The study of economic sanctions has so far focused on the actual imposition and to some extent on the threats involved but it has left out the consequences of ending an economic sanction. Evidence from testing the effects of ending a threat and lifting a sanction show that the levels of imports to the United States increase significantly afterwards. However, the recovery varies depending on the type of sanction and which year after the lifting is being observed.

There is evidence of lagged effects, where the first year post-lifting shows a continuous negative effect which later disappears. The relevance of this research lies in filling the gap of studying ex post effects of economic sanctions while using product level data.

Keywords: economic sanctions, international trade, ex post effects.

JEL codes: F00, F14, F51

(3)

- 3 -

1. Introduction

It has been said before that economic sanctions are what stands between diplomacy and war.

They are the mechanism utilised when statements are insufficient, but military action is exces- sive. The current state of affairs in the global political economy has brought sanctions back to the forefront of international policymaking. New heads of states, a vulnerable European Union (EU), a rise in terrorism and military action, among other issues, have yielded an uncertain future for international relations. Given the ongoing sanctions imposed against Russia, China and other Asian countries by the Unites States (U.S.), it is of great value to understand if indeed there are effects of economic sanctions on trade, and if so, what happens once a sanction is no longer active. The latter remains unexplored by current literature.

At the turn of the 20^th century, international trade was well underway and growing despite a lack of economic stability throughout the period. Two World Wars and several economic down- turns led to an increase in the use of economic sanctions by international policy makers (Eaton and Engers, 1999). At the early stages of the 21^st century, actions by powerful heads of state prove that this strategy is still broadly used. Thus, failing to acknowledge the importance of economic sanctions in current times is unwise and should be explored further.

Previous research has mainly focused on the economic impact of sanctions on trade and whether they have fulfilled their initial goal i.e. its effectiveness¹. However, there is a considerable lack of empirical research on the ex post effects of sanctions on international trade. Since sanctions are essentially exogenous trade shocks, the target country incurs an economic cost. Thus, when dealing with a penalty which forcefully decreases the amount of trade between two or more countries, one would assume, that even after the artificial depressor has been removed, some remnants of its damage endure.

For this reason, this research focuses on attempting to determine if the trade reduction experi- enced during an economic sanction persists after the sanction has been lifted. I have chosen to focus on trade, rather than another indicator like Foreign Direct Investment (FDI), because most economic sanctions are targeted towards imports and exports (Morgan, Bapat and Kobayashi, 2014). Furthermore, trade is generally more shock sensitive than FDI which may have a lagged effect since divestment is much more complex than ceasing to export/import (WTO, 1996).

However, there is evidence that American investors tend to pull their investments out of targeted countries before the actual imposition of the economic sanction (Biglaiser and Lektzian, 2011), which can have implications for the power of threats. Subsequent research questions will focus on the nature of the sanctions and how long the effect of the trade downturn lasts. The premise behind the paper is to ascertain the consequences of sanction lifting on trade after its natural course has been altered by the sanction.

1 Effectiveness refers to the compliance of the target country with the requirements of the sender country after the sanction has been imposed. This can entail a change in the target’s policy, a decrease in investment going into the country, a travel ban, and so on.

(5)

- 5 -

This study will take into account all U.S. sanctions, whether they remained as threats or were imposed on any other country in the world. The period observed is from 1989 to 2016 and the main variables taken into account will be import values and economic sanction characteristics.

The focus is on the U.S. since it is, by far, the largest sanctioner in the world, accounting for over 51% of all sanctions collected in the Threat and Imposition of Sanctions (TIES) dataset (Morgan, Bapat and Kobayashi, 2014), which is used in the subsequent analysis. The innovation of this paper is twofold: it covers a topic which, to the best of my knowledge, has not been comprehensively researched before; but also observes annual trade data for 28 years with all traded products at a 10-digit HTS level, yielding a substantial database which allows for more precise estimations. Previous studies have used aggregate data, which can bias results as section 4.2 demonstrates.

Using a high dimensional fixed effects estimation model shows that lifting a sanction increases trade significantly. The estimations provide evidence that the removal of a sanction can increase imports by as little as 10.3% and as much as 79%. These effects can be much larger (or smaller), depending on the type of sanction lifted. Further, results show that threatened sanctions recover to a greater extent than non-threatened sanctions. Finally, when it comes to the recovery of imports, the per annum effects following the end of the sanction are ambiguous.

This study will be structured as follows: Section 2 explains the previous findings on economic sanctions, their threats and their effects on third parties; Section 3 is related to methodology and data; Section 4 presents the empirical results and interpretation and Section 5 concludes, and presents the limitations and avenues for future research.

2. Literature

2.1 Sanctions and Effectiveness

Economic sanctions are defined by Morgan, Bapat and Kobayashi (2014), henceforth MBK, as “actions that one or more countries take to limit or end their economic relations with a target country in an effort to persuade that country to change its policies” (p.3). Implicitly, a sanction must include (at least) one sender who officially imposes the sanction under a national or su- pranational authority and one target. Since the end of World War I, sanctions have become a more popular control method for the United States government over other (non-) sovereign states. President Wilson and the League of Nations had discovered a cost-effective mechanism to exert control over countries that were not playing by their rules, not only in the terrains of economics but also in manners of democracy, human rights, terrorism and disarmament (Cor- tright and Lopez, 2000).

Over the last 70 years, there has been an increase in the frequency with which sanctions are utilised. Until the 1970s, the number of impositions was relatively stable, but with the new wave of globalisation also came a new wave of economic sanctions. The fall of the socialist bloc, the lowering of transport costs and the political changes of the time, created a new playing field for

(6)

- 6 -

the rebirth of non-violent intervention. Figure 1 shows the evolution of sanctions over time, showcasing again the hegemony of the American state. Sanctions imposed by the U.S. comprise 51.2% of all the sanctions in the TIES dataset². Together with the top 5 sanctioners in the world³, they account for 69.7% of impositions. However, in absolute numbers, these four senders have only issued about one third of the sanctions the U.S. has in the same period of time (MBK). During the last decade of the 20^th century, the use of sanctions became one of the most utilised tools by the American government to seek compliance from another nation (Allen, 2005).

