Nationalism and Conflict: Lessons from International Sports

Abstract

Does nationalism make interstate conflict more likely? Many believe so, arguing that it led to conflicts such as the Spanish-American War, the two World Wars, and Russia’s recent intervention in the Ukraine. However, others contend that strategic constraints greatly limit the effects of nationalism on state behavior. Resolving this debate has proven difficult because of endogeneity and measurement issues. I overcome these problems by analyzing one of the most powerful sources of nationalism in the modern era—international sports. I first investigate several cases in which surges of nationalism from sporting events led to military or political conflict between countries. I then analyze a regression discontinuity created by the format of the World Cup qualification process from 1958 to 2010. The results provide strong evidence that World Cup nationalism increases state aggression, especially for countries where association football (soccer) is the most popular sport. I also explore a case from the dataset—Senegal in 2002—to illustrate how World Cup nationalism led to a specific dispute in my sample.

Scholars view nationalism as a cause of international conflicts ranging from the Napoleonic Wars to the US invasion of Iraq following September 11 (McCartney 2004, 400; Cederman, Warren, and Sornette 2011, 606). Researchers argue that it can increase enmity between countries (Schrock-Jacobson 2010, 25–28), undermine international cooperation (Walt 2011, 15), motivate societies to fight costly wars (Posen 1993, 81), and cause governments to overestimate their relative military power (Snyder 2000, 67). Moreover, the idea that nationalism makes conflict more likely plays an important role in many theories of war. Examples include the theory that war can result from surges of nationalism following revolutions (Mansfield and Snyder 1995, 19–20) and the theory that charismatic leaders who manipulate national sentiments make conflict more likely (Byman and Pollack 2001, 141).

Nevertheless, some scholars question whether fluctuations in nationalism actually cause conflict (Posen 1993, 121–22; Laitin 2007, 1–28). International constraints, such as considerations about relative capabilities and strategic interdependences, might dominate foreign policy decision-making, or nationalism might not fluctuate enough to have a noticeable impact on state behavior. Two major challenges make this debate difficult to resolve. First, no dataset exists that tracks nationalism across the international system in a comprehensive way, limiting what researchers can do quantitatively. Second, surges of nationalism often occur when tensions between countries are already high. This endogeneity problem makes it difficult to tell whether nationalism causes conflict or merely accompanies it (Posen 1993, 122).

To overcome these challenges, I look at surges of nationalism that were created by international sporting events. International sports get around the measurement problem because nearly everyone acknowledges that they strengthen national identities. Many studies show that international sporting competitions often increase feelings of national unity and antipathy toward other countries (Maguire, Poulton, and Possamai 1999, 68–85; Toohey and Taylor 2006, 80–86; Tzanelli 2006, 485–99; Vincent, Kian, Pedersen, Kuntz and Hill 2010, 209–18). International sports also resolve the endogeneity problem because they are largely exogenous to politics. That is, they exist outside the political realm, but they inject doses of nationalism into it. Thus, if we find that surges of nationalism from sports frequently spark conflicts between countries, we can attribute these conflicts to nationalism without worrying that the nationalism merely arose due to growing political tensions.

I begin my analysis by investigating a number of cases in which scholars argue that nationalism from sporting events caused military or political conflict between countries. These cases include the 1969 Football War between El Salvador and Honduras, the 2009 Egyptian-Algerian World Cup Dispute, and the 2014 Serbian-Albanian Drone Conflict. After explaining how nationalism triggered international crises in these cases, I draw several important lessons from them about the nature of nationalistic conflicts started by sports.

I then analyze a regression discontinuity created by the World Cup qualification process from 1958 to 2010. Over this period, many countries qualified for the World Cup by playing a round of games against other states and achieving a top position in the final standings. This format makes it possible to compare the group of countries that barely qualified to the group that barely fell short. These countries went to the World Cup or stayed home based on small differences in their records after many games. Thus, which ones participated in the World Cup should be close to random, making this research design very similar to a randomized experiment.

Using this approach, I construct a sample of countries that barely qualified or barely missed the World Cup. Specifically, I select pairs of countries that were separated by no more than two points in the standings, provided that the qualifier scored at least five points. I made these design choices prior to collecting the data, believing that they would lead to a sufficiently large sample under which qualification was as-if random. In total, the sample consists of 142 countries. The qualifiers and nonqualifiers are balanced across a wide range of political, economic, and demographic factors, which supports the as-if random assumption. They are also balanced on past levels of aggression, which I measure in the standard way as the number of militarized interstate disputes (MIDs) that a country initiates.

The results show that going to the World Cup increases aggression substantially. The countries that barely qualified experienced a significant spike in aggression during the World Cup year. The disputes they started also tended to be much more violent than the disputes started by the nonqualifiers. The results hold under various robustness checks, and the estimated treatment effect is much larger for countries where soccer is the most popular sport. Substantively, the estimates suggest that going to the World Cup increases state aggression by about two-fifths as much as a revolution does and that it resembles the effect of electing a leader with military experience.

I also replicate these results using the regional soccer championships, such as the European Football Championship and the African Cup of Nations. In total, this new sample consists of seventy-eight countries that barely made or barely missed their regional soccer tournaments. The qualifiers and nonqualifiers were again well-balanced on aggression levels prior to qualification, but the qualifiers became significantly more aggressive following qualification. I present the results from this analysis in the supporting information.

The Importance of Nationalism in Security Studies

Nationalism is generally defined as the practice of identifying with a nation-state and viewing other nations as fundamentally different—often in negative ways (Anderson 1983, 6; Hobsbawm 1990, 9; Greenfeld 1993, 7). For many people, this identification is so powerful that it can compel them to kill and die for their countries (Anderson 1983, 7). While many scholars find this devotion difficult to explain at the individual level, most nonetheless view nationalism as a widespread and enduring phenomenon that plays a central role in international relations (Finnemore and Sikkink 1998, 889; Robinson 2014, 711; Harris 2016, 244–47). It defines the boundaries of the modern international system (Moore 2001, 6), and it provides an important basis for mobilizing populations to support and participate in warfare (Rosato 2003, 595).

Building off this connection between nationalism and warfare, many scholars hypothesize that surges of nationalism increase the chances of state aggression (Mearsheimer 1990, 20–21; Van Evera 1994, 27–28; Greenfeld and Chirot 1994, 86–88). They argue that this process can unfold through several different mechanisms, but three that prove especially relevant to the cases that I discuss later on. The first is the spiral model of nationalistic conflict. In it, nationalism leads to violence between citizens of different nationalities that escalates to the international level (Kaufman 1996, 109). Second, nationalism can make the public more hawkish by causing them to view other countries as “paper tigers” that are implacably hostile but also likely, as inferior societies, to crumble when challenged (Snyder 2000, 67). This shift in public opinion, in turn, may create opportunities and incentives for leaders to use military force. Third, nationalism can make leaders more hawkish by affecting their beliefs and attitudes independently of public opinion (Woodwell 2007, 32). Thus, it can prime them to view military aggression as a more legitimate option to resolve political disputes.

