Friends by Sanctions: International Relations, Trade, and Welfare

Friends by Sanctions: International Relations, Trade, and Welfare Yong Suk Lee, Stanford University August 2015 Abstract Despite the wide spread implementation, and debate surrounding the e cacy of sanctions, we have little empirical evidence on the impact of sanctions. This paper examines the impact of sanctions on trade and welfare. Sanctions primarily intend to punish target countries by inhibiting them from the gains from trade. The e ect of sanctions depends on not only the direct bilateral impact on trade but how much trade is diverted to other countries. I decompose international trade by the types of trade partners when countries are sanctioned, and empirically examine the impact of sanctions in a gravity framework. I construct a bilateral panel of all sanctions and trade between 1950 and 2005. To mitigate endogeneity, I focus on sanctions motivated by non-economic causes, such as sanctions imposed due to humanitarian or sovereignty issues unrelated to the target-sender trade relations. I find that sanctions reduce bilateral trade by 30%. This average impact almost exactly o sets the impact of regional or free trade agreements that the literature has found. For non-democratic countries the impact is larger at about 47%. Despite the large bilateral trade impacts, trade diversion mitigates the net impact of sanctions. Sanctioned countries increase trade with all countries that do not sanction them by 9%. The net e ect of sanctions is a 14% reduction in trade. However, this impact is asymmetric. Sanctions substantially reduce aggregate trade of non-democracies but do not impact democracies. When both countries are sanctioned by a third party, their trade volume increases substantially. Trade between two countries increases over two-fold when both are sanctioned by a third party country compared to when both are not sanctioned. The results imply that there is a common enemy e ect in trade. I further probe the international relations aspect of trade. I construct the absolute di erence in the democracy level between country pairs and examine the impact of this political distance in a structural trade gravity framework. I find that closer the two countries are on the political spectrum, the larger the trade amount is. This result even holds in the very tight specification where I focus on the variation within country pairs. International relations role in trade diversion generates heterogeneous welfare consequences when countries are sanctioned. I find that sanctions reduce welfare by about 0.5% in non-democracies using a method that summarizes welfare change from the change in trade openness. When I directly estimate the impact on real GDP per capital, which incorporates changes beyond trade openness, the impact becomes substantially larger with sanctions reducing welfare by 7%. Lastly, I find that target countries are willing to su er a 20% decline in GDP per capital before conceding to sanctions. Freeman Spogli Institute for International Studies, Stanford University, 616 Serra St. Stanford, CA 94305. Email: yongslee@stanford.edu. I thank Kyle Bagwell, Dave Donaldson, and participants at Stanford International Trade Workshop and the Western Economics Association International Conference for comments. Yoonsang Bae and Danielle Dobos provided excellent research assistance. 1

Keywords: Sanctions, Trade Diversion, International Relations, Welfare JEL Codes: F14, F51, F59, F69 2

1 Introduction Economic sanctions have become an increasingly popular foreign policy tool. Between 2001 and 2005 there were 138 di erent sanction cases, and 94 countries, almost half of the countries in the world then, were targets of sanctions. The list of target countries includes not only the usual suspects like North Korea, Iran, and Syria, but many of the non-democratic and poor countries around the world. 1 Sanctions predominantly aim to hinder or cut o the target country s access to international trade and finance through various measures, including embargoes, asset freezes, aid and travel restrictions and trade bans. In other words, sanctions leverage by inhibiting countries from the gains from trade. However, sanctions have generally been ine ective in changing the target country s behavior. Hufbauer et al. (2009) find that only 34 percent of the sanctions from 1915 to 2000 were at least partially successful, and that most of the successes happened before the 1970s. Either sanctions are not doing a good job at inhibiting trade or target countries are bearing substantial welfare losses to maintain their political objectives. One reason that sanctions might not be e ective in changing behavior is because sanctions often aim to hinder bilateral trade, whereas trade is multilateral. Globalization and the reduction in trade costs over the decades have enabled countries today to trade with a wide range of countries, and countries may substitute trade partners with another when one becomes less amicable. Despite the wide spread implementation of sanctions, and debate surrounding the e cacy of sanctions, we have relatively little empirical evidence on the impact of sanctions. This paper aims to fill this gap and examines the impact of sanctions on trade and welfare. The e cacy of sanctions depends on the direct impact of sanctions on trade and how well target countries are able to divert trade to other countries. I use an accounting framework to decompose trade into bilateral trade, trade with other sanctioned countries, and trade with countries that do not sanction the target country to examine trade patterns and diversion. I consider sanctions as one element of the bilateral trade friction that changes over time and empirically examine the impact of sanctions within the structural gravity framework which has widely been used in the 1 In contrast, 50 years ago between 1951 and 1955, well into the cold war, there were 41 sanction cases and 31 countries (out of 100) were targets of sanctions. 3

