Political psychology is enjoying a renaissance in international relations, as the papers in this special issue show. The introductory and concluding essays, however, raise a long-standing criticism: that psychological approaches identify a list of biases without specifying when and how these biases matter. Part of the problem is the unit of analysis. Studies of individual preferences and beliefs rarely address how biases aggregate, but foreign policy decision making often happens in groups.Footnote 1 Many theories of group decision making that could help bridge this gap, such as “groupthink” or the bureaucratic politics model, do not adequately address how politics can affect groups themselves.Footnote 2
What explains why biases matter in certain cases rather than others, even within the same institution? To address the “aggregation problem”—highlighted by both Powell's and Stein's conclusions to this special issue as a vital area of future research—I link individual-level variation in bias with a political model that relates individuals within a group. In doing so, the study addresses two still-unanswered questions raised by new behavioral research in IR, particularly on elites. The first question is how elite experience influences decision making, a crucial issue given that most foreign policy decisions are made by elites with varying degrees of experience. Although research has demonstrated the importance of leaders’ experience,Footnote 3 it also suggests that experience is a double-edged sword. Experienced elites think more strategically and are less prone to certain kinds of bias.Footnote 4 Yet experience can lead to other biases like overconfidence, and may not lead to better judgment.Footnote 5 We still know little about when the benefits or drawbacks of experience will dominate.
This leads to a second question: how do groups made up of individuals—for example, a leader and a team of advisers—with varying levels of experience and political power make decisions? This question is important not only for understanding how biases aggregate but also how they might be mitigated, since accountability can reduce some of the more pernicious effects of experience.Footnote 6
These aspects of decision making are intertwined. The balance of experience between leaders and advisers shapes how effectively leaders oversee group decision making, and whether they can harness the benefits and minimize the risks of an experienced advisory team. I draw on advances in the principal-agent (PA) literature—which has generated insights in many political settings, but usually within a rationalist framework—as well as in behavioral economics and political psychology. The PA model and individual-level approaches can be fruitfully combined, relaxing the rationalism of the former while providing a firmer political footing for the latter. The behavioral economics and business literature has begun to incorporate individual-level bias (such as overconfident CEOs or financial professionals), suggesting that exploring the characteristics of principals and agents is promising.Footnote 7
This paper examines how the balance of experience influences the assessment and mitigation of risks in war.Footnote 8 Given experienced advisers (i.e., agents), the experience of the leader (i.e., principal) affects decision making through three mechanisms. First, a leader's experience influences how effectively she or he can monitor experienced advisers. Second, experience in a leader affects the credibility of delegation to experienced advisers, and in turn, the nature and extent of information gathering. Third, experience affects whether leaders are able to diversify the advice they receive, as well as their preference for policies that appear certain. These processes affect how leaders gather and process information, and assess and mitigate the risks associated with conflict.
Arguments about group decision making face empirical challenges. I illustrate the argument using two cases, the 1991 and 2003 Iraq Wars, which hold an unusual number of factors constant, especially the identity and experience level of many of the advisers themselves. The 2003 Iraq War has been the subject of several analyses emphasizing psychological bias, including Herrmann's paper in this issue, which focuses on preferences and beliefs.Footnote 9 Yet failures arose despite a team with high levels of experience. My framework suggests that decision-making bias depends not only on the advisory team, but also how their experience interacts with that of the leader. George W. Bush's inexperience exacerbated the biases of his advisers, whereas his father's experience cast a long shadow over many of the same officials.
Ultimately, does it matter whether a leader has previous experience with the substance of international issues and problems, even if the advisory team is already experienced? In recent decades, the United States has tended to elect leaders without significant foreign policy experience—and those leaders, in turn, often try to reassure voters that they will be surrounded by experienced advisers. But experience is not fungible: an experienced team cannot compensate for a lack of experience at the top.
The Politics of Political Psychology in International Relations
In the revival of international relations scholarship on how individuals shape decisions, the role of experience has attracted particular attention. One line of observational research focuses on political leaders, especially how variation in the beliefs and experience that leaders acquire before attaining office affects their decision making and behavior—including the initiation and conduct of war—once ensconced in power.Footnote 10 On the experimental side, studies drawing on samples of experienced elites have complemented existing research documenting systematic biases in how individuals perceive and process information.Footnote 11
Much of this research demonstrates that experience correlates with many cognitive traits in elites, but it also shows that experience can have both positive and negative effects on decision making. On the one hand, as Hafner-Burton and colleagues summarize, experienced elites exhibit more strategic and cooperative behavior, more effective use of heuristics, more effective playing of iterated games, and greater awareness of other players.Footnote 12 On the other hand, experience can be correlated with certain forms of bias, such as overconfidence, which can lead to “overestimating one's capabilities and/or underestimating an opponent, the difficulty of a task, or possible risks.”Footnote 13 Tetlock's work has shown that experts do not make more accurate judgments than novices on many political questions.Footnote 14 He finds that experts tend to be overly attached to their own judgments, suggesting that overconfidence may reinforce inappropriate or inapplicable heuristics. Additionally, power can also exacerbate biases, including overconfidence and risk taking.Footnote 15 But much work remains to understand the political or institutional conditions that might magnify or diminish the effects of experience.Footnote 16 The default assumptions in the rationalist and institutionalist literatures are that these biases cancel or are aggregated out by institutions, and thus can be largely ignored.
Theories of group decision making would seem to be natural candidates to address the aggregation problem. But the literature on group decision making remains curiously apolitical. One of the most often-invoked theories remains Irving Janis's “groupthink,” a set of symptoms such as “pressures toward uniformity” that stem in part from group cohesiveness.Footnote 17 One possible remedy for groupthink is Alexander George's notion of “multiple advocacy,” or a diversity of viewpoints that can “improve the quality of information search and appraisal.”Footnote 18 George argued for “structured, balanced debate among policy advocates” with “no major maldistribution among the various actors.”Footnote 19 But neither the groupthink nor “multiple advocacy” approaches adequately incorporate the politics surrounding advisers, which can make the balance George advocates unrealistic. A second approach, the bureaucratic politics tradition, emphasizes the “pulling and hauling” among decision makers (sometimes, though not exclusively, in small groups), but likewise does not adequately address differences in power and influence.Footnote 20 A third approach has emphasized the political psychology of small groups. Some research in this vein focuses on leadership style and its effect on the group; other work explores advisory processes, or “openness of debate.” Footnote 21 Although descriptively rich, this work has also tended to give less attention to the politics, selection effects, and hierarchical relationships that shape the structure of debate itself.