Before an economic sanction is imposed, over half of the cases observed by MBK displayed a existent threat to sanction. Previous literature has shown that threatening to impose a sanction can be just as effective as an imposition under the right circumstances (Lacy and Niou, 2005).

Furthermore, the U.S. is a country that has enough political and economic power to influence the target country into changing their policies simply by threatening to impose a sanction. More- over, threats are being used emphatically as a show of authority, even if the intention is not to go through with an actual imposition (Morgan, 2015). Thus, it can be expected that threats are a viable policy tool for sender states to invoke compliance from the target⁴. This yields the first preliminary hypothesis:

H1: Threatening to impose an economic sanction will have a negative effect on trade levels.

Over time, researchers have focused primarily on whether economic sanctions meet their de- sired target, i.e. its effectiveness. A sanction is considered effective if the target state changes their policy after the sanction has been threatened or imposed. If, however, there is no change in the target’s behaviour after the imposition, the sanction is considered ineffective (Elliott and

2 A detailed explanation of the TIES dataset will be explained in section 3, Methodology.

3 Canada, the European Union, Russia and the UK, in order of number of sanctions sent. The European Union is considered as a unique sender representing all 28 countries, while the UK has also imposed sanctions unilaterally. These cases are considered separate from each other.

4 More information on the effects of threatening to impose a sanction in section 2.3.

0 100 200 300 400 500 600

1945 1955 1965 1975 1985 1995 2005

US Top 5 World

Figure 1. Sanction impositions by decade. Source: Morgan, Bapat and Kobayashi (2014)

(7)

- 7 -

Hufbauer, 1999). So far, scholars have studied sanctions’ effectiveness based on the number of countries sending a sanction (henceforth, senders), the sternness of the penalties, the third party effects⁵, policy changes, etcetera. For a sanction to occur, the sender usually identifies some action by the target country which is deemed to need correction, but not dangerous enough to require military action (Davis and Engerman, 2003).

Research on the effects of economic sanctions has been rigorous on the topic of international trade, the consensus being that sanctions put downward pressure on all types of economic ac- tivity in the target country: trade, income, investment, travel, and others. Hence, the effectiveness of the sanction lies in its ability to create economic costs such that the target state has the incentive to change its behaviour to recover the economic costs lost to the sanction. Thus, the receiving state needs to be susceptible to economic force. Marinov (2005), however, argues that even if the sanction does not have a strong economic effect, it remains proficient in destabilising the leaders they target; this could be why some sanctions are imposed even when they are not expected to reach their goal (Lacy and Niou, 2005).

However, scholars have found that economic sanctions do not provide a sufficient incentive for change. Elliot (1997) explained, in a speech to the House of Representatives, that since 1970 unilateral sanctions from the U.S. have had a success rate⁶ of only 13%, while multilateral sanctions have been considered effective in almost 35% of cases and only under certain conditions (namely a significantly smaller, weaker and less politically stable target).

Consequently, Hufbauer and Oegg (2003) measured the effectiveness of sanctions based on the severity of their penalties⁷. The study shows that sanctions depress bilateral trade with the target country by 99% if they are of an extensive nature. However, if the sanctions are of moderate or limited scope they are insignificant and pose no threat to the receiving state in terms of international trade. Thus, the level of comprehensiveness of a sanction increases its likelihood of success, that is, to depress trade levels (Allen, 2005). Caruso (2003), finds similar results using the same model, however, extensive sanctions reduce trade by 89%. The remaining types of sanctions do not have a significant effect on trade. Taking this into account, the following two preliminary hypotheses can be formulated:

H2: An imposed economic sanction will have a negative effect on trade levels.

H3: An economic sanction which has been threatened and then imposed depresses trade to a larger extent than those only threatened or those imposed without a threat.

5 Third party effects refers to sanction busting, which will be discussed later in section 2.4.

6Success rate refers to the ratio of sanctions which have accomplished the intended change in behaviour in the target state over those who have not. One of the most famous examples of a success is the use of economic sanctions by the U.S. towards South Africa with the objective of ending Apartheid. The sanction fell under the Comprehensive Anti-Apartheid Act of 1986 which was repealed after South African authorities starting taking measures to end segregation in 1991 (Thomson, 2012).

7 Hufbauer and Oegg (2003) categorise sanctions based on their severity into limited, moderate and extensive. Limited was left for minor penalties, moderate when they covered 5 or more restrictions and extensive was reserved for comprehensive prohi- bitions and embargos, like the case of Cuba, Iran, Iraq or North Korea.

(8)

- 8 -

2.2 Threatening to Impose a Sanction

The current literature has neglected the role threats play when it comes to economic sanctions. As early as the Edwardian era, speeches given by world leaders have been followed me- ticulously (Eaton and Engers, 1999). Nowadays, thanks to technological developments, the ver- nacular of global figures is quickly moving around the world and being painstakingly analysed to determine whether a threatening message exists. If that is found to be the case, markets tend to react immediately. In the past few days of writing this piece, several Gulf countries have threatened Qatar with economic sanctions bringing an almost immediate plummeting of Qatari financial markets (Beaumont, 2017). Shortly after, impositions related to access to trade, bank- ing, and media as well as travel restrictions were imposed, further disrupting the economic standing of the country (Feteha, 2017). Thus, it stands to presume that there could be a significant effect on trade emerging from threatening to sanction a country. In fact, Lacy and Niou (2005) find that threatening to impose a sanction may be just as effective as the imposition itself. Moreover, research on the topic of FDI has shown that U.S. investors are likely to divest from a country which is targeted before a sanction is imposed (Biglaiser and Lektzian, 2011).

Though not explicitly stated, this effect might be influenced by a threat to sanction which causes a change in investor behaviour.

In the dataset used for this research, a threat was issued on all occasions when there was not an imposition afterwards and in 57% of cases when a sanction was imposed subsequently (MBK).

Even so, few articles have focused on the consequences on trade of threatening to impose a sanction.

Kohl and Klein Reesink (2016) study the effect of threatening to impose a sanction and find that, surprisingly, upon the establishment of a threat, trade increases by 8% after the announce- ment. However, in the cases when a threat was followed by a sanction, the sole effect of a threat is no longer significant, while their interaction depresses trade by around 16% to 20%. Further- more, Barrett (1997) shows that when a country fails to provide a global public good⁸, a threat is only as strong as the power of the punishers is at enacting this penalty.

