H-Diplo | ISSF Article Review 81
H-Diplo/ISSF Editors: Seth Offenbach and Diane Labrosse
H-Diplo/ISSF Web and Production Editor: George Fujii
Alexander B. Downes and Lindsey A. O’Rourke. “You Can’t Always Get What You Want: Why Foreign Imposed Regime Change Seldom Improves Interstate Relations.” International Security 41:2 (Fall 2016): 43-89. DOI: http://dx.doi.org/10.1162/ISEC_a_00256
Review by Barry Hashimoto, American University of Sharjah
Published by ISSF on 14 June 2017
“Poisoned Chalice or Just Poison? Foreign-Imposed Regime Change and Peace”
Reviewed by Barry Hashimoto, American University of Sharjah
Using a cross-national data set covering the end of the Napoleonic Wars to the year 2000, Alexander Downes and Lindsey O’Rourke have produced a fascinating analysis of foreign-imposed regime change in international relations replete with new quantitative results. The question these two political scientists address is this: When a state tries to change the leadership and domestic institutions of target states does this policy of foreign-imposed regime change lead to more peaceful relations or more conflictual relations between the intervener and the target? Downes and O’Rourke argue that “overt and covert [foreign-imposed regime changes] do not improve relations between intervening and target states. Often, they become worse” (85). This essay examines this argument and raises questions about the statistical evidence supporting it.
The research question addressed by Downes and O’Rourke is one of enduring relevance. Since the end of the Cold War the United States has embarked on crusades of overt and covert regime change against states posing a threat to centerpieces of the rules-based international order such as the general prohibition on the use of force between states, outright bans on genocide and other forms of mass atrocity, the frozen Cold-War status quo in the possession and use of unconventional weapons, and the sponsorship of terrorists operating globally. During the Cold War both the U.S. and the Soviet Union intervened regularly within their spheres of interest and in the non-aligned world to build and defend friendly regimes—and to undermine inimical regimes—in strategically significant locations. Europe’s former colonial powers also inflicted regime change on newly independent colonies in this period.
Foreign-imposed regime change was also frequent during the “thirty years’ crisis” of 1914-1945. Peacetime autarkies emerging from the wreckage of the classical gold standard and liberal internationalism saw foreign-imposed regime change as a tool for prosperity and survival. In Europe, Africa, and in Asia they used it to acquire new markets and new land, labor, and capital. Belligerents of the First and Second World Wars also resorted to such interventions in order to ensure a lasting peace, but also to build networks of common interest and common vision with which to advance fascism, democracy, capitalism, and communism. And although European powers maintained a cooperative pact in the long nineteenth century, they turned to foreign-imposed regime in a few intense great-power wars, in wars of national unification, and against the peripheral states of Portugal and Afghanistan. Far from Europe, states in Latin American and East Asia used foreign-imposed regime change liberally against one another. And before this, European states imposed regimes on “uncivilized” polities during the vast period of mercantilist colonial expansion of the fifteenth through the eighteenth centuries. European states also imposed regimes on each other as ideological polarization set Catholics against Protestants from 1520 to 1690, and monarchists against republicans from 1770 to 1850. Ensuring survival, peace, and prosperity, advancing interests and ideologies, and protecting the rules-based international order have long been the goals of this controversial practice.
Yet it is not obvious that such interventions have worked as their designers intended them to. Foreign-imposed regime changes are expensive and have delivered unexpected results. This is evident from the latest U.S. state-building efforts in Afghanistan and Iraq, and from the efforts of Middle Eastern states to cultivate friendly regimes in Syria and Yemen. Quantitative analysis shows that the process of foreign imposed regime change greatly increases the risk of civil war, presumably because it destroys domestic institutions and security forces. Democracies imposed from abroad have also been prone to failure. Foreign-imposed regime changes intended to enforce postwar peace agreements have had pacific effects on international relations—but often only temporarily. When a state merely installs a puppet regime in that of its adversary at war’s end—when it fails to democratize its adversary—inevitable changes in leadership in the target state and at home result in misaligned interests and disputes re-emerging over time, which are not contained by the pacifying effects of democracy. The evidence bears this out. All this suggests that foreign-imposed regime change is a poisoned chalice, and policymakers ought to consider it last after deterrence, diplomacy, and dispute resolution through international institutions. Yet a poisoned chalice is still a chalice: installing rulers with aligned interests and imposing institutions to lock-in a pacific foreign policy has greatly prolonged peace between the former adversaries of the twentieth century’s wars.