Recent scholarship on leaders, however, explores how individual traits are expressed within institutions or groups. For example, certain regime types, such as authoritarian regimes, may select for individuals with traits like risk acceptance.Footnote 22 From another perspective, as I will discuss, approaches to reducing the effects of biases center on accountability, which might occur in some group settings or regime types rather than others. These arguments highlight the promise of exploring how individual-level tendencies interact with the politics of groups and institutions.
Theoretical Framework: Experience, Principals, and Agents
Individuals and the Riskiness of Policy Choices
Before turning to the model itself, I first define the dependent variable, which concerns the riskiness of policy options in wartime decision making, specifically the gap between the latent riskiness of the option chosen and the way leaders assess and mitigate that risk. Following Vertzberger, I assume that risk has both a subjective and an objective component. Vertzberger distinguishes between “real risk” and “perceived risk,” which is subjective and can vary across individuals. Although it is impossible to define “absolute risk,” he notes that “perceived risk need not be, and often is not, congruent with actual risk” as a result of “unavailable information, misperception, and misinterpretation.”Footnote 23 Two dimensions of risk may influence this potential gap between latent riskiness and the perceived risk: assessment and mitigation.Footnote 24 First, as Kertzer's paper in this issue suggests, there is a perceptual element of risk assessment. This risk perception, which affects estimates of risk itself, is distinct from risk preferences (i.e., those who are risk acceptant or risk averse for a given gamble).Footnote 25 Biases like overconfidence can also lead to severe underestimation of risk.Footnote 26 Perceiving (or misperceiving) risk to be low for a given option is not the same as holding risk-acceptant preferences or consciously choosing a risky option; rather, decision makers may believe that risk is low.
Second, the capabilities and actions of policy-makers can diminish risk. Setting aside the subjective assessment of risk for the moment, policy choices come with a distribution of risks, often described in terms of the outcome variance; for example, war is commonly seen as a higher-variance option than peace.Footnote 27 Planning, information gathering, and investing in capabilities can minimize downside risks and lower the probability of realizing the most costly outcome. One could think of this reduction as lowering the variance of a given option, or of expanding the choice set to include lower-risk options (or less risky versions of the same option).Footnote 28 A given policy like war can be undertaken in more or less risky ways depending on whether policy-makers have adequate capabilities, plan and mobilize them deftly, and acquire and process information (to accurately assess costs and the probability of success). Furthermore, if risk assessments themselves are biased, decision makers may not take mitigating steps. Risky projects can be valuable for firms; while high-variance strategies can sometimes be beneficial for states, managing risk would presumably be desirable.Footnote 29 Notably, however, risk is not solely the province of hawks. Decisions to preserve the status quo can involve risk.Footnote 30
Foreign Policy Experience
By foreign policy experience I mean substantive expertise about particular foreign policy areas, often (though not exclusively) acquired prior to taking office. Despite recent advances, it remains challenging to specify what kind of experience matters. There are empirical hurdles to studying experienced elites, especially in the context of “ill-structured” problems like those in political science, where experts may not agree on a solution.Footnote 31
Existing research suggests two important points about foreign policy experience. First, a base of substantive, domain-specific knowledge is important, and is distinct from procedural experience and acumen (such as good organizational or bargaining skills). A substantive knowledge base is critical for helping an expert convert a problem into a manageable task, break a problem down into component parts, and analyze it at a deeper level than a novice could.Footnote 32 This knowledge base also allows an expert to use information differently, and gives policy-makers the cognitive architecture and context to assimilate new information.Footnote 33 Experts can use heuristics to assess what information is important, use their previous experience with patterns to make connections between pieces of information in ways that a novice could not, and gather and assess information more quickly and efficiently.Footnote 34
Research in the political and international realm confirms the importance of a knowledge base. Jervis argues that it is a feature of rational foreign policy analysis to sort and process information in a way that is consistent with existing beliefs, and that cognitive context is essential to effective, rational decision making.Footnote 35 Even Tetlock, who is skeptical about the value of expertise, acknowledges that a knowledge base is important.Footnote 36 Researchers with the Good Judgment Project found that political knowledge was a key predictor of better forecasting.Footnote 37 Furthermore, Hafner-Burton, Hughes, and Victor show that domain-specific experience with a particular policy area is an important factor in crisis behavior.Footnote 38 In the bargaining game explored by Rathbun, Kertzer, and Paradis in this special issue, high-cognition egotists are able to think more strategically from the start of the game, while their low-cognition counterparts catch up only after successive rounds because they “need the experience and reinforcement provided by the game.”Footnote 39 These findings dovetail with public opinion research, where political knowledge provides survey respondents with the cognitive context to express opinions.Footnote 40 The findings of Rho and Tomz in this issue illustrate the importance of this kind of domain-specific knowledge—in this case, about trade policy.Footnote 41
A second feature of existing research is that there is broad consensus that expertise is highly specific, and is not easily transferred across different domains or subject areas.Footnote 42 This finding stretches back to research on chess, where even masters were flummoxed by the random placement of pieces on the board.Footnote 43 This domain specificity applies to ill-structured problems: as Voss and Post conclude, methods are not enough and “the applications of the methods per se will lead to inadequate solutions unless the individual has and employs substantial knowledge of the domain in question.”Footnote 44 Expertise in one area does not readily equip a policy-maker to grapple with issues in another domain.
Thus existing research suggests that substantive knowledge of a policy domain is crucial and distinct from executive experience, especially in other domains. One could further speculate that for making decisions on national security, experience with or exposure to issues of security and defense policy would be important, and that exposure to foreign and international economic issues is insufficient. Studies suggest a variety of ways that future leaders acquire experience, including education, military experience, and previous governmental experience.Footnote 45 Among recent presidents, truly broad-reaching experience—as George H.W. Bush had—is rare. But leaders with significant knowledge of foreign policy are more likely to know what questions to ask and what information may be missing from the discussion and thus are better equipped to assess the quality of proposals.