Moreover, Lacy and Niou (2005) find that economic sanctions are likely to be imposed even when there is no expectation that they will work in the first place. The authors show that a state which has not adapted their behaviour after receiving a threat, is unlikely to do so after a sanction, as well. As a result, governments knowingly sanction without expecting success but do so to maintain the credibility of their threats (Morgan, 2015). A similar approach is pursued by Eaton and Engers (1999) where they suppose that a threat is only issued if the sender has clear expectations that the target will change its behaviour immediately. Thus, an imposition only

8 The author analyses a link between policies that are meant to provide global public goods, like signing environmental care agreements, and the power of sanction impositions if compliance is not achieved. In this case, a threat may be enough to ensure that the target country follows the established protocols, but only in the case that the threat is backed by credible institutions or countries capable of inflicting a significant economic shock to the target.

(9)

- 9 -

occurs under two scenarios: (1) the sender has overestimated the compliance costs to the pun- ished state or (2) the target underestimated the determination of the punisher. Taking into account the findings of these scholars, the fourth hypothesis is:

H4: In the event of a sanction being lifted, sanctions that were threatened before they were imposed will recover less strongly than sanctions that were not threatened.

This expectation is in line with Eaton and Engers (1999) who suppose that a threatened and imposed sanction occurs because the resolve of the sender is strong. Thus, if a sending country has gone through the trouble of threatening a sanction, observed no change, and then imposed a sanction, it means that they stand more to gain from the imposition than from withdrawing the “attack” (Lacy and Niou, 2015). This would also mean that the sending government is less likely to allow a free flow of trade after the sanction ends without interfering. The lack of co- operation at the time of the threat may displease the sending state, which could echo after the sanction ends in fewer efforts to cooperate or taking longer is re-establishing business ties.

Thus, I expect a threatened sanction would have a lower recovery level post lifting.

2.3 Sanction Imposition

The action of imposing a sanction occurs, in most cases, because the threat of imposition has not been sufficient to incite changes in the behaviour of the target state. In the case of the US, almost half of all sanctions imposed between 1945 and 2005 have been due to disagreements in trade practices. In this period, the U.S. has threatened to impose full embargos in 119 occasions, yet have only enacted 11 such punishments. More often than not, the U.S. will endorse an import/export restriction or the termination of foreign aid to the target country but will avoid their total commercial isolation (MBK).

As discussed before, sanctions are strongest in the economic costs they trigger by forcefully decreasing the amount of trade between the sender and its target. The premise behind the sanction is that it will increase the costs to such an extent that the target will comply to recover the economic gains. Thus, many authors have asked the question of just how large is this effect on trade. Table 1 provides a short overview of some of the more popular papers regarding the effects of economic sanctions. The results of previous scholars indicate that sanctions indeed have a severe effect on trade, the initial consensus surrounding the 80% to 99% level of trade depression for extensive sanctions. However, if that were the case, should not the success rate of sanctions be higher than 13%⁹? If indeed, sanctions can nearly wipe out trade between two countries, then the incentives to comply should be higher.

This could imply that there are some biases in previous research. In this case, I consider two:

(1) initial research has been, while extensive in the scope of countries, not comprehensive over time. The observations are often a discrete number of dispersed years, while economic shocks

9 This estimate was presented by K. Elliot in 1997 as her testimony before the Committee on Ways and Means at the House of Representatives.

(10)

- 10 -

tend to have prolonged lagged effects. (2) These researchers have not accounted for country and time fixed effects (henceforth, FE). The first researchers focused on Ordinary Least Square (OLS), which does not discern internal variations of the data. Once time FE entered the picture, the effects of sanctions on trade decreased significantly. Fixed effects models would be preferred to OLS since they treat the variation in the data as if it is nonrandom. That means, that the FE model expects there to be some correlation between the data points within a certain group (Hill, Griffiths and Lim, 2012). This is extremely relevant for the study of economic sanctions, since the changes in trade levels within a certain country, year, or even product, are co-dependent. It is expected that the trade shock will affect some of the observations and this effect may resonate within the group.

Some of the newest research on the topic by Kohl and Klein Reesink (2016) finds striking results. Their methodology makes use of Multilateral Resistance Terms (MRT*) given that their dataset accounts for all country pairs. What MRT does is control not only for the time-varying effects, but also for country-varying effects, while taking into account the interaction of all other country pairs. The precision of their model results in a significantly lower impact than most previous research: imposed sanctions decrease trade by less than 20%. This means that much of the evidence from the past, which has been used as a yardstick for sanction policy until now, may be biased.

A large strand of research has focused on the number of senders. A sanction can be sent from one state to the other (unilaterally) or have multiple senders with a common change they want

Table 1. Previous findings on the effect of economic sanctions on international trade.

Author (Year) Model Years/Countries Results Hufbauer, El-

liott, Cyrus and Winston (1997)

Gravity Model with OLS

1985, 1990, 1995.

88 countries.

Extensive sanctions decrease trade by 90%. In OECD countries, these findings fluctuate to 78% decrease for extensive sanctions, 33% for moderate and 21% for limited sanctions.

Elliott and Hufbauer (1999)

Unspecified 1970, 1980, 1990.

Unspecified.

Extensive sanctions limit trade by 91%, moderate sanctions by 37% and limited by 27%.

Hufbauer and Oegg (2003)

Gravity Model with OLS

1995 and 1999. 175 countries.

Extensive sanctions depress trade by 99%, moderate and limited sanctions have no significant effects.

Caruso (2003) Gravity Model with country FE

1960-2000.

50 countries.

Extensive sanctions depress trade by 89%, while moderate and limited sanctions are not significant.

Yang, Askari, Forrer and Zhu (2009)

Gravity Model with time FE

1980-2003.

EU countries.

Multilateral sanctions depress EU trade by 42%

and EU imports by 64%. Unilateral sanctions are not significant.

Kohl and Klein Reesink (2016)

Gravity Model with MRT*

1948-2005.

223 countries.

Imposed sanctions decrease trade by less than 20%.

Note: Extensive sanctions refer to embargos or blockades, moderate sanctions are less stringent restrictions while limited sanctions involve minor penalties to trade. Sources: see articles in bibliography.

