Blame Avoidance in Comparative Perspective: Reactivity, Staged Retreat and Efficacy Christopher Hood, All Souls College, University of Oxford Will Jennings, University of Southampton Paul Copeland, Queen Mary, University of London Acknowledgements: the authors wish to thank the ESRC Centre for Analysis of Risk and Regulation for providing financial support to this research. An earlier version of this paper was presented at the London School of Economics and Political Science. Thanks to Shaun Bevan for his advice on earlier versions of this article. 1 Abstract Building on blame-avoidance analysis, this paper develops a method to assess the reactivity, sequencing and efficacy of defensive responses by officeholders facing a crisis of personal blame, analysing cases drawn from four advanced democracies. It tests the hypotheses that officeholders: react by positive action rather than non- engagement when blame levels are high; respond in a ‘staged retreat’ sequence; and can reduce the level of blame they face from one day to another through choice of presentational strategies. The paper applies event history analysis to test the sequencing hypothesis and time series cross-sectional models to test the reactivity and efficacy hypotheses. The analysis shows that officeholders tend to respond actively when blame levels are high, that to some extent their responses tend to follow a staged retreat pattern, and their interventions have a systematic effect on the next day’s media blame level only if they take the form of personal statements. Keywords: Blame avoidance, blame, crisis, staged retreat, event history analysis 2 1. Blame Avoidance in Political Crises and Scandals: Is There Order Beneath the Apparent Chaos? This paper aims to contribute to a new generation of research on blame avoidance. It challenges the commonly-held view (as put forward by Barker 1994, for example) that the course of events in political crises and scandals is typically unpredictable, chaotic and unmanageable, with events running out of control and no underlying regularities. That view is reflected in a range of sometimes contradictory proverbs about political scandals. For example it is claimed that active media management is the key to success (as in the oft-quoted maxim of former Nixon aide Roger Stone, ‘Admit nothing, deny everything, launch counterattack’ (Labash 2007)) or alternatively that initial non-response is the best strategy given short media attention-spans (Hood 2011: 58-9). Others suggest that ‘the cover-up is worse than the crime’, or that efforts at spin are ultimately self-defeating (Kurtz 1998). The analysis of blame avoidance in politics is beginning to allow us to bring more systematic evidence to bear on such claims. Although ideas about the handling of blame in politics stretch back to Machiavelli and beyond, the modern analysis of blame avoidance is barely 30 years old, starting with Kent Weaver’s (1986) seminal contribution. It brought together observations about negativity bias (Lau 1985) and political behaviour (Fiorina 1981), to develop a perspective that cut across several political science subfields such as voting and political behaviour, public administration and public policy (see also Hood 2002; 2011). Weaver’s basic claim was that politicians tend to give greater weight to potential blame (for losses) than credit (for gains), consistent with insights from prospect theory (Kahneman and Tversky 1979). There have since been some theoretical explorations of blame avoidance, but 3 most empirical studies have been qualitative, particularly in the welfare state literature (e.g. Pierson 1994; Lindbom 2007). There has been some quantitative analysis using survey data, notably to determine how political and institutional context affects blame of government for economic performance (e.g. Powell and Whitten 1993; Anderson 1995) and how ministerial resignations absorb blame and raise government popularity (e.g. Dewan and Dowding 1995); observational and experimental studies of negativity bias (e.g. Dixon et al. 2013; Olsen 2015: 469-79); and some other experimental work on the handling of blame (notably by McGraw 1990 and 1991 and Sulitzeanu-Kenan 2006). Today, data available through digitally-searchable media make quantitative analyses of blame episodes and media negativity more practicable than in the past (Hood et al. 2009; Hood and Dixon 2010; Resodihardjo et al. 2015). That approach is developed here. Up to now, most research on blame avoidance has been based on single-country studies. Our paper aims to enhance the generalizability of the analysis of blame avoidance by observing regularities across countries. We argue that contrary to the common idea that the politics of crisis and scandal is marked by chaotic contingency, there may be some systematic order in underlying behaviour and dynamics. We compare how heads of government, in four different democracies, behaved when they faced media blame over their personal judgements or behaviour. Further, we assess the effectiveness of their efforts to manage blame. Our data are drawn from the U.S., U.K., Ireland and Canada, and involve scandals over personal conduct by four heads of government, namely a U.S. President (Bill Clinton), an Irish Taoiseach (Bertie Ahern), a U.K. Prime Minister (Tony Blair), and a Canadian Prime Minister (Jean Chrétien). Taking negative press 4 coverage as a proxy