While other cognitive traits may be amenable to the aggregation mechanisms I describe later, prior foreign policy experience, and especially the relative experience of leaders and advisers, offers several advantages for explicating the model. This relative balance is simple to observe, not only for scholars, but also for decision makers in the group. Experience is tied to events and characteristics in an individual's background, rather than a trait that may emerge only in crises. Thus the balance of experience is more likely to be known from the outset and to generate the kind of monitoring and delegation effects described in what follows. Additionally, observing experience prior to crises allows measurement independent from behavior (a significant challenge for nonexperimental research on individuals).
The Principal-Agent Framework
All leaders must rely on the expertise and information-gathering resources of others to help them make decisions, but their agents—who include bureaucratic officials and military leaders—may have different preferences or incentives.Footnote 46 In the rationalist literature, two problems stem from the identity and actions of the agents.Footnote 47 First, principals may have incomplete information about the agent; this adverse selection problem can be alleviated through selection and screening mechanisms. Second, moral hazard problems may arise, leading to riskier behavior or actions that deviate from the principal's preferences. Actions that seemingly accord with the principal's preference but that undermine the principal's authority or long-term interest, or tie the principal's hands, can constitute “shirking.”Footnote 48 Monitoring or punishment can help alleviate moral hazard, but entail costs.Footnote 49 On top of the problems that emerge in the rationalist literature, agents can have cognitive biases that the principal may wish to manage.
The relative experience of the principal with respect to the agents affects the principal's ability to manage agency problems, including managing the biases associated with experience. Thinking about a president and a team of foreign policy advisers, a useful starting point is to consider four possible combinations of the balance of experience, as Table 1 shows. First, in the upper left, an experienced president (principal) might appoint experienced advisers, as George H.W. Bush did. In the upper right, if the president himself is experienced, he may reserve the key foreign policy role for himself, as in the case of FDR or JFK (upper right). In the lower left, an inexperienced president might appoint experienced advisers, as George W. Bush did. Alternatively, in the lower right, both the president and his advisers might lack experience. A mild version of this scenario might be an inexperienced president whose party has been out of power for a long time, weakening the foreign policy “bench” (e.g., the Carter administration).Footnote 50 A more extreme version would be if experienced advisers refused to serve an inexperienced nominee (as many Republican foreign policy officials initially refused to support Donald Trump during the 2016 campaign, diluting the pool of potential advisers). Since most presidents appoint experienced advisers, I focus on the left-hand column, holding constant experienced agents and varying principal experience.
How the Principal's Experience Shapes Group Decision Making: Three Mechanisms
The basic problem for the principal is how to harness the agents’ experience while reigning in problematic biases. I identify three mechanisms through which the experience of the principal magnifies or diminishes bias: monitoring, delegation and information acquisition, and diversity of views. Table 2 summarizes these mechanisms and their effects on decision making.
Experience as a Monitoring Device
Monitoring is an important tool for principals to ensure that agents serve their interests. Although monitoring is often discussed within a rationalist setting, scholars of organizational behavior have explored how it affects agents with varying psychological characteristics and reduces certain forms of bias. For example, agents who feel more powerful (relative to other agents) or more responsible for decisions engage in more self-serving or risky financial investment decisions, but these effects are mitigated by monitoring.Footnote 51 Yet monitoring does not merely entail costs to the principal in terms of effort and attention. It can also be counterproductive because it can undermine trust. The costs of monitoring are thus partly psychological. Frey argues that such trust effects dominate the “disciplining effect” of monitoring when the relationship between principal and agent is personal—generating a “psychological contract”—rather than anonymous (or market driven).Footnote 52
For presidential principals managing a team of experienced advisers, experience itself can serve as a monitoring device that is both efficient and often implicit, so that it does not impose undue costs or undermine trust. Experienced principals are better able to undertake direct monitoring that evaluates the work of agents because their stock of knowledge allows them to take in and assimilate new information, or at least have a better sense of what questions to ask or where there are gaps in the discussion. If, in contrast, the principal is not experienced enough to understand and process information, then the revelation of hidden action by agents might not mitigate the moral hazard problem. Lay people face this problem in everyday life when hiring agents with specialized knowledge: a patient who gets a detailed explanation of treatment usually lacks the background to judge a doctor's actions.Footnote 53
Experienced principals may also be more effective at indirect or implicit monitoring. Agents who serve a principal with a reputation for expertise or experience in a given domain like foreign policy know that their work will face an independent check. As Lerner and Tetlock argue, accountability can be an important tool for de-biasing, but its effectiveness depends on the audience to whom decision makers are accountable.Footnote 54 Preemptive self-criticism—which would presumably prompt consideration of alternatives, as well as better estimates of uncertainty or the probability of success—is more likely when the “evaluative audience” is “perceived to be well informed (so that it cannot easily be tricked) and powerful (so that decision makers want its approval).”Footnote 55 Accountability can also attenuate overconfidence under certain conditions.Footnote 56
If advisers know that the principal will ultimately review their work, they may assess alternatives, probabilities, and risks more accurately. However, weak monitoring (either implicit or explicit) may effectively underwrite more risky policy options (given that advisers can frame the way options, information, and estimates of costs and uncertainty are presented). Even apparently successful policies can be based on suboptimal risk assessments or plans, so poor monitoring would allow such risky behavior to continue in future decision making. These arguments lead to the following hypothesis:
H1:
An inexperienced principal leads to less effective explicit and implicit monitoring of advisers.
