The use of military force is one of the most fiercely debated topics in politics. Opponents of military intervention often decry administrations for their callous disregard of human life, while administration supporters condemn critics as unpatriotic or naïve idealists. With such fierce passions aroused on both sides of the ideological spectrum, the ongoing U.S. presence in Afghanistan and Iraq, and the high political stakes involved in maintaining public support for military intervention, the political acrimony raised by this issue is unlikely to recede in the near future. This is particularly important given that scholars have found that public opinion can influence government policy even in least-likely cases involving the use of military force abroad.Footnote 1 Indeed in some instances, such as the case studied in this article, public opinion can be decisive in constraining governments' political options.Footnote 2 Consequently, it is imperative for political scientists to address the question underlying this debate. When does the public support the use of military force and under what conditions does public support decline?
This article puts forward a theory that attempts to answer this question and conducts a crucial case study of its explanatory power.Footnote 3 Drawing on insights from psychology and international relations, the Framing Theory of Policy Objectives (FTPO) contends that public support for military engagements will increase when the public perceives its principal objective as seeking to prevent losses (that is, confronting external threat), while support for even the exact same policy will decline if the public perceives its objective as seeking gain (that is, achieving something of value). This reversal of preferences coincides with what Kahneman and Tversky have found in their laboratory experiments and labeled a reflection effect.Footnote 4 In the case of public attitudes toward the use of force, it results because the framing of a policy's objective in media coverage: (1) sets the public's decision-making reference point and (2) determines whether the public's calculations are risk-seeking and cost-acceptant to prevent losses (domain of losses), or risk-averse and cost-sensitive to achieve gains (domain of gains).Footnote 5
However, knowing how the public will judge a policy's objective is a complicated and contentious matter. Because the mass public is not privy to the administration's internal deliberations or the president's decision-making processes, they cannot directly observe a policy's true intention. This allows various political actors to offer competing frames to identify and define the policy's goal. It also means that the public must rely primarily on the mainstream media for the political information they use to judge the policy's objective. Thus, in the same way that the question wordings of Tversky and Kahneman's experimental prompts cued their respondents' decision making to be risk-averse to achieve gains or risk-seeking to prevent losses, so the news media's coverage of a policy's objective cues the public's decision making.Footnote 6 As a result, when a policy's goal is framed as trying to prevent losses, aggregate support for the use of force will increase. Conversely, when even the exact same policy's objective is framed as seeking to achieve gains overall, public support for the policy will decline.
This article proceeds in the following manner. First, it briefly reviews and critiques the international relations literature on public opinion and military interventions. Second, it lays out my explanation by integrating the findings of prospect theory with insights from Jentleson's Principal Policy Objective (PPO) theory and the work of Druckman on the limits of framing effects.Footnote 7 Third, it explains the case selection (U.S. intervention in Central America during the 1980s), describes the data used, and elaborates the methodology employed for testing the FTPO's hypotheses. In the fourth section, I conduct statistical analyses and interpret the results of my findings. I conclude with an assessment of the results' broader implications.
Previous Explanations of Public Support for the Use of Military Force Abroad
International relations scholars have long sought to discover what leads the public to either support or oppose military interventions. While scholars have answered this question in various ways, their responses can be categorized into three broad paradigms.Footnote 8 Historically the most prominent of these is the policy-driven paradigm, which is made up of cost-and-risk aversion theories.Footnote 9 While these scholars often have major disagreements amongst themselves, they are united by the fact that their theories' main explanatory variables are inherently a function of a policy's outcomes, such as its likelihood of success or level of casualties.Footnote 10 In contrast, scholars working within the perception-based paradigm focus on the media's role in shaping the public's perception of a policy's merit to explain public attitudes toward military intervention (that is, perception of: elite-cues, media spin, and policy's costs and benefits).Footnote 11 Lastly, scholars working in the public-preference paradigm explain support or opposition to the use of force as the result of some intrinsic proclivity held by the public such as, for example, the U.S. public's preference for military action as a function of increasing geographic proximity to the United States or the principal objective for which force is used.Footnote 12
However, the explanatory potential of this literature is limited by two fundamental assumptions. The first is that (explicitly or implicitly) these theories are based on a normative model of human decision making grounded in some version of rational choice. Yet during the past thirty years, numerous psychological experiments have shown that human decision making deviates significantly and systematically from the predictions of expected utility theory.Footnote 13 One of the leading challenges has come from Kahneman and Tversky's prospect theory.Footnote 14 Their studies have demonstrated that in violation of the required norms of rationality the same issue or problem is found to elicit different responses depending on whether it is presented in terms of preventing losses or achieving gains from a reference point.Footnote 15
The second limitation affecting all these theories, except those in the rational perception paradigm, is that their independent variables (for example, casualties, likelihood of success, and a policy's principal objective) are not directly observable by the mass public. Instead the vast majority of the public relies on the mainstream news media for nearly all of their political information.Footnote 16 Despite the fact that these theories are characterized by causal variables that the mass public does not routinely perceive first hand, their proponents have operationalized them as if they were.Footnote 17 In reality, the public typically perceives a policy's likelihood of success, the number of casualties it is generating, and its principal objective only as filtered and contextualized through an intervening variable: the media. Therefore, using information not readily accessible to the public can lead to mistaken inferences because they will be based on inaccurate measures of the information the public was using to formulate its decisions.
