ECONOMIC SANCTIONS AND CULTURE

Abstract

We address empirically the question of why international economic sanctions are, or are not, chosen as instruments of foreign policy and the question of what determines their success. We hypothesize that cultural linkages between nations are an important factor in explaining both instrument choice and conflict outcomes. Countries that share significant cultural attributes are found to be less likely to apply economic sanctions against one another than countries lacking such cultural ties. However, it is precisely in the case of culturally similar sender and target nations that sanctions are most likely to succeed.

Keywords:

• Economic sanctions
• Culture
• International political economy
• Democratic peace
• Economic coercion

JEL Codes:

• F13
• F51
• F59

ACKNOWLEDGMENTS

We thank Wenhui Fan and other participants in a session of the Canadian Economics Association annual conference (Vancouver, BC, June 7, 2008) for helpful comments. We are also grateful for valuable suggestions from the editor and referees. All remaining errors, however, are the responsibility of the authors alone. We thank Mathew Dunn and Alexis Neal for research assistance.

Notes

¹The definition of ‘success’ is problematic. Debate over what precisely constitutes a successful exercise of economic power in international relations has dogged the empirical sanctions literature throughout, with some scholars arguing that a sanction can be viewed as successful if it merely imposes costs on the target without eliciting a change in the target’s behavior, while others have insisted that, to be deemed successful, a sanction must be the main factor contributing to the target’s compliance. See Baldwin (1985), Pape (1997) and Kaempfer and Lowenberg (2007: 870–871).

²The outcome of each episode is ranked from one (failure) to four (success); the contribution of sanctions to this result is likewise ranked from one (none) to four (significant). The overall success score of the sanctions episode is then the product of the policy result score and the contribution score.

³The explanatory power and predictive success of the HSE model is weak (Bonetti, 1997; Leitzel, 1987). Their regression equation explains only 21% of the variation in success scores and only three of the 18 coefficients used in the equation are statistically significant. One of the main reasons for these weak results is that HSE employ an ordinary least squares estimation technique that is not appropriate to the case of a limited dependent variable such as the sanctions success score (Bonetti, 1997; van Bergeijk, 1994). This problem has been addressed by various authors using discrete dependent variables estimators. For example, van Bergeijk (1989, 1994), Drury (1998) and Dehejia and Wood (1992) use logit models, Lam (1990) and Hart (2000) use probit, and Bonetti (1998) uses logistic regressions to identify the circumstances in which high levels of success or failure are probable. The HSE regression analysis has also been criticized by van Bergeijk (1994: 73) and by Lam (1990: 241) for using as a dependent variable a measure of sanctions success that includes the importance of the sanction to the attainment of the goal. This requires prejudging the extent to which a given sanctions episode caused a change in the target’s policy, which is precisely what is supposed to be estimated on the right‐hand side of the regression equation.

⁴For a survey of results, see Kaempfer and Lowenberg (2007: 892–898).

⁵For a contrary finding, see Jing et al. (2003).

⁶Allen (2008) observes that the ability of sanctions to successfully mobilize antigovernment activity in a target nation depends on the nature of domestic political institutions in that nation.

⁷A similar view is taken by Alerassool (1993), Smith (1995), Dashti‐Gibson et al. (1997) and Selden (1999).

⁸For clear statements of the democratic peace hypothesis and empirical tests that confirm the specific benefits of democracy, see Russett (1993), Dixon (1994), Oneal and Russett (1997), Mousseau (1998), Russett and Oneal (2001), Dixon and Senese (2002) and Danilovic and Clare (2007). For reviews of the literature, see Russett and Starr (2000) and Weede (2004).

⁹See Nooruddin (2002), Lektzian and Souva (2003), Cox and Drury (2006), Goenner (2007) and Hafner‐Burton and Montgomery (2008), whose contributions are discussed further in Section 2.

¹⁰Although generally perceived as relatively ineffective instruments of foreign policy, there is evidence that sanctions are actually quite successful in ending civil wars. Escribà‐Folch (2010) finds that sanctions imposed by both international institutions and by other senders have negative effects on intra‐state conflict duration, either because sanctions increase the likelihood of conflict resolution or because they increase the probability of a military victory.

