A growing scholarship suggests that humanitarian concerns determine contemporary foreign policy behavior to a greater extent than traditional explanations such as power, regime type, or domestic audience costs account for. In particular, values and norms are influential for emerging international entities such as the European Union (EU).Footnote 1 Although the idea of the EU's commitment to norm-based politics is commonly found both in official statements and in academic research,Footnote 2 it has not been systematically tested. Instead, scholars have primarily explored the coherence of EU statements on humanitarian objectives such as abolishing the death penalty, promoting minority rights, and strengthening the jurisdiction of the International Criminal Court.Footnote 3
Although this approach has yielded important insights about the EU self-image as a normative power, there remains little systematic empirical evidence about its de facto foreign policy behavior. In particular, cases of EU action have not been studied in comparison with cases of nonaction, making it impossible to confidently draw conclusions about whether humanitarian concerns are indeed influencing EU policy. As a result, the idea that human rights guide EU policy decisions has been challenged by scholars who argue that normative statements are mere rhetoric whereas the policy pursued is determined by conventional security and trade objectives. The view that geostrategic concerns are more important than humanitarian objectives has been supported by case studies focusing on EU grand strategy, democracy promotion, development aid, and counterterrorism.Footnote 4
I provide an empirical investigation of EU foreign policy between 1989 and 2008 to identify whether it is primarily determined by human rights violations or power politics. While the EU has an extensive selection of different foreign policy instruments to choose from, including both carrots and sticks,Footnote 5 I follow convention in the international relations literature and focus on military or economic interventions. Such measures carry greater political and economic risks than more symbolic and preventive actions such as diplomatic protests or human rights conditionality clauses in trade agreements, making them suitable for exploring whether human rights concerns actually determine policy action.
Determinants of EU Foreign Policy Action
The first challenge for studying EU foreign policy consists of defining the universe of events that constitute EU action. While the literature on the EU's nature as a primarily inter- or supra-governmental entity is too large to address in this research note, it makes theoretical sense to explore both EU-specific measures as well as actions initiated by member states outside the common institutional framework. Such an approach is also motivated considering the EU's self-professed emphasis on multilateral action while the member states still maintain the option of independent policy in areas relating to national security interests.Footnote 6 To capture both of these dimensions, I have collected data on all military and economic interventions initiated by the EU (subsequently referred to as EU only) as well as actions initiated by individual member states either separately or in alliance with non-EU states (EU plus) between 1989 and 2008.Footnote 7 My analysis explores both these dependent variables, with all cases listed in Appendix 1 online.
Figure 1 shows the incidence of EU plus external action over time, illustrating how Europe has become increasingly active and diversified in its use of military and economic measures. Most observations in the data set consist of economic sanctions, but there are also two military attacks against an external state (Afghanistan 2001, Iraq 2003), some military actions in support of a civil war belligerent, and the establishment of several peace operations.Footnote 8
Figure 1. EU external action over time
Official EU policy documents for these measures repeatedly refer to the importance of human rights and democracy. Indeed, ever since the 1973 Copenhagen Declaration, EU statements have reiterated that joint foreign policy practice is based on common values such as the respect for human rights.Footnote 9 While this has led to the famous depiction of the EU as a “normative power,”Footnote 10 there is limited systematic evidence about whether these policies influence de facto behavior.
The most telling gap in extant research is the lack of empirical testing that compares cases of EU action with cases of EU inaction. The literature has identified a correlation between examples of human rights violations and EU action—such as against China 1989, Yugoslavia 1991, Nigeria 1993, Sudan 1994, Afghanistan 1996, Indonesia 1999, Zimbabwe 2002, and Uzbekistan 2005—but few studies have compared these instances with situations where humanitarian atrocities are met with inaction. Unless all countries where human rights abuse occur are included in the analysis, it is impossible to draw conclusions about whether human rights indeed influence EU action or not. To my knowledge, there is only one previous systematic study that explores European (and US) economic intervention against human rights abuse and that focuses only on violations by authoritarian states.Footnote 11 Within this subset, findings suggest that human rights abuse is likely to be punished, but that the response seems to be conditional on alliance structures, trade interest, and the context of the atrocities.