Experience, Delegation, and Information Acquisition
A second mechanism through which an experienced leader can affect bias among experienced advisers is by facilitating delegation, a process that has long interested scholars of Congress and the bureaucracy and that affects information gathering. Delegation can help busy principals draw on the expertise and information of agents, but a significant question is how to motivate agents to invest in acquiring potentially costly information.Footnote 57 The principal may have difficulty committing not to use the information gathered by agents for his own purposes.Footnote 58 Delegation can give agents discretion over policy and therefore incentivize them to gather information—thus affecting both the objective and subjective empowering of agents—but credibly committing to delegation is a major challenge. Successful delegation may result in policies that diverge from the principal's preferences, or “bureaucratic drift.” Some models thus see a trade-off between this drift and the level of information applied to a decision.Footnote 59
These models of endogenous information acquisition are built on the strong assumption that all actors know that the relationship between policy choices and policy outcomes is simple and linear, and the difference between an expert and a nonexpert is a single piece of information about the state of the world. For example, all policy-makers might know that more defense spending might lead to more aggressive security policies, but only experts know the current level of an adversary's military capabilities.Footnote 60 As Steven Callander argues, existing models assume that if a layperson (or uninformed principal) acquired this information, the principal would instantly become an expert—and thus could appropriate the information to implement its preferred policy.Footnote 61 If, however, policy is highly complex, then expertise cannot be “inverted” (in Callander's terms)—merely passing on this information cannot render the principal an expert—and so it is much more difficult for principals to expropriate the information. Policy complexity thus functions as a form of de facto delegation, and “when delegated to, an expert agency acts as if the threat of oversight did not exist.”Footnote 62 The price, of course, is that the agent achieves its ideal outcome. As Callander notes, his model is informational, but a fruitful avenue would be to “view expertise as an ability … rather than a difference in knowledge.”Footnote 63
A principal's inexperience can enable credible delegation by rendering information non-invertible. Agents who serve an inexperienced principal may believe he is unlikely to fully grasp the complexities of policy options or ask probing questions, and thus these agents will be more likely to believe they have authority. An experienced principal, by contrast, may be at an informational disadvantage in any given crisis, but has a greater ability to detect low-quality proposals or gaps in information (or simply know when he is working without full information), as well as assimilate new information.Footnote 64
Some aspects of these effects, such as bureaucratic drift, could be accommodated by a rationalist framework but cognitive differences within a group can lead to more risky behavior. First, credible delegation to experienced agents may increase their real and subjective sense of power, leading to biased risk assessment or mitigation strategies. Second, delegation may lead agents to define the problem in specific, possibly more limited ways than the principal would.Footnote 65 Agents may invest in information only about a portion of the policy. The military, for example, might declare that it does not “do” nation building. Third, studies in business and finance suggest that overconfidence can affect information acquisition. For example, Goel and Thakor argue that overconfident CEOs underinvest in information about risky projects because they believe the initial signal they receive about the project to be more precise than a rational CEO would believe; overconfident CEOs thus believe less information investment is required to achieve the same level of precision.Footnote 66 These arguments lead to the following hypotheses:
H2:
An inexperienced principal is likely to make delegation more credible, increasing perceived and actual levels of power and overconfidence among advisers.
H3:
An inexperienced principal may decrease incentives for advisers to invest in information, or lead to information gathering on agent-defined alternatives.
Experience and Diversity in Decision Making
A third mechanism through which experience can shape decision-making bias is by affecting whether and how divergent viewpoints are incorporated into decision making. Although evidence for either the groupthink phenomenon or the benefits of multiple advocacy has been elusive, there is the phenomenon of “group polarization,” which suggests that when groups deliberate, the group response “will tend to be more extreme in the same direction as the average of the pregroup responses,” or in some cases (dubbed the “risky shift”), groups behave in riskier ways than their average member.Footnote 67 One mechanism driving group polarization is “the existence of a limited argument pool, one that is skewed … in a particular direction.”Footnote 68 Additionally, although a leader with well-developed views can be beneficial, a risk associated with an experienced principal is that those views can be resistant to change.Footnote 69
These problems can be mitigated by diversifying advice. There is some evidence that diversity can be helpful to group dynamics and to advisory systems in particular.Footnote 70 In a study of corporate boards, Zhu argues that one way to counter group polarization is to have a credible representative of a minority position; he finds that when directors with minority viewpoints have more experience with acquisitions, the degree of group polarization around subsequent acquisition decisions is reduced.Footnote 71
Thus presidents might wish to pursue “multiple advocacy” by appointing an authoritative, experienced individual who is more likely to take an alternative view, potentially diversifying the argument pool and endowing these alternative arguments with credibility in the eyes of others within the group. One route to such credibility is an independent power base or source of popularity, as in a “team of rivals.”Footnote 72 Yet empowering a credible source of dissent comes with potential costs, since it gives an adviser with independent standing access to inside information and a platform, and the risk of defection remains.
Although empowering a rival poses challenges for any president, they are likely to be greater for inexperienced leaders. Disagreement with an adviser with a strong reputation on foreign policy might confirm the image of the leader as inexperienced. Inexperienced leaders are thus more politically beholden to their experienced advisers, which might suggest that such advisers would be empowered, but could give them disproportionate voice for the wrong reasons. Alternatively, knowing the political risk, inexperienced leaders might choose to cut potential dissenters out of the loop, negating the benefits of their distinct viewpoint. The need to appoint someone to cover inexperience may thus result in a form of adverse selection, or undermine mechanisms to avoid adverse selection.
Inexperienced leaders may rely on more homogeneous advice not only because they face greater political risk from diversity, but also because they may incentivize proposals that are crafted to look more certain. Scholars have documented “ambiguity aversion,” or the dislike of outcomes with uncertain probability (as opposed to a risky choice with a known uncertainty).Footnote 73 Although there is little evidence on whether experience reduces ambiguity aversion, one argument might be that inexperienced elites are more likely to prefer certainty (and are more averse to ambiguity), leading to incentives to present inexperienced principals with choices that are framed in more certain terms. These arguments suggest two final hypotheses:
H4:
An inexperienced principal is more likely to marginalize potentially divergent viewpoints.
H5:
Advisers are more likely to present inexperienced principals with proposals that are framed with high degrees of precision.
Research Design
When we study group decision making, a significant methodological challenge is to find comparable groups and situations. My approach is to vary principal experience while holding adviser experience and the identities of the advisers as constant as possible. The repeated US involvement in Iraq provides a particularly useful comparison between two presidents who represent extreme values on the independent variable while holding an unusual number of factors constant. The comparison is imperfect—given the intervening factor of 9/11, the decay within Iraq, and the connections between the wars themselves—but nonetheless affords significant advantages. The two presidents involved were both Republicans from the same family. The elder Bush had an unusually high level of foreign policy experience, including service as the US envoy to China, ambassador to the United Nations, CIA director, and vice president. The younger Bush, in contrast, had virtually no foreign policy experience. Crucially, many of the same individuals served both the elder and younger Bushes, and were highly experienced in both cases: many advisers to the first Bush served at high levels of previous Republican administrations.