This idea that nationalism makes conflict more likely plays a central role in international relations scholarship. For instance, scholars use it to explain the conquests of Imperial Germany and Japan, the breakup of the Soviet Union, and the Balkan Wars (Mansfield and Snyder 1995, 6–7; Tuminez 2003, 81; Gollard 2009, 115; Bakke, Cunningham, and Seymour 2012, 272). It also operates as a key assumption in many theories of war. For example, it underpins the theory that revolutions make war more likely by creating surges of nationalism (Mansfield and Snyder 1995, 19–20), the theory that charismatic leaders who incite mass nationalism pose dangerous threats to international stability (Byman and Pollack 2001, 141) and the theory that grievances from past wars and atrocities make conflict more likely (Wang 2008, 799–803; Ciociari and Weiss 2012, 65–67). Fears about the destabilizing effects of nationalism also motivate many important arguments made by international relations scholars. Some examples include the idea that bombing civilians during wartime makes the target population less willing to surrender (Pape 1996, 21–25) and the notion that the United States can provoke backlash by stationing military troops abroad (Posen 2013, 120–21). Thus, the idea that nationalism causes conflict matters for how scholars interpret history, theorize about international relations, and formulate recommendations to policymakers.

Nevertheless, uncertainty remains about whether surges of nationalism actually affect state behavior. After all, strategic constraints might greatly limit nationalism´s effects, or nationalist sentiments might not fluctuate enough to influence foreign policy in meaningful ways. Endogeneity makes this subject particularly difficult to study. As Posen (1993, 122) explains,

[l]eaders use nationalism to mobilize public support for military preparation and sacrifices. When war seems imminent, for any reason, the intensity of propaganda increases. The same is true when wars last for any length of time. Thus, it will often be difficult to show that nationalism caused any conflict, because it will generally be accompanied and accentuated by other causes of the conflict.

Past quantitative efforts to investigate this question have remained limited, in part because no dataset exists that tracks nationalism in a systematic way. However, some recent studies have made noteworthy progress. First, several experiments find that people tend to think more hawkishly when they see their national flags (Hassin, Ferguson, Shidlovski, and Gross 2007, 19757–60; Kemmelmeier and Winter 2008, 864–71). These studies provide important microlevel evidence of a causal link between nationalism and conflict. Nevertheless, they fall short of showing that nationalism affects state behavior because they only look at the responses of individuals.

Second, Schrock-Jacobson (2012, 836–46) provides a major contribution to this research program by conducting the first large-N cross-national study that tests for a relationship between nationalism and military conflict directly. To deal with the measurement problem, she selects a random sample of state-years from 1816 to 1997 and codes for whether countries experienced a nationalistic movement in those years. Using a rare-events logistic regression model, she finds a correlation between nationalism and the onset of war that does not disappear after controlling for some baseline factors. This study provides a major step forward in establishing a causal link between nationalism and interstate conflict. Its only real shortcoming is that it cannot rule out the possibility of omitted variable bias. Unlike experimental methods, regression analysis deals with an endogenous treatment. This makes it susceptible to bias even after controlling for many important factors (Clarke 2005, 341; Pearl 2013, 6–15). This issue can be resolved with a research design that focuses on exogenous surges of nationalism, as I explain in the next section.

Using International Sports as a Source of Nationalism

International sports provide an excellent opportunity to test whether nationalism affects state aggression. First, they offer a way around the measurement problem. Analysis can be carried out without ever constructing a comprehensive dataset that tracks nationalism, because most scholars agree that international sporting events increase it (Cha 2009a, 1604–5; Markovits and Rensmann 2010, 207–70; Walt 2011, 15). Many studies show that international sports tend to make the national discourses within countries more hawkish (Maguire, Poulton, and Possamai 1999, 68–85; Toohey and Taylor 2006, 80–86; Tzanelli 2006, 485–99). Reporters often describe games in military terminology and compare wins and losses to past battles (Garland and Rowe 1999, 81). As Vincent et al. (2010) explain, “only warfare feeds the imagination and cements national identity more than sports.”

Many examples attest to international sports’ powerful ability to increase national unity within countries. For instance, Nelson Mandela used rugby to unite South Africa following the end of the apartheid regime (Steenveld and Strelitz 1998, 614–15; Jaksa 2011, 40). Similarly, qualification for the 2006 World Cup helped unify the Ivory Coast after four years of civil war (Mehler 2008, 99–103). Government leaders in Yemen also used the national soccer team to break down divisions between the northern and southern regions of the country following integration in 1990 (Cha 2009a, 1585–86). In short, international sports can evoke feelings of nationalism that spill over into the political realm.

Second, international sports provide a promising test because they are unlikely to increase the chances of interstate conflict in any way besides generating nationalism. Perhaps the only alternative means by which they might do so is by distracting the public and giving leaders more freedom to do what they want in foreign affairs. However, this mechanism would only explain a very short-term effect, which is inconsistent with the data presented later in this article. Other possible mechanisms unrelated to nationalism are hard to think of, in part because international sports and nationalism are inextricably linked. The vast majority of fans root for their countries because of national identification, and the very act of cheering for one’s side is an expression of nationalism (Hobsbawm 1990, 31).

Third, the most popular sporting event, the World Cup, features a qualification process based on a scoring system. As I will discuss more in the coming sections, this format creates a regression discontinuity in which countries receive surges of nationalism in an essentially random way. This process mirrors what an experimental researcher would do to test how nationalism affects conflict, if such an experiment were ever ethically permissible. This natural experiment provides a very clean test of how nationalism affects state aggression.

Fourth, the populations of some countries care much more about their national sports teams than others. Thus, the amount of nationalism that a sporting event produces varies across countries in a predictable way. Specifically, countries that take a sporting event more seriously should experience a larger surge of nationalism from that event. This variation leads to a valuable design check; the estimated treatment effect should be larger for countries that place more significance on the sporting event. I will conduct this design check in the results section.

Fifth, much historical research already exists linking sports nationalism to interstate violence across a wide range of cases. I turn to these cases in the next section. They provide some important qualitative evidence that nationalism from sporting events can increase state aggression, and they also illustrate some of the ways that this process can unfold.

Investigating the Historical Cases

Scholars point to many examples in which nationalism from sporting events sparked conflict between countries. These cases suggest that conflict often occurs as an unintended consequence of sports nationalism. However, they also show that leaders sometimes exploit international sporting events to generate support for their foreign policy ambitions. I discuss these cases below, and then I conclude by highlighting some important points to take away from these examples.

Football War

Many scholars blame this war on a series of soccer riots that exacerbated an already tense political situation between El Salvador and Honduras (Anderson 1981, 105; Hitchens 2010; Kapuscinski 2013, 157–59). In the late 1960s, agricultural workers in Honduras faced serious economic problems. To deal with this crisis, the Honduran government took land away from the Salvadoran immigrants living in its country and gave it to Honduran-born peasants. This policy led to skirmishes between Honduran and Salvadoran nationals and prompted a media war between the two countries that further encouraged xenophobia.

In this volatile political atmosphere, Honduras and El Salvador played three World Cup qualification games. The night before the first game in Honduras, a mob of Honduran fans surrounded the hotel where the Salvadoran players were staying and made as much noise as possible to prevent them from sleeping. The next day, El Salvador lost 1–0, but stories of the incident were widely reported throughout El Salvador and incited public outrage.