Looking beyond the postwar relations of ex-belligerents to the international relations of all states, Downes and O’Rourke supply the field with an ambitious and historically grounded view of the mechanisms by which foreign-imposed regime change might produce war or peace. Their argument is that these interventions generally fail to pacify relations between interveners and targets. At the heart of this are two familiar skepticisms in the theory of international relations: skepticism of the ability of states to alter interests and institutions abroad, and skepticism of the pacific effects of institutions.
Downes and O’Rourke argue first that it is difficult for a state to orchestrate even simple covert strategies like coups, assassinations, and electoral fraud from a distance. Overt institutional engineering is likely to be even harder. A reasonable guess is that a majority of foreign-imposed regime attempts over the last two hundred years failed. The result of this type of failure—especially a failure of covert action—is to turn “a tense relationship into an openly conflictual one” (46, 62-63). U.S. foreign relations with North Korea, Cuba, and Iran since the start of the Cold War are cautionary tales.
Even when they succeed, foreign-imposed regime changes may reform the target’s domestic politics in a superficial fashion. They mail fail to align the interests of the intervener and target even in the short run and generate principal-agent problems leading to disputes and conflict. Merely deposing foreign leaders does not change the target state’s external security constraints. It does not change the interests of powerful domestic actors within it. Nor does it install institutions that might permit more effective conflict resolution. This results in a fragile pact between puppet-master and puppet, such as that between Germany and France after 1871 or Rwanda and the Democratic Republic of the Congo after 1997 as shown in the authors’ study of the latter case. The authors note that “states’ foreign policy interests are often relatively stable and long-lived, and rarely boil down to the preferences of particular individuals,” while the civil war and regional conflict generated by foreign-imposed regime change can turn public opinion against the puppet-master (53, 56-67). By this logic, even foreign-imposed democratization may backfire, if the factions inimical to the intervening state take control of public office. In sum, foreign-imposed regime change is a gamble: failed attempts generate blowback, and the degree of possible success is limited.
Yet Downes and O’Rourke also theorize that there are scenarios where foreign-imposed regime change will result in more peaceful relations. When autocratic interveners impose repressive institutions on a target state with success, their puppet leader are more likely to survive and thus are able to avoid further conflict with the target, as was the case in the Soviet sphere of interest in the Cold War. When democratic interveners successfully democratize the target, the pacifying normative and institutional effects of democracy are available to help them resolve disputes peacefully. The long peace between the former Allied and Axis powers of the Second World War is emblematic of this cooperation. And when interveners merely restore to power deposed leaders—as France did in 1964 and 1995 in Gabon and Comoros, respectively—they are less likely to be seen as tools of foreign interests and thus more likely to survive in office while setting the target state back on the path to peace with the intervener. Acknowledging the myriad ways that foreign-imposed regime change might improve and pacify intervener-target relations leaves us with a nuanced view in which the target and style of interventions play a role in their success or failure.
Some of these arguments are disputable. For example, the core logic of the authors’ principal-agent theory suggests that when a state restores a deposed leader abroad, it will run into the same problem of diverging interests that arise when a leader installs a puppet abroad. The geopolitical and domestic factors that precipitated the deposed leader’s downfall in the first place remain unaffected by the intervention. In addition, while failed attempts at regime change might generate blowback, they also may credibly signal the intervener’s resolve to prevail in the disputes at hand despite paying the costs of regime change and sacrificing reputation in the intervention’s failure.
Nonetheless, the historical record appears to confirm much of what Downs and O’Rourke predict. Using a data set of directed dyads formed from all pairs of states in existence from 1816 to 2000, they estimate that regime changes that only install new leaders cause Militarized Interstate Disputes (MIDs)—threats, displays of force, minor clashes, or war—in the immediate ten-year or five-year period after regime change. Those which install new leaders and new institutions—democratic or not—are reported to cause neither more nor fewer MIDs, accounting for uncertainty, while regime changes to restore former leaders are shown to cause fewer MIDs. Covert regime changes of all types are shown to cause more MIDs (69-72).
The authors have done an impressive job in bringing to light new data on foreign-imposed regime change in the empirical sections of the paper. The analysis is thankfully free of many of the missing-data problems that have arisen in quantitative, historical work in the field. The findings are clearly presented. A thorough discussion of robustness checks is included in supplementary material online. In that document the authors explore the consequences of reasonable departures from their research design. Independent and dependent variables are recoded, tests for cross-sectional dyadic dependence are run, and alternative estimators are deployed to counter some of the potential biases of endogeneity and model-dependence. Some of these efforts could be extended to put the analysis on a more thorough footing. Explicit discussions of how the authors account for endogeneity (more on this later) and temporal dependence in a repeated-events framework are missing. But overall, the authors have produced an extensive analysis at a high standard.