Humanitarian atrocities can take many forms, ranging from genocide or mass killings to imprisonment, torture, forced labor, or restrictions on movements and press freedom. Because this study focuses on responses that constitute costly foreign policy action, I expect that the severity and visibility of the humanitarian atrocities are important. The following analysis uses data on human rights violations from two different sources that are collected independently of each other. The first consists of the physical integrity rights index from the CIRI Human Rights data project.Footnote 12 This index is made up of information on government use of torture, disappearance, political imprisonment, and extrajudicial killings for any given year based on the yearly reports from the US State Department and Amnesty International. My second indicator consists of the annual number of victims of one-sided violence taken from data collected by Uppsala Conflict Data Program (UCDP).Footnote 13 This data set includes information on intentional killings of civilians by the government of a state or by a formally organized group resulting in at least twenty-five deaths per year, excluding killings that occur in custody. Because the sources used in creating the UCDP one-sided violence estimates are drawn from publicly available material such as media and nongovernmental organization (NGO) reports, declassified archival material, and scholarly investigations, this measure captures visible and severe humanitarian atrocities. However, comparing UCDP estimates with the CIRI physical integrity (PI) scores reveals that the measures correlate strongly for country-years in the most repressive states. For countries with a PI score of 0 to 4 (the worst cases), this correlation is statistically significant at the 95 percent level; for countries with a PI score of 5 it's significant at the 90 percent level. There are no statistically significant correlation for states with a PI score of 6 to 8. The relationship between the two measures is illustrated in Figure 2.
Figure 2. Adjusted predictions of HR violations with 95% confidence levels based on yearly data 1988–2008
Empirical Investigation
Table 1 provides a first glance at the relationship between human rights abuse and EU action. Models 1a through 2b consist of pooled cross-national analysis of all countries between 1989 and 2008 on whether the EU acted against countries with poor human rights records. Models 3a to 4b are time-series cross-sectional to explore whether EU action was introduced in years when the humanitarian situation was particularly poor. The dependent variables are the onset of EU-only or EU-plus action in the form of military intervention, the deployment of peacekeepers, or the imposition of economic sanctions. Because the purpose of this investigation is to identify when and where the EU acts, subsequent years of continued EU intervention are excluded from the analysis.Footnote 14 All models are estimated using the RELogit procedure that corrects for the systematic underestimation in normal logit models of the probability of rare events.Footnote 15 The time-series cross-sectional models also include the year of the observation and use robust standard errors clustered on country.
Table 1. Humanitarian atrocities and EU action
Notes: Estimations performed using Stata 12. Standard errors are in parentheses. * p < .10; ** p < .05; *** p < .01.
The first measure for human rights abuse, used in Models 1 and 3, is the physical integrity rights index from the CIRI Human Rights data project. This variable measures governmental use of torture, extrajudicial killing, political imprisonment, and disappearance. I include the average country-score for the time period (Model 1) or the yearly scores (Model 3). The coefficient is negative and strongly statistically significant indicating that EU action is more likely against countries where the government has no or little respect for these human rights.
Models 2 and 4 include the other measures for humanitarian atrocities consisting of the log of yearly fatalities of one-sided violence in a country. Model 2 includes the country average fatality count for the time period, whereas Model 4 is disaggregated by year. The bivariate results using data on lethal humanitarian atrocities show findings similar to the CIRI index; when more civilians are killed, there is an increased likelihood of EU action. These findings indicate that there is a correlation between human rights abuse and EU foreign policy, as research on EU statements and case studies has suggested. Additional analysisFootnote 16 shows that EU action generally responds to situations of highly visible human rights atrocities—a finding in line with expectations, considering that both economic and military interventions are costly measures. By separating the different CIRI components and reestimating the model, I find that increases in government killings or political imprisonment motivate EU action and that, in particular, worsening human rights in states already holding political prisoners are swiftly punished by the EU. For torture and disappearances, there are no similar discernible effects.
Alternative Explanations for EU Foreign Policy
The criticism directed against the concept of a norm-based EU policy suggests that other factors influence the selection for where the EU chooses to act in support of human rights. As Wood puts it, “the EU's mission civilisatrice is susceptible to a relatively straightforward if unpleasant realpolitik that can expose a rhetoric-behaviour gap.”Footnote 17 In particular, studies have argued that geostrategic considerations influence EU overall strategy, as well as specific policies relating to democracy promotion, counterterrorism, development aid, and sanctions.Footnote 18 It is therefore necessary to explore whether traditional international relations theory provides more convincing explanations for EU interventions.