(11)

- 11 -

to impose (multilaterally). Many scholars agree that multilateral sanctions tend to be more suc- cessful than unilateral sanctions due to the additional pressure from the international commu- nity (Bapat and Morgan, 2009; Caruso, 2003; Elliott, 1997; Petrescu, 2016; Yang et al., 2009).

Neuenkirch and Neumeier (2015) test this by looking comparatively at sanctions imposed by the United Nations (UN) versus sanctions by the US, and find that impositions by the UN decrease trade by approximately 2% annually, amounting to 25.5% in a 10-year period. In the case of the US, the effect only aggregates to 13.4%.

In contrast, other scholars support the claim that a larger number of senders decreases the likelihood of success mostly due to the greater complexity of its enforcement and terms of trade, i.e. the “spaghetti bowl” effect¹⁰ (Allen, 2005; Kaempfer and Lowenberg, 1998; Miers and Morgan, 2002). In this specific research, sanctions are only considered if they are either sent or received by the U.S. Thus, the political and economic power of the U.S. makes their unilateral sanctions considerably stronger than a unilateral sanction imposed by any other one country, ipso facto, these are atypical unilateral sanctions. This implies that the support of other sanctioning partners is not as vital as it would be for a smaller nation. Taking into account this factor, and the research done previously, my fifth hypothesis follows as:

H5: Unilateral sanctions sent by the U.S. will depress bilateral trade flows more than multi- lateral sanctions.

Regardless, it must be noted that research done by the first set of scholars claims that multilateral sanctions are likely to cause a larger disruption in trade due to the scope of senders (Bapat and Morgan, 2009; Caruso, 2003; Neuenkirch and Neumeier, 2015; Petrescu, 2016; Yang et al., 2009). Following this train of thought, once the sanction is lifted, it would be expected that restoring trade relations with multiple sanctioners to be slower than with one sanctioning country. I would also argue that this restoration comes at a larger cost than dealing with one sole partner, i.e. renegotiations with government authorities, suppliers, distributors, etc. The added complexity of the rebuilding of trade flows yields the sixth hypothesis:

H6: The recovery of trade will be higher in a case involving a multilateral sanction than in the case of a unilateral sanction, once the sanction has been lifted.

2.4 Sanction Busting

Sanction busting is defined as “the deliberate disregarding of sanctions that are in force against a state, organization, etc.” (Collins Dictionary, 2017), and is argued to be one of the main reasons why sanctions fail to realise their objectives. For instance, Yang et al. (2009) have found evidence that countries sanctioned by the U.S. trade more with the EU after the sanction has been imposed. Thus, trade does not simply stop, it is rerouted. This redirection of trade does

10 The “spaghetti or noodle bowl” effect is often used with relation to trade agreements, where the proliferation in the number of treaties (or, in this case, sanction senders) creates overly complex and hard to apply conditions in each subsequent country (Kawai & Wignaraja, 2011).

(12)

- 12 -

not come without a price, and several scholars have tried to determine the lingering effects of economic sanctions.

Some researchers have concentrated on the duration of sanctions. Allen (2005) finds that the longer the sanction spell lasts, the less likely it is to succeed. Therefore, sanctions are preferred to be short and sturdy, since each additional year in the duration of a sanction increases the probability of sanction busting by a factor of 1.08 (194%) (Early, 2009).

The legitimacy of this premise can have important implications for the consequences of sanction lifting. Firstly, because longer sanctions are more likely to be undermined by third party states (Early, 2016; Hufbauer and Oegg, 2003; Van Bergeijk, 1995). There is supporting evidence that sanction busting is motivated by profit seeking, rather than political gain (Early, 2009). At the time of writing, there are ongoing sanctions sent by the EU against the Russian Federation due to the destabilisation of Ukraine and annexation of Crimea. The sanction imposed includes export restrictions, restricted access to capital markets, and other (European Council, 2017). Russia then imposed an embargo on all European foods products (European Commission, 2017). Nonetheless, inquiries have found that EU products are reaching the Rus- sian markets by being circumvented through Turkey and other traders, who are not part of the sanction (Szczepański, 2015). Nevertheless, there is a still a significant impact on Russia who has lost billions in potential trade flows due to the sanctions (Crozet and Hinz, 2016).

Secondly, because the longer a sanction lasts, the more severe consequences will follow for the reinstatement of trade relations. This is particularly important if there is a strong degree of trade dependence. Nonetheless, evidence shows the longer a sanction is in place, the lower the per annum effects, thus decreasing the probability of success (Neuenkirch and Neumeier, 2015).

On a similar note, research performed by Chen and Garcia (2016) related to trade barriers ap- plied to Norwegian salmon by China¹¹ suggests a new perspective on sanction busting. It was found that the circumvention of the sanction came from the inside of the economy, rather than other nations. Chinese importers utilised newly established trade relations to acquire the product by using less centrally observed airports and by smuggling cargo via Vietnam and Hong Kong.

These findings are deeply relevant, in that they cause a recalculation of expectations on the recovery period of trade following sanction lifting. In this case, the recovery to pre-sanction trade levels may not be complete due to informal trade relations undermining the sanction while it is still active.

Altogether, this could provide evidence on potential reasons why trade levels will not go back to pre-sanction levels. Four potential exogenous¹²reasons are identified from the information

11 The non-tariff barriers were imposed after Norway awarded the Nobel Prize to a Chinese dissident. Moreover, the country even refused to welcome the Dalai Lama fearing that Chinese retaliation would worsen given the strained relationship between the Chinese authorities and the Tibetan leader (Chen and Garcia, 2016).

12 Exogenous effects are mainly due to external forces related to the sanctions like the power of the sender country, sanction busting, international regulations, etc. While endogenous effects are consequences arising from the inside of the targeted economy such as corruption, informal markets, etc.

(13)

- 13 -

above: (1) trade-busting practices could undermine the official trade flows of the U.S., (2) the deterioration of business relationships entail time and pecuniary costs to the target traders, (3) a longer sanction implies a higher chance that international regulations have changed increasing new trading costs in adapting products or services, and (4) a “wait and see” effect, where ven- dors from both sides would rather wait and determine the new status quo before re-investing to avoid incurring in unnecessary costs¹³.