¹¹Even if democracies do become embroiled in disputes with other democracies, they tend to resolve these conflicts more quickly and more peacefully than in situations where non‐democracies are involved (Dixon, 1994; DeRouen and Goldfinch, 2005).

¹²It has also been hypothesized that the absence of liberal‐democratic norms among leaders of non‐democratic states creates an expectation of aggressive intentions and a presumption of enmity on the part of democratic leaders. See Farnham (2003). Gartzke (2000) attributes much of the lack of militarized disputes between democracies to preference similarity across democratic states, while Siverson and Emmons (1991) document a high rate of alliance formation among democracies.

¹³For empirical tests of various competing theories of the democratic peace, see Lektzian and Souva (2009). It is of course conceivable that the direction of causality might be quite the reverse of that postulated in the democratic peace literature, i.e. that the observed rarity of war between democracies might be explained by a negative effect of war on democracy rather than vice versa. For example, it is plausible to suppose that nations become more autocratic as they prepare for impending wars. Mousseau and Shi (1999) test this hypothesis but find that disputant countries are in fact equally likely to become more democratic as they are to become more autocratic in periods leading up to the outbreak of wars. James et al. (1999, 2000) reject single equation estimates of the relationship between democracy and war as merely ad hoc reduced forms, lacking in causal inference. In their view, a more appropriate methodology would be to identify a structural equation as part of a simultaneous system in which both democracy and conflict are treated as endogenous. Using a simultaneous equations model, Reuveny and Li (2003) find that dyadic militarized disputes reduce joint democracy while at the same time joint democracy reduces the probability of militarized disputes.

¹⁴Bueno de Mesquita and Siverson (1995) find empirical evidence that engaging in war is hazardous to the survival in office of all types of leaders, but especially democrats.

¹⁵For critical perspectives on the democratic peace, see Hess and Orphanides (2001), Rosato (2003, 2005), Henderson (1999, 2002) and Ward et al. (2007).

¹⁶As Kaempfer and Lowenberg (1988) point out, majoritarian democracies are especially vulnerable to pressure from interest groups to impose sanctions, sometimes motivated by moral outrage against the objectionable behavior of a foreign state, but often driven by little more than thinly disguised protectionist interests in trade restrictions.

¹⁷Nevertheless, as Lektzian and Sprecher (2007) demonstrate, sanctions implemented by democracies are frequently followed by the use of military force.

¹⁸This finding is disputed by Hafner‐Burton and Montgomery (2008) on the grounds that it is derived from a flawed empirical method that fails to take account of the disproportionate weight of the United States among sanctioners.

¹⁹In an empirical test of Fearon’s theory, Partell and Palmer (1999) find that domestic political audiences do indeed exert a strong influence over which countries in a crisis are likely to achieve successful outcomes, although relative national capabilities are also found to be important.

²⁰Hart (2000) suggests that democracies are generally more successful sanctioners than non‐democracies because the audience costs confronted by democratic governments ensure that they will choose to apply sanctions in the first place only if they are committed to holding out for success.

²¹This rally‐around‐the‐flag effect has captured the attention of many scholars in the sanctions literature. See, for example, Mayall (1984: 631) and Mack and Khan (2000: 282). Moreover, as Galtung (1967) points out, sanctions can be counterproductive by giving rise to a new elite in the target country that benefits from international isolation. On such unintended consequences, see also Doxey (1980, 1996) and Selden (1999).

²²Drawing on a large panel of cross‐country and time‐series data, Marinov (2005) finds empirical evidence for the destabilizing effect of foreign economic pressure on the leadership of target countries.

²³Pape (1997) cites the case of Iraq in the 1990s, which, despite facing the most damaging economic sanctions in history, with 48% of its GNP destroyed, did not acquiesce.