Many commentators have invoked individual biases or some aspect of group decision making—usually groupthink—to explain the lack of postwar planning in the 2003 Iraq War.Footnote 74 Existing accounts suffer from several drawbacks, however. First, Janis's groupthink argument focused on group cohesiveness, but the Bush administration was laden with tension.Footnote 75 Second, key officials also had their own agendas that did not always align. Third, there remains the aggregation problem. Johnson persuasively argues that overconfidence shaped the conduct of the Iraq War, and that “openness of debate” affected the level of overconfidence itself.Footnote 76 But given that not all US cases display this degree of overconfidence, what are the political antecedents of the nature of debate? Furthermore, why was a group that included many veteran decision makers and a well-respected skeptic (Colin Powell) unable to test its assumptions?
It is important to specify which aspects of the Iraq decision making my argument addresses. The 2003 Iraq decision can be thought of as the evaluation of the expected utility of (1) intervening with a “light footprint”; (2) intervening with a large footprint; or (3) not intervening at all. The choice of strategy—that is, between (1) and (2)—was directly influenced by George W. Bush, who was averse to nation building prior to taking office and appointed advisers who shared this aversion. The debate thus reduced to a light-footprint invasion, or not intervening at all. Yet the evaluation of the risks of a light-footprint invasion was cursory. Had it been more thorough, the administration might not have initiated the war.Footnote 77 Notably, George H.W. Bush's team was also not enamored of nation building, but left Saddam Hussein in power in 1991.
Some cognitive or ideational tendencies may have affected the comparison between invading with a light footprint or staying out, especially after the shock of 9/11.Footnote 78 But an increase in the estimated cost of doing nothing does not explain the poor estimates for the costs and probability of success of a light-footprint invasion.Footnote 79 Furthermore, many accounts also stress George H.W. Bush's emotional response to Saddam Hussein's aggression in Kuwait, which may have raised the perceived costs of doing nothing in that case as well.Footnote 80 In the 2003 case, there were also differential costs to the participants. Fallows argues that “the President must have known that however bright the scenarios, the reality of Iraq eighteen months after the war would affect his re-election.”Footnote 81 Lake notes that administration officials “paid dearly” in terms of their reputations and the Republican majority.Footnote 82 But blame often attaches to presidents. Those serving Bush did not face re-election and may have seen their careers as nearing an end.
In the following section, I examine how the balance of experience affected decision making in 2003, focusing on the postwar planning. Given space constraints, I then briefly discuss the 1991 case. Ultimately, the decision making in 1991, particularly with respect to the postwar phase, resulted in a smaller gap between the latent and actual risk than in 2003.
George W. Bush, Experience, and the 2003 Iraq War
Accounts of the Iraq War often focus on the dominance of members of George W. Bush's administration, including Vice President Dick Cheney and Defense Secretary Donald Rumsfeld. Bush was comfortable with invading Iraq with a light footprint, which reflected his own stance against nation building and shared beliefs with his advisers. But the inadequate estimates of the costs and risks of this approach arguably stemmed at least in part from the mechanisms outlined earlier.
First, Bush's inexperience led to poor monitoring of his subordinates (consistent with H1). There was little explicit monitoring, in the sense of presidential probing and questioning of plans. For example, Fallows notes that “in several months of interviews [he] never once heard someone say ‘We took this step because the President indicated…’ or ‘The President really wanted.’”Footnote 83 Fallows concludes that to “fully understand how intelligent people convinced themselves” that a “successful occupation would not require any more forethought than they gave it,” Bush's leadership was a key factor, including his “lack of curiosity about significant details.”Footnote 84
Bush's inexperience also contributed to an atmosphere in which subordinates would not perceive themselves as accountable to a well-informed leader. For example, in a meeting with Iraqi exiles in January 2003, the “very notion of an Iraqi opposition appeared to be new to him,” and Bush also seemed “unfocused on the key policy questions of the future of the Iraqi army, debaathification, and an interim government.”Footnote 85 The inadequacy of Bush's probing of his subordinates was on display when he asked National Security Adviser Condoleezza Rice, “A humanitarian army is going to follow our army into Iraq, right?”Footnote 86
One of the few times Bush did step in forcefully illustrates the difficulties of an inexperienced principal. Dyson argues that Bush was on an “intellectual journey on questions of stability and governance in post-conflict states,” from his position against nation building to a more expansive vision of democracy in Iraq “that would require a good dollop of nation-building.” Bush thus “risked setting a target of creating democratic institutions without fully committing to the ‘armed social work’ necessary to achieve them.”Footnote 87 Even if we assume that Bush had shifted to a democratization position, he still had to live with his initial policy choices, which strongly deemphasized nation building and reduced resources available for such missions.Footnote 88 When Bush personally empowered L. Paul Bremer to take over following the initial post-invasion stumbles, he did not take into account how difficult it would be to change US policy on a dime, nor did he take steps to mitigate those risks. Available capabilities were insufficient for the shift toward a more transformative vision for Iraq. Furthermore, Bush gave Bremer ambiguous instructions and failed to resolve existing tensions, exacerbating postwar problems.Footnote 89
Bush's inexperience also enabled credible delegation to his subordinates (H2). Bush “styled himself as an MBA president and believed good management was to pick good people and then delegate to them.”Footnote 90 But Bush's inexperience in foreign policy also empowered those who worked for him. Bush lacked the context for asking the right questions when absorbing new information. For example, in a briefing on 4 March 2003, Undersecretary of Defense Douglas Feith briefed Bush and the National Security Council (NSC) on postwar planning. As Bob Woodward puts it, the briefing was “a lot of abstract political science, and the president didn't have much to say other than to remark that he wanted to see information on how they would deal with the military and intelligence services.”Footnote 91 On 5 March CENTCOM Commander Tommy Franks made his final pre-war presentation to the NSC, a meeting in which Bush asked about postwar plans. Franks asserted that there would be “lord mayors” in major cities and towns, a response “which seemed to satisfy Bush and there was little follow-up.”Footnote 92 Policy information was thus not “invertible”—Bush could not be simply briefed into a full understanding of the risks of invading Iraq.