2.5 Sanction Lifting

As for the ex post effects of economic sanctions, Hufbauer and Oegg (2003) find that they affect global value chains and trigger additional costs to the economy in the re-establishment of business ties. Moreover, these ties may remain broken long after the lifting, due to fear that the country may be sanctioned again. Yet, their regressions find very limited evidence that sanctions continue to suppress trade after they have ended. Caruso (2003), on the other hand, counterfactually estimated that target countries would have traded 59% more with G7 countries, had the U.S. not sanctioned them. Thus, the compounded effect over time is paramount to the economic costs of the target country. Neuenkirch and Neumeier (2015) estimate that over a 7 year period, U.S. sanctions cost the target nation 13.4% of GDP growth.

After a sanction is lifted, there is a multitude of concerns for the economy ranging from damaged business relationship (Hufbauer et al., 2003), to rising inequality (Afesorgbor and Mahadevan, 2016; Peksen, 2016). Petrescu (2016) focused on the informal economy and deter- mined that sanctions also increase criminality. These effects tend to persist because of the higher wages in the informal sector which remove the incentives to move to taxable work. Cor- ruption too is resilient and countries which have been sanctioned are perceived to be more corrupt than those which have not. Evidence shows that nations that have had comprehensive sanctions imposed are more corrupt than nations with partial sanctions (Kamali et al., 2016). These scholars show that there are other endogenous¹² factors which can affect the extent to which bilateral trade will recover, arising from the inside of the target economy. The consequences of sanctions are not limited to trade flows per se, but also to internal issues and external percep- tions of the country after a sanction ends. Yielding my main and final hypothesis:

H7: After the lifting of an economic sanction, the target country will experience a lagged effect on trade recovery.

2.6 Relevance

The existing literature has explored extensively the effectiveness of sanctions (Chesterman and Pouligny, 2003; Malloy et al., 1990; Neuenkirch and Neumeier, 2015), they have delved into the issue of multilateral vs unilateral sanctioning (Allen, 2005; Bapat and Morgan, 2009;

13 This is related to point (2) where the re-establishment of business ties comes at a cost of negotiations, abiding by regulations, contract making, finding new suppliers, etc.. Investing in these relationships could be considered a potential sunk costs if a new sanction was sent afterwards, making the investment a loss.

(14)

- 14 -

Yang et al., 2009), they have even considered the effects on the informal economy (Petrescu, 2016). Many others have based their research on a single country like Iran (Ianchovichina, Devarajan and Lakatos, 2016), Myanmar/Burma (Bünte and Portela, 2012; Kubo, 2014) or, more currently, Russia (Christie, 2015; Crozet and Hinz, 2016; Menon and Rumer, 2015;), yet, there is a significant gap in the topic that remains unaddressed.

To my knowledge, there is no published article which comprehensively addresses the consequences of the lifting part of economic sanctions. The strength of this paper lies in its purpose and its data. First, this article tries to understand the ex post effects of an economic sanction in the target and host economies. My research question is to determine if the trade depression experienced during an economic sanction persists after its lifting. This study will take into account all sanctions which have been imposed by the US and have ended, from 1980 until 2012.

Trade data will continue until 2016 to observe the after-effects. Second, it is also innovative in using annual trade flows at product level, using 10-digit HTS codes. It is clear that not all sanctions fall under the category of a full embargo; therefore, studying aggregate trade data will inevitably bias the results forgoing alternative sanction types.

Further, the relevance of the present research can be separated into several channels. For policy makers, empirical testing of the consequences of economic sanctions can help in the better de- sign of the sanction itself. Some researchers (Elliott and Hufbauer, 1999; Hufbauer and Oegg, 1997; Malloy et al, 1990) have called upon the need for better, targeted sanctions to ensure that the brunt of the punishment falls on the offenders and not on the blameless. Secondly, for researchers, the use of an extensive data set combined with a state-of-the-art gravity model can help in new research of a similar nature to obtain more reliable results. The use of product data will also set an interesting precedent in the study of international sanctions. Finally, the number of impositions has increased over time, particularly in the U.S.

The current sanction tide against countries like Russia (due to the annexation of Crimea), the Middle East (due to its inaptly labelled Arab Spring) and China to a lesser extent (due to their claims on the South China Sea), have brought the issue back to the forefront of retaliatory policy. The regime of the newly appointed President Trump will most likely impose more sanctions as his mandate continues. Thus, it will be useful to know what the long term consequences of lifting an economic are, now more than ever.

3. Methodology

3.1 Gravity Model

International economics has made extensive use of the gravity equation as an estimator for bilateral trade flows, and now, also for sanctions (Caruso, 2003; Hufbauer et al., 1997; Hufbauer and Oegg, 2003; Yang et al., 2009; etc.). The rationale behind its utilisation is that it takes into account the size of the trade flows while accounting for other factors that could influence the trade relationship. The researchers mentioned used the standard gravity model with Ordinary

(15)

- 15 -

Least Squares (OLS) or with standard Fixed Effects, leading to very large, significant results.

However, with time, other scholars have proven that this method leads to the overestimation of trade effects and thus, have modernised the gravity equation. One of the most significant changes came from the study of Anderson and van Wincoop (2003), and later Baier and Berg- strand (2007), who revolutionised the gravity model by introducing fixed effects in the form of multilateral resistance terms (MRT) (Hear and Mayer, 2014). These MRT are used when work- ing with multilateral trade flows data. What they do is explain the variation in trade flows between a specific country pair, while still accounting for trade interactions of these two countries with all other nations in the world. Thus, the effects are much smaller since there is a better grasp of the data. In the topic of sanctions, Kohl and Klein Reesink (2016) made use of this method and encountered much lower results than previously research. Their version of the gravity equation is as follows:

𝑙𝑛𝑇𝑅𝐴𝐷𝐸_𝑖𝑗𝑡 = 𝛽₀+ 𝛽₁𝐺𝐷𝑃_𝑖𝑡+ 𝛽₂𝐺𝐷𝑃_𝑗𝑡+ 𝛽₃𝐷𝐼𝑆𝑇𝐴𝑁𝐶𝐸_𝑖𝑗+ 𝛽₄𝑇𝐻𝑅𝐸𝐴𝑇_𝑖𝑗𝑡 (1) + 𝛽₅𝐼𝑀𝑃𝑂𝑆𝐼𝑇𝐼𝑂𝑁_𝑖𝑗𝑡+ 𝛽₆𝑧_𝑖𝑗𝑡+ 𝑒_𝑖𝑗𝑡

where TRADEijt shows the bilateral trade flows between countries i and j at time t, and it is regressed on their respective gross domestic products (GDPit and GDPjt) and the DISTANCE between both countries. THREATijt and IMPOSITIONijt are both dummy variables taking the value of one in the years t when a threat or sanction is active between countries i and j,. Next, zijt represents specific variables for each country pair on common language, border sharing, land locked, etc. Finally, eijt is the error term.