A plethora of empirical findings emerge when one considers the article and supplemental material together. They defy a simple characterization. Many of the article’s reported findings are fragile, and whenever foreign-imposed regime change is estimated to cause more conflict, the effects are substantively small. They range from one-half to two percentage points per year on average. A reasonable conclusion to draw is that foreign-imposed regime change has virtually no effect on intervener-target conflict. Or to put it differently, a robust finding that foreign-imposed regime changes reduce conflict between interveners and targets fails to emerge. It bears mentioning that the modal condition of international relations in the authors’ data set is peace, and these interventions do little to change that. All this begs the question: Might the poisoned chalice just be poison—a costly adventure with no clear payoff?
There are a few reasons to pause before jumping to conclusions. Prior to producing their statistical estimates, the authors first discarded hundreds of thousands observations from the data set because they belong to pairs of states that are neither geographically contiguous nor contain at least one major power. This is likely to dramatically alter the joint distribution of the independent and dependent variables, biasing causal effect estimates and their standard errors. The rarity of both foreign-imposed regime changes and violent conflict in history certainly makes it challenging to infer the effects of one on the other, but that is a feature of history rather than a technical nuisance. Second, the authors lack data on failed overt foreign-imposed regime changes, and so they measure only the successes (66). The consequence is systematic measurement error in the variable measuring all foreign-imposed regime changes, which could lead to substantial bias. Third, the authors select MIDs as the dependent variable because it is “a more sensitive barometer” of conflict than the outbreak of war is (66). But the sensitivity of this barometer is offset by the insensitivity introduced by the author’s assumption that the effect of foreign-imposed regime changes lasts at most ten or five years (66). The consequences of this research-design choice are again unclear, although it is possible that it biases the study against detecting a persistent and pacific effect of foreign-imposed regime change. Lastly, the authors recognize the danger of post-treatment bias that arises when controlling for variables that are consequences of foreign-imposed regime change: for example, domestic political institutions, political similarity, and national capabilities. Little is typically done to avoid this sort of bias, but acknowledging it is important.
Each of these problems is surmountable by analyzing the full data set with variables measuring foreign-imposed regime changes corrected to account for failed overt attempts, by allowing foreign-imposed regime change to exert effects beyond ten years, and by omitting plausibly post-treated variables from the right-hand side of the regression equations. Perhaps none of these alterations would disturb the central null finding of this work. Yet this question is an empirical one. It is not satisfactorily answered by assumption or argument.
A problem that is more difficult to surmount may ultimately be what disturbs the central null finding in Downes and O’Rourke. It relates to a third familiar skepticism in the study of international relations: skepticism about our ability to do causal inference with historical data. A typical problem here is endogeneity as understood by the potential-outcomes framework. Potential outcomes are the values the dependent variable would take under “treatment” and “control” conditions regardless of whether units were in fact under treatment or control. An unmistakable conclusion of this framework is that correlation between the actual treatment status of units and their potential outcomes under treatment and control implies that a comparison of the observed outcomes of units in the treatment and control conditions will yield a biased estimate of the true causal effect of treatment on the outcome. One way to obtain a zero-correlation between treatment status and potential outcomes is to randomly assign treatment status across units. Another way to obtain this zero-correlation is to adjust the comparison of observed outcomes for treated and control units using pre-treatment confounding variables that are thought to cause changes in both the endogenous treatment status and levels of the potential outcomes. Mathematically, the use of control variables “subtracts” the bias introduced by a correlation between treatment status and potential outcomes.
This second strategy is clearly what Downes and O’Rourke adopt. They specify control variables on the right-hand side of probit regressions where the authors’ foreign-imposed regime change variables indicate treatment status and measures of MIDs are the observed outcomes. The authors rightly do so for the obvious reason that pairs of states in which one state imposes regime change on the other are highly unusual in international relations. They are adversaries with a history of entanglements and difficult issues under dispute. The target’s regime stands in the way of key goals of the intervening state. Crises between them have escalated to the overt or covert use of force. One or both are possibly insensitive to the consequences of serious breaches of international norms and law. Whether an intervening state ultimately attempts to impose a new leader or institutions in the target or not, the intervener and target are far more likely to end up in a fight than are other pairs of states. Yet such intensely adversarial states are also more likely to experience foreign-imposed regime change. In other words, highly adversarial pairs of states ‘select into’ foreign-imposed regime changes. This correlation in treatment and potential outcomes means that a naïve comparison of observed conflict levels in dyads that experienced foreign-imposed regime changes with dyads that did not will not yield an unbiased estimate of the causal effect of the interventions on conflict.
