The effect of economic sanctions on
remittances in developing countries
Contents
1 Introduction 1 2 Literature Review 3 2.1 Sanctions . . . 3 2.1.1 Sanctions as a shock . . . 3 2.1.2 Types of sanctions . . . 3 2.1.3 Mitigating sanctions . . . 4 2.2 Remittances . . . 6 2.2.1 Motives . . . 62.2.2 Cyclicality and shocks . . . 6
2.2.3 Remittances and other capital flows . . . 7
2.3 Sanctions and Remittances . . . 8
3 Methodology 9 3.1 Ordinary Least Squares (OLS) . . . 9
3.2 Synthetic Control Method (SCM) . . . 10
3.2.1 Formalisation . . . 10 3.2.2 Significance Testing . . . 11 4 Data 13 4.1 Variables . . . 13 4.2 OLS sample . . . 14 4.3 SCM sample . . . 15 5 Results 17 5.1 OLS Estimation . . . 17 5.2 SCM Estimation . . . 18 5.2.1 Latin America . . . 19 5.2.2 Sub-Saharan Africa . . . 20 5.3 SCM robustness tests . . . 22
5.3.1 Individual OLS Estimation . . . 22
5.3.2 Limiting the donor pool . . . 23
5.3.3 Placebo tests . . . 23
6 Conclusion 25 6.1 Discussion . . . 25
6.2 Limitations . . . 25
6.3 Suggestions for further research . . . 26
7 Appendix 32 7.1 Other descriptive statistics . . . 32
7.2 Case selection . . . 33
7.3 Robustness checks . . . 36
1
Introduction
Economic sanctions; a foreign policy tool that is stronger than diplomacy, yet weaker than military intervention. They are meant to coerce the target governments’ policy through inflicting economic damage (Hufbauer et al., 2007). Even though the debate whether these sanctions are actually effective in reaching their desired objectives is still inconclusive, sanctions remain a popular tool. For instance, this year we have seen a series of sanctions being imposed on Iran by the United States (US) and several other countries. In the name of stopping Iran from their alleged nuclear proliferation the US tries to inflict economic damage by cutting them off from the international financial system, among other sanctions (Wong, 2019). The critics of such statecraft emphasise that more often than not the ones hurt by sanctions are the civilians rather than the government; to illustrate, detrimental effects have been found on the level of poverty (Neuenkirch and Neumeier, 2016) and income equality (Afesorgbor and Mahadevan, 2016).
Even though much literature is devoted to confirming the impaired internal circumstances or researching whether the objectives of the sanction have been accomplished, less attention has been paid to the effect the economic sanction imposed by the sender state (i.e. the state imposing the sanction) has on foreign capital flows to the target state (i.e. the state receiving the sanction). As foreign capital flows are a very important source of income for developing countries the effect sanctions have on their volume and the possible role they can play in mitigating the negative effects is one that must not be overlooked. This relationship is not heavily researched in the empirical literature, however some studies have look at FDI (Mirkina, 2018; Lektzian and Biglaiser, 2014, 2013), ODA (Early and Jadoon, 2016; Early, 2015), and FPI flows (Shin et al., 2016). These studies find mixed results for FDI, an increase in ODA, and no significance for FPI flows following a sanction episode.
What is striking is that apart from one specific case study by Spadoni (2010) the effect on remittances has been overlooked up to date. Remittances are especially relevant nowadays considering that The World Bank (2019) forecast expects annual remittance flows to developing countries to be larger in 2019 than FDI and ODA flows. Apart from their volume, remittances are seen as an important capital flow to help smooth consumption for households and dampen the effects of economic shocks. The money if often invested in human capital and spent on productive investment, which spurs economic growth and alleviates poverty (Chami et al., 2008). Remittances pave the way for such objectives, because they are person-to-person transfers and are thus tied strongly to family bonds. Remittances receiving countries also struggle with a few several drawbacks of this capital flow too These are, among others, deterioration of (governmental) institutions (Abdih et al., 2012), income substitutes, moral hazard problems (Chami et al., 2005), the Dutch disease effect (Chami et al., 2008) and that they are only being used for consumption rather than investment (Barajas et al., 2009).