Salvadoran fans retaliated when the Honduran team traveled to El Salvador to play the next game. They threw rocks through the windows of the hotel where the Honduran players were staying and used drums and horns to keep the team up all night. The mob posed such a threat that the next morning the Honduran players had to be driven to and from the stadium in armored cars. The Salvadoran army also had to provide security at the game. Outside the stadium, Salvadoran fans assaulted many Hondurans, burned their cars, and in some cases chased them to the border. The Honduran team lost the game 3–0. When news of what happened reached Honduras, it prompted widespread violence against Salvadoran nationals living in the country, forcing thousands to flee to El Salvador.

Tensions escalated again when the countries played a tie-breaking game thirteen days later in Mexico City that required supervision by the Mexican military. That day, the two sides broke diplomatic ties, and El Salvador formally accused the Honduran government of human rights violations for not protecting the Salvadoran nationals living in its country. Less than three weeks later, El Salvador invaded Honduras, starting a war that caused about two thousand casualties.

This case provides an excellent illustration of the spiral model of nationalistic conflict. Nationalism from the games led to violence between citizens that created a political crisis, and leaders had strong incentives to respond with military force. The next case that I will discuss follows a similar storyline, although it did not lead to interstate violence on such a large scale.

Egyptian-Algerian World Cup Dispute

In November of 2009, Egyptian fans attacked a bus transporting the Algerian team and injured three players. Algerians retaliated against Egyptian companies and property in Algeria. In response, Mubarak removed the Egyptian ambassador from Algiers, gave an impassioned speech about Egypt’s insulted national pride, and accused the Algerian government of staging the attacks on its players (Lindsey 2009). Egypt also sent a plane into Algeria to rescue Egyptian nationals from the soccer violence, although the Algerian government refused to allow it to land (Shenker 2009). As Hitchens (2010) describes, “[b]efore the match in Khartoum, Egypt and Algeria had no diplomatic quarrel. After the game, perfectly serious people in Cairo were saying the atmosphere resembled that following the country’s defeat in the June 1967 war.” In fact, the Arab League responded to this incident by proposing that politicians from its member-states should not attend sensitive sports games because these events could make them more hostile toward other countries (Belmary 2009).

Serbian-Albanian Drone Conflict

At a Union of European Football Associations (UEFA) Cup qualifying game in Serbia on October 14, 2014, an Albanian fan flew a small drone into the stadium that was carrying an Albanian flag. When the drone got close to the ground, a Serbian player ripped the flag off, sparking a fight between the two teams. Serbian fans rushed the field to attack the Albanian players, who were barely able to get back to their locker room. Amid all this action, the Serbian crowd chanted, “Kill! Kill! Kill the Albanians” (Bouchaud 2014). The entire scene was captured on video, available on the New York Times website (Mackey 2014). This incident greatly increased tensions between the Serbian and Albanian governments (Bilefsky 2014). Both sides blamed each other for the conflict. In addition, the prime minister of Albania, who was scheduled to take an important diplomatic trip to Belgrade on October 22 that would be the first of its kind in seventy years, had to postpone his visit to allow tensions to cool (Bilefsky 2014). When he finally arrived on November 2, the visit quickly turned adversarial, and the opportunity for improved relations between the two countries slipped away. Meanwhile, the quarrel over the drone incident played out in international courts over the next ten months. Albania eventually litigated its way to victory.

English-Russian Euro Riots

England and Russia played in the first round of the UEFA European Championship in June of 2016. Three days before the game, a Russian attack submarine ventured into the English Channel and was intercepted by a British frigate. The defense editor of the British Newspaper The Sun describes the incident as “an attempt to intimidate the UK ahead of the Euros” (Willetts 2016). The next day, major clashes broke out between English and Russian fans that injured more than thirty-five people. While English politicians condemned the violence, the response differed in Russia. Igor Lebedev, a top Russian football official, tweeted “Well done, lads. Keep it up!” (quoted in de Menezes 2016). High-ranking British officials claimed that Putin had secretly orchestrated the violence, but they could not find concrete proof (Reuters 2016). While Putin denied this accusation, he also added, “I don’t know how [two hundred] Russian fans could fight several thousand of the British” (Hargreaves 2016). Recently, English hooligan groups vowed to get revenge by sending thousands of men to fight at the 2018 World Cup in Russia (Holloway 2016). Thus, this case remains a hot spot that could trigger an international crisis if not handled properly.

1934 Italian World Cup

Throughout the 1930s, Mussolini intentionally used sports to increase support for the Italian war machine. As Tunis (1936, 606) describes, the main purpose of sports for Mussolini was “the mass production of cannon fodder.” He writes, “sport ceased to be a free activity and became a function of the government… . The results are watched, collected, catalogued, and exploited, at home and abroad” (Tunis 1936, 612–14). Similarly, Martin (2004, 189) explains that the 1934 World Cup “was more like a fascist rally than a sporting contest.” Goldblatt (2008) draws a clear link between this competition and Italian aggression toward other countries. As he describes, “[t]he preparations for the tournament coincided with a steadily more expansionist and aggressive Italian foreign policy that would culminate after the World Cup in the invasion of Abyssinia [Ethiopia], intervention in the Spanish Civil War, and relentless pressure on Albania and Central Europe” (Goldblatt 2008, 253–54).

This case stands out from the others for two main reasons. First, it shows that nationalism can influence public opinion, even when no actual violence between citizens occurs. Second, it featured a leader who held a sporting event with the goal of increasing public support for his aggressive foreign policy agenda. Therefore, the sporting event was endogenous to politics, making it harder to draw conclusions about its effects. Nevertheless, the fact that a leader held a sporting event because he believed that it would make the public more hawkish and militaristic should count as evidence that surges of nationalism from sporting events can make conflict more likely. The 1936 Nazi Olympics provide similar evidence.

Nazi Olympics

Following Mussolini’s example, Hitler used the 1936 Nazi Olympics to provoke feelings of Aryan supremacy and victimization in Germany prior to World War II (Bachrach 2000; Cha 2009b, 12). Large (2007, 2) describes the games as “a crucial part of the Nazi regime’s ‘spiritual mobilization’ to win the hearts and minds of the German people.” In fact, Hitler did not care about sports until his advisers convinced him that they could increase support for the German war machine (Krüger 1998, 35). Of course, these scholars are not suggesting that World War II would not have happened without the 1936 Olympics. However, like the 1934 World Cup, this case illustrates how a leader with military ambitions can use international sports to generate a surge of nationalism that increases public support for future conflict.

Other Cases

Table 1 lists the examples discussed above along with some other notable cases in which scholars have argued that nationalism from sporting events led to conflict between countries. Bodyline (1932) involved a dispute between Britain and Australia over cricket that resulted in rioting, vandalism, boycotts, and other economic fallout. Many Australians consider this controversy to be one of the two main reasons that Anglo-Australian relations deteriorated in the 1930s, the other being the Great Depression (Frith 2013; Swan 2013).

Table 1.