To date, scholars studying public support for military engagements have not incorporated these insights into their explanations. This disconnect is unfortunate in light of: (1) the extensive media coverage that military interventions receive;Footnote 18 (2) the substantial literature that has documented the news's impact in shaping public opinion; and (3) experimental findings that show that changes in the news media's framing of a policy alters the publics' preferences on that issue.Footnote 19
Public Support and Opposition to Military Interventions
To explain when the public supports the use of military force abroad, while addressing the two criticisms noted in the previous section, the FTPO integrates developments from psychology and political psychology with an insight from the international relations literature.
Prospect Theory
Starting with the psychological literature, I adopt the model of human decision making developed by Tversky and Kahneman.Footnote 20 Their framework, called prospect theory, synthesizes experimental evidence of various observed individual behavior patterns that systematically deviate from the predictions of rational choice into an integrated theory of decision making under conditions of risk. They begin with the empirical observation that people make decisions based on changes (gains or losses) from a reference point rather than based on total assets. Next, they note that people consistently respond in different ways to losses than to gains, overvaluing losses relative to gains of comparable objective value. Third, they find that people tend to be more willing to accept risky options to prevent losses (that is, risk-acceptance in the domain of losses) than they are to accept risky options to achieve comparable gains (that is, risk-aversion in the domain of gains). Finally, prospect theory notes that due to the differences in how people process gains as opposed to losses, as well as the reference point's role in determining these decision-making domains, the identification and framing of a choice's reference point can critically impact people's decisions—even causing preference reversals.Footnote 21 Specifically, when a problem is framed in terms of preventing loss it induces a risk-acceptant response, and when the exact same problem is framed in terms of achieving gains it provokes a risk-averse response.
International Relations
Yet, one of prospect theory's central limitations is that it cannot specify what the reference point will be at any given time or for any particular decision. As Levy states, “A knowledge of the actor's reference point is absolutely essential for any empirical application of prospect theory, however, and the absence of a theory of framing is the single most serious limitation of prospect theory and the most important task for future research.”Footnote 22 To help in this regard I draw on the work of international relations scholar Jentleson, who has convincingly argued that public support for or opposition to the use of force will depend on the policy's principal objective.Footnote 23
While accepting the underlying premise of PPO theory—that the public has an inherent preference for using military force to restrain aggressive foreign adversaries, rather than for ousting unfriendly regimes or maintaining friendly regimes in power—I make two claims that fundamentally reformulate the process and logic by which a policy's principal objective influences public opinion. The first claim is about the public's judgment and the second about its decision making.Footnote 24 First, given that the public judges a policy's objective based on information received from the media, a policy's principal objective is not as important as what the public perceives it to be.Footnote 25 But the presentation of a policy's objective is not always a straightforward issue. In fact, the framing contestation surrounding the objective of military intervention is often highly controversial, politicized, and acrimonious. Thus, I contend that the framing contestation (or lack thereof), played out in public view through the media, provides the information that the public uses to judge what the objective of a policy is at any given point in time.
Second, the framing of a policy's objective in the news impacts the public's decision-making process, not in the rationalist ways proposed by Jentleson, but rather by setting the reference point from which people evaluate the policy.Footnote 26 The result is that in the same way that Tversky and Kahneman's experimental prompts provided their subjects with a decision-making reference point (that is, the wording of their experiments framed questions to cue risk-seeking or risk-averse decision making in their subjects), the mainstream media's coverage of a policy's objective performs an analogous function for the mass public.Footnote 27 As the public's primary source of political information, the news media's coverage (possibly unknowingly and inadvertently) put their “subjects” in the domain of gains or losses at different points in time as the framing that dominates reporting changes over the duration of a conflict.Footnote 28 This leads the public to judge even the exact same policy, as well as the costs and risks associated with it, in different ways depending on whether they perceive it to be designed to prevent losses or achieve gains.
Political Psychology: Media, Elites, and Framing
Lastly, I incorporate political psychologists' work on the limits of framing effects and the impact of competing frames. As Chong and Druckman have noted, by their very nature political debates are unlikely to correspond to the single frame setting found in experimental conditions.Footnote 29 Rather policy discussions are characterized by the existence of competing frames of varying strengths and in varying quantities to describe a particular issue or choice facing the public. Moreover, Druckman's work has found that framing effects are limited by exposure to alternative frames.Footnote 30 His findings show that when a particular frame minimizes or maximizes a certain consideration it has the inverse effect on the persuasiveness of the consideration raised by a counter frame.Footnote 31 In other words, when respondents are exposed to competing frames the impact of framing effects are significantly diminished or eliminated.Footnote 32 Thus, capturing the framing contestation as it actually unfolded in mainstream news coverage at the time is essential for accurately testing the empirical impact of framing effects on public attitudes.