According to realist theory, the foreign policy behavior of international system actors who are primarily concerned with their own survival can largely be understood as a function of their respective power. Based on this view, economic and military intervention is less likely against powerful states or alliance partners. To explore this, I include variables for economic and military power status: gross national product (GDP) per capita in 2009 US$ and military expenditure in 2010 US$.Footnote 19 Joint alliances are captured by a count of a country's membership in the North Atlantic Treaty Organization (NATO), the Commonwealth of Nations, or the Organisation international de la Francophonie.Footnote 20
Liberal international relations theory suggests that regime type or economic interdependence will influence the likelihood of military or economic intervention. Because all EU member states are considered democratic, action is more likely against less democratic statesFootnote 21 or against countries with little economic openness. Regime type is measured on an autocracy-democracy continuum,Footnote 22 while trade is measured as a country's share of world trade in 1988.Footnote 23
A similar set of foreign policy explanations focuses on domestic opinion. Because political survival is the prime concern for leaders, they will be sensitive to public opinion regarding whether to engage in certain policies.Footnote 24 Although there is no EU-wide constituency to consider, concern for prestige (at least partly) contributed to EU willingness to act as a peace broker in the conflicts in former Yugoslavia.Footnote 25 To control for domestic audience concerns, I include a measure of whether the public is positive to the EU, taken from the annual Eurobarometer survey findings. I also include two measures for election years: the first indicates elections to the European Parliament, while the other indicates election years in major EU powers, defined as Germany, France, or the UK.
Table 2 shows the findings from the multivariate analyses. Models 1 to 4 consist of RElogit estimates of determinants for EU action 1989–2008. Across all models, there is a positive and statistically significant correlation between increased violence against civilians and the probability of EU action. Postestimation analysis of the full model for EU only (Model 4a) show that the probability of EU action against a country increases from 0.5 percent (95 percent confidence interval 0.2 to 1.4 percent) to 1.4 percent (0.5 to 4.1 percent) when moving two standard deviations from the median value that is no one-sided violence; an increase of more than 200 percent holding all other variables at their mean.Footnote 26 Two standard deviations in the data represent the killing of approximately thirty-two victims so this should be considered a relatively moderate change. The effect is similar for EU plus action (Model 4b), moving from 0.6 percent (0.3 to 1.3 percent) to 1.6 percent (0.9 to 3.0 percent) after this relatively small increase of civilian killings. While these probabilities may seem small, it is important to remember that this analysis focuses on the rare event of EU action onset within its global data over a period of twenty years.
Table 2. Multivariate analysis of EU foreign policy action
Notes: Robust standard errors clustered on country (in parentheses). Estimations performed using Stata 12. * p < .10; ** p < .05; *** p < .01.
The findings are partly in line with expectations. The EU is less likely to act against its allies, even though the statistical significance of this correlation is sensitive to model specification. Additional estimations show that it is not a specific alliance (NATO, Commonwealth, Francophonie) driving this result but that the effect becomes visible only when all alliances are pooled. Table 2 also shows that negative measures are more likely to be imposed on nondemocracies, which corresponds with the EU's stated policy of promoting democracy as well as protecting human rights.
However, contrary to what realist and liberal theories can predict, the EU is more likely to act against more economically powerful countries and those more integrated in international trade. Because the dependent variable is the onset of EU measures rather than the incidence, this may actually be an underestimation of this effect because, for example, China is dropped from the data from 1989 onward. There are two possible explanations for why economically strong countries are more likely to be targeted by EU action. The first is a strategic calculation that the protrade nature of the EU means that economic interventions are perceived as more credible and thus can be expected to be more effective against countries relying on trade. Because the imposition of sanctions is also costly for the sender country,Footnote 27 such calculations support the view of the EU as willing to bear costs to pursue normative goals. The second explanation is that these correlations indicate the importance of geostrategic factors that are not sufficiently modeled in the global country-year setup. Because countries bordering the EU are the relatively most economically powerful states in the international system, this finding could then simply confirm that the EU is more likely to act in its immediate vicinity. Because spatial proximity has been found to increase both the level of cooperation and the risk of conflict between states, it is necessary to control for possible geostrategic influences on EU action.Footnote 28
Figure 3 describes the geographic characteristics for where the EU has imposed military and economic interventions. The different regions are coded in accordance with EU praxis contained in the 1993 Copenhagen criteria, the 1995 Barcelona Declaration, and the 2004 European Neighborhood Policy. Non-EU Europe includes countries that are theoretically eligible for accession including Turkey. The EU Near Abroad consists of southern and eastern Mediterranean states and the non-European former Soviet Union. I also distinguish separate regions for sub-Saharan Africa, the Middle East, Asia/Oceania excluding the Soviet Union successor states, and the Americas. Figure 3 indicates that geostrategic considerations have influenced the prevalence of EU action. Although the EU has acted most often in its immediate vicinity, this is not a consequence of a linear distance-to-target trend. Following European states, the EU is most likely to act in sub-Saharan Africa and then in the Asia/Oceania region, while it generally has adopted a careful approach toward countries in its Near Abroad and the Middle East.