Notable cases in which scholars claim that sports nationalism led to interstate conflict

1. Bodyline (1932)

2. Italian World Cup (1934)

3. Nazi Olympics (1936)

4. Moscow Dynamo Soccer Trip to Britain (1945)

5. Football War (1969)

6. Croatian War of Independence (1991)

7. Egyptian-Algerian World Cup Dispute (2009)

8. Serbian-Albanian Drone Conflict (2014)

9. English-Russian Euro Riots (2016)

The Moscow Dynamo soccer trip to Britain (1945) featured a series of games intended to strengthen relations between Britain and the Soviet Union following World War II. However, they were marred with fistfights and foul play (Kay 2008). Britain ended the tour early when it became clear that the Soviets were exploiting the games to generate nationalism at home (Kowalski and Porter 1997, 100–2). In response to this incident, Orwell (1945) published an essay about the dangers of sports nationalism, in which he describes international sporting events as “war minus the shooting” (see the supplementary material for further details).

I also include the Croatian War of Independence because it was prefaced by a substantial amount of soccer violence. In fact, many Serbs and Croats consider a battle in the Croatian soccer stadium to mark the unofficial beginning of the war (Ðorđević 2012, 205). While the two sides had major underlying political disagreements at the time, many scholars argue that the soccer riots played a critical role in starting this conflict (Boniface 1998, 93–94; Sack and Suster 2000, 310–13; Ðorđević 2012, 201–11). As Sack and Suster (2000, 316) explain, “it would be a mistake to view these matches as mere epiphenomena mirroring larger social and political events but having no power to influence them.”

Key Points

The above cases illustrate several important points. First, nationalism can be either unintended or driven by government leaders. In many cases, especially the Football War and Bodyline, international tension arose simply from the passions that sports evoke. Emotions over the sporting event changed public opinion, and leaders then had to respond. However, in other cases like the 1934 Italian World Cup and the 1936 Nazi Olympics, leaders deliberately exploited international sports to generate public support for their aggressive foreign policy agendas. The sporting event gave them a way to create a surge of nationalism in their country that could change public opinion. Thus, these cases suggest that nationalistic conflict may be accidental or elite-driven. Both causal paths appear to be valid ways that nationalism can intensify international rivalry.

Second, sports-instigated conflict sometimes occurs between countries that compete against each other head-to-head, but not always. As the 1934 World Cup and the 1936 Olympics demonstrate, nationalism from sporting events can increase the likelihood of conflict between countries even when they do not play against each other directly. Ethiopia did not attend the 1934 Italian World Cup, but it became the first target of Italian expansion in 1935. Similarly, the Soviet Union and the Jewish community boycotted the 1936 Nazi Olympics, yet they found themselves the victims of German nationalism in the coming years. Iraq’s experience at the 2007 Asian Cup also supports the idea that indirect sports nationalism can heighten tensions between countries. Stephens (2007) claims that this event increased Iraqi nationalism and feelings of resentment toward the United States. Similarly, when countries play games against states that they do not have international rivalries with, fans often burn flags of their traditional rival countries (Associated Press 2001; Traynor 2010; Barlow 2016).

This distinction between the direct and indirect effects of sports nationalism is important to highlight because the main statistical test in this article focuses on the indirect effect. Since countries play only a few other states at the World Cup, usually from other continents, my natural experiment primarily relies on indirect conflicts. Thus, it investigates whether state aggression can increase from general nationalism (like in the case of the French Revolution) rather than from nationalism directed at a specific rival country (like in the case of the Football War). However, I also show in the results section that the pairs of countries that played against each other at the World Cup became much more likely to engage in military disputes afterward.

Third, nationalism from sports appears capable of having downstream effects. In cases like the 1934 World Cup and Croatian War of Independence, tension from sports did not immediately lead to conflict at the international level. Rather, it appears to have primed people to think more aggressively toward other countries, thereby influencing their behavior when disputes did arise in the future. The analog case in US history might be how nationalism from September 11 made many Americans much more willing to support the Iraq War over a year-and-a-half later (McCartney 2004, 400). Thus, a major surge of nationalism seems capable of having downstream effects that appear a year or two afterward.

Fourth, sports nationalism usually seems to act as a contributing cause, rather than the main source, of conflict. In almost every case, nationalism from international sports exacerbated an already-tense political situation. Therefore, sports nationalism does not appear to be a deep cause of conflict. However, it may still function as an important one if some of these conflicts would not have occurred without the sports nationalism. For example, if the Football War would not have happened without the soccer riots, then sports nationalism would be a key cause of that war.

It is, of course, difficult to know what would have occurred in the counterfactual world where the sporting event did not take place. This inherent limitation of qualitative analysis makes it difficult to draw strong conclusions from case studies. The research design that I present in the next section will help overcome this problem. Specifically, it will use the group of countries that barely missed the World Cup as counterfactuals for the group of countries that barely made it. This approach provides a fair and straightforward test of how sports nationalism affects state aggression.

Design

I estimate the impact of World Cup participation on state aggression by using a regression discontinuity design. This method has become an increasingly popular research technique in the social sciences over the last decade because of its ability to identify causal effects (Hyde 2010, 73; McDermott 2011, 505–18; Dunning 2012, 63–84). It can be used when a treatment is given to units that surpass an important cut-point in a scoring system. The idea is to compare the group of units that scored just above the cut-point to the group that scored just below it. For example, if some students took a test for which everyone who scored a 50 percent or higher received a scholarship, researchers would compare the students who scored 50 percent to the ones who scored 49 percent. Provided the scoring process includes some randomness, it should be close to random who ended up on either side of the cut-point. Thus, this method should approximate a randomized experiment.

There are two main approaches to analyzing regression discontinuities: the local randomization approach and the continuity approach. The local randomization approach treats the regression discontinuity like a natural experiment. Thus, it involves comparing the units that scored just above the cut-point to the units that scored just below it using normal experimental tests, such as t-tests or randomization inference (Dunning 2012, 84; Cattaneo, Frandsen, and Titiunik 2015, 1–3). Recently, randomization inference became the preferred method because it performs better than t-tests on small- and medium-sized samples (Cattaneo et al. 2015, 2).

When using the local randomization approach, researchers do not need to make many design choices. They only need to decide how to define barely making and barely missing the treatment. In the test example, they might look at the students who scored 49 percent and 50 percent, or they might also want to include students who scored 48 percent and 51 percent.

The continuity approach involves estimating the difference in the potential outcomes at the cut point using two regression lines (Voeten 2014, 301–6; Cruz and Schneider 2017, 406; Lopes da Fonseca 2017, 8). This approach does not assume as-if randomness in a small window around the cut-point. Instead, it requires that the potential outcomes are smooth at the cut-point and that they can be estimated reasonably well using regression. Researchers using this approach must make many design choices, including which smoother and bandwidth selection procedure to use and how to estimate the confidence intervals for the regression lines.

My results are significant for both approaches, although I focus on the first one in this article. I do so because the standard procedures for the second approach were not designed for scoring systems with discrete values, such as the World Cup qualification process (Lee and Card 2008, 655–56; Cattaneo et al. 2015, 20). In addition, the first approach involves fewer design choices and statistical assumptions, since it does not require us to model the relationship between the outcomes and the scores using smoothing functions.