This lack of invertibility empowered Bush's subordinates, particularly Rumsfeld and Cheney. Peter Baker notes that in considering the vice presidency in the first place, Cheney “surely understood that a president with as little knowledge or interest in details as Bush would leave him plenty of room to maneuver.”Footnote 93 Though Bush relied on Cheney, as Baker describes, after the 2002 midterms there were “Bush people and Cheney people,” with “fissures … that reflected profound differences in policy and personality.” Bush “allowed a fractious struggle to play out beneath him without resolving it firmly one way or the other.”Footnote 94 Ari Fleischer, Bush's first press secretary, posited that Bush “would almost always agree with Cheney and Rumsfeld about what the objectives should be, which was a hawk, but what people miss is that he would agree with Condi Rice and Colin Powell about how to achieve it.”Footnote 95 Cheney clearly felt empowered. In August 2002, for example, Bush was reportedly displeased with Cheney's aggressive speech to the Veterans of Foreign Wars, but “chose not to confront Cheney, instead telling Rice to do it.”Footnote 96 In discussing Iraq at one of their weekly lunches, Cheney asked Bush, “Are you going to take care of this guy or not?”Footnote 97
Rumsfeld was also disproportionately empowered. Dyson argues that Bush's inexperience gave Rumsfeld significant control over policy. Dyson reports that Kim Holmes, an assistant secretary of state in the first Bush term, saw the imbalance between Bush as a “novice” on foreign policy and those who “had been around for decades” as problematic because he did not “have an independent base of knowledge that enables you to be able to choose between [different positions], to stop the fighting and make a decision that sticks.”Footnote 98 As Baker notes, a significant problem was that “Bush generally left it to Rice to manage the rivalries,” but she was herself far less experienced than Cheney and Rumsfeld, who frequently circumvented her.Footnote 99
This delegation to Cheney and Rumsfeld also biased the search for and production of information (H3). First, information acquisition was largely limited to Rumsfeld's goal of toppling Saddam Hussein with a small force and the optimistic assumptions of the postwar phase were not tested. Rumsfeld obtained control over the postwar planning in the Pentagon, but he was able to define the mission in accordance with his agenda to transform the military into a leaner force.Footnote 100 But no measures were taken to guard against downside postwar risks.Footnote 101 Although Bush's views influenced the removal of nation building as an option, the president, as Fallows notes, had incentives to manage risks more effectively.
Second, Dyson reports that Rumsfeld in particular saw bureaucratic infighting through the lens of power and control.Footnote 102 This power may have contributed to the failure to acquire information; Fallows notes that at the “zenith of his influence” during the planning phase, Rumsfeld “was not careful about remembering his practical obligations.”Footnote 103 Third, specific biases have been documented among those who shaped information acquisition: availability bias from the apparently successful light-footprint operation in Afghanistan;Footnote 104 motivated bias to ignore contradictory evidence, especially about expected costs;Footnote 105 and overconfidence.Footnote 106 But Bush's credible delegation and failure to monitor gave these biases extended life.
The Iraq decision making was also notable for its marginalization of alternative views (H4). There was one insider with the stature and visibility to credibly voice an alternative perspective: Colin Powell. Shortly before Bush's inauguration, James Traub predicted that “if there is to be a regent in foreign affairs, it will … be Colin Powell … thanks to his standing in the world—and Bush's lack of it—Powell will be able to operate with tremendous latitude … It's not easy to see how Bush could fire him or perhaps even overrule him.”Footnote 107 Yet at the press conference during which Bush announced the appointment, Powell “dominated the event,” raising the possibility of “a secretary of state with the potential to overshadow his boss.”Footnote 108 Rumsfeld was selected in part to counterbalance Powell, who was almost immediately marginalized.Footnote 109 In the end, Powell's main influence was to get Bush to involve the United Nations. Indeed, the administration exploited Powell's skepticism, choosing Powell to make the now-infamous presentation to the UN. Bush asked him explicitly in a private Oval Office meeting on 13 January 2003, “Are you with me on this? I think I have to do this. I want you with me.” When Powell expressed his support, Bush told him, “Time to put your war uniform on.”Footnote 110
Another factor contributing to the lack of diversity in the Iraq decision making was the drive for certainty, stemming in part from the desire to present an inexperienced president with a consensus view (H5). As Jervis argues, there was a widespread psychological need to downplay the costs of the administration's limited postwar approach because probing further might prompt awareness that the war was too risky.Footnote 111 This tendency may have stemmed from Bush's decision-making style, widely seen as instinctual, stubborn, risk acceptant, and resistant to new information.Footnote 112 But why did this administration fail to draw on the benefits of an experienced team and instead suffer from its drawbacks? Apart from Bush's style, a contributing factor may have been his inexperience and desire that his advisers present him with a consensus. Packer suggests that “the advisers around the president understood his strengths and what he needed to know in order to make decisions.” A senior administration official told Packer in an interview that “no one ever walks into the Oval Office and tells them they've got no clothes on—and persists … I think it's dangerous that we have an environment where our principal leader cannot be well-informed. It's part and parcel of the office,” but more so in this administration, which was “scary, because of the president and the atmosphere and the people there.”Footnote 113 Further research with elite subjects might explore whether less-experienced individuals are more likely to feel ambiguity aversion.
It is important to consider at least three potential alternative explanations. First, perhaps motivated bias accounts for most of the failure: knowing too much might have impeded the war itself. Second, there is the argument that postwar planning would undercut the political case that the war would be easy. I do not argue that motivated bias or fear of political fallout played no role, only that they are incomplete explanations. As Fallows notes, the failure to consider the domestic political consequences of postwar problems is puzzling, especially given that many Bush-era decisions were politicized. The Bush administration was also notoriously secretive, so officials might well have believed it possible to keep postwar plans secret (and there was some successful planning for humanitarian contingencies). Third, perhaps the hawkish Bush administration preferences mattered most, an explanation I consider after first briefly discussing George H.W. Bush's decision making.