The media's role is fundamental for another reason. Not only is the media the public's primary source of political information, but the press also serves as the principal conduit of communication between political actors and the mass public.Footnote 33 Thus, if the news media does not report on an issue, then even the most high-profile elites, including the U.S. president's most impassioned appeals—let alone the casualties or successes on some distant and remote battlefield—fail to impact public discourse or mass opinion. However, this does not mean that the news media are a monolithic source of information or that they are framing strategically.Footnote 34 Rather, scholarship on the mainstream media acknowledges that there are multiple communications streams (elite, nonelite, journalists, experts, etc.) that often provide the public with contradictory information.Footnote 35
While elites often dominate public discourse (and much of the scholarship on media and public opinion), they are not the only political actors that can influence public attitudes.Footnote 36 Even on foreign policy issues transnational advocacy networks, grassroots activists, and social movements can promote alternative frames that affect public opinion.Footnote 37 So although elites can change their position or explanation of a problem suddenly, this does not mean that a policy's overall framing in the media will necessarily follow suit because other political actors and information sources also contribute to the media's coverage. In fact, elite rhetoric is only one of many sources that make up the totality of information in the press, which together construct the framing of the policy's objective.
At the same time, these different communication streams have various intentions. Some are explicitly trying to sway the public to support or oppose a policy (for example, the president and congressional or grassroots opponents), while other communication streams are putatively reporting objective facts and descriptions surrounding a policy and its objective (for example, media actors—reporters, journalists, anchors—and human rights organizations, etc.).Footnote 38 Yet, even ostensibly neutral actors often operate under professional norms or rely on particular source types that lead them to choose particular ways of describing an issue.Footnote 39 While all these political actors use their narrative scripts with diverse intentions, when they do so their descriptions of the conflict inevitably frame the policy's objective, implicitly helping set the public's decision-making reference point.
Specifically, proponents of the use of force justify it—both offensive and defensive deployments—by identifying its objective in one of two ways. The first justification is to confront a threat and prevent potential losses (loss frame). Presidents and their supporters advocating military force use various narrative scripts implicitly evoke or explicitly construct this frame.Footnote 40 For instance, they can describe the threat as being derived from the external behavior of an adversary, such as aggressive foreign policy or threatening troop movements. In addition, they can also describe the threat as arising from a previous attack—as in the U.S. justification for attacking Afghanistan after the 2001 terrorist attacks on the United States—the loss of previously held territory, resources, or prestige. Or the narratives can describe the threat as arising from the adversary's internal characteristics—such as being communist, terrorist, or undemocratic.Footnote 41
Conversely, proponents of using military force can and often do employ various narrative scripts that inadvertently set the public's decision-making reference point in a way that decreases support for intervention. The second type of justification for using force is to achieve some gain (gain frame). For example, supporters of military intervention can describe the gains as being altruistic, such as bringing democracy, human rights, or a better life to another nation. Or they can be described as egoistic, such as seeking to acquire new territory, resources, prestige, or markets. They can also be described as seeking a combination of altruistic and egoistic goals, such as supporting one side as “the good guys,” who are also seen as allies, in a civil war or internal conflict.Footnote 42
Similarly, opponents of military engagements also employ these two frames or their negations using various narratives that also set the public's decision-making reference point, sometimes with unintended consequences. For example, even while seeking to oppose intervention, opponents evoke the loss frame when they use a narrative that negates the threat posed by another country because of its actions, capabilities (for example, weapons of mass destruction), or associations (that is, with communists, terrorists), or negates the notion that the other country is threatening because of its undemocratic, communist, or terrorist nature. Likewise, opponents of intervention evoke the loss frame when they accept that the other country is threatening either because of its actions or internal characteristics, but nevertheless advocate meeting the threat with means other than force.Footnote 43
Alternatively, opponents of the use of force evoke the gain frame when they use narratives that challenge the assertion that the adversary is a violator of human rights or undemocratic, or argue that the internal situation is not as bad as it is made out to be. Or if they acknowledge that the country's internal characteristics are deplorable, they can still argue that the cost in U.S. lives that would be lost to change the situation (or to acquire the resource, territory, etc.) outweigh the benefits, or simply that better alternatives to using force should and can be found to achieve the same end. Additionally, they might argue that the side being supported is not worthy of that support (for example, because they are an undemocratic or human-rights-violating regime), or that the side being opposed is not as bad as made out to be and may even be the “good guys” in the conflict.
Framing Theory of Policy Objectives
To date scholarship on media framing has focused on these narratives' explicit content—operationalized by a range of cognitive heuristics identified in the news content such as elite cues, moral values, positive or negative media spin, or description of the policy's costs and benefits—to explain the press's impact on public opinion.Footnote 44 Moreover, this scholarship expects these scripts to have an impact on public attitudes in line with the predictions of expected utility theory. That is, the quantity of positive or supportive narrative characteristics—as measured by the authors' preferred cognitive heuristic—minus the quantity of homologous negative or oppositional narrative characteristics is expected to determine the level of public support for military action. In the process these theories have neglected how the various scripts implicitly identify and frame the public's decision-making reference point.