²⁴Peksen (2009) reports empirical evidence that sanctions are more likely to bring about a deterioration in human rights conditions in target countries than an improvement. A similar finding is obtained by Wood (2008), who argues that the negative impact of sanctions on human rights is due to the fact that target states frequently respond to sanctions by tightening repression in an attempt to stabilize their regimes, protect core supporters, minimize opposition threats and quell popular dissent.

²⁵On the tendency of sanctions to strengthen the ‘hawks’ and weaken the ‘doves,’ see Willett and Jalalighajar (1983/84).

²⁶Lektzian and Souva (2007) point out that the rents generated by sanctions typically benefit non‐democratic leaders to a greater extent than they benefit democratic ones. On the capture of sanctions rents by autocratic rulers, see also Kaempfer and Lowenberg (1999) and Kaempfer et al. (2004).

²⁷For theoretical treatments of the political effects of sanctions on non‐democracies, see Wintrobe (1990, 1998), Kaempfer et al. (2004) and Mulligan and Tsui (2006, 2008).

²⁸This finding does not necessarily refute the notion of an economic peace among democracies, however. The economic peace is a dyadic phenomenon (Lektzian and Souva, 2003: 647); it requires that pairs of democratic countries be less likely to be involved in sanctions than pairs of non‐democratic countries or mixed pairs. Although Nooruddin finds that democracies are more likely to be targets of sanctions, he does not test whether democracies are also more likely than non‐democracies to impose sanctions. In fact, Nooruddin’s data include only US‐imposed sanctions (Nooruddin, 2002: 63, 70).

²⁹There is some evidence, however, that democracies favor the use of positive economic incentives rather than sanctions, especially in dealing with other democracies (Davidson and Shambaugh, 2000; Drezner. 2000).

³⁰See, for example, La Porta et al. (1999, 2002, 2008), Beck et al. (2003) and Levine (2005).

³¹Grier (1999) and Feyrer and Sacerdote (2009) show that time spent under European colonial rule is positively associated with economic development. See also Bertocchi and Canovab (2002) on the effects of colonization on Africa’s development performance and Acemoglu et al. (2002a) on colonial institutions as the main cause of Botswana’s impressive growth record.

³²See Knack and Keefer (1995), Olson (1996), Clague et al. (1999), Rodrik (2000a, 2000b), Feng (2003), Acemoglu et al. (2004), Doucouliagos and Ulubaşoğlu (2008) and Mijiyawa (2008). Helpman (2004) provides a review of research on the growth effects of institutions, including property rights protection, legal systems, customs and political systems.

³³See also Tabellini (2008).

³⁴Cultural embargoes include restrictions on educational, artistic and sporting links with a target country.

³⁵It is often argued that smart sanctions are not only more effective than broad, GNP‐reducing sanctions in achieving their objectives but are also less likely to impose ‘collateral damage’ on innocent groups within the target country, such damage having the undesired effect of potentially strengthening the regime and retarding the emergence of a middle class and civil society (Haass, 1998: 202). On smart sanctions, see also Weiss (1999), Cortright and Lopez (2000, 2002a, 2002b), Lopez (2001), Cortright, Lopez and Rogers (2002) and Cortright, Millar and Lopez (2002).

³⁶Kaempfer and Lowenberg (1999) offer the example of the sports embargo on apartheid South Africa, imposed by a handful of rugby‐ and cricket‐playing nations, which arguably had a greater impact on whites in that country, without hurting blacks much, than did many of the severest economic sanctions that were applied. On sanctions that entail proscriptions on cultural activities and sporting events, see Hanlon and Omond (1987: 225).

³⁷On the propensity of the United States to resort to sanctions, see Hufbauer (1998).

³⁸According to Drezner (1998, 1999), sanctions outcomes are determined in part by expectations of future conflict. The reason for the greater success rate of sanctions against allies than against adversaries is that allies are less concerned about the relative gains of holding out for a better bargaining position in the future, due to a low expected likelihood of future conflict.

³⁹Spolaore and Wacziarg (2009) examine the effects of genealogical ‘relatedness’ between populations on inter‐state conflict, while Guiso et al. (2009) find that bilateral trust between European nations, which is driven partly by cultural similarities, influences the amount of trade and investment flows between them.