George H.W. Bush, Experience, and the Gulf War
One of the often-noted ironies of the 2003 Iraq War is that so many of the participants had chosen not to remove Saddam Hussein in 1991. Three features of the 1991 case are notable. First, the elder Bush was directly involved in the details of the crisis. As Woodward reports (channeling Powell's perception), “Bush wanted the details, all the details. He wanted to be the player, the guy who made as many of the calls as possible.”Footnote 114 Steve Yetiv's study of decision making in the Gulf War highlights the elder Bush's experience and skill in foreign affairs, and quotes Sandra Charles, Bush's Deputy National Security Advisor, as observing that Bush “was attached to and aware of Gulf politics and acted on the basis of knowledge.”Footnote 115 In the NSC meeting following Saddam Hussein's invasion of Kuwait, Bush “immediately took charge” from his own National Security Advisor, Brent Scowcroft (who had himself served in that position in the Ford administration). Woodward notes that Bush, “the former UN ambassador,” asked about the UN diplomacy, and as “the former [envoy] to China,” asked about China's response.Footnote 116 Woodward reports another moment in October 1990 when Bush told his chief pollster that he “felt he knew more than anyone about the region, and also about the diplomacy, the military, the economics, and the oil,” and that “it was important that he had served as United Nations ambassador, US envoy to China, CIA director and Vice President. Those experiences allowed him to see all the pieces. Now he could put them together.”Footnote 117
Second, the shadow of Bush's experience loomed large over his subordinates (H1). Yetiv argues that “others around [Bush] understood that he was willing and able to play a predominate role. They must have factored that into their calculations.”Footnote 118 Stephen Wayne's analysis of Bush's decision-making style suggests that “although Bush was reported to be a good listener in the strategy sessions of the group, his personal experience and knowledge in foreign affairs, which exceeded that of most of the participants with the possible exception of Scowcroft, may have worked to intimidate those who might have fundamentally disagreed with the president.”Footnote 119 Wayne notes that on military issues, “where the president lacked detailed knowledge and experience,” Bush tended to be more deferential and open to suggestions.Footnote 120 Although Bush did delegate, especially to the military, he retained significant control over decisions (H2). Woodward reports that “often Powell and Cheney returned from [principals meetings] and said to each other, now what did that mean? What are we supposed to do? Frequently, they had to wait to hear the answer later from Scowcroft or the television.”Footnote 121 Bush also influenced decisions through implicit accountability. Yetiv notes that “even when he did not participate in group meetings … [Bush's] influence worked its way down to the deputies group serving the group of eight [principals]. In this sense, indirect influence obviated a direct presidential presence.”Footnote 122
Bush's expertise on foreign policy affected how the principals, including Cheney (then serving as secretary of defense), gathered and produced information (H3). In October 1990, for example, when Bush asked for a briefing on an offensive option against Saddam Hussein, Cheney and Powell went over the plans at the Pentagon, asking “question after question” and deeming parts of the plan “unwise.” Cheney “concluded that an attack with the US forces now in place and based on this plan would be a risk of a high order.” As Woodward reports,
Cheney recognized that he had an obligation to present this brief to President Bush. The President needed to know exactly where [Norman] Schwarzkopf was, the status of the deployment, and what might happen if offensive operations were ordered. The President, Scowcroft, and [White House Chief of Staff John] Sununu at least had to be educated on the magnitude of the task. Cheney did not want to walk over to the White House one day, months down the road, to say, “Here's the plan, bang, go.” The President had to comprehend the stakes, the costs and the risks, step by step.Footnote 123
As Cheney had put it in a speech the previous month, Bush was part of “the ‘Don't screw around’ school of military strategy.”Footnote 124 When Bush received the briefing the following day, “he and Scowcroft had many questions on various subjects,” and came to a conclusion “similar to Cheney's,” that the military was “not ready for an offensive operation.”Footnote 125
Third, dissenting voices were not completely marginalized (H4). Ironically, groupthink dynamics, in terms of the collegiality and cohesiveness highlighted by Janis, seem to have been more operative in the 1991 case than in 2003. In Woodward's account, Powell at several points complains about the collegiality of the elder Bush's inner circle amid his attempts to bring his opposition to the use of force to the table.Footnote 126 But Stanley Renshon argues that “one cannot assume that the decision was necessarily flawed or that no differences were present or expressed.”Footnote 127 Indeed, in October 1990, Cheney suggested that Powell should bring his proposal for containing Saddam to a private meeting with the president to which Powell was not usually invited.Footnote 128 Additionally, “Bush was in frequent personal contact with a range of world leaders, not all of whom shared his views or favored his plans.” Renshon concludes that “alternative points of view could easily have been presented and forcefully pressed.”Footnote 129
Perhaps most pertinent for purposes of comparison with the 2003 case, on the question of when and how to end the war, Powell—who seems to have expressed dissatisfaction with the decision making in the run-up to the war—did have his voice heard. As Rick Atkinson writes, “George Bush was depicted as the locomotive engineer who drove the war train … Colin Powell was the brakeman.”Footnote 130 One could imagine the momentum of the war leading to mission creep. But Bush heeded the military's concerns that the war was becoming so lopsided as to tread on morally dangerous ground. Bush's political goals were also limited to ejecting Saddam from Kuwait. Even before the war began, Bush was worried about the ending, dictating into his tape recorder: “How do we keep from having overkill? … I think we need to watch and see when our military objectives are taken care of in Baghdad and Iraq.”Footnote 131 Of the decision not to take out Saddam—later criticized by some for failing to finish the job and for allowing the Iraqi uprisings to lead to a humanitarian crisis—Bush and Scowcroft wrote:
Trying to eliminate Saddam, extending the ground war into an occupation of Iraq, would have violated our guideline about not changing objectives in midstream, engaging in “mission creep,” and would have incurred incalculable human and political costs … We would have been forced to occupy Baghdad and, in effect, rule Iraq … Had we gone the invasion route, the United States could conceivably still be an occupying power in a bitterly hostile land.Footnote 132
The decision was taken by many to be a prudent assessment of risks.Footnote 133 Cheney and others who would later serve the younger Bush in 2003 supported the 1991 decision.