Yet the strategy of controlling for confounding variables only works if the control variables appreciably cause both treatment and outcome—or at least correlate with the causes without being consequences of the treatment. Downes and O’Rourke use a pragmatic approach to identify such variables. Select variables that are thought to cause the outcome: conflict (67). Variables causing conflict are also likely to cause states to attempt foreign-imposed regime changes. Yet the in-sample predictive power of the nine or ten variables used as controls in the authors’ many analyses is quite poor. This reflects both the well-known fact that predicting conflict is hard, and the fact that many good predictors of conflict are likely to be consequences of foreign-imposed regime changes, and thus useless as control variables. Indeed, the lack of a good argument about confounding—let alone good measures of the confounders—is a shortcoming of this work.
The plain fact is that learning the effects of foreign-imposed regime changes from the historical record is not easy. Fixed-effects regressions and matching fail in the presence of unobserved and potentially unknown confounders that vary across dyads and over time—confounders such as the entry and exit of policymakers with heterogeneous interests, beliefs, levels of resolve, and foreign policy doctrines. Sub-setting the data to limit the causal complexity in large data sets is a feasible but imperfect approach, one in which the researcher might “hold constant” key unobserved confounds by analyzing only pairs of states with a high propensity for conflict, such as states that just finished major wars. But this approach entails a shift in the research question and estimand. Instead of learning the effect of foreign-imposed regime change on conflict, we learn the effect of post-war foreign-imposed regime change on conflict among ex-belligerents. These two things are clearly different from the perspective of the policymaking community to whom Downes and O’Rourke recommend caution. Other approaches such as instrumental variables and bivariate estimators may give unpersuasive results due to their stringent parametric assumptions and exclusion restrictions. There is room for innovation in the identification and measurement of confounding variables, and in imagining clean research designs for causal inference on this front.
Whether foreign-imposed regime change is a poisoned chalice or just poison is still debatable. The literature of the past several decades was perhaps too sanguine that such interventions made for peaceful relations by aligning interests and imposing democracy. Downes and O’Rourke have advanced it by suggesting why foreign-imposed regime change might fail to produce peace between interveners and targets. Attempts that fail to take root or are done covertly may inflame hostile public opinion in the target state. Successful attempts still face a principal-agent problem in which a residual or growing divergence of interests causes disputes to re-emerge. Imposing democracy enfranchises a potentially hostile public and could give rise to new disputes. Yet states have imposed regimes abroad in the pursuit of diverse goals. Some states have indeed been intent on enforcing postwar peace treaties and preventing the outbreak of new wars. The evidence suggests that they succeeded. But others sought new markets and factors of production. They may have paid little regard to intervening in a way that would lead to peaceful relations with the target. Other states largely meant to enforce international law by building reputations as retaliators, again without much regard for the future of peace with the target. What states want from foreign-imposed regime change depends on the case at hand. Whether it results in war or peace is just one question of many that may matter to those who consider it. And as this essay has argued, there are substantial challenges involved in learning its effects on subsequent politics from the historical record.
Barry Hashimoto is Assistant Professor of Political Science at the American University of Sharjah in the United Arab Emirates. He held the positions of Global Postdoctoral Fellow at New York University Shanghai and Assistant Professor/Faculty Fellow in the Department of Politics at New York University. He received his Ph.D. in Political Science from Emory University and his A.B. in Government from Dartmouth College. His areas of focus are international security, international law, human rights, and the International Criminal Court.
©2017 The Authors | Creative Commons Attribution-NonCommercial-NoDerivs 3.0 United States License
 The phrase comes from Jeffrey A. Frieden. Global Capitalism: Its Fall and Rise in the Twentieth Century, (New York: W.W. Norton, 2006).
 See, e.g., Dan Reiter, How Wars End, (Princeton: Princeton University Press, 2009). John M. Owen IV. The Clash of Ideas in World Politics: Transnational Networks, States, and Regime Change, 1510-2010, (Princeton: Princeton University Press, 2010).
 Goran Peic and Dan Reiter, “Foreign-Imposed Regime Change, State Power, and Civil War Onset, 1920-2004,” British Journal of Political Science 41:3 (July 2011), 453-475.