Figure 3. Regional characteristics of EU foregin policy action 1989–2008
A Multilevel Model of Geostrategic and Human Rights Concern
A reading of EU policy documents over time reveals two common themes: the emphasis on the aim to protect democracy and human rights, and the promotion of intraregional governmental cooperation. The latter can be identified with regard to accession rules, as part of the drafting of trade agreements, and notably in order to prevent poor human rights practices through conditionality clauses in aid agreements.Footnote 29
To explore how human rights violations lead to EU action while properly controlling for regional effect, I model the dependent variable—eu action—using mixed-effect multilevel modeling to analyze country-level characteristics and regional-level properties simultaneously.Footnote 30 The advantage of this procedure is that all the different kinds of effects can be considered jointly in a single multilevel model. The downside is that it implicitly assumes that the variation on the macro level (the region level) is captured by the included macro variables.Footnote 31 However, since I have no ambition to explain the difference across regions but use the model to explore the effect of unobserved region specifics, this should not be a problem. The model has two components: a logistic regression for the country-years predicting the binary outcome given country-level predictors with an intercept that can vary by region, and a linear regression with six data points predicting the region intercept from region-level predictors. The regression at the country-level is as follows:
$${\rm Pr}\lpar y_{i} = 1\rpar = {\rm logit^{-1}}\lpar X_{i} \beta + \alpha_{\,j[i] }\rpar \comma \; {\rm for} \;i = 1\comma \; \ldots \comma \; n\comma \;$$where X is the matrix of country-level predictors and j[i] indexes the region where country i is located. The second part of the model is the regression of the region coefficients as follows:
$$\alpha_{\,j}\sim {\rm N}\lpar U_{\,j}\gamma\comma \; \sigma^{2}_{\sigma}\rpar \comma \; {\rm for} \;j = 1\comma \; \ldots \comma \; 6\comma \;$$where U consists of a matrix of region-level predictors; γ is the vector of coefficients for the country-level regression; and σσ is the standard deviation of the unexplained group-level errors. While the country-level regression is estimated with fixed effects, I allow the intercept α to vary across regions, assuming a normal distribution.
Table 3 shows the findings from the multilevel analyses of both humanitarian and regional influences on EU action. The models are estimated with the realist and liberal controls as earlier specified since domestic audience constraints arguably should have less influence on the interregional variance of targets. Across all models, EU external action continues to correlate strongly with situations of civilian victimization, even when accounting for the influence of geostrategic concerns.Footnote 32 In fact, the explanatory power for some of the indicators from realist or liberal theories decreases when spatial dimensions are explicitly modeled. As these results indicate, the EU should be considered an actor with global ambitions in its foreign policy, meaning that studies focusing on just its relations in the near neighborhood may be too narrow. That said, there is still interregional variation in the EU's foreign policy behavior that suggests there may be different thresholds for EU sensitivity in reacting to humanitarian atrocities. This may in part be explained by the regional variation in human rights abuse because this type of violence has been found to cluster in certain areas. According to Eck and Hultman, Africa has suffered almost 50 percent of global fatalities of one-sided violence in 1989–2004 even when excluding the genocide in Rwanda (including it pushes the figure to above 90 percent).Footnote 33
Table 3. EU action; multilevel model
Notes: Estimations performed using Stata 12. Standard errors are in parentheses. * p < .10; ** p < .05; *** p < .01.