To address the fact that which countries barely qualified and which barely missed was not perfectly random, I use a difference-in-differences estimator to reduce any possible bias. Thus, I compare the change in aggression for the countries that barely qualified to the change in aggression for the countries that barely missed. As a robustness check, I also control for baseline differences between the qualifier and nonqualifier groups using linear regression. The findings remain significant under all specifications, which are primarily detailed in the supporting information. I also show that the results are insensitive to the limited number of design choices that I made, such as how I defined barely qualifying and barely falling short. I explain these choices in the next section.

Constructing the Treatment and Control Groups

The World Cup takes place once every four years during the summer, with the qualification process ending the prior winter. The first qualification round was held in 1934, and it has been in place ever since. The host country qualifies automatically, as did the winner of the previous World Cup up until 2010. All other countries must play their way in. They do so in one of two ways: (1) by playing a round of games against other states in their region and earning a certain place in the standings or (2) by winning a play-in game or several playoff games. Either way, the format is set well in advance.

This study focuses on the standings format, which is illustrated in Table 2.¹ Under this format, countries that finished very close to the qualification cut-point went to the World Cup or stayed home based on small differences in their records after many games. Thus, qualification should be close to random for these countries. This as-if randomness should make the group of states that barely qualified similar to the group that barely missed across observable and unobservable factors.

Table 2.

Example of the final standings from a 1994 qualification round in Europe

Rank	Country	Score	Qualified
1	Italy	16	Yes
2	Switzerland	15	Yes
3	Portugal	14	No
4	Scotland	11	No
5	Malta	4	No
6	Estonia	1	No

Note: The sample consists of pairs of countries like Switzerland and Portugal that barely made and barely missed qualification.

The playoff format cannot be analyzed as easily with regression discontinuity analysis. Countries make or miss the World Cup based on their performance in the final round, which often features states from different regions with large disparities in terms of gross domestic product and population, along with many other factors. Thus, which countries win and lose in the final round likely depends on factors related to future aggression. However, the findings in this study remain significant when close playoff games are included, so the results do not hinge on whether the outcomes of these games are considered random or not.

Using data from the standings format from 1934 to 2010, I select pairs of countries that were separated by no more than two points in the standings. This decision is optimal according to a method recently developed by Cattaneo, Titiunik, and Vazquez-Bare (2016) for constructing regression discontinuity samples when using the local randomization approach. They designed their procedure to maximize balance between the treatment and control groups. I also exclude pairs in which the qualifier scored less than five points. In these cases, the teams played only a small number of games, typically three or fewer each. I do not feel that qualification can be considered as-if random for these countries, so I drop them from the analysis. The results are very robust to changing both the two-point window and the five-point minimum score requirement (see the supporting information for full details).

In total, the sample consists of 142 countries, which are listed in Table 3. There are no pairs before 1958 because in all previous World Cup qualification rounds the qualifier scored less than five points. For instance, in 1938 Sweden scored four and Estonia scored two, which probably cannot be considered as-if random. After dropping these cases, no remaining pairs have a difference of 5–3, so the largest disparity in this dataset is 6–4.

Table 3.

Countries That Barely Made and Barely Missed the World Cup

Qualifier	Non-Qualifier	Year	Qualifier	Non-Qualifier	Year	Qualifier	Non-Qualifier	Year
Yugoslavia	Romania	1958	Yugoslavia	Spain	1974	Czechoslovakia	Portugal	1990
France	Belgium	1958	East Germany	Romania	1974	United States	Trinidad	1990
Austria	Netherlands	1958	Poland	Britain	1974	UAE	Qatar	1990
Soviet Union	Poland	1958	Uruguay	Colombia	1974	Ireland	Denmark	1994
Hungary	Bulgaria	1958	Argentina	Paraguay	1974	Switzerland	Portugal	1994
Britain	Ireland	1958	Haiti	Trinidad	1974	Bulgaria	France	1994
Paraguay	Uruguay	1958	Italy	Britain	1978	Netherlands	Britain	1994
Argentina	Bolivia	1958	Austria	East Germany	1978	Bolivia	Uruguay	1994
Bulgaria	France	1962	France	Bulgaria	1978	Cameroon	Zimbabwe	1994
Switzerland	Sweden	1962	Poland	Portugal	1978	Nigeria	Ivory Coast	1994
Portugal	Czechoslovakia	1966	Sweden	Norway	1978	Morocco	Zambia	1994
Bulgaria	Belgium	1966	Spain	Romania	1978	South Korea	Japan	1994
West Germany	Sweden	1966	Tunisia	Egypt	1978	Nigeria	Guinea	1998
Chile	Ecuador	1966	France	Ireland	1982	Jamaica	Costa Rica	1998
Czechoslovakia	Hungary	1970	Austria	Bulgaria	1982	Chile	Peru	1998
Romania	Greece	1970	Britain	Romania	1982	Senegal	Morocco	2002
Bulgaria	Poland	1970	Peru	Uruguay	1982	Nigeria	Liberia	2002
Italy	East Germany	1970	El Salvador	Mexico	1982	Ivory Coast	Cameroon	2006
Sweden	France	1970	New Zealand	China	1982	Tunisia	Morocco	2006
Belgium	Yugoslavia	1970	Portugal	Sweden	1986	Togo	Senegal	2006
Peru	Bolivia	1970	Soviet Union	Switzerland	1986	Angola	Nigeria	2006
Morocco	Nigeria	1970	Bulgaria	East Germany	1986	Algeria	Egypt	2010
Sweden	Austria	1974	Romania	Denmark	1990	Nigeria	Tunisia	2010
Netherlands	Belgium	1974	Austria	Turkey	1990

Seventeen of the seventy-one pairs tied in the standings. Nine of these ties were broken by a playoff game, seven were decided by looking at which country had the larger average margin of victory, and one was awarded to the team that scored more goals. I include these pairs in the analysis for two reasons. First, the playoff games seem more like toss-ups because they were played between teams of comparable skill. Second, I believe that comparable teams would not sort based on margin of victory or total goals scored. Nevertheless, the results remain significant whether ties are included or not.

Lastly, I remove pairs in which one of the teams did not represent a country. I exclude Scotland, Northern Ireland, and Wales from this analysis, and I count England as Britain. I also exclude the Representation of Czechs and Slovaks, which played from 1992 to 1993. No other changes are necessary.

Measuring Aggression

Similar to past studies (Leeds and Mattes 2007, 184; Melin 2011, 701), I measure aggression using the number of militarized interstate disputes (MIDs) that a country initiates. These disputes are instances in which states explicitly threaten, display, or use military force against other countries (Ghosn, Palmer, and Bremer 2004, 133). This measure is commonly used in security studies, since wars happen too infrequently to be a useful measure in most statistical tests. While many MIDs are low-level and not very interesting in their own right, they provide a good proxy for the likelihood of major interstate conflict. They are indicative of the foreign policy stance of a country and its willingness to start conflicts with other states. I also show in the next section that the results remain significant for high-level disputes that involved a direct attack, clash, or the start of interstate war.

Checking for Balance

The design should create balance across observable and unobservable pretreatment characteristics. The qualifier and nonqualifier groups should look very similar, except that the qualifiers went to the World Cup and the nonqualifiers did not.

