venue wherein the source s opinion is conveyed. This was constructed of nine categories: (a) opinions that were clearly made during an interview with the source (structured as a series of questions and responses) or an official press conference (n=86); (b) opinions by journalists themselves (n=21); (c) Knesset proceedings (n=62); (d) court proceeding (n=30); (e) formal correspondence with the BSD (n=19); (f) governmental or semi-governmental report (n=15); (g) conference (n=45); (h) miscellaneous ad hoc (n=27); or (i) verbatim quotes of sources, where we could not code with full confidence whether or not these were made in the course of an interview because the quotes were not structured as a sequence of questions and responses (n=130). Summary statistics are presented in Table 1A of the Appendix. RESULTS When do regulators choose to publicly deny a problem, admit a problem, or altogether refrain from responding to criticism? In order to consider these alternatives simultaneously and explicitly model them, we employ Multinomial logistic regression that estimates three separate contrasts: (1) problem denial relative to no response, (2) problem admission relative to no response, and (3) problem admission relative to problem denial. Table 2 presents the beta coefficients from three models examining the effects of level of regulation, agency salience, issue severity, reputation reserves across regulatory functions, negativity of source judgment and opinion venue on the above three outcomes, controlling for the year when the article was published. [Table 2 about here] 18

19 Since the models are nonlinear, coefficients do not easily lend themselves to substantive interpretation. Therefore, we exponentiate the coefficients for the purpose of presentation, converting them to relative risk ratios, and presenting the results in Table 3. The relative risk describes the ratio of the odds for the outcome occurring before and after the change in a specific predictor, holding all other variables constant (Long, 1997). [Table 3 about here] First, in line with H 1a, we find that when facing claims of under regulation (compared to the claims of adequate or mixed regulation), the odds of admitting the problem relative to not responding increase by a factor of about 4, and the odds of admitting the problem relative to denying it increase by a factor of about 3, both of these changes being statistically significant (with the latter significant in a directional - that is, one-tail - test). Contrary to our expectation (H 1b ) there is no significant change in the odds of denial (versus no response) when the regulator is accused of excessive regulation. We find mixed support for H 2. As expected, the agency is found to be more inclined to deny a problem rather than remain silent under both medium and high issue salience (compared to low salience), with the odds for a denying response vs. silence increasing by a factor of and 2.613, respectively, holding all else constant. However, it was also hypothesized that there is a higher chance for publicly admitting a problem vs. denying it under high-level salience. Still, the detrimental effect of high vs. low level salience on the odds for problem admission vs. problem denial falls short of statistical significance under acceptable confidence level (although approaches marginal significance, p=. ). In addition, media coverage intensity does not significantly affect the probability for admitting a problem vs. not responding at all. As posited in H 3, the severity of the event underlying the allegation, with incidents of bank collapses employed as a proxy, was found to increase the probability of publicly admitting a problem relative to either denying it or keeping silent. Ceteris paribus, the odds for admitting vs. denying a problem increase by a factor of and the odds for admission vs. no response increase by a factor of for severe events. 19