Within days after hitting the Gulf Coast of the United States, Hurricane Katrina had become a powerful symbol of the Bush administration's inability to act. The hurricane marked one of the deadliest natural disasters in American history, and its initial magnitude and subsequent flooding caught government officials at all levels totally unprepared. Nonetheless, it was the executive, President George W. Bush, that experienced the greatest political fallout (Malhotra and Kuo 2008). In the end, Bush's own advisors came to see the event as a pivotal moment in which the president lost credibility with the public. (1) Although reflecting failures at both local and national levels, criticism for the hurricane's stunning destruction was directed overwhelmingly, and without delay, at the executive branch, specifically President Bush.
What drives this sort of political blame? Was it merely the hurricane, or did the nature of the reporting exert an independent effect? Short of randomizing the content of media coverage, we have no way of knowing whether events or news coverage underlie these assessments of blame and accountability. Yet, a growing body of work finds that voters react to negative events by punishing or rewarding specific political figures (Achen and Bartels 2002; Atkeson and Maestas 2012; Bechtel and Hainmueller 2011; Gasper and Reeves 2011; Malhotra and Kuo 2008). Such studies often identify a link between exogenous events and subsequent electoral responses but are less explicit as to when and why individuals connect an issue to the politician. We utilize a unique event, the BP Deepwater Horizon oil spill of 2010, to study this intervening step whereby specific events become tied to presidents.
If voters blindly attribute blame for a disaster, then what accounts for the delayed reaction in response to the Deepwater Horizon oil spill? Compared to President Bush's experiences with Hurricane Katrina, President Barack Obama's handling of the crisis did not receive widespread criticism until well after the initial incident. Some pundits went so far as to label the spill as Obama's Katrina Moment, yet the explicit connection came over a month after the explosion. This noticeable delay highlights the need for a richer understanding as to how blame is attached to a specific individual. What is more, the peculiar features of this case provide an opportunity to examine the dynamics by which blame is attributed to political actors.
To tackle these questions about opinion shifts directly, we exploit a natural experiment in which roughly half of the General Social Survey's (GSS) respondents were questioned before the April 20 spill, and the remainder were interviewed in the following months. (2) We can then examine how respondents differed in their perceptions of the president depending on when they were interviewed. Rather than looking at this crisis as simply pre- and postspill, as other scholars of natural disasters and exogenous events have done, we break the postevent period into two separate treatment phases. By dividing the period according to the substantive focus of the spill's reporting, we can more accurately examine the extent to which events and media coverage separately influence individual assessments of confidence in President Obama.
We begin with a brief discussion of recent works on blame attribution and media influence, noting a relative lack of attention in the literature to issue applicability. (3) We then examine coverage of the oil spill more closely to show how the media's framing of the spill shifted over time and how these alternative frames influenced public opinion of Obama in turn. Paying attention to the shifts in media coverage, we break the postspill sample into two treatment periods: the Crisis Phase and the Accountability Phase. After addressing the spill's reporting, we then briefly discuss the individual-level data and empirical design we use to test the study's hypotheses. In the end, we find that presidential confidence does decrease, but this effect is only visible during the Accountability Phase when the media is explicitly targeting the president's inaction. That is, the drop occurred after the media reframed the story of the spill as a failure by President Obama to effectively take charge of the situation. Rather than blindly attributing blame to the president, respondents do not link the oil well disaster to the president until after the media makes the connection for them.
Blame Attribution and the Media
In analyzing the media's response to the oil spill, this study speaks to both the long-standing literature on media effects as well as more recent work on blame attribution and natural disasters. We unite these works in studying the process by which blame is linked to political figures through the media environment. By focusing on temporal variation in the disaster's applicability to President Obama, we can examine how the media communicates blame to the public and thus attaches responsibility to a given target. Although some studies have used exogenous events to study blame attribution, none have examined shifts in news coverage to examine the media's role in this process.
In recent work on natural disasters, for instance, Gasper and Reeves (2011) find that electorates punish presidents for severe weather damage, but the gains attributed to a timely response of relief spending greatly exceeds this punishment. While Gasper and Reeves posit media as a potential mechanism for how the public gains the necessary information, they do not explicitly discuss how media coverage connects disaster relief, or lack thereof, to a president. The literature on media effects helps provide an answer to this important mechanism.
Over five decades after Klapper (1960) declared that the media had minimal effects on political opinion, few scholars of the media would accept this reductionist view. A long line of research has sought to revise this original claim in favor of a more nuanced account that conceives of the media's influence not necessarily as persuading, but as agenda setting, priming, and framing of an issue (Iyengar and Simon 1993). The media has a powerful influence over which issues citizens think are important (agenda setting), what issues citizens use to judge politicians (priming), and how citizens think about issues (framing) (Peters and Kinder 1982; Iyengar, Peters, and Krosnick 1984; Krosnick and Kinder 1990). Valuable accounts of these effects have come from experimental tests, macro stories of media coverage, and quasi-experimental studies alike.
In early experimental work, Iyengar, Peters, and Kinder (1982) and Iyengar, Peters, and Krosnick (1984) demonstrated the existence of priming effects, but only recently have more nuanced experimental tests been able to parse out the underlying cognitive mechanisms by which priming occurs. Building off notions of accessibility from earlier cognitive science work (e.g., Tversky and Kahneman 1973), these more recent accounts have modified it in developing a more solid micro foundation. For instance, Miller and Krosnick (2000) argue that unlike previous accounts, citizens are not mere victims of the media and that attitude change is moderated by trust in the media and political knowledge. That is, knowledgeable citizens with high levels of trust in the media are more likely to exhibit priming effects. Furthermore, they find that while accessibility is thought to be a mediator of the effect of priming, accessibility has little influence on the weight people placed on an issue when making judgments of the president. Drawing on these experimental works, others have attempted to test how the media affects evaluations of governmental officials in response to real-world events. As Table 1 shows, many of these published works have focused on relatively narrow types of events, only a few of which are exogenous shocks. Additionally, only a few examples highlight areas where the applicability to political figures varied over time or across information, which we discuss in the next several paragraphs. Finally, most, but not all, studies controlled for respondent characteristics to assess heterogenous effects across subgroups.
Despite using well-identified designs, many of the media studies in Table 1 suffer from a common issue, namely, that the events are inherently political in the sense that they have a close or obvious link to the target politician. These works are unable to answer questions about how issues are connected to politicians and, if so, how such an association drives evaluations. Stoker (1993) discusses revelations of Gary Hart's extramarital affair, which although unexpected, is intimately connected to Hart from the outset. While Druckman, Jacobs, and Ostermeier (2004) examine how Richard Nixon varied the presentation of different issues and character traits over the course of a campaign, this variation in framing is not exogenous. Indeed, as Druckman and his coauthors convincingly show, the issues and character traits were strategically chosen by Nixon's campaign. Malhotra and Kuo (2008) provide the best test of blame attribution so far, utilizing a survey experiment after Hurricane Katrina. They examine how party and other cues affect blame placement on a wide range of political officials, conditional on the actors' ties to the disaster. The authors do not, however, distinguish whether this link is generated by the events themselves or by the discussion of these actors in the media environment. In another work on Hurricane Katrina, Atkeson and Maestas (2012) further explore this media link as well as the public's response. Unfortunately, their study cannot address this issue of applicability, as the media blamed President Bush almost immediately following Hurricane Katrina.
Building on Althaus and Kim's (2006) examination of public opinion from the Gulf War (1990-91), we argue that changes in issue applicability...