trade literature. I also examine the welfare consequences from sanctions. Arkolakis, Costinot, and Rodriguez-Clare (ACR hereafter, 2012) recently illustrated that in a general class of models, the change in welfare due to trade shocks can be represented succinctly by the change in the domestic share of consumption and the trade elasticity. I first estimate the welfare change from sanctions using this method. Since, sanctions could impact welfare through channels other than trade, 2 I also examine the impact of sanctions on real GDP per capital. The validity of the reduced form analysis relies on the assumption that sanctions are exogenous to the countries trade volumes. The diverse causes of sanctions present opportunities for a plausible identification strategy. There are indeed economically motivated sanctions such as, sanctions that arise from trade disputes. However, many sanctions are motivated by humanitarian causes unrelated to the economic relationship between the sender and target. For example, genocide, weapons proliferation, infringement of sovereignty motivate certain countries, often third party western countries, to punish the target country by imposing sanctions. 3 By focusing on these set of sanctions, I am able to minimize endogeneity and better identify the impact of sanctions on trade and welfare. Using the universe of sanctions and bilateral trade data between 1950 and 2005, I find that sanctions reduce bilateral trade by 30%. This average impact almost exactly o sets the impact of regional trade agreements (RTAs) or free trade agreements (FTAs) that the literature has found. For non-democratic target countries the impact of sanctions is larger at about 47%. Despite the large impact of sanctions on bilateral trade, trade diversion mitigates the overall impact. Sanctions increase trade with countries that do not sanction the target country. Because the share of trade with non-sanctioning countries is substantially larger than the share of trade with sanctioning countries, the overall net e ect of sanctions reduces to about a 14% reduction in trade, still a significant impact. This net impact is highly asymmetric between democratic and non-democratic target countries. Sanctions reduce overall trade of non-democracies but have no impact on democracies. The heterogeneous trade diversion by democratic level indicates that country relations and 2 For instance, sanctions may have domestic political-economy and inequality consequences that impact welfare, may impact worker health and productivity, may impact the improvement and maintenance of infrastructure and ultimately firm productivity. 3 In some events, the sender often does not have much or any meaningful trade relationship with the target. For example, the US imposed bilateral sanctions on North Korea when it conducted nuclear tests, but trade between the two is very low. 4

a nity may impact trade patterns. I examine how bilateral trade changes when a country pair s sanctions status changes. When both countries are sanctioned by a third party, their trade volume increases substantially. Being sanctioned brings the two closer in terms of trade. This pattern is stronger when we focus on the set of non-democratic countries. Trade between the two increases over two-fold when both are sanctioned by third parties compared to when both are not sanctioned. The results imply that there is a common enemy e ect, where substantial amount of trade is diverted towards countries that face similar hostilities. In other words, sanctions bring target countries, especially non-democratic ones, closer. I further probe into how sanctions impact the international relations aspect of trade. I find that when non-democracies are sanctioned their trading partners become even less democratic - the average polity index of trade partners drops by almost one. However, when democracies are sanctioned, sanctions have no impact on the average democracy level of their trading partners. Sanctions elucidate how international relations are an important factor in determining trade patterns. But are international relations only important in situations that involve hostilities like sanctions and war, or is international relations a general factor that impact trade patterns? I examine whether international relations can explain the patterns of international trade in the structural gravity framework that has done very well in explaining trade patterns. Measuring international relations is challenging and international relations is endogenous to trade. The strategy I pursue is to use the absolute di erence in the polity index between country pairs as a proxy for the a nity between countries. I find that closer the two countries are on the political spectrum, the larger the trade volume is. This result even holds in the very tight specification where I focus on the variation within country pairs. The fact that I get a statistically significant impact on this distance variable in a fully specified structural gravity model indicates that international relations in general impacts the pattern of international trade. International relations role in trade diversion generates heterogeneous welfare consequences when countries are sanctioned. I find that sanctions reduce welfare by about 0.5% in non-democracies when I use ACR s method. When I directly estimate the impact on real GDP per capital, the impact becomes substantially larger with a 7% reduction. Lastly, I examine how large the welfare loss 5

is in cases where the target country concedes to sanctions. The reduction in real GDP per capita is about 11% on average when target countries concede. This reduction in welfare is substantially higher for non-democracies. Their autocratic leaders are willing to su er along the lines of a 21% decline in GDP per capital before they concede to the senders. I believe this is the first paper to present a full empirical examination of how the universe of sanctions in the past century has impacted trade and welfare. This paper relates to the literature that examines international relations and trade. Berger et al. (2013) find that CIA interventions during the Cold War increased imports from the US, because foreign governments increased purchases of US products. Martin et al. (2008) examine the relationship between trade and military conflict and find that bilateral trade reduces the probability of conflict but multilateral trade openness increases the probability of conflict. Glick and Taylor (2010) examine the impact of war on bilateral trade. A few papers examine sanctions and trade patterns. Haidar (2015) examines trade diversion in Iran using firm level data, and Yang et al. (2004) find evidence that countries divert trade to the EU when imposed by US sanctions. Lee and Lee (2015) use US sanctions since 1992 and examine how trade patterns change during the threat stages of sanctions and show that international relations uncertainty can impact trade patterns. Also related is the literature that examines the impact of sudden trade barriers created between nations or regions. Atkin and Donaldson (2015) examine the impact of the border between Pakistan and India on trade patterns, and Etkes and Zimring (2015) examine the impact of the blockade along the Gaza Strip. Lastly, I further contribute to the literature that studies the e cacy of sanctions. Eaton and Engers (1992, 1999) presents a game theoretical model that examines the conditions under which sanction threats and imposition occur and when sanctions might be an e ective tool to influence foreign policy. Whang et al. (2013) empirically examine when sanction threats can successfully extract concession from target countries. Lee (2015) examines how autocrats counter sanctions domestically and the subsequent distributional consequences. Using satellite lights data he shows that that North Korea counters the impact of sanctions on the elites by diverting resources to the urban area, which further reduces the welfare of the marginalized population. 6