⁴⁰In the words of Eaton and Engers (1992: 902), ‘sanctions can be effective even if, in equilibrium, they are not actually used.’ See also Blake and Klemm (2006).

⁴¹The finding that threats are normally more effective than actually‐implemented sanctions is disputed by Krustev (2010), who contends that this result is an artifact of exogenous policy objectives on the part of sanctioners. If the demands of sanctioners are treated instead as endogenous, determined strategically by the capabilities and interests of target states, then the apparently paradoxical tendency of actually‐implemented sanctions to be selective of the most difficult disputes is explained and it is no longer necessarily the case that threats are more effective than sanctions. Of course, sanctions are often threatened or applied not to achieve a given policy objective in the target country but for expressive or demonstrative purposes, e.g. to signal disapproval or resolve, to establish a reputation in the eyes of domestic and foreign audiences, or merely to be seen to be ‘doing something.’ See Galtung (1967), Renwick (1981), Leyton‐Brown (1987), Lundborg (1987) and Tsebelis (1990).

⁴²Another two‐stage estimation approach that has been used to deal with the problem of simultaneity of instrument choice and effectiveness is the censored probit model, exemplified by Nooruddin (2002).

⁴³For a comparison of results, see Kaempfer and Lowenberg (2007: 892–898).

⁴⁴The HSE dataset has since been updated to include sanctions episodes after 1990. See Hufbauer et al. (2007). Although we did not test our hypotheses with the newer data, which were released only after we had commenced our study, we have no a priori reason to believe that the fundamental role of culture was any different after 1990 than it was before, despite the fact that the degree of similarity of political and economic culture within some dyads might have changed with the end of the Cold War.

⁴⁵1972–1973 civil liberties data were used for any sanctions occurring before 1972. Pre‐1972 sanctions arise in only 2.6% of the 1,972 dyads used in the first set of tests in the study, and in 21% of the 138 dyads used in the second part of the study. However, in the latter case, some of the pre‐1972 sanctions are very close to 1972, with only 18% of sanctions occurring before 1970. Omitting this variable from the analysis did not significantly affect the estimates for the other variables.

⁴⁶1972–1973 political rights data were used for any sanctions occurring before 1972. See previous note. Again, omitting this variable from the analysis did not significantly affect the estimates for the other variables.

⁴⁷1950 GOVGDP data were used for any sanctions occurring before 1950. Pre‐1950 sanctions arise in less than 1% of the 1,972 dyads used in the first set of results, and in none of the 138 dyads used in the second part of the study. The same holds for GDPPOP and TRADEGDP.

⁴⁸Note that TARGETDEM is excluded from the test of proportions in Table III. TARGETDEM is a monadic, not dyadic, variable: its value is a measure of regime type only in the target country and therefore cannot be interpreted to reflect the degree of similarity or dissimilarity between the two countries in a dyad.

⁴⁹This result holds if the one sanction in which the United States was a target is included.

⁵⁰Examining the correlation matrix for the independent variables reveals one possible multicollinearity problem, as indicated by the 0.74 correlation coefficient of CIVILLIB and POLRIGHTS. If POLRIGHTS is removed from the regression, CIVILLIB becomes significant at a 1% error level, with the rest of the results coming out about the same. If POLRIGHTS is included and CIVILLIB is omitted, the results for the other independent variables are similar to those shown in Table IV.

⁵¹ http://en.wikipedia.org/wiki/List_of_countries_by_GDP_(nominal)_per_capita.

⁵²The finding that sanctions are more likely to occur between countries with similar income levels than dissimilar might be due to the fact that sanctions applied by a very poor country against a very rich country would likely be futile, while sanctions applied by a very rich country against a very poor country might be perceived as untoward bullying.

⁵³Table V differs from Table II in that only observations that are used in the regression analysis (observations for which no independent variables are missing) are included.

⁵⁴For a general discussion of this type of bias, see Judge et al. (1985: 779–785).