Ask yourself if you believe that Hillary Clinton would have pursued the same course at the helm of American foreign policy as Donald Trump. As candidates, one advocated for the widespread use of torture, “carpet bombing” ISIS, abandoning the Paris Agreement on climate change, and nuclear weapons expansion in Asia. The other offered diametrically opposed views on these and countless other issues. But are you convinced that once in office, the strategic realities of the international system or domestic U.S. institutions would push both toward equivalent policies, with similar implications for America’s place in the world? If, as we suspect, the answer is “no,” then you have effectively conceded the core argument of Why Leaders Fight by Michael Horowitz, Allan Stam, and Cali Ellis.

In this compelling volume, the authors set out to restore the place of the leader in the study of international relations. They correctly argue that the emphasis on states and the structure of the international system, owing to the predominance of system level theory, has obscured the role of individuals in world affairs. Even the more recent turn toward domestic institutional explanations for international outcomes tends to gloss over the attributes and experiences of heads of state. More importantly, they argue that this imbalance has compromised the explanatory power of the prevailing theories of international politics. Seen in this light, the book is the logical end game of the decades-long process of unpacking “black boxes” and, in so doing, crawling back from third to first image theory.

Against this backdrop, Why Leaders Fight derives and tests a series of hypotheses concerning when, how, and why the characteristics of individual leaders influence their decision-making in arguably the most high-stakes aspect of international politics—the decision to go to war. Their starting point is that some leaders are simply more risk prone than others. They then endeavor to unpack the observable precursors of risk acceptance by mining existing work in history, psychology, sociology and other social sciences for plausible variables.

In the course of this review, Horowitz, Stam, and Ellis identify a series of personal characteristics and experiences associated with risky behavior, ranging from previous endeavors, like serving in a military force in a combat versus noncombat role, to childhood experiences such as the loss of a parent or birth order, to simple demographic characteristics including age and gender.

For instance, some research suggests that leaders’ experiences with combat can increase or decrease their risk propensity when it comes to armed conflict. Leaders who have personally experienced the horrors of war, the argument goes, are likely to be hesitant to pursue it. The archetypal example here, as recounted by the authors, is Dwight Eisenhower, who commanded U.S. troops in World War II before entering politics, and who famously commented that he “hate[d] war as only a soldier who has lived it can” (quoted on p. 136). In contrast, leaders who never saw combat may draw opposing lessons, focusing more on the glory, pomp, and circumstance surrounding the military, and consequently be more prone to engage in wars. The example offered by the authors here is Francisco Solano Lopez of Paraguay, who toured Europe in peacetime as part of his military education and later led his country into ruinous wars against Brazil and Argentina, during which he repeatedly rebuffed offers to end the conflict short of the near-total destruction of the nation and its people.

Deriving hypotheses about how life experiences influence leaders is one thing, but rigorously testing them is quite another. Here is where the authors arguably make their greatest contribution by developing the Leader Experience and Attribute Descriptions (LEAD) data set—a unique compilation of personal characteristics and life experiences of all leaders of all countries between 1875 and 2004. They measured and cataloged each leader’s military background, upbringing (e.g., birth order, parental status, legitimacy, etc.), educational status, family (e.g., marital status, number of children, age, gender, etc.), preleadership occupation, and subjective experiences (e.g., mental and physical health, wealth). This herculean undertaking represents a major contribution to the study of foreign policy and international politics. It will surely also prove valuable to scholars of comparative politics interested in the effects of individual leader characteristics on domestic decision-making.

Horowitz, Stam, and Ellis, however, go one step further by comparing their individual-leader risk factors—that is, those background characteristics that are hypothesized to make a leader more or less risk acceptant—with prevailing explanations from international relations scholarship concerning states’ decisions to go to war. These include traditional system- and state-level variables, like the relative power of actors in the system, system polarity, regime type (democracy vs. autocracy), geographic proximity of states, and others. They find that while many of the traditional variables emphasized by scholars of international relations, such as democracy, do matter, so too do some of the individual characteristics and experiences of leaders, particularly prior military and childhood experiences. In some instances the individual-leader characteristics appear to have more explanatory power than those of their competitors. They also find, as one might expect, that individual-leader characteristics interact with state-level factors, like regime type. For instance, among leaders with very low risk propensities (that is, leaders who are highly risk averse), differences in regime type matter hardly at all. Conversely, among leaders with very high risk propensities, those leading autocracies are considerably more likely to engage in militarized interstate disputes than their democratic counterparts.

Horowitz, Stam, and Ellis balance this first-of-its-kind large-n analysis with a series of case sketches of individual leaders located at relative extremes in terms of their predicted risk profile (that is, leaders whose life experiences should, per their hypotheses, have made them particularly war prone or averse). They profile such noteworthy leaders as Nikita Khrushchev, John F. Kennedy, Woodrow Wilson, Margaret Thatcher, Franjo Tudjman, Golda Meir, and several others. In most instances, the leaders they profile behaved in manners consistent with expectations. In those cases where war outcomes did not fit individual-leader profiles, they found evidence that systemic or institutional factors overrode individual-level factors.