The next section of the paper describes the data and present background and descriptive evidence on sanctions. Section III presents the accounting framework to decompose trade when countries are sanctioned. Section IV presents the empirical strategy. Section V presents the empirical result on sanction and trades. Section VI further explores how political distance between countries explain the trade patterns. Section VII examine the welfare consequences from sanctions. Section VIII concludes. 2 Background on Sanctions and the Data Information on sanctions comes from the Threat and Imposition of Sanctions (TIES) data, a comprehensive data of all sanctions cases from 1945 to 2005, constructed by Morgan, Bapat, and Kobayashi (forthcoming). The TIES data includes all sanctions cases from 1945 to 2005, and has information on the sender and targets, the causes that triggered the sanctions, and the actions taken. It also codes whether the target countries conceded to the sanctions. I focus on the post 1950 periods as sanctions incidences were rare before 1950. Sanctions have been increasingly used as a foreign policy tool since the end of the World Wars. Figure 1 presents this pattern over time. The solid line represents the number of sanction cases in e ect each year and the dashed lines indicate the number of targeted countries. The frequency of sanctions has been increasing quite steadily. By 2002 there were over 140 sanction cases in e ect. The number of countries, on the other hand, was relatively stable before the 1990s and then jumps up after. The timing coincides with the end of the Cold War. In short, sanctions have been popular since the World Wars ended, but especially increases in both the frequency of cases and target countries once the World hegemony shifted to capitalism and democracy. The countries that impose sanctions, the senders, and that receive sanctions, the targets, are quite di erent. Table 1 lists the top 15 senders and targets by each decade. The top target countries are diverse and changes over time. The most frequently sanctioned country in the 1950s was the US, followed by Yugoslavia and North Korea. In the following three decades South Africa s apartheid regime was the most frequently targeted, among countries like Cuba. In the 1990s the Middle 7

Eastern countries and in the 2000 s Israel became frequent targets. Contrary to the list of target countries the list of top sender countries are much more homogenous. The United States has always been the dominant sender of sanctions by a far margin, followed by the UK, Canada, and other west European nations. The senders of sanctions are predominantly richer western democracies. Table 1 shows that there is substantial asymmetry between the senders and targets of sanctions. Figure 2 summarizes the asymmetry by plotting the average polity index, i.e, a measure of democracy which I describe below, of senders and targets by year. Consistently, there are gaps between the two groups, and if any has become wider in recent years (maybe show gap instead). Figure 3 presents the causes of sanctions by decade. I group the causes into 5 groups. 4 The validity of empirically analyzing the impact of sanctions on trade and welfare hinges upon whether sanctions are exogenous to these economic variables. The diverse causes of sanctions present opportunities for a plausible identification strategy. There are sanctions caused by disputes over trade practices or economic reforms and examining these sanctions su ers from endogeneity. However, as Figure 3 indicates, many sanctions are caused because of non-economic reasons: terrorism, weapons proliferation, human rights, territorial disputes etc. In such cases, the sender is often a third party country not directly involved with the underlying issue, or often does not have significant economic relationship with the target. 5 By focusing on these set of sanctions, I am able to minimize endogeneity and better identify the impact of sanctions on trade and welfare. Figure 4 presents the type of sanctions, i.e., the action of sanctions, by decade. I also group the types in to five categories - suspension of economic agreement, travel ban, termination of foreign aid, asset freeze, or blockade. All of these actions aim to inhibit the target country to international trade and finance, with an ultimate goal of reducing the gains from the trade of goods, money, or people. To construct the panel of sanctions and trade, I merge the TIES data to the the Correlates of Wars (COW) bilateral trade data. I merge in country macro economic variables from the Penn World Tables data and the country democracy level information from the Polity IV data. The 4 Describe the original classifications and groups here. 5 For example, the US imposing on South Africa s apartheid regime, or imposing sanctions on North Korea for conducting nuclear weapons tests. 8

Polity IV data, constructed by Marshall, Gurr, and Jaggers (2014), quantifies whether a country is more close to a democracy or an autocracy. The polity index ranges from -10 (strongly autocratic) to 10 (strongly democratic). I will refer to a country as democratic if the polity index takes a positive value and non-democratic otherwise. This is a obviously a rough categorization. I will perform analysis on finer categories in the robustness tests and the appendix. Table 2 presents summary statistics on some of the main variables by democratic status. The final data set is a one-directional country pair level data that spans the universe of countries between 1950 and 2005. The trade data often report missing trade or zero trade. The literature has tried di erent adjustments to deal with such cases. The main analysis will take the data as given, i.e., consider zero as zero trade and missing as missing information, but I will also examine results when missing is considered zero trade. To deal with the high frequency of zeroes in the trade data, the literature has used di erent methods. For now I use OLS but will examine how sensitive results are to di erent estimation methods, such as, tobit or poisson regressions. 3 Sanctions and Trade Diversion: Decomposing International Trade If the target country can divert trade to other countries then the overall impact of sanctions will depend on the direct impact of sanctions on trade with the sender but also on how much diversion occurs. A large amount of diversion would most likely mitigate the net impact of sanctions. In examining diversion it will also be important to examine to whom trade is being diverted to. If the degree of diversion di ers by the type of target countries, e.g., democratic vs non-democratic countries, then there could be overall distributional consequences from sanctions. Trade models and structural gravity equations focus on the economic, geographic, and international organizational determinants of trade. However, sanctions are simply another form of trade friction between countries, and because the causes of sanctions were motivated by humanitarian, security, sovereignty issues and so on, the pattern of divergence would likely reflect the trading 9