However, the dataset used in this paper is different, in that it uses product level U.S. imports coming from all other countries in the world. Since they are not bilateral flows, there is no need for MRT. Instead, I will make use of the following empirical specification and combine it with an alternative regression method which can account for high-dimensional fixed effects. Effec- tively, the fixed effects absorb the heterogeneity the control variables would be accounting for and thus, making the use of standard gravity model control variables unnecessary. The method was developed by Sergio Correia (2017) and is an extension to Stata under the name REGH- DFE. This software is able to run panel regressions while absorbing several types of fixed ef- fects at once, making some of the control variables obsolete. Thus, the specification changes to the following:

𝑙𝑛𝐼𝑀𝑃𝑂𝑅𝑇𝑆_𝑗𝑡𝑝= 𝛽₁𝑇𝐻𝑅𝐸𝐴𝑇_𝑗𝑡+ 𝛽₂𝐼𝑀𝑃𝑂𝑆𝐼𝑇𝐼𝑂𝑁_𝑗𝑡+ 𝛽₃𝑧_𝑡+ 𝐹_𝑗+ 𝐹_𝑡+ 𝐹_𝑝+ 𝑒_𝑗𝑡 (2) where Fj, Ft and Fp are fixed effects for target country, time and product, respectively. Their role will be to absorb the variations which the control variables are meant to account for. These three fixed effects, along with the combinations they make, were chosen due to the nature of the dataset used, which is described next. The result section will specify which type was used in each case and why. Under this estimation method, the constant is also dropped since we are performing a within transformation (Correia, 2017). For information on all the variables used, a thorough description of them and their source, see Appendix A.

(16)

- 16 -

3.2 Data

The data used is divided into two strands: (1) economic sanctions, threats and their characteristics and (2) general imports. For the first, sanction data has been sourced from the Threat and Imposition of Economic Sanctions (TIES) database v4.0, developed by MBK. This dataset is extremely precise in its sanction information including specific dates, types of sanctions, carrots¹⁴, institutions involved and others. For this research, the information extracted consists of dates surrounding the different events within one sanction; if there was a threat and/or imposition and in the event that there was, which type¹⁵. For the extension to the results, the resolu- tion of the sanction is also extracted. TIES covers sanctions starting from 1948 until 2005¹⁶. For the latter strand of data, general import values have been extracted from the United States International Trade Commission (USITC, n.d.) for all countries that trade with the U.S. The data consists of 238 nation states and regions, described in Appendix B. Making use of the following specification: general imports were extracted at their customs value on annual basis using 10-digit Harmonized Tariff Schedule (HTS)¹⁷ codes for all traded products. This sample of the USITC data ranges from 1989 until 2016 with a total of 6,855,454 observations¹⁸. Sum- mary statistics can be found under Appendix C. Due to the restrictions in the USITC database, we only consider sanctions which have ended between 1980 and 2012. Since the USTC data only runs from 1989, all sanctions that have ended between 1980 and 1988 are included to observe the lagged effects of sanctions ending. Thus, the data covers 238 countries from 1980 to 2016. This provides a good overview of the ex-post effects of sanctions over time.

In the extension to the core research section, I have also considered the case based on monthly data. For this, general imports were extracted from USITC at monthly values for two countries:

Brazil and Côte d’Ivoire. In this case, the data is at a 2-digit HTS product code (USITC, n.d.).

These two cases were picked since they are the latest in the TIES dataset. The case of Brazil is only a unilateral threat, while Côte d’Ivoire had a multilateral threat and an imposition.

In this research, the focus is on sanctions sent by the U.S. to any other country/countries, in the world. As was evident in Figure 1, the U.S. is by far the most avid sender of economic sanctions in history. Thus, in the time frame considered there are 407 economic sanction observations in which the U.S. takes part as the main sender or within the top 5. Out of these, there are 280

14 A sanction has carrots if the sender has offered some concession to the target once a threat has been issued.

15 There are 10 types identified and are described in Appendix A.

16 Although the start date of sanctions is up until 2005, the end dates go as far as 2012. Given that the data lasts until 2016, the latest sanctions will still allow 4 years of lagged effects.

17 HTS codes are used to categorise internationally traded products under the same classification scheme. For a good to be tradable it must be assignment an HTS code. 6-digit HTS are congruent worldwide, with the first two digits (01-99) being broad chapter classifications and reaching over 21 thousand 10-digit codes (Datamyne, 2017) In this research, about 6 thousand of the 10-digit codes are used.

18 The structuring of the dataset under one sole file was done using the Stata script written by Dr. Tristan Kohl. The first was to transform the output tables from USITC into a single column so that Stata can read iand interpret the data. The second script was used to create the variable SanctionLifted which was not part of the TIES dataset.

(17)

- 17 -

unique sanctions codes¹⁹. For the sake of comparison, in the same period of time, the U.S. had been a target only 50 times by 24 countries.

Before regressing the data some tests were performed, included in Appendix D. The data has shown to be heteroskedastic in nature. To counteract these effects I will make use of robust standard errors clustered by country in all regressions. This is done to control for between country heterogeneity while allowing within country variation to be correlated. Furthermore, the skewness and kurtosis of the data are not within the normal margins, so it stands to say that the data is not normally distributed. Finally, the coefficients do not show extremely high levels of correlation between each other. The independent variable General Imports does have outliers in the minimum and in the maximum levels. Since this research tests the effects of sanctions, which can be the reason for some of the minimum values in particular, these have not been removed from the dataset.