In contrast, I contend that what matters most is not the manifest narrative descriptions of a policy or conflict per se, but rather the often-latent ways in which the policy's objective is framed by these narratives. This determines the public's decision-making reference point and sets their risk orientation, leading them to calculate even the same costs and risks differently depending on whether they are trying to achieve gains or avoid losses. This means that for a framing effect to result one must not only take into consideration the quantity of a particular framing of the policy's objective, but also the absence or presence of an alternative framing of its objective.Footnote 45
Hypotheses
Therefore, I argue that the best predictor and explanation of the public's support for the use of force abroad is found in what the public judges a policy's principal objective to be. This is a function of whether the framing of the policy's objective as seeking to prevent loss (loss frame) or trying to achieve gains (gain frame) is hegemonic or contested in the mainstream media. By hegemonic I mean that only one framing of the policy's objective is presented, and is unchallenged by a counter-frame that offers the public an alternative decision-making reference point.Footnote 46 Thus, I expect that when a policy's objective is framed exclusively as seeking to prevent losses this should induce risk-seeking and cost-acceptant decision making. Alternatively, when the exact same policy's objective is framed as seeking to achieve gains, the public should become risk-averse and cost-sensitive. Two hypotheses follow:
H1: When a loss frame is hegemonic in media coverage, the public will respond in a risk-acceptant manner, driving up public support for the use of force.
H2: When a gain frame is hegemonic in media coverage, the public will respond in a risk-averse manner, driving down public support for military intervention.
In contrast, when the public is exposed to both the gain and loss frame, the literature suggests that one should see a significant reduction in or even elimination of the framing effect.Footnote 47 Thus, I derive my third hypothesis:
H3: When a loss frame is contested by a gain frame, the challenge will have a dampening effect, leading to less support than when a loss frame is hegemonic, but more support than when a gain frame is hegemonic.
If the framing of a policy's principal objective does not matter, one should find that public attitudes are not correlated with news coverage in the ways I've hypothesized. This would lend support to Jentleson's PPO theory, which argues that the public is prudent enough to discern what a policy's principal objective is despite media coverage and political elites' attempts to manipulate it and them.Footnote 48
In addition, my FTPO predicts that the impact of the reference point's framing should not be dependent on the positive or negative valence of news coverage. Instead the impact of a news story's valence on public opinion should be trumped by the way in which the policy's objective is framed. Therefore, I hypothesize that the impact of positive or negative descriptions of the policy will be outweighed by the framing of the public's decision-making reference point in news coverage. In other words,
H4: The reference point framing effects will not be eliminated or substantively diminished by the positive or negative valence of news coverage.
This hypothesis allows one to test the expectations of theories based on the rational calculation of preferences directly against a theory based on a prospective model of human decision making. Whereas theories based on rational choice expect the quantity of positive or negative news coverage to determine public support, the FTPO—based on prospect theory—expects that the way in which the public's decision-making reference point is framed will be a more important determinant of public support.
Case Selection, Data, and Methods
I test this theory on U.S.–Central American relations during the 1980s. Throughout the decade the administration of President Ronald Reagan sought to defeat the Farabundo Marti National Liberation Front (FMLN) guerrillas in El Salvador and “roll back” the Sandinista Revolution in Nicaragua (Sandinista Front for National Liberation—FSLN), claiming both to be major communist threats to U.S. national security.Footnote 49 There are several reasons that make the Central American case ideal for testing the FTPO.
First, U.S. intervention in Central America was characterized by both offensive and defensive deployments of force.Footnote 50 Consequently, this case allows one to rule out the type of military intervention (offensive versus defensive) as an explanation for low levels of public support. In the case of El Salvador, the United States provided defensive support to the Salvadoran military with material, training, and (what is now known) combat support to defend the right-wing government against leftist rebels.Footnote 51 Meanwhile, in Nicaragua the United States supplied offensive military aid, training, material, and even operational direction and combat support to the right-wing Contras trying to overthrow the Nicaraguan government.Footnote 52 Yet, throughout the decade both these cases were inextricably linked to one another—as the Reagan administration justified its Central America policy by claiming that the Nicaraguans were seeking to export communist revolution to El Salvador.Footnote 53 Because the Central American case includes both offensive and defensive deployments, one can rule out the type of military intervention as an explanation for low levels of public support.
Second, because less than one hundred U.S. military servicemen were killed throughout the duration of the conflict, the Central American case controls for the negative impact of large numbers of U.S. casualties on public opinion, which has been cited as a leading explanation for low or declining support for military conflicts.Footnote 54 Additionally, this policy occurred during a time when the public supported analogous military interventions in other cases around the world with similar or higher casualty levels.Footnote 55 Likewise, because U.S. Central America policy generated stiff opposition from Congress at a time when other cases of U.S. military intervention did not, this case allows me to control for the partisan role that Congress plays in generating opposition to the use of force.Footnote 56
Third, U.S.–Central American relations constitute a crucial case in the sense that it tests the theory's explanatory power on a “must fit” case that other prominent theories have previously used to support their explanations.Footnote 57 Specifically, because the conflict took place over a prolonged period of time one is able to observe changes in media framing and fluctuations in public support that would not be feasible in military actions of short duration. Thus, if changes in the media's framing of the policy objective are found to drive public opinion it will undermine Jentleson's assertion that the public is prudent enough to discern a policy's principal objective in spite of elites' attempts to influence them through the media.Footnote 58
Finally, there are methodological reasons for testing the FTPO on an empirical case. Although previous deductive tests of the prospective model of human decision making have consistently generated strong, significant, and robust results these have largely been conducted experimentally. Yet more recent scholarship has noted that typical political debates are often intensely competitive, meaning that exposure to only one frame as found in laboratory settings is highly unlikely.Footnote 59 Likewise, there is strong evidence that when people are exposed to competing frames their effects are diminished or eliminated.Footnote 60 These two findings constitute major challenges for the empirical application of prospect theory in political science. Thus to accurately test the empirical impact of framing effects' on public attitudes, methodologically it is essential to capture the framing contestation, or its absence, as it was actually presented to the public via the mainstream press. At the same time, by deductively testing the external validity of a prospective model of decision making in a real-world setting, the study's methodological research design constitutes a hard case for the FTPO.Footnote 61
Data and Methods
To test the hypotheses derived from the FTPO, I use ordinary least squares (OLS) regression analysis.Footnote 62 I measure the dependent variable, public opinion, using nationally representative surveys, which ask about support for U.S.–Central American policy in the 1980s. To operationalize media framing of the policy's principal objective, I conducted content analysis of ABC News and the New York Times reporting on this issue to produce proxies for mainstream news coverage during the time period in question.