partner s value on the issues involved. 6 In other words, some political preference between countries, or international relations could impact the patterns of international trade in trade diversion but also in general. I later explore this further in Section 6. I use a simple accounting framework to empirically examine trade diversion. The accounting framework easily allows one to decompose trade patterns based on the trade shock of interest, in this case, sanctions. Since an accounting equation is an identity reflecting general equilibrium outcomes, the elasticity of each component of the accounting equation to a trade shock can potentially be estimated using the standard gravity equation. To ease notation I drop time indexes but consider trade at any year. I use the terms sender and target when referencing countries in relation to both sanctions and trade. S ij refers to sanctions imposed by i on j and X ij refers to the trade flow from i to j. X j refers to all trade flows into j, i.e., total imports by j. The sender country i is interested in how the sanction it imposes on j impacts j s trade. Not only is i interested in the impact on bilateral trade with j, but is concerned of any trade diversion that might negate the impact of the sanctions it imposes. Hence, I examine both the direct impact of sanctions on bilateral trade, but also the impact of sanctions on trade with other countries. Given that multiple countries can sanction j, I further decompose import from other countries into import from other sanctioning countries and import from non-sanctioning countries. The following identity describes the total quantity of imports by j. X j = X ij + X is j + X i ns j (1) where X is j denotes j s total imports from sanctioning countries other than i and X i ns j denotes j s total imports from non-sanctioning countries. Di erentiating with respect to sanctions, I derive the following expression for the impact of country i s sanction on trade. d ln X j ds ij = r ij d ln X ij ds ij d ln X i + r s j d ln X i i s j + r ns j i ds ns j (2) ij ds ij 6 Note that I am implicitly assuming that the states influence firms s trade partners. 10

In this expression, s ij is the fraction of total imports accounted for by imports by i, s i s j is the fraction of import from other countries than i that sanction j, and s i ns j is the fraction of import from countries that do not sanction j. Note that for now, I have sanctions as a continuous variable but will consider both the above equations can easily accommodate a discrete sanction variable. This decomposition amounts to a series of regressions, which one can estimate given convincing identifying variation of sanctions. I am ultimately interested in how sanctions impact welfare of the target country which can be examined using the more aggregate target country level data. Under the ACR framework the change in welfare is expressed as d ln W j = 1 1 d ln jj (3) where W j refers to the welfare of country j and that a country consumes that are produced locally, and 1 jj is the domestic share, i.e., the fraction of goods is the trade elasticity. Hence, the welfare impact of sanctions can be summarized, given the trade elasticity, by the change in the country s domestic share. jj can be estimated as 1 X j Y j from the data and takes on values from Head and Mayer (2014). 7 The above is based on ACR s assumption, which are generally accepted assumptions in the literature, but nonetheless, I also examine the a-theoretical and reduced form impact of sanctions by examining how sanctions impact the target country s real GDP per capita. The following section describes the identifying sources of variation that I use and the resulting estimation equations. 4 Empirical Strategy I estimate the e ect of sanctions on the di erent trade flows in equation (2) based on the gravity equation. The general gravity equation, where X ijt denotes country j s import from country i, can be represented by 7 Discuss measurement of domestic share in more detail. 11

X ijt = G t M ex it M im jt ijt (4) G t is a common year-specific factor determining trade, M s it and M t jt represent exporter i and importer j s multilateral terms. These are indexes that represent i s exporting and j s importing capabilities based on structural gravity equations. ijt represents the trade friction or bilateral accessibility of j to i. Economic sanctions are also a part of ijt. I approximate the log of ijt as a linear combination of factors that a ect trade costs between i and j. ln ijt = DSanc ijt + ij + u ijt. (5) DSanc ijt is a dummy variable equal to one if country j is sanctioned by country i. As before, in the context of sanctions country i is the sender and j is the target. ij captures the unobserved bilateral trade costs terms that are fixed over time, such as, distance, border, etc. u ijt is assumed to be random noise. I estimate the gravity equation by taking the log of (5) and substituting in (6) which returns ln X ijt =lng t +lnm s it +lnm t jt + DSanc ijt + ij + u ijt. (6) Note that I now denote the multilateral terms with superscript s for sender and t for target to focus on the sanction context. I use year dummies to capture ln G t, sender-year and target-year fixed e ects to capture ln Mit s and ln M jt t, and target-sender pair (dyad) fixed e ects to capture ij. I will estimate equation (7) as linear regression with high-dimensional fixed e ects. The high dimensional fixed e ects often make estimation challenging and the literature has used various methods, including ratio s, triads, etc.. to estimate (3). Head and Mayer (2013) perform a Monte- Carlo of the di erent methods and find that OLS with fixed e ects when implementable to well. The estimation of the first partial derivatives in equation (2) will be based on the empirical gravity equation (7). The high-dimensional fixed e ects imply that I will be using only the sanctions variations within a country pair over time while controlling for any sender and target specific 12