This book is an impressive contribution, and the most comprehensive and rigorous analysis to date of the effects of leaders on interstate war. The combination of a thorough and highly informative review of individual-level theorizing, derivation of specific hypotheses concerning the relationship between individual-leader characteristics and war proneness, the development of a large-scale original data set making it possible to systematically test those hypotheses, the comparing of their predictions with extant explanations offered by international relations theory, and finally the integration of a wide array of individual biographical sketches of important leaders over the past century and a quarter make this an important book that belongs on the syllabus of any graduate-level course on international relations theory.

All that said, we do have some concerns and suggestions for extensions. First and perhaps most important, the theoretical underpinnings of the hypotheses are somewhat uneven, raising the possibility that the microfoundations of the argument are insufficiently strong to drive some of the finer-grained predictions. For instance, the notion that military experience matters to leaders’ risk profiles seems to be solidly supported by prior work. However, the more refined expectation that such experience can make a leader more war prone absent combat experience in a national military or given combat experience as a rebel (as opposed to a national military), yet less war prone given combat experience in a national military, find their bases in biographical studies of just a few high-profile leaders. This is not to say that these expectations are not intuitively appealing—for the most part they are—but the shakier theoretical footing leaves the door ajar for a rebuttal that the relationship emerges from the mining of a very large data set rather than from something more solid.

Similarly, when the authors turn to the effects of age, they empirically find that older leaders start more wars. They again offer intuitively appealing explanations—that older leaders have more experience bending state bureaucracies to their will and/or shorter time horizons for achieving their goals. But these explanations are again somewhat ad hoc, and the same problem therefore emerges. The evidence seems to support the proposition, but absent a clearer theoretical logic it is difficult to draw strong and generalizable conclusions. We should note, however, that to an important extent these issues arise through no fault of the authors—there is simply a quite uneven level of theoretical and empirical work on the underlying individual-leader risk factors in the literatures that they mine for their hypotheses.

Another issue that the authors might address in a follow-up study is the problem of censored data. In particular, their data set only includes individuals who successfully rose to become heads of state. This exacerbates the potential problem of selection. The disproportionate magnitudes of effects among leaders over the 90th and 99th percentiles highlight this concern that a small number of outlier leaders could drive the findings. The authors, to their credit, consider at some length the possibility that leader selection might produce biased estimates of the effects of individual factors on war proneness. For instance, might citizens in a period when war seems inevitable seek out leaders who are willing to employ military force? The authors do about as good a job as seems possible given the limitation posed by excluding unselected potential leaders. For instance, their secondary investigation of leaders who entered office without being directly elected (e.g., vice presidents who came into office because a president passed away or resigned)—thereby ruling out the possibility that citizens elected them because of their war proneness—is ingenious.

Still, a better fix would be to extend the analysis to include a random sample of political figures who competed for leadership but were not selected. Obviously, it would be difficult (more probably, impossible) to produce a comprehensive data set on all such failed candidates for leadership. After all, historians rarely write books about losers and also-rans. But it would surely be possible to do so for a subset of such leaders, particularly if this effort were focused on the more recent side of the book’s period of analysis. Doing so would have the added benefit—albeit perhaps primarily of value to a different set of scholars—of facilitating an investigation of the individual factors associated with coming to power in the first instance.

We would also like to plead for another addition to the data. The biographical sketches presented in the book (e.g., the relationships among Thatcher, Ronald Reagan, and Mikhail Gorbachev) make clear that interpersonal relations between leaders matter a lot for policy outcomes, at least in some instances. The authors note this in the conclusion but do not account for it in their data. Doing so could be a valuable addition to the LEAD data set, expanding its potential explanatory power and usefulness to a wide group of scholars.

Finally, we also wonder whether it would be revealing to follow up more seriously on the role of outliers in the relationships that are uncovered in this book. The authors tend to compare the behaviors of leaders at the 10th and 90th percentiles along their scale of risk propensity. This seems reasonable enough, but we found ourselves worrying that there might be such overwhelming outliers that they skew these percentiles. More broadly, this raises the possibility that those out at these extremes are different in type rather than simply degree. Put differently, there seems to be something more to the extraordinary risk propensity of a leader like Muammar Qaddafi than that he checked all the boxes on the index where most others checked just a few.

In sum, Why Leaders Fight is an important book and a valuable contribution to the international relations literature. As the authors clearly acknowledge, it does not “close the deal”; questions remain as to the plausibility of attributing causality to the individual factors compiled and investigated by Horowitz, Stam, and Ellis, and much further work can be done to extend the findings. However, it clearly succeeds in reminding scholars of international politics that they need not, and more importantly should not, check their instincts as individual voters and citizens that the choice of leaders matters at the door (recall the Clinton vs. Trump example). Rather, such factors should, to the extent possible, be assessed alongside the explanatory variables that dominate the literature in international relations. In this groundbreaking book, the authors have made a large and important first step toward making such assessments possible.