There are several reasons for choosing these news sources. First they are both prestigeous press outlets with national audiences. Additionally, many local and regional newspapers draw on the New York Times's coverage, especially on foreign issues where they do not maintain their own correspondents. Similarly, ABC News has local affiliates around the country who draw on the network's coverage of international affairs. Moreover, during the 1980s 66 percent of the U.S. public read a newspaper on a daily basis and 72 percent watched the evening television network news regularly, according to the Times-Mirror national survey.Footnote 63 These statistics indicate that these sources are reasonable proxy representations of overall mainstream news coverage.
Second, while it may be argued that these media sources are not representative of the mainstream press because of their limited audiences (that is, because most people get information from local news sources), they nevertheless provided information that was available nationwide. Thus, most people in all regions of the country had potential access to these sources, making these outlets better choices than geographically restricted and distinct local news sources. Even more importantly, because of these outlets' limited but national audiences they actually a priori bias the results against my expectations—systematically underestimating framing effects nationally. So findings supportive of my hypotheses using these sources lend strong support for the FTPO a fortiori.
Dependent variable selection
Next, I collected a total of fifty-five national public opinion polls conducted between 1981 and 1988, which asked respondents about U.S. policy toward Central America. All the surveys used similar question wordings, specifically asking whether the respondent supported or opposed U.S.–Central American policy. This was done to ensure that they were capturing public support versus opposition for the policy, rather than some other related policy dimension. This entailed excluding surveys that asked respondents whether they would support the policy given some hypothetical scenario, whether the policy was a mistake, or how to improve the policy. Similarly, question ordering in the survey itself was taken into consideration.Footnote 64 In particular, several polls were excluded because their Central America questions were asked immediately following questions cueing respondents to think about presidential approval.Footnote 65
In the end, the surveys used were drawn primarily from two sources, the Gallup organization and ABC-Washington Post. These were supplemented by several New York Times-CBS and NBC-Wall Street Journal surveys.Footnote 66 All the survey questions were asked in a format that allowed respondents to answer either: favor/approve/agree, oppose/disapprove/disagree, or don't know/not sure. My dependent variable is the percentage of the public who supported U.S. policy toward Central America. In other words, it is the percentage of each of these polls' respondents who respectively answered: favor, approve, or agree in each successive poll. Overall the mean of public support across all the surveys is 27.96 and the standard deviation is 6.215.
Independent variable construction
I then collected every article written in the New York Times and all the ABC News transcripts on El Salvador, Nicaragua, Central America, or U.S. policy toward the region between 1980 and 1990. Articles were identified with the Proquest online database archive's search engine of the New York Times, and the Lexis Nexis online database for ABC News, using “El Salvador,” “Nicaragua,” “or” “Central America” as headings in a keyword search. The stories were then screened to ensure that they were in fact relevant to the subject. If they were even tangentially related to the subject the articles were saved. The collection process for the New York Times produced about eight thousand relevant articles. The ABC News transcripts were similarly screened and produced approximately one thousand different stories. A sample of approximately two thousand articles was drawn from the ABC News transcripts and the New York Times articles by choosing every fifth transcript or article between 1981 and 1988.
Once the sample was drawn, a team of about twenty advanced undergraduate research assistants, who had been trained for more than a year to code the articles and news transcripts, were assigned a set of approximately one hundred transcripts and articles apiece.Footnote 67 Each of these news stories served as one unit of analysis and were coded for the article or transcripts' “framing” and its “valence.”Footnote 68 The articles' and transcripts' framings were coded into three categories: loss frame, contested, and gain frame (see Table 1). A final category, not applicable, was included to capture any stories that mentioned El Salvador, Nicaragua, or Central America but offered no framing of the policy's objective. Similarly, articles and transcripts were coded based on their valence to capture the positive or negative tone of media coverage on this issue (that is, whether the news coverage cast a positive or negative light on the use of force). The news stories' valence were coded either very positive, balanced, or very negative, but also included the possibility of a not applicable category (see Table 1). The valence category measures several dimensions of news coverage that previous scholars have used to explain changes in public support for the use of force: a policy's cost and benefits, its likelihood of success, and elite cues.Footnote 69 The total number of stories in each of these categories is summarized in Table 2.