factors that can vary each year. These are tight controls but nonetheless one may be concerned that sanctions maybe endogenous even within country pairs. Such concern is valid if one considers sanctions triggered by trade disputes. For instance, dumping of certain goods by one country may trigger sanctions by another. However, endogeneity concern is substantially reduced if we focus on sanctions that are triggered by the target country s rogue behavior towards its domestic population, e.g., South Africa s apartheid regime, Syria s civil conflict, and has no direct relation with the senders, often the western democracies. This is why in the empirical analysis I focus sanctions triggered by humanitarian or political issues. US s decision to sanction on Syria or Russia are not motivated by trade concerns. In some sense, despite the economic and trade consequences, the US imposes sanctions to show that such rogue behavior can not be left unpunished. Similarly, I estimate the other partial derivatives that focus on trade diversion in equation (2) using equation (7) s bilateral framework and constructing trade values that correspond to X i s j, j s import from countries other than i that sanction j in year t, and X i ns j, j s import from countries that do not sanction j in year t. Since, my focus is empirical, rather than making theoretical assumption of balanced trade, I examine exports separately in addition to imports when the target country is sanctioned. The above gravity equation and equation (2) examines the impact of sanctions on bilateral trade from the sender s perspective. This is what the sender of sanctions would be interested in. I am also interested in the target country s perspective, i.e., how do sanctions impact overall trade and welfare of the target country? Countries are often sanctioned by multiple senders when sanctioned because of humanitarian or sovereignty issues. Hence, examining the target s perspective requires using data at the target country level. In practice, I use the following specification ln X jt = Sanc jt + Z jt + j + t + u ijt. (7) Sanc jt can be a dummy variable indicating whether country j was ever sanctioned in year t, or the total number of sanctions, or the set of indicators for each number of sanctions. Z jt is the set of target country variables, such as log population and log GDP. j and t are country and year 13

fixed e ects. Similarly, equation (8) serves as the base equation when I estimate trade by type of partners, domestic share, or real GDP per capita. 5 Results on Bilateral Trade and Trade Diversion Table 3 present results for the three elasticies from equation (2) : the direct impact of sanctions on bilateral trade, the impact on trade with other sanctioning countries, and the impact on trade with countries that do not sanction the target country. The latter represents the trade diversion that I am particularly interested in. The estimation framework is the structural gravity framework presented in equation (7), which includes the full set of fixed e ects. I cluster standard errors by country pairs. Panel A present results for imports and Panel B for exports. Column (1) indicates that sanctions of any type on average reduce import from the sender by about 27%. As discussed previously, focusing on sanctions motivated by non-economics related causes alleviates endogeneity. In column (2) I find that sanctions triggered by non-economic causes have a larger impact and significantly reduce import from the sender by 33%. Columns (3) and (4) examine the impact of sanction on imports from other sanctioning countries. This specification represents a general equilibrium outcome since, by definition the other countries are sanctioning the target. As one can expect the coe cient estimates in columns (1) and (2) are similar to that of columns (3) and (4). Columns (5) and (6) present the trade diversion from sanctions. Sanctions increase import from non-sanctioning countries by about 9%. The magnitudes are about 3.5 fold smaller than the impact on imports from sanctioning countries. However, the fraction of imports each group comprises will jointly determine the overall impact as derived in equation (2). If the fraction of imports from non-sanctioning countries is large enough, then the impact of sanction on overall imports would be neutralized due to the diversion in trade. Panel B presents results when I examine the impact of sanction on the target country s export. Overall results show very similar patterns from that of Panel A. 14

Table 4 presents the impact by sanctions type. Sanctions restrict trade but takes various forms. Columns (1) and (2) present results when I run regressions on the di erent types of sanctions individually, and columns (3) and (4) are results when I include all sanction types in one regression. As before, the estimation framework is the structural gravity framework presented in equation (7), which includes the full set of fixed e ects, and I cluster standard errors by country-pairs. When I examine each sanction type separately, the coe cient estimates are all negative and the majority of the estimates are statistically significant at the 1% level. However, in columns (3) and (4) economic embargoes stand out with large e ects. Asset freeze has a significant impact on reducing imports, and termination of foreign aid significantly reduces exports. Total economic embargoes have the strongest impact, reducing bilateral trade by about 50% (imports by 48% and exports by 67%). Given that each sanction type has the expected impacts, I will not focus on this margin further. Rather, I will focus on the heterogeneity of sanctions based on their causes, and primarily focus on sanctions due to non-economic causes, since those are less subject to endogeneity. The degree of trade diversion and how much diversion neutralizes the impact of sanctions will likely di er by the target country. I examine this heterogeneity by focusing on the democracy level of the target countries. I focus on the democracy level because depending on the level of democracy the target government would respond di erently from international pressure and internal hardship. An autocracy may be more determined to deter sanctions with less concern about the damage it causes to it s population. Also, because the causes of sanctions were motivated by humanitarian, security, sovereignty issues and so on, the pattern of divergence would likely reflect the trading partner s value on the issues involved. The level of a country s democracy is one simple way to proxy for these values. Table 5 presents the bilateral trade and trade diversion results by the target country s level of democracy. I focus on sanctions due to non-economic causes. I divide target countries into two categories, non-democratic and democratic, based on whether the polity index is less or greater than zero. Obviously, this encompasses a wide range, but the results are substantially di erent between the two. Panel A indicates that sanctions reduce bilateral trade by almost 50% in non-democracies, and the impact is statistically significant at the 1% level. However, the impact of sanctions on 15