This research sets out to fill the gap in the empirical literature by answering the research question ’Do remittances mitigate the effect of sanctions?’. This research is founded in the belief that remittances are able to do so, especially in the context of developing countries, as they are expected to increase following an economic sanction due to the negative effects the sanctions have on the target country. An increase in remittances is often associated with economic growth (Chami et al., 2008) and it is through this channel that remittances are expected to be able to mitigate the effect of sanctions; this part of the relationship is not researched empirically in this paper, due to the confines of this research, however might be interesting to look at for further research.
Squares (OLS) estimation. Especially interesting is the second research method, which employs the synthetic control method (SCM) for a sample of 9 coercion episodes in developing countries from different regions. This method constructs data-driven counterfactuals, which pose close resemblance to the sanctioned country (Abadie et al., 2010). In this way a realistic idea can be attained of what the target country’s remittance to GDP ratio would have been when a sanction had not been imposed. Also, this method allows us to look at each sanction case individually and over the whole sanction period, thus accounting for differences across country and over time. Especially the long-run effects are generally ignored by only doing an aggregate empirical analysis since they only look at the year of imposition and do not consider lagged or early effects. In doing this, this study takes the dynamic nature of remittances in consideration. On the one hand, the OLS estimation finds a negligible negative effect, therefore no effect can be concluded from this method. On the other hand, the SCM clearly shows that the remittances to GDP increases following a sanction when compared to its synthetically constructed counterfactual and therefore highlight the mitigating possibilities this capital flow can have.
All in all, this study contributes to the body of literature regarding sanction mitigation through capital flows (for example, Early (2015)), by looking at a capital flow not before researched; namely remittances. Bear in mind that the novelty of this research lies in that it is an exploratative research, as it is the first time that the effect of sanctions on remittances is tested empirically and thus no theoretical base has been established so far. Remittances are an especially relevant capital flow to be researched in this context as they are considered to be more stable than other foreign capital flows are almost the largest capital flow nowadays. In addition, a lot of research has been done analysing the cyclical nature of remittances by looking at what happens in the event of several adverse economic shocks, which is especially important for developing countries as they are very sensitive for to such shocks. Currently, none of these studies have looked at sanctions as a possible shock in their analysis. Thus, this research can contribute to the on-going debate on the cyclicality of remittances in developing countries by looking at the question from a new perspective; namely a sanctions shock.
2
Literature Review
The first section will start by providing relevant information on sanctions. First, sanctions are conceptualised as exogenous shocks, then the different types of economic sanctions are briefly explained. Finally, ways in which sanctions can possibly be mitigated are explored. The second section is about remittances and contains essential background information for understanding the way remittances operate. First, the motives to remit are touched upon as they are vital in explaining the arguments regarding the cyclicality of remittances. The cyclicality of remittances, in turn, explains their reaction to shocks, which provides the foundation for the hypothesis formulation. Lastly, insights in how remittances compare to other capital flows in their role as stabiliser as compared. The third and final section shows builds upon the preceding sections to formulate the theorised relationship between sanctions and remittances.
2.1
Sanctions
2.1.1 Sanctions as a shock
In the empirical literature, economic sanctions are increasingly recognised as exogenous shocks (Bali, 2018; Durmaz and Akku¸s, 2019; Dizaji and Van Bergeijk, 2013; Eyler, 2007; Hatipoglu and Peksen, 2018; Kholodilin and Netsunajev, 2016; Pestova and Mamonov, 2019). Generally, when categorising shocks, a distinction is made between endogenous and exogenous shocks. An endogenous shock comes from within the country and an exogenous shock is a ”sudden event beyond the control of the authorities that has a significant negative impact on the economy” (IMF, 2003). A sanction can be categorised as an exogenous shock when referring to the aforementioned definition since it is a sudden policy implemented by a state or organisation on a target state, beyond the control of the latter. Also, economic sanctions are treated as shocks because of the possible detrimental effects on the economy of the target country. For example, Neuenkirch and Neumeier (2015) established a link between sanctions and a negative GDP growth, Hufbauer et al. (2007) find that sanctions lead to increased unemployment and inflation, and Hatipoglu and Peksen (2018) find that sanctions can lead to banking crises. Thus, sanctions can be classified as an exogenous shock; this principle will serve as a basis for formulating the hypothesis.