Article framing and direction variables' range
Overall number of stories in each coding category
Following Jordan and Page, variables were then constructed from the content analysis of New York Times articles and ABC News transcripts found in the one month immediately prior to the dates on which the surveys were conducted.Footnote 70 Specifically, the number of stories in each of the frame and valence categories (listed in Table 1) appearing in the one month immediately prior to the date of every survey were summed. The summed total in each of the different categories are my framing and valence variables.Footnote 71
In Figure 1, I present a graphic display that is suggestive of the relationship existing between public support and the loss and gain frames. For illustrative purposes I graph the average yearly level of public support alongside the net number of yearly frames (loss minus gain frames). This is done simply for ease of presentation and to highlight the statistical relationship, but does not convey the full complexity of the relationship or the control variables that need to be accounted for. Nevertheless, underlying this figure is a statistical correlation between public support and the framing variables. Specifically, I find that increases in the loss frame are positively correlated with increased public support for military intervention (.202, standard error .07, p < .10) and that increases in the gain frame are negatively correlated with the level of public support for the use of force (−.306, standard error .012, p < .05).
Average public support and loss-gain frames
Control variables
Next I control for several possible confounding variables. First, despite the careful selection of polls I still control for possible question wording effects, because even slight changes in question wording have been found to shift public responses.Footnote 72 Therefore, I include dummy variables for three different survey questions whose language is likely to generate wording effects (that is, they ask respondents whether they support “overthrowing” the government of Nicaragua). Second, controlling for time has been found to be important for determining the public's support for the president in foreign policy crises.Footnote 73 Moreover, it is also necessary to control for time to distinguish durable attitude changes from temporary opinion shifts responding to major events.Footnote 74 Thus, I include a variable to control for time, which is operationalized as the number of days from the first poll. Third, because several polls repeat the same question over time, creating a small time series, I also control for the possibility of auto-correlation. In other words, it may be that the level of public support in the previous time period (T−1) was strongly influencing support in the present time (T1). To account for this possibility, I include the level of public support in the previous time period (T−1) as a control.
Fourth, another factor that could explain public support for the use of force is the president's popularity. There is a substantial literature on the famous rally-around-the-flag effect.Footnote 75 This well-documented phenomenon occurs when foreign policy crises emerge that lead the president's popularity to rise in the public's estimation. This rise in popularity in turn increases support for the president's policies, especially foreign policies involving the use of force. I operationalize presidential popularity using the percentage of public approval for the president's overall job performance at the time. These are drawn from Gallup survey questions that were conducted at the same time or immediately before my measure of public support.
Modeling Statistical Tests of Competing TheoriesFootnote 76
To test the FTPO's hypotheses I specify two regression models that test the framing effect of the media coverage on public support for the use of force (see Table 3). The first model includes the framing variables, dummy variables for the different question wordings, and controls for time, presidential popularity, as well as the level of public support in the previous time period. This model tests my FTPO's hypotheses against the claims of Jentleson's PPO theory, which expects that media framing should not have an impact on public support. The second model is designed not only to test my FTPO's three hypotheses but also add a further hurdle. It tests whether the framing of the policy objective's impact is dependent on or qualitatively influenced by the positive or negative tone of media coverage as predicted by theories that explicitly or implicitly adopt rational choice assumptions. The second model explores this possibility by incorporating the news valence variables.Footnote 77
Regression models: The impact of media framing on public support for the use of force abroad
Notes: The dependent variable is public opinion support for the president's Central American policy. Standard error are in parentheses. Statistical significance is based on two-tailed tests.
*** P < .01 (>99%);
** .01 < P < .05 (>95%);
* .05 < P ≤ .10 (>90%).
Statistical Results
The impact of the media's framing on public opinion is presented in Table 3. As these results indicate, the FTPO's first two hypotheses are strongly supported by the evidence. Despite the introduction of the numerous control and alternative explanatory variables, these findings are statistically significant and robust in both models. To illustrate these results, I focus on the second model, but the findings in the first model are qualitatively the same. The loss frame variable's coefficient is in the expected direction (positive) and is statistically significant below the .05 level using a two-tailed test. The size of the coefficient (.310) is also relatively large, indicating that one loss frame article appearing anytime in the month prior to the poll increased public support by about a third of a percentage point at the national level. Given that the number of loss frame articles varied between zero and nineteen in the month immediately before each survey, even an increase on the low end of the spectrum (one standard deviation or five articles) results in nearly a 2 percent change in national public opinion. A reasonable increase in articles (two standard deviations or ten articles) where the loss frame goes unchallenged is found to provoke a change of greater than 3 percent. A dramatic increase (maximum frequency or 19 articles) in coverage using solely the loss frame is found to increase support for intervention by just under 6 percent.