bilateral trade for democracies are smaller in magnitude and not significant at the 5% level. The amount of trade diversion to non-sanctioning countries are also higher for non-democracies at 11% compared to 8% for democracies and the two estimates are statistically di erent. The results for exports in Panel B mirror the results for imports in Panel A. The fact that there is di erential impact and diversion from sanctions by the level of democracy could imply several things - senders may impose harsher sanctions on non-democracies, the institutional capacity to deal with sanctions may di er by the level of democracy, and the network of countries, or allies di er between democracy levels. I explore this further in later sections. The previous tables indicate that there is a significant impact of sanctions on reducing bilateral trade, but also that there is substantial trade diversion due to sanctions. Then what is the net impact of sanctions on trade? Table 6 examines the aggregate impact of sanctions on the target country s import and export. I will focus on Panels C and D, which examine the non-economically motivated sanctions, for the discussion. Panel C column (1) indicates that sanctions reduce aggregate import by 14% on average, but as columns (2) and (3) indicate the reduction only comes from non-democratic countries. In other words, despite significant trade diversion, non-democracies see a substantial reduction in aggregate trade because of sanctions, in the order of 22% for both imports and exports. On the other hand, democracies see no reduction in aggregate trade from sanctions. In Panel D, I semi-parametrically examine the impact of sanctions by including dummy variables representing the number of sanctions each year. Being sanctioned by one country significantly reduces aggregate import for non-democracies and the magnitude generally increases with the number of sanctions. When sanctioned by five or more countries the impact almost doubles. However, for democracies sanctions have no significant impact on both imports and exports. The only coe cient estimate that is weakly significant is on imports when the target is sanctioned by five or more countries. Table 6 implies that sanctions have welfare impacts only on non-democracies. 16

6 International Relations and Trade In this section, I further explore the patterns of diversion. If sanctions cause diversion then who are the target countries diverting trade towards? In Table 7, I examine how bilateral trade changes when the country pair s sanctions status changes. Specifically, I ask how trade volumes di er when both are sanctioned by a third party country, when only one is sanctioned by a third party country, and when neither are sanctioned, compared to situations when one is sanctioning the other. The results paint an interesting picture. When both countries are sanctioned by a third party, their trade volume increases substantially more compared to all other situations. This pattern is stronger when we focus on the set of non-democratic countries. Trade between the two increases over twofold when both are sanctioned by third parties compared to when both are not sanctioned. Table 7 provides evidence that there is this common enemy e ect where substantial amount of trade is diverted towards countries that face common threats. In other words, sanctions bring target countries, especially non-democratic ones, closer. I further probe into how sanctions impact the international relations aspect of trade. In Table 8, the dependent variable is trade weighted polity index of the target country s trading partners. Results are similar across the di erent specification. As before, I focus on Panel C for the discussion. On average sanctions reduce the polity index of trading partners by 0.5 and this e ect is statistically significant at the 1% level. Again, when I divide the sample, we can see that this result is driven by non-democracies. When non-democracies are sanctioned their trading partners become less democratic - the polity index drops by almost one. However, when democracies are sanctioned sanctions have no impact on the average polity index of their trading partners. With sanctions as the forcing mechanism, tables 7 and 8 show us how the a nity between countries impact trade patterns. In other words, sanctions elucidate that international relations is an important factor that impacts trade patterns. But is international relations only important in situations that involve hostilities like sanctions and war? Or is international relations a general factor that impact trade patterns? Table 9 examines whether international relations can impact international trade in the structural gravity equation which has been very successful in explain- 17

ing trade patterns. Measuring international relations is challenging and international relations is endogenous to trade. The strategy I pursue is to use the absolute di erence in the polity index between country pairs as a proxy for a nity between countries. Democracies tend to like democracies and autocracies tend to like autocracies. Alliances are formed between countries to counter hegemony. The post World War years have been shaped by alliances forged along ideological lines during the Cold War and by the Western democratic dominance post-cold war. The dependent variable in Table 9 is total trade between two countries, i.e., export plus import. Columns (1) to (3) present estimates on absolute di erence in the polity index. In the full sample result of column (1), a decrease in the absolute distance by one increases trade by 3%. The impact becomes smaller when I split the sample, but is still statistically significant. In other words, political distance between countries is important in determining trade patterns. Columns (1) to (3) control for the target-year and sender-year fixed e ects, but not the country pair fixed e ects. Endogeneity is a concern, since geographic distance which is omitted may be correlated with political distance. I control for such endogeneity in columns (4) to (6), by additionally including country-pair fixed e ects. Hence, I am now examining the impact of the change in absolute political distance within country-pairs. This is a tight specification as one can imagine that democratic levels are often quite persistent and changes are slow. Nonetheless, column (4) indicates that closer the two countries are on the political spectrum, the larger the trade amount is. The impact is statistically significant but the magnitude becomes smaller. When I split samples between democracies and autocracies, the estimates are similar but the standard errors increase quite a bit. This is natural given that the variation in the polity index between country pairs will likely come from country pairs with one above and one below the zero polity cuto I use. However, the fact that I get a statistically significant impact on this distance variable even in a fully specified structural gravity model indicates that international relations plays an important role in international trade. 18