The evolution of key players’ views and behavior from 1991 to 2003, especially those of Cheney, has been the subject of much debate. In Cheney's case, even accounting for 9/11, some commentators have seen the environment in which he was embedded as an important factor. Although it comes from Powell, an assessment reported by Peter Baker is illuminating. Powell's theory was that Cheney had not changed, but that
in the first Bush White House he was serving a president who was very sure of himself on foreign policy and national security. As defense secretary, Cheney had often expressed ideas similar to those he would advance as vice president. But back then, he was surrounded by other adults in the room with experience and gravitas … In this White House, Powell believed, Cheney was not effectively contained by anyone. He had a much freer hand with a president whose background gave him little real preparation to be commander in chief.Footnote 134
In 1991, the elder Bush's strong foreign policy background helped keep the risks associated with the experience level of his team under control. In a biography with which he cooperated, George H.W. Bush noted that in his son's administration, Cheney's views had become “very hard-line and very different from the Dick Cheney I knew and worked with.” Cheney “had his own empire there and marched to his own drummer,” and the elder Bush felt that “the big mistake … was letting Cheney bring in kind of his own State Department.” He explicitly, however, noted that “it's not Cheney's fault, it's the president's fault.”Footnote 135
Principal-Agent Dynamics Beyond Iraq
Although the research design holds constant the Republican partisanship and hawkish preferences of the presidents, it remains difficult to disentangle the effects of hawkishness from biases such as overconfidence that are associated with experience. It is natural that presidents try to appoint like-minded advisers, so we might expect that biases point in the same direction as policy preferences like hawkishness. On the other hand, it is impossible to be sure of an adviser's preference on every issue, and one would expect experience to mitigate some biases. Brief illustrations from other cases can help isolate the role of experience and show how the argument generalizes.
First, there is evidence from the same administration over time, especially as leaders gain experience. Take George W. Bush himself. Many accounts stress the different decision-making dynamic by the time of the Iraq “surge” in 2007, with Bush significantly more willing to ask questions, challenge his military and civilian advisers, listen to a variety of outside experts, and ultimately overrule powerful figures who opposed the surge.Footnote 136
Furthermore, historically the principal-agent dynamics identified here manifested on both sides of the aisle. Returning to the principal-agent combinations in Table 1, all combinations come with risk, but this risk may take different forms. In the upper left, the case of experience in both the principal and the agents might be termed “controlled risk.” In addition to George H.W. Bush, Dwight Eisenhower is an exemplar in this category. Eisenhower both directly and indirectly managed his secretary of state, John Foster Dulles, who tended to take a more aggressive stance and was prone to bluster; one could argue that the administration's “brinksmanship” depended on the ability to manage risk.Footnote 137 In the lower left, the inexperienced principal/experienced agent combination leads to “excess risk.” Notably, inexperienced Democratic presidents have also found themselves magnifying biases that tend toward more aggressive stances. When Harry Truman took over the presidency in 1945, for example, he was both inexperienced and “totally unbriefed” on the complex policy FDR had been pursuing toward the Soviets. Truman's inexperience and desire to “appear decisive and in command” led him to rely excessively on his advisers and accelerate (though not cause) the confrontation with the Soviets.Footnote 138 In the upper right, if an experienced principal appoints less-experienced agents, perhaps to maintain control or play key roles himself, there may be fewer efficiency gains from delegation, and there is a risk of overreliance on the principal. Franklin Roosevelt's advisory system, for example, was famously chaotic, with most of the authority vested in the president. The dominant role of John F. Kennedy in resolving the Cuban Missile Crisis also illustrates what we might call “centralized risk.” Finally, in the lower right, if both the principal and the agents are inexperienced, there is “incompetent risk,” since information gathering and processing will be weak at all levels.
Conclusion
The new behavioral revolution has confirmed that individuals’ experience matters for decision making. But the interplay among elites with different levels of experience is less straightforward. Chief executives are undoubtedly important, but are not on the front lines of information gathering and planning. The preferences and beliefs—and resulting biases—of subordinates, as well as how those biases aggregate, also shape decision making.
Moving from individuals to war outcomes reminds us that merely demonstrating individual-level variation and bias is not enough. But political frameworks, such as the principal-agent model and theories of delegation, can help bridge from individuals to groups and institutions, and ultimately, to international outcomes. I have argued that the balance of experience between a leader and a group of advisers can affect how biases in the group aggregate. As the introductory essay suggests, one could imagine greater use of institutional theory in understanding how biases are expressed, or how institutions are designed to minimize or perhaps harness bias. Existing theories, in turn, can benefit from new individual-level insights, as this paper's exploration of the traits of principals and agents (and related work in finance and organizational behavior) illustrates. More concretely, in the realm of bureaucratic oversight and delegation, the argument suggests areas for future research. Diminished foreign policy experience in Congress, for example, may reduce the effectiveness of oversight of the national security bureaucracy.Footnote 139 While reduced oversight might aid delegation, it may also increase risk taking by national security agencies.
As the concluding essays stress, there remains much work to do on the aggregation front. Given the experimental advances highlighted by papers in this special issue, as well as Tetlock and Lerner's work on accountability agents, a next step might be to study the balance of experience in a laboratory setting. Are groups with a leader who has a reputation for expertise in a game more likely to use available information, engage in pre-emptive self-criticism, and reign in risks, even when group members have played many rounds and gained experience?
Finally, there is a sobering lesson for the politics of leadership. A team of seasoned veterans cannot fully compensate for an inexperienced leader, and inexperience in a presidential principal may enable or underwrite risky behavior by advisers. More generally, deviations from rationality depend not only on who you are and where you sit, but also who is in charge. The same experienced adviser may exhibit bias in some settings but not others. It is thus not enough to understand the biases of any one individual—the balance of characteristics within a group is also important. A fruitful next step in the new behavioral revolution would be to revisit questions about how powerful individuals interact, using theoretical and empirical tools to understand the political and institutional forces that bridge from the minds of decision makers to outcomes on the international stage.