Of course, we should expect to find some differences between the two groups that simply arose by chance. In real experiments, the p-values for baseline factors should be distributed uniformly between 0 and 1. Thus, statistically significant differences should appear for about one out of every twenty pretreatment characteristics, simply because of chance variation. However, the results of this study will be most persuasive if we find balance on all observable characteristics that might influence future aggression, as well as on past levels of aggression.

Figure 1 provides a comparison of the two groups. For each variable, I report the means for the qualifiers and nonqualifiers, and I plot the two-sided p-value, which I computed using randomization inference. Researchers frequently use this type of balance test when analyzing natural experiments to verify that which units received the treatment appears to be random (Jones 2014, 688; Dell 2015, 1749). I also plot the p-values for aggression levels in the years leading up to the World Cup.

Figure 1.

Balance between the qualifiers and nonqualifiers

Overall, the balance between the qualifiers and nonqualifiers looks similar to what we would expect in a randomized experiment. The p-values seem to be distributed uniformly between 0 and 1, which supports the idea that this data resembles experimental data.² Most importantly, the qualifiers and nonqualifiers are balanced on levels of aggression in the years leading up to the World Cup. In fact, this comparability extends much further back. If you compare the aggression levels of the qualifiers and nonqualifiers in each of the fifty years prior to qualification, statistically significant differences only appear twice at the 5 percent level, in year ten and year thirty-six. Aside from these years, the difference in aggression levels between the two groups never reaches significance at even the 10 percent level. Over the entire fifty-year period, the qualifiers average just 0.04 disputes per year more than the nonqualifiers (p = 0.46).

Thus, the data passes a key balance test. Figure 1 probably did not leave out some factor that would make one of the two groups behave much more aggressively than the other after qualification, since this factor should have also caused a difference in the two groups’ aggression levels before qualification. So aside from the treatment effect, there is little reason to suspect that the two groups would behave differently after qualification.

Findings

Figure 2 tracks the aggression levels of the qualifiers and nonqualifiers. Prior to qualification, they had very similar records of aggression. However, the qualifiers became much more aggressive following qualification, and they remained so until about the second year after the tournament. The fact that this increase begins after qualification accords with historical evidence showing that many countries experience a surge of nationalism when they qualify for the World Cup (Ralph 2007, 98; Mehler 2008, 99–103).

Figure 2.

Comparing aggression before and after the World Cup

Note that the aggression levels of the two groups drop the year before qualification, and then the qualifiers spike while the nonqualifiers stay low. This trend fits with the theory that the World Cup causes conflict. Recall that four years before qualification, some of the countries from both groups went to the previous World Cup. These effects appear to wear off fully in the year before qualification. In fact, in the years following the previous World Cup, the countries that went initiated about 50 percent more disputes than the countries that did not go.

The treatment effect seems to wear off after the second year following the tournament. This trend fits with the qualitative evidence presented earlier suggesting that nationalism from sporting events can have downstream effects. It also disconfirms the idea that the public distraction mechanism discussed earlier could explain these results, since it would only account for a short-term effect right around the World Cup. It should be noted, however, that this long-term trend might also partly be explained by the fact that the qualifiers started many conflicts in the first two years that reoccurred in the third year. In fact, roughly 50 percent of the disputes started by the qualifiers between year two and year three targeted countries that they had initiated disputes against at least once since qualification. Put simply, when a group of countries experiences a large spike in aggression, it should affect their aggression levels for several years, since some of the disputes that they start will likely lead to additional conflicts. Thus, this feedback mechanism could partly explain the longevity of the effect.

The qualifiers not only took military action more often than the nonqualifiers, but the actions they took tended to be more violent. The MID dataset codes for the highest level of action taken by each country. These levels range from the threat to use force (= 1) to the start of interstate war (= 20). In the two years after the World Cup, the median for the qualifiers was fifteen, whereas the median for the nonqualifiers was eleven. Similarly, the qualifiers initiated seven disputes that resulted in fatalities, whereas the nonqualifiers initiated one. The MID data set also codes for whether countries intended to revise the status quo in each case. In the two years following the World Cup, the qualifiers aimed to revise the status quo in 72 percent of their disputes, compared to only 54 percent for the nonqualifiers (p = 0.001).

Table 4 shows the results for a number of statistical tests. The estimated effect is significant at the 1 percent level for the entire sample. Moreover, the results are entirely driven by countries where soccer is the most popular sport.³ In fact, countries where soccer is not the most popular sport experienced no change in aggression from the World Cup. These countries include the United States, Japan, New Zealand, Ireland, Australia, the United Arab Emirates, Jamaica, and Trinidad. Compared with findings from past research, these results suggest that going to the World Cup increases state aggression about two-fifths as much as a revolution does (Colgan 2010, 682) and that it resembles the effect of electing a leader with military experience (Stam, Horowitz, and Ellis 2015, 134).

Table 4.

Estimating the Effect of the World Cup on State Aggression

	Estimate (SE)	p-value	n
Entire sample	0.38** (0.14)	0.007	142
Subgroups
Countries where soccer is the   most popular sport	0.41* (0.16)	0.011	132
Countries where soccer is not   the most popular sport	0.00 (0.00)	1.00	10
Shifting the regression discontinuity window	0.37*	0.023	92
Countries that qualified/missed   by one point or less	(0.17) 0.49**	0.001	162
Countries that qualified/missed   by three points or less	(0.15)
Entire sample (no ties)	0.43* (0.17)	0.012	102
Other statistical tests
Linear regression with all control   variables	0.40** (0.14)	0.004	142
Difference-in-differences t-test	0.38** (0.14)	0.006	142
Post-treatment outcome alone   (not dif-in-dif)	0.44** (0.17)	0.009	142
Tests that are insensitive to outliers
Signed-rank test	—	0.009	142
Dummy for increase in disputes   initiated	0.15** (0.05)	0.007	142
Removing the US and Soviet   Union	0.33* (0.14)	0.021	139
Other outcomes
Revisionist disputes initiated	0.38** (0.12)	0.001	142
Disputes initiated that involved   the use of force	0.28* (0.13)	0.036	142
Disputes initiated that involved a   direct attack	0.23* (0.10)	0.026	142
Military participation	0.6%* (0.3%)	0.013	138

Notes: (1) The standard errors and p-values in this table were computed using randomization inference (except for the t-test). All tests are two-sided. Unless otherwise specified, the outcome is the change in aggression between the qualifier and nonqualifier groups (difference-in-differences). I define change in aggression as the difference in the number of militarized interstate disputes initiated between (1) the period ranging from qualification to the second year after the World Cup and (2) the period of the same length prior to qualification. I use these time intervals to account for conflicts that may have been caused by the residual effects of nationalism. Nevertheless, the estimates presented here are similar for other choices of interval length. See the supporting information for a full summary of the robustness checks. (2) Statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001.

Table 4 also shows that the findings hold under various robustness checks. They remain significant after controlling for baseline differences between the two groups using linear regression, as well as for tests that are insensitive to outliers. Similarly, they hold when the two-point regression discontinuity window is set at one point or three points, as well as when ties are dropped. The results are also insensitive to shifting the five-point minimum score requirement and adjusting the time interval. I provide a full summary of these results and other robustness checks in the supporting information.