7 Welfare Implications International relations role in trade diversion would generate heterogeneous welfare consequences when countries are sanctioned. In this section I estimate the welfare consequences from sanctions. How do sanctions impact welfare and how large of a welfare loss do sanctions need to generate for target countries to concede? Recall that in Table 6 we saw that despite significant trade diversion, non-democracies still see a substantial reduction in aggregate trade when sanctioned, in the order of 22%, whereas democracies see no reduction in aggregate trade. The reduction in trade has direct implications for welfare. ACR (2012) shows that the welfare change from trade shocks can be succinctly presented using the change in the domestic share of consumption and trade elasticity. I use their formula and the trade elasticity of 5, the median value from Head and Mayer s (2013) meta analysis of gravity estimates. I first estimate the impact of sanctions on log domestic share in Table 10. Domestic share is calculated as one minus the import share of GDP. If sanctions reduce trade, the domestic share would increase, thereby reducing welfare. Focusing on Panel C, I find that sanctions increase domestic share by 0.027 in column (1). I further control for the country s trend in openness by including country specific time trends in column (2). The impact drops to about a half to 0.014, but is still statistically significant at the 5% level. I next examine democracies and non-democracies separately. In specifications with target country specific time trends, the increase in domestic share due to sanctions is significant for non-democracies only. A sanction event increases domestic share by 0.02 in non-democracies. How does this translate into welfare changes? Using equation (4) and -5 for 1, the decrease in welfare amounts to about 0.5%. This result measures the welfare changes due to the reduction in trade based on the ACR framework. However, there may be other channels than trade where sanctions can impact welfare. For instance, sanctions would likely impact the internal reallocation of resources for production as documented in Lee (2015). Also, sanctions could impact the budget and negatively a ect the fundamental factors of development like health, education, infrastructure, etc. Hence in Table 11, I directly examine how sanctions a ect real GDP per capital measured in 2005 dollars. Focusing 19

on Panel C column (4), sanctions reduce the target country real GDP per capita by 7%. This is a substantially larger impact than what we get based on the ACR method. This discrepancy may represent the channels other than trade by which sanctions impact welfare. Lastly, I examine how big of a welfare loss there is in cases where the target country concedes to sanctions in Table 12. The variable concedes to sanctions denotes indicates situations where the target country unconditionally obliges to all or part of the sender s request Focusing on noneconomically motivated sanctions in Panel B, GDP per capita loss amounts to about 11% on average when target countries concede. The welfare loss is substantially higher for non-democracies at 21%. Since there are many targets that never concede, I examine the additional welfare loss from conceding to sanctions, in addition to the imposition of sanctions. Column (4) indicates that sanctions reduce welfare by 11% in non-democracies and that they see a further 11% reduction in GDP per capital when targets concede. Target countries, or more like their autocratic leaders, are willing to su er along the lines of a 20% decline in GDP per capital before they give in to the senders. 8 Conclusion Despite the wide spread implementation, and debate surrounding the e cacy of sanctions, we have little empirical evidence on the impact of sanctions. This paper examines the impact of sanctions on trade and welfare. Sanctions primarily intend to punish target countries by inhibiting them from the gains from trade. The e ect of sanctions depends on not only the direct bilateral impact on trade but how much trade is diverted to other countries. I decompose international trade by the types of trade partners when countries are sanctioned, and empirically examine the impact of sanctions in a gravity framework. I construct a bilateral panel of all sanctions and trade between 1950 and 2005. To mitigate endogeneity, I focus on sanctions motivated by non-economic causes, such as sanctions imposed due to humanitarian or sovereignty issues unrelated to the target-sender trade relations. I find that sanctions reduce bilateral trade by 30%. This average impact almost exactly o sets the impact of regional or free trade agreements that the literature has found. For 20

non-democratic countries the impact is larger at about 47%. Despite the large bilateral trade impacts, trade diversion mitigates the net impact of sanctions. Sanctioned countries increase trade with all countries that do not sanction them by 9%. The net e ect of sanctions is a 14% reduction in trade. However, this impact is asymmetric. Sanctions substantially reduce aggregate trade of non-democracies but do not impact democracies. When both countries are sanctioned by a third party, their trade volume increases substantially. Trade between two countries increases over twofold when both are sanctioned by a third party country compared to when both are not sanctioned. The results imply that there is a common enemy e ect in trade. I further probe the international relations aspect of trade. I construct the absolute di erence in the democracy level between country pairs and examine the impact of this political distance in a structural trade gravity framework. I find that closer the two countries are on the political spectrum, the larger the trade amount is. This result even holds in the very tight specification where I focus on the variation within country pairs. International relations role in trade diversion generates heterogeneous welfare consequences when countries are sanctioned. I find that sanctions reduce welfare by about 0.5% in non-democracies using a method that summarizes welfare change from the change in trade openness. When I directly estimate the impact on real GDP per capital, which incorporates changes beyond trade openness, the impact becomes substantially larger with sanctions reducing welfare by 7%. Lastly, I find that target countries are willing to su er a 20% decline in GDP per capital before conceding to sanctions. 21