To explore interregional differences in size of the effect of human rights abuse on EU foreign policy, I estimate the predicted likelihood of EU action toward a hypothetical scenario played out in different regions. The basis for this hypothetical case is Egypt before the “Arab Spring.” Following protests against the regime in neighboring Tunisia in January 2011, there were calls for antigovernment demonstrations at the central Tahrir Square in Cairo with a turnout that was unprecedented in Egypt. As demonstrations continued, and President Hosni Mubarak's offers of concessions were rejected, the Egyptian police started attacking protesters in the square. The EU response was initially subdued—EU foreign ministers produced a statement encouraging peaceful change. It took until March, when the revolution in Egypt was complete, before the EU acted by offering humanitarian aid and, following a request from the new Egypt regime, placing targeted sanctions against the Mubarak family.
But what if Egypt had been located in another region of the world? How likely is it that the EU would have reacted? Using Clarify,Footnote 34 I estimate the predicted probability of EU action in a hypothetical Egypt where I manipulate the region and the severity of violence. The demonstrations in Egypt before President Mubarak's resignation on 11 February 2011 reportedly led to between 131 and 846 deaths, and these two estimates are used as benchmarks for severity.Footnote 35
Figure 4 shows the change in predicted probabilities of EU action depending on region and regional power when all other covariates are set to the observed values of Egypt before the Arab Spring.Footnote 36 The change is positive and statistically significant at the 95 percent level for all regions using both the low and high estimate for violence severity. This indicates that there is a consistent global increased likelihood of EU action when civilians are being killed. In the Egypt example, state repression of demonstrators in Cairo in January 2011 increased in the predicted probability of EU action in Egypt with at least 2.8 percent (low estimate, EU Near Abroad; 95 percent confidence interval 0.5 to 16.2 percent). If the higher estimate of fatalities is used instead, this predicted increase in probability would be 5.6 percent (0.7 to 33.1 percent). Had the hypothetical Egypt instead been part of a different geographic region, the size of the change with statistical confidence at 95 percent is at the lower bound from 0.2 percent (Middle East, 135 deaths) to 1.7 percent (Sub-Saharan Africa, 846 deaths). Because these estimates combine two relatively rare phenomena in the form of civilian killings (especially the higher estimate) and the onset of EU action, the confidence intervals for predicted change are large. This exercise shows that the likelihood of EU action consistently increases following humanitarian atrocities, and not necessarily against excessive atrocities. The EU is, however, particularly sensitive to events in non-EU Europe, followed by sub-Saharan Africa, the Middle East, Near Abroad or Asia, and the Americas.
Figure 4. Change in predicted probablity of EU action after civilian killings
As a way of assessing model fit, I compare these findings with out-of-sample post-2008 EU foreign policy actions.Footnote 37 According to the UCDP one-sided violence data set (version. 1.4-2012), government actors in fourteen countries killed at least twenty-five civilians in at least one year between 2009 and 2011. Six of these (Democratic Republic of Congo, Cote d'Ivoire, Sudan, Libya, Myanmar, and Somalia) were already subjected to either UN or EU action. Among the remaining countries, none were located in Europe and just one (Syria) was in the EU Near Abroad.Footnote 38 Because Syria had the most severe humanitarian atrocities in the time period, it is in line with predictions that the EU responded by economic intervention in May 2011, specifically targeting members of the regime “responsible for the violent repression against the civilian population.”Footnote 39 However, the EU activity in its Near Abroad also included economic sanctions against Egypt in 2010 and Tunisia in 2011, and support to antigovernment forces in Libya 2011. Although the extent of EU action in this region increased compared with the 1989–2008 time period, this can be explained by the unexpected events of the Arab Spring and government repression of these prodemocracy movements. In sub-Saharan Africa, the EU acted in response to one-sided violence by the government in Guinea 2009, but not in Rwanda 2009, Nigeria 2011, or Madagascar 2009. Of these, only Guinea and Rwanda led to more than 100 fatalities, and while the EU specifically responded to the repressive actions in Guinea, those in Rwanda occurred while UN sanctions were winding down. The EU has, however, subsequently frozen development aid to Rwanda, which suggests consistent sensitivity to civilian victimization in this region. As before, the EU has been less active in response to human rights violations in the Middle East with no intervention either in Yemen or Bahrain 2011. However, economic sanctions imposed against Iran 2011 were explicitly motivated by extrajudicial killings even though those are not captured by the UCDP one-sided data. Finally, there was a single case of government one-sided violence in Asia (Vietnam 2011) that drew no EU response and no cases in the Americas.