for blame,1 we develop a method for analysis and comparison of the behaviour of officeholders facing blame, analysing the day-to-day level of blame those individuals faced in the press over periods ranging from a month to two years, and the day-to-day actions they took in response. The paper is organised as follows. First, it outlines three general hypotheses about presentational strategies of blame avoidance by senior public officeholders (i.e. attempts to deflect, avoid or mitigate blame through public rhetoric, argument or news management). Second, it briefly describes the four episodes of scandals relating to personal or political misconduct by heads of government, as mentioned above, and explains how media blame and officeholders’ presentational responses were measured. Third, it uses this data, in time series cross-sectional and event history format, to analyse the link between blame levels and the incidence of officeholder responses, the underlying structure of the sequencing of responses and the impact of categories of response on the level of blame in the media. The paper concludes by discussing the findings and their significance for comparative analysis of blame avoidance and future possible applications of the method. 2. Sequencing, Reactivity and Efficacy of Presentational Blame Avoidance Behaviour: Three Hypotheses Theories of blame avoidance are concerned with how holders of public or political office seek to avoid blame for their decisions or actions. Blame (B) is t 1 Our underlying assumption is either that: (a) levels of blame in the press are a good proxy for latent public blame; or (b) even if (a) does not apply, public officeholders have sufficiently good reasons to fear high levels of blame in the press to wish to avoid or minimize it. 5 conventionally defined as a combination of perceived avoidable harm (PAH) and t perceived responsibility (PR) at any given point in time (Sulitzeanu-Kenan and Hood t 2005; Hood 2011: 6). The developing literature on blame avoidance, mentioned earlier, identifies various ways for organizations and officeholders to try to reduce or avoid blame (for summaries, see Bovens et al. 1999: 127-128; Hood 2002; 2011; Sulitzeanu-Kenan and Hood 2005).2 Figure 1 depicts the sequence of choices faced by officeholders over how to respond to actual or expected criticism that are explored in this paper. For the sort of scandal considered here, involving allegations of personal misconduct or misjudgement by senior political officeholders, presentational strategies of blame avoidance seem to be the most appropriate focus for two reasons. One is that direct attempts to manage media coverage will be more readily applicable in responding to claims of personal misconduct or individual misjudgement by top officeholders, compared to alternative approaches to blame avoidance (such as delegation of responsibility or the use of standard decision protocols). The other is that in such cases officeholders have little option but to engage in reactive crisis management after a scandal has erupted, obviating the need for pre-emptive strategies for limiting perceived responsibility redundant (Hood 2011: 148-50). FIGURE 1 HERE Existing studies offer some limited insights on a number of key aspects of blame avoidance behaviour, including (i) the effect of different levels of blame (B ) on t0 decisions of officeholders whether to take an active or a passive stance; (ii) the temporal sequencing of presentational strategies; and (iii) the effectiveness of different strategies in affecting the relationship between blame on one day (B ) and t0 2 This approach necessarily puts factors such as individual personality and emotion into the background when modelling senior officeholders’ behaviour. 6 the next (B ). Drawing on earlier studies, we accordingly present three general t1 hypotheses about blame avoidance behaviour that we later test. (a) The reactivity hypothesis The reactivity hypothesis refers to stage 2 in Figure 1, and suggests that when levels of blame are high, officeholders will be more likely to take an active stance (AS) than remaining passive (PS). H : Higher levels of blame will be associated with a greater likelihood of 1 presentational responses by officeholders. That is, higher levels of blame will be likely to lead to active attempts to reduce or remove it than ‘do nothing’ or ‘no comment’ responses. The hypothesis here, following the tenor of the active media management approach, is that modern media and politics create pressures for the pursuit of strategies of active denial and counterattack (as advocated by spin doctors such as Roger Stone and Alastair Campbell) when blame levels are high. Still, it could be that there is some level of blame at which any given officeholder or organization’s credibility is so low that any active response is likely to produce a worse outcome than a passive stance, by exacerbating the blame level. (b) The sequencing hypothesis Our second hypothesis concerns the sequencing of actions in the active stance branches of presentational strategies of blame avoidance, depicted in stages 3 and 4 in Figure 1. This derives from earlier work (Hood et al. 2009) building on the analysis of Schütz (1996: 116ff). (For alternative frameworks encompassing a wider array of 7 strategic choices, see Bovens et al. 1999; Branström and Kuipers 2003; Hearit 2006; Hood 2011.) Problem denial (PD) responses operate on the perceived avoidable harm element of blame, involving denials, justifications or excuses (McGraw 1990; 1991) intended to show there is no significant avoidable harm, loss or risk. If that strategy is successful, at stage 3, the issue of blame cannot arise and the question of responsibility is therefore redundant (see for example Schütz 1996: 122-3). Alternatively, presentational strategies can focus on the responsibility element of blame, focusing on responsibility denial (RD) in the next branch of the decision tree, stage 4 in Figure 1. Responsibility denial comprises ways of persuading potential blamers that the officeholder or organization concerned was not responsible for any avoidable loss or harm that could have occurred. When problem denial is not a tenable presentational approach for some reason, officeholders must choose problem admission (PA). If they do so, they then face a choice between the strategy of problem admission combined with responsibility denial (PA+RD) and of problem admission combined with responsibility admission (PA+RA). The former involves admitting the existence of harm or loss combined with denial of responsibility (for example by asserting that no-one is responsible, that others are responsible or that responsibility has yet to be determined). The latter involves an admission of both problem and responsibility. The sequencing hypothesis to be tested here, implicit in the logic of blame avoidance, is that officeholders normally accept responsibility (PA+RA) only if problem denial (PD) and/or problem admission combined with responsibility denial (PA+RD) are untenable. 8 H : officeholders will tend only to admit problems (PA) after problem denial 2 (PD) is no longer tenable, and only accept responsibility after problem denial (PD) and responsibility denial (PA+RD) strategies have become untenable. (c) The efficacy hypothesis Our third hypothesis concerns the effectiveness of presentational strategies in changing the relationship between blame at one point in time (B ) and another (B ). . t0 t1 This issue is important because there are conflicting proverbs about the merits of active media management versus non-response (as mentioned earlier), and because, even though the term ‘blame avoidance’ is commonly used both to denote the activity of seeking to avoid blame and the outcome of blame avoided, it is vital to distinguish the two in analysis (Sulitzeanu-Kenan and Hood 2005). Of course the problem of making causal evaluations in the face of numerous intervening variables must be acknowledged (see Bovens et al. 1999: 146), but the method used here enables us for the first time in a comparative study to establish the causal effects of framing as a presentational strategy in a way that controls for statistical noise and the inherent dynamics of events. We therefore test the hypothesis that officeholders’ presentational responses, in the form of problem denial (PD), responsibility denial (PA+RD) and responsibility admission (PA+RA), are effective in making a difference to the level of blame in future time periods and we look more specifically at the effects of the different sub-types of presentational approach, to be described in more detail later. H : presentational strategies of blame avoidance reduce the level of blame in 3 subsequent time periods. 4. Episodes, Cases and Data 9 We analyse here 106 cases of officeholder responses to blame in the media drawn from four episodes of political scandal which are summarized in Table 1. Each episode involved allegations of serious personal misconduct, misjudgement or abuse of office by heads of government, raising issues whose handling would be crucial to the career or reputation of any such officeholder. Each of the scandals was arguably among the most serious threats to the political career of each of the respective heads of government during their term in office – where substantial blame and political consequences were present. So these scandals represent a key test of how officeholders behave when faced with blame in contemporary democracies, given the small number of such scandals available. The purpose of the comparison, as explained earlier, is to establish whether there are common patterns of behaviour during blame firestorms across different leaders and political systems. All four officeholders concerned were male, white, centrist or centre-left heads of government in wealthy liberal democracies. The allegations of misconduct they faced consisted of improper sexual behaviour linked to perjury in one case, and improper acceptance of funds for personal or party use in the other three – all issues with sufficient potential to damage reputations or end careers to make blame- avoidance a serious concern for the officeholders concerned. The scandals broke at different points in the officeholders’ careers – at the start, indeed during the ‘honeymoon,’ of one officeholder’s tenure as head of government (Tony Blair), but much later in the term of two others (Bill Clinton and Bertie Ahern, who had been in office some five or nine years respectively when the scandals considered here began to break), and continued well after the individual concerned had left office in the other case (Jean Chrétien). The political systems over which these officeholders presided varied in their scale and institutional form (two unitary, two federal, three 10