At the same time, the gain frame variable's coefficient is in the expected direction (negative) and statistically significant (p < .05). Similarly, the magnitude of its coefficient (.300) is also relatively large, indicating that an increase in one gain frame story led to nearly a third of a percentage decrease in national support for Reagan's Central American policy. The number of gain frame articles in the month immediate before each survey also varied between zero and nineteen, and had a standard deviation of about five articles. So a change on the low end of the spectrum (one standard deviation) is found to result in a 1.5 percent decrease in support. Similarly, a reasonable increase in gain frame articles (two standard deviations) is found to decrease public support by 3 percent. Meanwhile a dramatic increase of gain frame articles in the month immediately preceding a poll (maximum frequency or nineteen articles) is found to decrease support for military intervention by slightly less than 6 percent across the nation.Footnote 78
The variable testing the third hypothesis also received support. The contested framing variable's coefficient is negative in both model specifications and statistically significant below the .05 level using a two-tailed test. The magnitude of the coefficient is smaller (−.20), indicating that a contested article appearing in the month prior to the poll is correlated with a decrease in public support of about a fifth of a percentage point at the national level. In line with Druckman's expectations, this result suggests that the impact of exposure to both frames have muted each other's effect.
Two control variables' impacts are found to be statistically significant and robust: time and the question wording of the second ABC-Washington Post (abc/wp 2) survey. The time variable's coefficient is slightly positive and is statistically significant below the .10 level using a two-tailed test. The size of the coefficient (.003) indicates that for every one day that the conflict lasted public support increased by three thousandths of a percentage point. While this is individually small, taken over the eight years of the Reagan administration it indicates a slight increase in public support for the policy over time. Conversely, the coefficient for the second ABC-Washington Post survey is negative and significant below the .01 level. The size of the coefficient (−11.3) means that the question wording of this poll, which asked respondents whether they supported the overthrow of the Sandinista government, is correlated to an 11 percent drop in overall support for the policy. Surprisingly, the valence variables are all slightly negative and none of them approach statistical significance. Moreover, including the valence variables does not qualitatively change the impact of the framing variables.Footnote 79 This result suggests that the impact of reference point framing trumps the impact of news coverage's valence.Footnote 80
Discussion of Findings
These findings indicate that when the objective of military force is framed in media coverage as seeking to safeguard the public from potential threats, public attitudes are more likely to support risky and costly policies.Footnote 81 In stark contrast, when even the exact same policy's objective is framed as trying to achieve some direct or indirect benefit or gain, public support for hawkish policies declines. The evidence also indicates that this is true regardless of whether an article or political actor's intention is to support or criticize the policy. In other words, the framing effect of the public's decision-making reference point is not the result of, nor is it dependent on, positive or negative media coverage. Inversely this means that simply criticizing or opposing a policy is not enough to drive down public support when a loss or gain frame is hegemonic. Rather for support to decline among the mass public, they must receive an alternative framing of the policy's objective that shifts their decision-making reference point—redefining the issue in public discourse and undermining the policy as the appropriate response.
Once the implication of these findings is understood there is the question of the framing effect's strength. The magnitude and salience of the above findings become apparent, especially when one considers two factors. One factor is particular to this case and the other has to do with a general characteristic typical of news coverage. In this specific case it is important to keep in mind the actual observed fluctuation of public support. The largest fluctuations in this case were swings of 15 percent in either direction, while more typical fluctuations were in the 6 percent range from month to month. Therefore even at their lowest level, reference point framing effects (2 percent) would account for 13 percent of the largest observed changes of support in this case, but in more typical cases would account for a third (33 percent) of the fluctuation.
However, one must also consider a general characteristic of news coverage: that framing effects can be concurrent and additive. In other words, the rise in the number of gain frame stories can be correlated with declines in loss frame articles and vice versa. Therefore, in months that the number of gain frame articles rises the number of loss frame stories might also decline. This would mean that in any one-month period fluctuations on the low end of the spectrum could entail not just swings of 3 percent public support, but rather 6 percent swings in either direction. Likewise on the high end, these fluctuations of public opinion would therefore entail either increases or decreases of up to 13 percent support for the use of force in one month. Under this scenario the largest changes in the media's framing of the public's decision-making reference point would explain 87 percent of the largest swings of public opinion in this case. In fact, the shifts seen in Figure 1 are indicative of this phenomenon.
Finally, contrary to Jentleson's assertion that the public is prudent enough to discern a policy's principal objective, the evidence suggests that this may not be the case. Instead, the media's framing of a policy's objective is found to have a strong, significant, and robust impact on what the public perceives that policy's objective to be, and directly impacts whether the public judges that policy to be worthy of support or not.
Conclusion
This article asks the important question of when the public supports military action and under what conditions it opposes the use of force. It proposes a framing theory of policy objectives (FTPO) to answer this question. The theory argues that public support for military engagements depends crucially on the public's decision-making reference point, which is determined by what the public perceives the policy's objective to be. However, because members of the public cannot directly observe the policy's objective, they must judge it based on media coverage, which permits influential political actors to offer competing frames to define its goal. This framing contest, in line with the expectations of a prospective model of human decision making, determines the public's risk orientation: risk-seeking to avoid losses and risk-averse to achieve gains. Accordingly, the FTPO predicts that support for the use of force increases when the public perceives its objective as seeking to prevent losses and declines when the public judges it to be seeking gains.