References Arkolakis, C., A. Costinot, and A. Rodriguez-Clare. 2012. New Trade Models, Same Old Gains? American Economic Review. Atkin D., and D. Donaldson. 2015. Berger, D., Easterly, W., Nunn, N., and S. Satyanath. 2013. Commercial Imperialism? Political Influence and Trade During the Cold War. American Economic Review. 103(2): 863-896. Davis, Lance, and Stanley Engerman. 2003. History Lessons Sanctions: Neither War nor Peace. Journal of Economic Perspectives, 17(2): 187-197. Dashti-Gibson, Jaleh, Patricia Davis, and Benjamin Radcli. 1997. On the Determinants of the Success of Economic Sanctions: An Empirical Analysis. American Journal of Political Science, 41(2): 608-618. Eaton, J. and M. Engers (1992), Sanctions. Journal of Political Economy, 100, 899-928. Eaton, Jonathon, and Maxim Engers. 1999. Sanctions: Some Simple Analytics. American Economic Review Papers and Proceedings. 409-414. Etkes and Zimring. 2015. Glick, R. and A. M. Taylor (2010), Collateral: Damage: Trade Disruption and the Economic Impact of War. Review of Economics and Statistics, 92(1), 102-127. Haidar, J. 2015. Head, K., T. Mayer, and J. Ries (2010), The Erosion of Colonial Trade Linkages after Independence. Journal of International Economics, 81(1), 1-14. Hufbauer, Gary C., Je erey J. Schott, Kimberly Ann Elliott, Barbara Oegg. Economic Sanctions Reconsidered, 3rd Ed. Washington, DC: Institute for International Economics, 2009. Lee Y. 2015. Countering Sanctions: The Unequal Geographic Impact of Economic Sanctions in North Korea. Stanford Center for International Development Working Paper No. 519. Lee, S. and Y. Lee. 2015. Political Influence and Trade Uncertainty: Evidence from Sanctions Threats and Impositions. mimeo. Levy, Philip I. 1999. Sanctions of South Africa: What Did They Do? American Economic Review Papers and Proceedings. 415-420 22

Marshall, M. G., T. R. Gurr, and K. Jaggers (2014), Political Regime Characteristics and Transitions, 1800-2013. Polity IV dataset version 2013. Distributed by Center for Systemic Peace. Martin. P, T. Mayer, and M. Thoenig (2008), Make Trade Not War? Review of Economic Studies, 75(3), 865-900. Mayer, T. and K. Head. 2013. Gravity Equations: Workhorse, Toolkit, Cookbook. Handbook of International Economics Vol. 4. Morgan, T. C., N. A. Bapat, and Y. Kobayashi (forthcoming), Threats and Imposition of Economic Sanctions: Updating TIES 1945-2005 Conflict Management and Peace Studies. Nunn, N. and N. Qian. 2014. U.S. Food Aid and Civil Conflict. American Economic Review, 104(6): 1630-1666. Whang, Taehee, Elena V. McLean, and Douglas W. Kuberski. 2013. Coercion, Informaiton, and the Success of Sanction Threats. American Journal of Political Science, 57(1): 65-81. 23

Figure 1. Sanctions and target countries over time 140! 120! 100! 80! 60! 40! 20! 0! 1950! 1952! 1954! 1956! 1958! 1960! 1962! 1964! 1966! 1968! 1970! 1972! 1974! 1976! 1978! 1980! 1982! 1984! 1986! 1988! 1990! 1992! 1994! 1996! 1998! 2000! 2002! 2004! Sanctions! Target!countries!! 24!

Figure 2. Polity index of sender and target 12! 10! 8! 6! 4! 2! 0! 82! 84! 86! 1950! 1952! 1954! 1956! 1958! 1960! 1962! 1964! 1966! 1968! 1970! 1972! 1974! 1976! 1978! 1980! 1982! 1984! 1986! 1988! 1990! 1992! 1994! 1996! 1998! 2000! 2002! 2004! sender! target!! 25!

Figure 3. Causes of sanctions by decade 100! 90! 80! 70! 60! 50! 40! 30! 20! 10! 0! 1950s! 1960s! 1970s! 1980s! 1990s! 2000s! Other! Trade!Practices/Economic! Reform! Weapons!Proliferation/ Terrorism/Drug!TrafGicking! Improve!Human!Rights,! Environmental!Practices! Release!Citizens,!Property/! Territorial!Dispute/!Deny! Strategic!Material! Figure 4. Type of sanctions by decade 100! 90! 80! 70! 60! 50! 40! 30! 20! 10! Suspension!of!Economic! Agreement! Travel!Ban! Termination!of!Foreign!Aid! Asset!Freeze! Blockade! 0! 1950s! 1960s! 1970s! 1980s! 1990s! 2000s!! 26!

Table 1. Top senders and targets by decade! 27!