With the exception of the Arab Spring, the out-of-sample evidence corresponds with the findings from the statistical analysis: EU foreign policy is influenced by humanitarian concerns. This practice does not seem to be influenced by a country's relative importance for trade or the risk of overspill in the form of refugee flows—such effects would be more pertinent for action in the Near Abroad or the Middle East. Instead, the EU is particularly attentive to human rights violations in sub-Saharan Africa. One possible explanation for this is that sub-Saharan Africa, despite its relative geographic distance from Europe, has a long history of common EU policy approaches, including institutional commitments to the promotion of human rights. Already the 1957 Treaty of Rome included provisions for African colonies to be included in the free trade area while aid became partly coordinated through the European Overseas Development Fund. As decolonization gathered speed, the relationship was revised through the first international agreement ever signed by the EU, the 1963 Yaoundé Convention.Footnote 40 This trade arrangement and its subsequent renegotiations (Yaoundé II and Arusha in 1968, Lomé in 1975) led to EU promotion of regional organizations as counterparts in Africa (Organisation of African Unity (OAU), Economic Community of West African States (ECOWAS) and the Caribbean (Caribbean Community [CARICOM]), with the aim to stabilize future economic interdependence.Footnote 41 During the 1990 renegotiations of the Lomé Convention, the EU introduced a political element into these economic agreements in the form of so-called conditionality clauses through which aid can be suspended following human rights violations by the recipient state.Footnote 42 While these clauses are primarily intended as preventive measures, they also institutionalized the EU's commitment to protecting human rights by delegating responsibility from the Council to the Commission.Footnote 43 As a consequence, EU policy should arguably be more consistent now its response to humanitarian atrocities is no longer influenced by short-term political fluctuations such as the current political leader's personality traits or domestic audience demands.
To illustrate EU policy sensitivity to human rights violations, consider the EU policy vis-à-vis Zaire (now Democratic Republic of Congo) in the early 1990s. During the Cold War, the West provided extensive support for the Mobutu Sese Seko regime and the country was one of the main recipients of European development aid, primarily provided by Belgium and France.Footnote 44 Following four months of partial political liberalization in early 1990, Mobutu became concerned with increased opposition activism and sought to reassert his hold on power. Repression of the opposition increased and reports of a massacre of more than 100 students at the Lubumbashi University emerged in late May 1990. The Zairian government denied the reports and rejected demands for an impartial investigation that led to decisions by Belgium, France, and the EU to cut assistance to Zaire and strong statements of condemnation.Footnote 45 In the subsequent year, the European reduction of development assistance was followed by similar US measures. Large-scale violence erupted again in Kinshasa in September 1991 when mutinous troops went on a looting spree, leading to French and Belgian soldiers deployed to protect Europeans, although it was made officially clear that this should not be interpreted as support for the regime. The international pressure did succeed in establishing a process for a transition toward democracy, but Mobutu renounced the process in December 1992, which was followed by widespread popular protests across Zaire. On 28 January 1993, after signs that some members of the army were defecting to the opposition, Mobutu-loyal elite troops massacred several hundred regular soldiers and civilians in Kinshasa. The death toll was estimated at between 259 and 1,000 victims, and included the French ambassador who was shot while standing at the window of his office.Footnote 46 Condemning the violence, France, Belgium, and the United States jointly demanded on 3 February that Mobutu resign and hand over power to the opposition. Mobutu instead appointed a new prime minister but as government repression continued in February and March, the EU Council on 7 April 1993 imposed an arms embargo on Zaire accompanied by travel restrictions on senior Mobutu-loyal officials.Footnote 47
Extending the Analysis: Human Rights, the EU, and the World
The investigation has indicated how the importance of human rights expounded in EU rhetoric corresponds with empirical evidence but that the effect differs across regions. A possible limitation of these findings is the absence of controls for the behavior of other international actors. As the example from Zaire shows, the EU does not act alone in a geostrategic vacuum and, according to official documents, relies on multilateralism and alliances to a greater extent than individual states in its foreign policy activity.
Table 4 replicates Model 3a in Table 3, adding controls for UN and US activity in the preceding and the same year. For the UN, I code mandatory sanctions imposed on a state or the deployment of peacekeepers, while US action consists of economic and military intervention.Footnote 48 As could be expected by EU proclamations, it is difficult to disentangle EU-specific foreign policy action from efforts involving other actors in the international system. A first glance at the findings in Models 1, 2, and 4 suggests that EU action occurs either in countries that already have received UN attention, or occurs at the same time as UN and US action. However, a closer look at the data and the within-year temporal ordering of events reveals that the empirical story is somewhat different.