Bruce Bueno de Mesquita and George W. Downs. “Intervention and Democracy.” International Organization 60:3 (Summer 2006), 627-649. Alexander B. Downes and Jonathen Monten, “Forced to Be Free? Why Foreign-Imposed Regime Change Rarely Leads to Democratization,” International Security 37:4 (Spring 2013), 90-131.
 Nigel Lo, Barry Hashimoto, Dan Reiter, “Foreign-Imposed Regime Change and Postwar Peace Duration, 1914-2001,” International Organization 62:4 (Fall 2008), 720.
 Dan Reiter, “The Poisoned Chalice of Foreign Imposed Regime Change,” E-International Relations (February 2011).
 Lo, Hashimoto, and Reiter 2008.
 Table 1 of Downes and O’Rourke 2016 (67) lists 32 failed covert foreign-imposed regime change cases and 25 successful covert foreign-imposed regime change cases from 1816-2000. While there are 95 successful overt foreign-imposed regime change cases, the authors did not assemble a list of failed overt foreign-imposed regime change cases for the same period, citing the large number of such failures presumed to exist (66).
 Bueno de Mesquita and Downs 2006.
 Bruce Russett, Grasping the Democratic Peace: Principles for a Post-Cold War World (Princeton: Princeton University Press, 1993).
 Ranjit Lall. “How Multiple Imputation Makes a Difference,” Political Analysis 24:4 (Autumn 2016) 414-433.
 See Replication Data and Supplemental Materials for: Alexander B. Downes and Lindsey A. O’Rourke, “You Can’t Always Get What You Want: Why Foreign-Imposed Regime Change Seldom Improves Interstate Relations,” International Security 41:2 (Fall 2016), 43-89. Accessed 10 March 2017 at https://dataverse.harvard.edu/dataverse.xhtml?alias=isec.
 See Supplemental Materials, 76: “The dependent variable in this analysis [of a data set matched on the incidence of a foreign-imposed regime change] is a dummy variable signifying whether a MID occurred at any point during the ten years after the year in which a FIRC occurred. It is necessary to use this variable since we can use only the actual FIRC-year for matching or else risk post-treatment bias.”
 Gary King and Langche Zeng, “When Can History Be Our Guide? The Pitfalls of Counterfactual Inference,” International Studies Quarterly 51:1 (March 2007): 183-210.
 Guido W. Imbens and Donald B. Rubin, Causal Inference for Statistics, Social, and Biomedical Sciences, (Cambridge: Cambridge University Press, 2015).
 The pseudo R-squared fit statistics reported in the Supplementary Materials are generally below 0.2 with the exception of the fixed-effects models reported in section K. These fit statistics are derived from models in which the foreign-imposed regime change variables are also on the right-hand side.
 Michael D. Ward, Randolph M. Siverson, and Xun Cao, “Disputes, Democracies, and Dependencies: A Re-examination of the Kantian Peace,” American Journal of Political Science 51:3 (July 2007): 583-601.
 Fixed-effects regressions as reported in section K of Supplementary Materials control only for the net effects of dyad-specific unobserved variables that are assumed not to vary over time. Moreover, accounting for these effects in probits results in a loss of about 70% of the data set (dyads in which there are never MIDs drop out). This problem is compounded by the fact that a number of other fixed effects are imaginable: leader effects, yearly effects, etc. Matching is a form of data pre-processing intended to reduce the dependence of statistical results on the specific functional form of the estimator. The researcher uses a statistical model predicting the treatment variable (i.e. foreign-imposed regime change) to produce a matched sub-sample in which the distributions of control variables achieve some measure of statistical balance in the observed treated and untreated samples. The intuition is that this mitigates bias resulting in the mathematical function used to correct causal effect estimates using these control variables. Matching does not reduce biases arising from unobserved confounding variables. There is, moreover, no conventional method to conduct matching with the sort of cross-national time-series that authors analyze. See the authors’ Appendix M of the supplemental materials for the matching results.
 Christopher H. Achen, “Toward a New Political Methodology: Microfoundations and ART,” Annual Review of Political Science 5 (June 2002): 423-450.
 This is an alternative view of the research design in Lo, Hashimoto, and Reiter 2008. There, we analyzed the effect of foreign-imposed regime change on postwar peace only for dyads concluding war with a peace agreement. In doing so we substantially amended extant data on the incidence of war, correcting incorrect classifications states at war and peace in the First and Second World Wars. The data set analyzed in that work is thus significantly different than the one analyzed in Downes and O’Rourke 2016.
 Lo, Hashimoto, and Reiter 2008.