The final test in Table 4 shows that the qualifiers experienced a significant increase in military participation compared to the nonqualifiers. Specifically, the qualifiers jumped from an average participation rate of about 0.85 percent the year before the World Cup to 0.91 percent the year after, whereas the nonqualifiers dropped from an average participation rate of 0.82 percent the year before to 0.81 percent the year after (p = 0.013). This finding provides further evidence for the nationalism mechanism and against the public distraction mechanism. If the World Cup merely gave leaders more freedom to take military action abroad by distracting people from politics, the countries that went to it should not have experienced an uptick in military participation.

Figure 3 shows the results for the continuity approach. In each graph, the points on the right represent the means for the countries that qualified, and the points on the left represent the means for the countries that fell short. The estimated treatment effect is simply the difference between the two lines at the cut-point. The graphs show that the qualifiers and nonqualifiers had very similar aggression levels prior to qualification. However, the qualifiers experienced a significant increase in aggression in the two years following the World Cup. The difference at the cut-point is statistically significant for the middle graph (p = 0.033) and the one on the right (p = 0.008).⁴

Figure 3.

Using smoothers to estimate the treatment effect

Note: The shaded regions represent the 95 percent confidence intervals, which were computed using nonparametric bootstrapping.

Lastly, the data also indicates that direct competition between countries at the World Cup increased the chances of conflict between them. The pairs of countries that played against each other found themselves on the opposite sides of military disputes twenty-one times in the two years following the World Cup, compared to only nine times in the two years before it (n = 758, p = 0.040, confidence interval = [1,19]). The number of these pairs with at least one dispute jumped from nine to fourteen (a 56 percent increase). This trend does not merely reflect a broader spike in global conflict. In fact, over these years the total number of disputes not involving World Cup participants only rose by 5.1 percent. These results suggest that the World Cup increased the likelihood of conflict between the pairs of countries that competed against each other head to head.

Exploring a Case from the Dataset

I will now examine a case from the dataset to look for a possible link between World Cup nationalism and a specific dispute in my sample. As discussed earlier, the challenges inherent to counterfactual reasoning often make it difficult to establish a causal relationship in a single case. Nevertheless, causal processes do leave traces of evidence behind. To maximize the chances of finding credible evidence, it makes sense to focus on a case that occurred recently, since there will probably be more reliable data on recent cases. We also want to make sure that the country in the case took the World Cup very seriously, since those countries drove the results in the previous section. Lastly, we need a case that featured a military dispute, since we want to check for a link between World Cup nationalism and a dispute in the sample.

Given these considerations, I believe that Senegal (2002) makes the most sense to investigate. Besides the fact that this case occurred recently, it stands out for three main reasons. First, Senegal had never gone to the World Cup before. Thus, it probably experienced more excitement and energy around this event than countries that qualify on a more regular basis. Second, Senegal performed surprisingly well. In the group stage, it defeated France, the defending World Cup champion and its former colonizer. It then beat Sweden in the knockout stage to become only the second African team ever to reach the quarterfinals. Third, Senegal initiated a military dispute against Gambia about two weeks after its final World Cup game. In fact, it marked the only time that Senegal engaged in a military dispute between 1993 and 2010. This timing would be an incredible coincidence if it was unrelated to the World Cup.

Thus, the preliminary story line suggests that World Cup nationalism may well have triggered this dispute. However, the question remains whether the details of the case provide further substantiating evidence. Specifically, if World Cup nationalism made Senegal more likely to initiate this dispute, we should expect a careful analysis of the case to yield the following observations:

1. Qualifying and playing in the World Cup caused a powerful surge of nationalism in Senegal.

2. The leader of Senegal took the World Cup seriously and made public statements claiming the event marked an important moment in the country’s history.

3. The World Cup clearly affected the leader’s decision-making in important ways.

4. World Cup nationalism caused the people of Senegal to view questions related to foreign policy more hawkishly.

5. This shift in public opinion was plausibly related to the specific dispute that Senegal initiated.

The historical record certainly confirms (1). When Senegal qualified for the World Cup, celebrations erupted throughout the country and carried on for three days (Brockes 2002, 3). This mayhem was just a preview of what was to come once the tournament started. One reporter described the scene in the capital, Dakar, as “a joyous ritual [that lasted] two weeks” (LA Times 2002). After Senegal won its first World Cup game against France, “fans poured into the streets of Dakar gravitating, significantly, around Le Place de l’Independance and the presidential palace. Red, yellow, and green Senegalese flags, hats, scarves, t-shirts, and African-style boubous were the only acceptable attire to commemorate the occasion” (Ralph 2007, 201). A similar celebration followed Senegal’s defeat of Sweden. The LA Times (2002) described, “[t]housands streamed into the streets of Dakar… . Men, women, and children ran full speed to the main boulevards, joining crowds dancing and bouncing up and down waving flags.”

This euphoria makes sense when one recognizes that the World Cup is about far more than just sports. Much of Senegal saw it as an affirmation of their nation. The editor of the largest newspaper in the country, Le Soleil, explained, “A successful football team is the expression of the confident nation, one in which there is democracy and stability and human rights. You do not see Zimbabwe or Cameroon producing a winning team” (Brockes 2002, 3).

The leader of Senegal, President Abdoulaye Wade, also placed a great deal of significance on this event. When the country qualified, he was visiting Jacques Chirac in France. However, he returned home early to celebrate Senegal’s qualification. As he explained, “it’s the most important thing that can happen to any country, and I will join the team and the nation in celebrating by reducing the amount of time I was expected to stay in Paris” (Ralph 2007, 201). To reward his players, he invited them to his palace and presented each of them with bonuses of about $15,000. He also actively encouraged celebrations of Senegal’s victories in the tournament. For instance, after the win over France, he declared a national holiday and paraded around the capital in a vehicle with the top open so that fans could see him juggling a soccer ball (Ralph 2006, 308).

Whether President Wade truly cared about the World Cup or simply exploited it to increase his popularity is difficult to know. However, the historical facts clearly confirm that (2) he took the event very seriously and (3) he allowed it to affect his decision-making in important ways. He formulated domestic policy around it by declaring national holidays and organizing celebrations. It also clearly influenced his foreign policy decision-making by causing him to return home early from his trip to France.

Several pieces of evidence suggest that (4) the World Cup caused people in Senegal to view foreign policy matters more hawkishly. First, between 2002 and 2003, Senegal experienced its largest increase in military participation since 1980. This notable shift toward greater militarism at the mass level suggests that World Cup nationalism made the public more inclined to think hawkishly. It also reflects the broader trend in the data mentioned earlier.

Second, many Senegalese championed the World Cup victories as proof that their nation had emerged as an important country in world politics that could now boldly advance its interests with success (Ralph 2006, 308). Leading into the tournament, the government described the team as, “[t]he Lions that hail from the Senegal that wins.” Meanwhile, the Senegalese band Pape et Cheikh popularized a song with the following chorus: “You should win, win some more, win a lot, always win.” This message was captured in a slogan that emerged after the World Cup: “[t]he Senegal that wins.” Fans chanted these words when the team returned home from competition, and President Wade frequently used this line in his speeches.