Table 4. EU, UN, and US action: multilevel model
Notes: Control variables (nor reported) include economic and military power, alliance, regime type, regime type (sq), and trade. Estimations performed using Stata 12. Standard errors are in parentheses. * p < .10; ** p < .05; *** p < .01.
First, the vast majority of cases where EU measures have been introduced in the aftermath of UN attention consist of situations where UN sanctions have been removed but where the EU has retained them for longer. These decisions to keep sanctions were motivated by the risks of a resumption of humanitarian atrocities with the typical case being all the successor states of former Yugoslavia in 1995.Footnote 49 Looking at the within-year temporal ordering for measures introduced in the same year, it becomes clear that the United States or the EU often acted quicker than the UN, which hardly is surprising considering that the UN can act only if there is a threat to or endangerment of international peace and security. The EU was the first to act following human rights violations in China 1989, the Iraqi invasion of Kuwait 1990 (two days before the UN), Yugoslavia 1991 (two months before the UN), Indonesia/East Timor 1999, Zimbabwe 2002, and Uzbekistan 2005. However, the EU followed the lead of the United States in Myanmar, Nigeria, Sudan, and Afghanistan, while the intervention against the Democratic Republic of Congo 1993 was a joint French-US initiative. Thus, while these findings indicate that there clearly is a multilateral aspect to EU foreign policy, this does not constitute a systematic practice of following the lead of only the United States or the international community at large.
Table 4 also shows that even when controlling for the behavior of the UN and the United States, humanitarian atrocities correlate with EU action with a statistical significance at the 95 percent confidence level. In Models 5 and 6, I estimate the determinants for UN and US economic and military intervention, and find that these also correlate strongly with situations of humanitarian atrocities. This provides tentative support for the general argument that foreign policy can be based on normative concerns rather than realist objectives and suggests that the EU may not be unique in this respect. In this way, this study complements and adds to recent studies that find that humanitarian motivations are important for at least some types of international engagement, such as the allocation of development aid, the offer of mediation in civil war, and the deployment of peacekeepers in postconflict situations.Footnote 50
Conclusion
The first contribution of this study is the presentation of unique data on EU foreign policy action 1989–2008, covering fifty different military or economic interventions against thirty-one different countries. To my knowledge, there is no similar comprehensive data available for researchers interested in furthering the research of the EU as an international actor. Because much of the extant empirical work has focused on the internal decision-making processes in Brussels, there is a need to pay attention to the actual deliverance of policy and policy outcomes on the ground. By making these data accessible to the scholarly community, this research note may open new avenues for comparative research on the current international systems, including to what extent EU foreign policy is different from other actors.
Empirically, this research note performs the first systematic cross-national investigation of EU foreign policy behavior. Following statistical analysis through cross-sectional, time-series cross-sectional, and multilevel modeling techniques, I find that EU action correlates with human rights violations, regardless of whether this is measured by outright killings or through the use of torture, imprisonment, and disappearances. These findings are robust to controlling for competing explanations drawn from realist, liberal, and domestic audience cost theories. That said, I also find that geostrategic considerations influence the size of the effect because the EU is likely to act against primarily non-EU countries in Europe, followed by Sub-Saharan Africa, EU Near Abroad, the Middle East, Asia, and finally the Americas. I extend the analysis to explore whether the EU is taking the initiative for responses to human rights violations or following the lead of the UN or the United States. Although this exercise indicates that multilateralism is influential for EU action, qualitative evidence suggests that the EU has on many occasions been the first to react to atrocities.
Finally, this piece also contributes to scholarship on whether military or economic intervention is effective in stopping humanitarian atrocities. In assessing the effectiveness of international action, it is important to consider the possible selection bias relating to where the measures are imposed. As my findings show, interventions are primarily aimed at the worst situations where, arguably, they can also be expected to have the least chances of success. Accounting for this selection bias is becoming standard practice in studies on peacekeeping,Footnote 51 and the findings from this study suggest it should also be considered in future scholarship when focusing on human rights compliance. An important challenge for future scholarship is exploring whether it is possible to prevent or stop human rights abuse. The stated commitment of humanitarian concern goes beyond rhetoric and is visible in EU's foreign policy behavior.
Supplementary material
For supplementary material for this article, please visit http://dx.doi.org/10.1017/S0020818314000368.