4. Results and Discussion

4.1 Annual Data Results

Before discussing the effect of lifting sanctions on trade levels, it must first be proven that the threatening and imposition of economic sanctions has a negative effect of trade. Appendix E shows the results for the specifications for hypotheses 1 to 3. The estimation in Table E.1 shows that threats lower imports to the U.S. from 2.8% to 11% (all highly significant). These results are robust along three types of fixed effects (FE/F): country-product-time (Fjp + Ft), country- time (Fj + Ft) and country (Fj). Each FE controls the heterogeneity between the groups, while allowing correlation inside of them. In this case, the strongest effect is when only con- trolling for country heterogeneity, showing that the more specific the FE used, the smaller the coefficients. Table E.2 shows the effects of impositions which vary from -3.4% to -10.3% trade.

However, when accounting for Fjp + Ft the results were not significant. Finally, Table E.3 con- siders the interaction between threats and impositions, where the results are ambiguous. Signif- icance is only present when looking solely at country fixed effects, where, strangely, the interaction of both characteristics yields a positive effect on imports of 8.8%. Moreover, the same pattern is evidenced in the other, non-significant, estimations.

These results lead me to fail to reject the preliminary hypotheses 1 and 2, proving that threats and impositions are indeed harmful to trade levels. However, these estimations are much lower than those of previous literature, which is most likely due to the use of disaggregated product level import flows. As mentioned before, there are many different types of sanctions, so it stands to infer that not all of products will be affected to the same extent at the same time. To see this effect more clearly, section 4.2 will observe two sanction cases using monthly observations.

Hypothesis 3, on the other hand, is rejected since the evidence shows that the interaction of

19 About half of the observed sanctions have more than one target and have been separated into individual sanctions. They retain the same code, but the target country differs. These codes are not generalizable and were created by the author.

(18)

- 18 -

threats and impositions yields positive results for import levels.

Now, moving on to see the effects of lifting an economic sanction. The results will be presented in different tables to reflect the hypotheses stated in the literature review. Table 2 shows the results of an imposition being lifted in the event of a previous threat compared to a lifted imposition that was not threatened. Hypothesis 4 claims that a sanction that was threatened before it was imposed would depress trade by a larger degree after the sanction has been lifted than a non-threatened sanction. From Table 2, columns (1) and (2), it becomes clear that a non-threatened sanction that has been lifted increases imports from 10.8% to 73%²⁰. On the other hand, a threatened sanction increases imports from 10.3% to 73.2%. All coefficients are significant, but too close together to be able to form a conclusion, therefore we reject hypothesis 4. What is interesting in these coefficients is that, once again, the role of FE is extremely important in the effect of a lifted sanction. The estimations using Fjp + Ft have significantly lower results that in the case of Fj + Ft. This shows that the product dimension of the dataset is detrimental to determine the impact of a sanction, implying that not all products react the same to an imposition and therefore also to its lifting. Evidence of this effect will be presented in section 4.2.

Additional testing was done to see the effects based on the type of sanction threatened on the years following the lifting of a threat (see Appendix F). Though the impacts are volatile, there is clear depressing effect on most types of sanctions threatened one year after the threat was made, all being highly significant. After the first year, the effects and their significance vary based on the type of sanction threatened. The results indicate that some types of threats, like a travel ban, have a considerable effect on the economy which lasts for more than one year,

20 To calculate the percentage change on the dependent variable, ln(GeneralImports), caused by any of the independent variables, the following formula was used: Percentage change = 100*(ecoefficient - 1).

Table 2. Effects of Sanctions being Threatened before Imposition (Hypothesis 4)

Variables (1)

ln(Imports)

(2) ln(Imports)

(3) ln(Imports)

(4) ln(Imports)

Threat -0.0255

(0.0359)

0.0052 (0.0395) Imposition 0.0123

(0.0274)

0.0631***

(0.0232)

0.0192 (0.0238)

0.0614**

(0.0239) ImpositionLifted 0.1023**

(0.0502)

0.5486***

(0.0511)

0.0980*

(0.0523)

0.5490***

(0.0495) Fixed Effects -Country Product

-Time (Fjp + Ft)

-Country -Time (Fj + Ft)

-Country Product -Time

(Fjp + Ft)

N 6,593,430 6,855,454 6,593,430 6,855,454

Adj. R² 0.6793 0.1999 0.6793 0.1999

Robust standard errors are in parentheses clustered by country.

* p<0.10, ** p<0.05, *** p<0.01.

(19)

- 19 -

while others, like an import restriction threat, have short term effects which later disappear.

Hypotheses 5 and 6 are both related to the number of senders of the sanction. The former states that unilateral sanctions will depress import flows more than multilateral sanctions while they are active. The expectation comes from the fact that the unilateral sender is the U.S. which is a key trading partner to many countries around the world. The sixth hypothesis states “the recov- ery of trade will be higher in a case involving a multilateral sanction than in the case of a unilateral sanction, once the sanction has been lifted”. Columns (1) and (2) of Table 3 corre- spond to hypothesis 5, with the results showing that unilateral sanctions depress import flows an additional 4.7% to 7.3%. Multilateral sanctions, on the other hand, have an additional effect ranging from 9.6% to 13.1%, all being significant. Thus, the results show that hypothesis 5 is also rejected since it is not in line with the expectations previously set. Multilateral sanctions seem to have double the additional effect on trade than unilateral sanctions. Evidently, there is power in numbers.

These regressions were tested with both country fixed effects (Fj) and country-time fixed effects (Fj + Ft). The latter were used to determine if the combination of country and year could hold explanatory power. Examples are cases where a particular regime was influential in the political/economic status of the country, like the years following the Cold War in Eastern Europe.