2.1.2 Types of sanctions
Sanctions can be imposed on different dimensions of the target state, for instance, they could be related to military, diplomatic, sport, environmental, or economic facets of the target country. However, only the last type of sanction is relevant for this research and is therefore discussed in this section. Following the broad definition of economic sanctions, identified by Hufbauer et al. (2007), these can be trade, financial, and asset sanctions. Trade sanctions are those limiting imports and exports, financial those ”delaying or denying credit or grants” (Hufbauer et al., 2007, p. 94), and sanctions targeted on assets are mostly asset freezes, which occur when the target country’s assets held in the sender country are frozen as a result of the sanction.
and sset sanctions are known to have relatively high sender costs accompanying them (Ibid.).
They also have different economic impacts; trade, financial, and asset sanctions have a deteriorating effect on GNP, respectively, 0.7%, 1.7%, and 4.3% (Hufbauer et al., 2007). These types of sanctions are not mutually exclusive, as often a mixed strategy is deployed. For instance, often a combination of trade and financial sanctions are imposed, with a total negative effect of 2.9% on the GNP of the target country and a political success rate of 34% (Ibid.).
2.1.3 Mitigating sanctions
The successes of sanctions are highly susceptible to the extent to which the detrimental effects of sanctions can be mitigated. The importance of mitigation can especially be seen in the first stages of the sanction, seeing as anticipation plays an important role (Van Bergeijk, 2014). Dizaji and Van Bergeijk (2013) show that mitigation mainly takes place in the first phase since sanctions are only effective at that time, as the economic structures in a country adjust to the situation in the long-run. Nonetheless, sanction mitigation can take place through factors relating to the sending country, the target government, or third-party engagement. Each of these possibilities will be discussed hereafter.
Counter-intuitively, the sending country itself can contribute to the mitigation of negative effects of the sanction, through either their firms or government. First, the firms of the sending country themselves can undermine the effect of the sanction by continuing trade relations. As Afesorgbor (2019) shows, when merely a threat of a sanction is in place (so not yet imposed) this leads to an increase in trade between the sending and target country. Stockpiling is given as main explanation to this preliminary increase, however, Afesorgbor (2019) finds that after the imposition of the sanction trade decreases. Nevertheless this might just be pretence, as Barry and Kleinberg (2015) show in their US case. They show that after the sanction is imposed on a trading partner country of the sender, firms in the sender country shift their investments to states that can provide indirect access to the sanctioned country. Even though this form of mitigation is created by the sender country’s domestic firms, the government themselves are also take part in the blame, as they often do not sufficiently enforce or monitor whether the sanction is adhered to nationally following the imposition (Hufbauer et al., 2007).
As for mitigation by the target country, governments can impose domestic regulations to circumvent the detrimental effects of sanctions. Many of these policies are usually for their own and the elites benefit and will not gain a lot of support from the rest of the country. For example, by implementing predatory policies, such as the redistribution of capital to the elite or regime supporters, undermining property rights, nationalising industries, or transferring resources (Peksen, 2016). Governments can also generate lost revenue by, for example, raising taxes (Escrib`a-Folch and Wright, 2010). Protests can upsurge as a consequence of the preferential treatment given to elite and the sanction and its effects on the domestic population. However, sanctions creates optimal circumstances to oppress their population through limiting political and human rights (Jeong, 2019), thereby suppressing resistance and unrest.
known as ’sanction busting’ and can be reached through several forms of economic engagement with the target state. Especially in the case of economic sanctions, which are trade and capital flow driven, the intervention of third-parties has the largest detrimental effect on the impact of sanctions; and vice versa when the sanction is supported (Early, 2015) .
On the one hand, McLean and Whang (2010) showed that sanctions can have a greater chance of success with third-party intervention when the main trading partner of the target state supports the sanction of the sender state. Also, when other states enforce the sanction too, such multilateral sanctions often achieve their intended goals sooner (Bapat and Morgan, 2009). Spadoni (2010) offers an explanation for this in that in a highly integrated and liberalised world it should be easier to mitigate unilateral sanctions, but this might pose a difficulty with multilateral sanctions as the options to mitigate through alternative cooperation become limited.
On the other hand, Early (2015) provides two ways in which third-parties could possibly negatively influence the outcome of an economic sanction: trade-based or aid-based. The former is motivated by profit-seeking behaviour, done so by exploiting the consequential economic opportunities that arise following a sanction imposition. A well known example is Dubai (U.A.E), who, following the US sanctions imposed on Iran, acted as a middleman for firms who engaged in trade activities with Iran (Early, 2015).