The empirical results provide support for the FTPO. The evidence indicates that when the framing of the policy's objective as seeking to prevent loss was uncontested in news coverage, support for the use of force increased. Conversely, when the framing of the policy as seeking to achieve gains was unchallenged in the media, public support for military intervention decreased. When the framing of the policy was contested support for the use of force decreased, but not to the degree that it did when a gain frame was hegemonic in media coverage. Moreover, the evidence suggests that the framing of the reference point matters more than the news coverage's valence for moving the public to support or oppose the use of force. These results are strong enough to warrant further investigation on other cases and lead me to be optimistic about the FTPO's potential explanatory power, not only for explaining the movement of public opinion in other cases of military intervention, such as those with high casualty levels, but also on other theoretically interesting international relations questions. For instance, it provides a potential explanation for what international relations scholars have identified as entire eras of hawkish or dovish trends in public opinion in the U.S. during the past fifty years (that is, Cold War consensus and post-Vietnam eras). These eras may simply be the result of how the objectives of military interventions were framed in the media and perceived by the public.
Of more contemporary relevance, the FTPO also offers a theoretically plausible explanation for why the U.S. public originally supported the war in Iraq despite some elite opposition. Preliminary analysis of how the objective of U.S. policy was framed in the media points in this direction. Content analysis of ABC News coverage, using the same techniques I've outlined, in the month before five national Gallup polls conducted between March 2003 and October 2004 finds that early reporting almost exclusively used a loss frame and is correlated (.80) with high levels of public support.Footnote 82 However, news coverage increasingly became more dominated by gain frames, which is correlated (−.88) with decreased public support.Footnote 83 It also potentially explains why support for this policy has been steadily declining and is unlikely to rebound, regardless of how much the president touts the success that his policy is achieving or how often he repeats his confidence in the mission's eventual success.
However, this article may ultimately be seen as a story about the limits of even elites' framing. It shows how during the 1980s a popular president sought to mobilize public support for his policy by framing U.S. military intervention in Central America as a response to a communist threat in the region. Yet President Reagan was unable to rouse support for his policy because of the existence and prevalence of an alternative frame, which was heavily promoted by a transnational social movement and transnational public diplomacy campaign by Central American revolutionaries.Footnote 84 What would public opinion have looked like without this counterframe? It is unlikely that Reagan would have faced such strong opposition—instead public attitudes toward this policy would likely have resembled other cases of military intervention of that period. Additionally, the findings presented here offer an implicit critique of international relations theory's great power bias. The article's theoretical insight was generated by understanding how materially weak actors confronted the global hegemon in an asymmetric conflict and specifically their use of contentious politics to do so.Footnote 85 It also offers insight into how national security is a socially constructed concept created and contested in civil society, as well as how that ideational process plays out materially.
Of more general significance, the FTPO modifies prospect theory in a way that makes it empirically applicable to a broader set of social science questions than previously possible. Specifically, the FTPO provides a theory about what sets the public's decision-making reference point for policy choices—how the objective of that policy is framed in the mainstream media. Because prospect theory offers no explanation of what determines the reference point for any particular choice option, but requires one to be applicable in real-world political settings, identifying what the public uses as its decision-making reference point in foreign policy choices involving the use of force is in itself an important theoretical contribution.
At the same time, the FTPO provides well-defined scope conditions. First, it should be applicable whenever and wherever the mainstream media serves as the public's primary source of political information about the policy's objective. Therefore, one should be able to use it to explain support or opposition to the use of force in other countries and in other time periods where this is also the case. Theoretically, there is no reason to believe that the FTPO would not be applicable in other countries, as long as the public: (1) is exposed via the media to alternative framings of why force is being used, (2) has the ability to express their opinion on their government's use of force, and (3) public surveys of opinion are possible to measure the public's attitudes.
Second, the theory should also be applicable in other policy realms. However, it should work best in policy realms that are furthest away from the average individuals' personal experience and where the public is most heavily dependent on the media for the information that it uses to form its judgments and make decisions. Conversely, the theory does not help one predict or explain the decision making of presidents or political elites generally, because they are not as reliant on the mainstream media as the public for making judgments and decisions. Therefore, one does not know what information is setting their reference point and cannot expect an analysis of media framing to be reliable predictors of their policy positions.
Finally, by deductively testing a prospective model of human decision making in a real-world setting, this study provides a way to overcome what have been noted as prospect theory's major external validity problems, especially for questions relevant to political scientists. The first is that on most political issues the public is exposed to competing frames, something not typical in Tversky and Kahneman's experimental tests of prospective decision making and which has been found to diminish or eliminate framing effects.Footnote 86
The second, noted by Levy, is that “it is extremely unlikely that empirical data in international politics will provide a convincing test of these competing theories of choice [prospect theory and rational choice]. It is simply too difficult to operationalize and measure the key variables and to control for confounding variables.”Footnote 87 However, using content analysis of mainstream media coverage and national public opinion polls, the FTPO's research design operationalizes not only the public's decision-making reference point ex ante, the gain and loss frames, and captures the framing contestation but also operationalizes measures of valence used to test theories based on rational choice assumptions. Thus, the research design used here provides a way to test a prospective model of human decision making against theories based on the rational calculation of preferences. In sum, not only does the FTPO help explain why the public supports or opposes the use of force, but its research design provides a method for testing a prospective model of human decision making on empirical cases of public attitude formation in a broad array of policy issues.