For hypothesis 6, we test the recovery of trade flows comparing unilateral to multilateral sanctions. To observe the consequences for import flows at a more disaggregated level, I will make use of lag variables. These variables will show how import levels adjust to the lifting of the sanction in the years thereafter while discerning on whether there was one sender or more. In Table 3, columns (3) and (4) show three consecutive lags from the ending of the sanctions, while column (5) includes also lags for 5 and 7 years after the end of it. The coefficients of Imposition Lifted (IL) are positive, but only significant in the second specification, with an effect of 15.5% recovery of import flows after the sanction has been removed. As for the lags, one year after the sanction has been lifted shows that unilateral sanctions keep depressing trade in all three columns ranging from 62.5% to 77% less trade. On other hand, multilateral sanctions have a larger range of negative effects from 9.41% to 69.9%. Two years after the lifting shows a recovery on all specifications for both unilateral and multilateral sanctions, though the recovery for multilateral is considerably larger. This could be the “wait and see” effect discussed before, where some exporters prefer to hold off on investing in import flow restoration until the market has stabilised. The third year after the sanction being lifted the effects for unilateral sanctions are negative but not significant, while multilateral sanction have a negative effect on trade on both significant specification of around 34.5% to 63.4%.

Now looking only at column (5) and moving to 5 years post lifting, the results are not significant for both unilateral and multilateral sanctions. Finally, the 7 year mark shows a weak recovery in both cases, with a higher percentage in unilateral sanctions. Despite the variations in the coefficients, the overall trend shows that multilateral sanction have indeed recovered to a

(20)

- 20 -

Table 3. Effects of Unilateral and Multilateral Sanctions on Import Levels (Hyp. 5 & 6)

Variables (1)

ln(Imports)

(2) ln(Imports)

(3) ln(Imports)

(4) ln(Imports)

(5) Ln(Imports)

Threat -0.0228

(0.2960)

0.0155 (0.0264) Imposition -0.0418*

(0.0220)

-0.0150 (0.0159) Unilateral -0.0758***

(0.0268)

-0.0486*

(0.0276) Multilateral -0.1408***

(0.0500)

-0.1014**

(0.0423) ImpositionLifted

(IL)t

0.0549 (0.0635)

0.1442***

(0.0532)

0.0518 (0.0704)

(IL)(Unilateral)t-1 -0.2608***

(0.0990)

-0.4700***

(0.0894)

-0.3587***

(0.1337)

(IL)(Unilateral)t-2 0.2832**

(0.1265)

0.4049***

(0.1019)

0.3604**

(0.1707)

(IL)(Unilateral t-3 -0.0147

(0.0685)

-0.0792 (0.0913)

-0.0783 (0.1229)

(IL)(Unilateral)t-5 -0.0758

(0.1229)

(IL)(Unilateral)t-7 0.1828**

(0.0722)

(IL)(Multilateral)t-1 -0.3579**

(0.1688)

-2.8432***

(0.1200)

-0.6918***

(0.1856)

(IL)(Multilateral)t-2 0.6227***

(0.0346)

3.2640***

(0.0404)

1.9736***

(0.0688)

(IL)(Multilateral)t-3 0.0266

(0.0622)

-0.4242***

(0.0638)

-1.0059**

(0.0702)

(IL)(Multilateral)t-5 0.0737

(0.0620)

(IL)(Multilateral)t-7 0.0012**

(0.0741)

Multiple 0.0475*

(0.0267)

0.0673**

(0.0295)

0.0300 (0.0319)

0.0674**

(0.0329) Fixed Effects ^-Country

(Fi)

(Fjp + Ft)

N 6,855,454 6,855,454 3,726,781 3,825,527 2,986,098

Adj. R² 0.1929 0.1963 0.7058 0.1905 0.7148

Robust standard errors are in parentheses clustered by country.

* p<0.10, ** p<0.05, *** p<0.01.

(21)

- 21 -

greater extent than unilateral sanctions after the lifting and therefore I fail to reject the sixth hypothesis. The U.S. appears to be able to continue to exert power even after the sanction has been lifted, by a greater extent than a multilateral sanction.

Finally, all regressions included the binary variable Multiple, which took the value of one in the years where more than one sanction was actively imposed in the country. The results are quite contradictory. In all instances, having more than one sanction imposed has a positive effect on import levels with an average effect of 5.8% more trade with the U.S.

The adjusted values of the coefficient of determinations are quite low in the specifications of Table 2 and Table 3 showing that the variation in the data is not well explained by the model.

The exceptions are when thorough fixed effects are used (Fjp + Ft), but then the coefficients are also lower. Given the size of the dataset and the fact that is panel data, the R² cannot be expected to be very high without using FE. The highest value is observed in column (5) of Table 3, which takes country, time and product fixed effects and shows an adjusted R² of 71.48%, which is the percentage variation in the data explained by the model.

The last hypothesis is shown in Table 4 and states that “after the lifting of an economic sanction, the target country will experience a lag effect until trade levels start to increase again”. Two types of regressions were run for this hypothesis: the first only includes lags of the variable ImpositionLifted for up to 7 years, while the second makes the lags interact with the type of sanction previously imposed. According to columns (1) and (2), the first, third and fifth year after the end of an economic sanction have no effect on import levels, with all coefficients being insignificant. Nonetheless, column (2) provides evidence that during the second year trade continues to be depressed by 6.7%, where country and time fixed effects were used. In the 7^th year post-lifting trade continues to decrease with 7.6%. Thus, the lack of significance proved to be insufficient to corroborate the hypothesis.

A second specification is included which is shown in columns (3) and (4), and is using different fixed effects of country-time-product and country-product, respectively. In this case, each lag variable of ImpositionLifted interacts with the variable SanctionType which is a categorical var- iable explaining 10 types of economic sanctions²¹. The first year post-lifting is highly significant when taking country-product fixed effects showing that there is heterogeneity in the effects of the sanction per country per product that needs to be taken into account. The largest increase in import levels comes from the lifting of a Partial Economic Embargo skyrocketing levels of 4,336% more trade, while the lowest in the first year comes from Import Restrictions which continues to decrease trade by 41.3%. It is also relevant to note, that after the ending of a Sus- pension of an Economic Agreement or Protocol, trade recovers after the first year by a very substantial margin. The second year after a lifting is highly significant in both specifications

21 According to the TIES database, these are: (1) Total Economic Embargo, (2) Partial Economic Embargo, (3) Import Re- striction, (4) Export Restriction, (5) Blockade, (6) Asset Freeze, (7) Termination of Foreign Aid, (8) Travel Ban, (9) Suspension of Economic Agreement, and (10) Unspecific (Morgan, Bapat and Kobayashi, 2014).

Table of Contents

Abstract