This new mind-set was directly tied to politics. As one spectator told a reporter from the Guardian after the win against France, “[i]n one sense, we have already won, because we have beaten the world champions. But what use is winning if the victory is not translated into political action?” (Brockes 2002). Similarly, Ralph (2006, 302) argues that this new confidence reinforced a belief that Senegal could also win in international relations.

Third, sports were closely linked to Senegal’s primary security threat at the time: a separatist movement in the southwestern part of the country known as the Casamance region. This movement began in 1982 and persisted as a low-level civil conflict throughout the 1980s and 1990s. It also had an international dimension. The separatists received assistance and safehaven from the bordering country Gambia, which sought to undermine Senegalese power (Foucher 2003, 105).

Throughout this conflict, soccer acted as an important social force. As Deets (2016, 356–65) explains, it played a key role throughout the struggle by reinforcing local and national identities. Subnational rivalries between teams from the Casamance region and the capital created tensions that fueled the conflict. However, the World Cup had the opposite effect. It strengthened the notion that the Casamance was part of Senegal, principally because some of the players on the national team came from the Casamance region (Deets 2016, 368–70). Thus, to many Senegalese, World Cup nationalism strengthened their belief that the conflict in the Casamance needed to be resolved, a goal that Gambia had impeded in the past.

It was in relation to this conflict that (5) Senegal started its only military dispute from 1993 to 2010, against its neighbor Gambia. This dispute occurred on July 7, just fifteen days after Senegal lost to Turkey in the knockout stage. However, it can be traced back to a military operation that Senegal launched in the Casamance on June 21, the day before the game against Turkey (Davenport 2002). This operation aimed to capture the rebels and establish control over the region. A BBC (2002a) report described this engagement as “the largest military operation ever launched since President Abdoulaye Wade came to power.”

During this operation, the Senegalese government viewed Gambia as a problem. Another report (BBC 2002b) described “[t]he mounting anger of Senegalese authorities against Gambia is perceptible, as they claim that Gambian officials have been reportedly complacent toward Casamance rebels.” On July 7, after many of the rebels had fled into Gambia, Senegal deployed troops along the border to ensure that they would not reenter the country (BBC 2002b). By making this move, Senegal registered a military dispute against Gambia, which qualifies as a display of force.⁵

No doubt, soccer nationalism was not the deep cause of the conflict, which predated the 2002 World Cup by at least twenty years. Nevertheless, the details of the case suggest that the World Cup triggered this dispute. It caused a major surge of nationalism in Senegal. It affected President Wade’s actions, including his foreign policy decision-making. It coincided with a notable increase in military participation. It encouraged the Senegalese to view their nation as a powerful country that should boldly pursue its agenda on the world stage and should win. It reinforced the notion that Senegal needed to resolve the Casamance conflict, a decades-long struggle that was historically linked to soccer. In sum, the timing of this dispute should not be dismissed as a mere coincidence. The details of the case suggest that World Cup nationalism played an important role.

Conclusion

This study provides strong evidence that surges of nationalism can increase the chances of interstate conflict. In doing so, it confirms a causal relationship of central importance to the study of international relations. Put simply, the history of international sports provides enough qualitative and quantitative evidence to conclude that fluctuations in nationalism can affect state aggression in powerful ways.

Moving forward, more work is needed to identify other potentially dangerous sources of nationalism. One promising avenue would be to look at whether major national achievements and tragedies tend to increase state aggression in the short run. While some scholars have already explored this topic (see McCartney 2004), there remains a need for broader theoretical analysis and more comprehensive empirical study.

This article also has some important policy implications. Since international sports are a powerful source of nationalism, we should oppose bids to hold major sporting events in countries where leaders show a penchant for using nationalist sentiment to increase support for aggressive foreign policies. For example, allowing Putin to host the 2014 Winter Olympics and 2018 World Cup was a poor decision; international sports organizations should not make similar mistakes in the future. They should also try to prevent avoidable games between countries with underlying political tensions. Doing so could help avert conflicts like the 1969 Football War.

Policymakers might also consider creating sporting events where countries play as small regional blocks—like a Scandinavian team and Balkans team. This format would encourage spectators to identify with their regions rather than their nations. Neighboring states would be allies rather than adversaries. While this format could increase animosity between the competing regional blocks, practitioners could minimize any negative political fallout by refraining from pitting neighboring regions against each other. These competitions would likely have broad appeal, since they would feature an even higher level of play than games between countries. If they became popular, they could encourage a type of transnational identification that could prove much more beneficial to international relations than nationalism.

Supplementary Information

Supplementary Information is available at https://www.andrewbertoli.com and the International Studies Quarterly data archive.

Footnotes

1

Before 1994, countries earned two points for a win, one point for a tie, and nothing for a loss. Starting in 1994, the value of a win was increased to three points.

2

One concern here is that the window selection procedure developed by Cattaneo et al. (2016) selects the window that creates the best balance, which could weaken this design check. However, I initially chose the two-point window before these scholars had developed their procedure. Fortunately, my initial choice matched the one deemed optimal by their procedure. Therefore, this design check is valid, since I originally selected the window size without any reference to the covariates.

3

The p-value for countries where soccer is the most popular sport jumps to 0.011 because subsetting to this group of countries breaks the paired structure of the data, which decreases statistical power. If you subset to the pairs of countries where soccer is the most popular sport for both states (thus maintaining the paired structure), the p-value is 0.007 (n = 124).

4

I construct the regression lines for these graphs using local linear regression with a triangular kernel that down-weights units the further they are from the cut-point. The results for the change in aggression are also significant for many other smoothers. These include local linear regression for any bandwidth greater than or equal to four and kernel regression for any bandwidth greater than or equal to 1.5.

5

Senegal also conducted other military operations in the Casamance prior to this incident, and at times it had increased tensions with its neighbors. Most notably, a crackdown in 2000 resulted in the alleged bombing of a town in Guinea-Bissau (Evening Times 2000). However, Senegal denies that the incident ever happened, and it was working with Guinea-Bissau at the time to subdue the Casamance rebels (African News Service 2000).

Author notes

Author’s note: I am grateful to everyone who helped me develop this project, especially Jasjeet Sekhon, Allan Dafoe, Ron Hassner, Aila Matanock, Thad Dunning, Michaela Mattes, Vinod Aggarwal, David Broockman, Robert Trager, John Henderson, David Kang, Brian Rathbun, Patrick James, Gerardo Munck, Arthur Stein, Jason Lyall, Kelly Greenhill, Nicholas Sambanis, Hein Goemans, Keir Lieber, Michael Desch, Neil Malhotra, Andrei Markovits, Andrew Eggers, Nathaniel Beck, Benjamin Graham, Andrew Coe, Jonathan Markowitz, Brett Carter, Erin Baggott Carter, James Lo, Pablo Barberá, Cali Ellis, Thomas Zeitzoff, Rocio Titiunik, Paul Avey, Daniel Sargent, Alexander Theodoridis, Baobao Zhang, Josh Kalla, Jason Klocek, Benjamin Bartlett, Caroline Brandt, and Tara Buss. I would also like to thank the University of Southern California´s Center for International Studies, the participants of workshops at Stanford, the University of California–Berkeley, and the University of California–Los Angeles, the four anonymous reviewers, and the International Studies Quarterly editors. Replication materials are available in the supplementary information.