Forms of sanction busts that are aid-based, are driven by politics (Early, 2015). This is primarily done by states that, for example, have the same ideologies as the target state and want to counteract the sanctioning state. A prominent example is when the US imposed sanctions against Cuba during the Cold War, after which the Soviet Union increased their aid to offset the negative effects caused by the sanctions (Early, 2015). Early (2015) finds that sudden sharp increases in foreign aid flows can have a significant negative effect on the succession of sanctions against target states and vice versa. Later research by Early and Jadoon (2016) empirically researched the effect of sanctions on development aid. Looking at sanctions imposed by the US and the consequential ODA flows to 133 recipient countries from 1960 to 2000 they find that when the US is the primary imposer this leads to an increase in ODA. They then research the motive of this increase in ODA (altruistic or self-interest) for a limited amount of case studies and find convincing evidence for the self-interest motive.
negatively in the short-run for sanction episodes taking place in the 1990s, however none of the other decades or long-run effects yielded significant results.
The only research into remittances has been a case study by Spadoni (2010), who explains that following the US sanctions imposed on Cuba, remittances inflows have significantly decreased the adverse economic effects. The main reason given for this were family ties to mitigate negative economic effects. Also, albeit to a lesser extent, to increase physical investments in Cuba. Even, after the US imposed sanctions directly targeting remittances in 2004, the remittances flows to Cuba even increased.
2.2
Remittances
2.2.1 Motives
The literature generally distinguishes between two main approaches that can be used to classify the reasons migrants send remittances: the endogenous approach and the portfolio approach. Whilst the former is an act of altruism the latter is an act of self-interest. However, be aware that a combination of both motives is a possibility too. This section will briefly discuss the theoretical foundations for both approaches and their accompanying motives as given in the renowned paper by Rapoport and Docquier (2006).
On the one hand, the endogenous approach is based on altruistic intentions, implying that remittances are not sent for the migrant’s personal gain. This means that in times of hardship in the migrant’s country of origin they increase their flow of remittances, in order to help their relatives or friends. On the other hand, the portfolio approach is based on the migrant’s self-interest. This could be achieved by benefiting from the macroeconomic conditions of the receiving country, for example, through exchange rate differentials, interest rate differentials, or inflation rates. The remitter can exploit this market inefficiency to his or her benefit and realise profits from it. Second, the exchange motive is an informal agreement to send remittances in exchange for taking care of relatives or property back home. Inheritance is another self-interest motive to be considered; the migrant sends money back home to secure their share of the bequest. In the case of inheritance and exchange improved economic circumstances lead to higher values of goods and services, which in turn induces the migrants to increase their share of remittances. Investment is also motive of self-interest that increases with better economic prospects; here the remitter lets the family members invest in (profitable) assets in their name. The last motive to remit insurance, meaning that the remitter wants to diversify their risk related to the macroeconomic situation in their country of origin by migrating. In times of hardship in their home country the relatives left behind expect an increase remittances and vice versa. 2.2.2 Cyclicality and shocks
The empirical literature has mostly pointed to counter-cyclicality in the event of shocks. Bettin et al. (2015) find evidence that remittances increase in the event of adverse economic shocks, thus in favour of counter-cyclicality. More spefically, Bettin and Zazzaro (2018) find evidence for the insurance motive and the counter-cyclical nature of remittance flows for a broad scheme of natural disasters. They also find that remittances become more important in this context in countries with a less developed financial system. Other, more case specific, empirical studies regarding natural disasters support this result (Yang and Choi, 2007; Attzs, 2008). Also remittances are seen as a hedge against food price shocks (Combes et al., 2014) or oil price shocks (Lueth and Ruiz-Arranz, 2007).
Also, a recent paper by De et al. (2019) showed convincing evidence for the a-cyclicality of remittance flows. The study analysed how remittances behaved during large macroeconomic shocks, such as sudden stops and financial crises, and found that actually remittances are a-cyclical in 80% of the countries. It is important to note that a-cyclicality does not mean nothing happens in the event of a shock; it means that they have the potential to provide stability, especially seeing as other flows are more volatile (De et al., 2019).
Significantly less evidence has been found for advocates of the pro-cyclicality of remittance flows following shocks, for example, Giuliano and Ruiz-Arranz (2009) find evidence of the investment motive. Khodeir (2015) finds that remittances do not act as a hedge against macroeconomic output shocks in Egypt, and are in fact more pro-cyclical. Lueth and Ruiz-Arranz (2007) find a similar result in Sri-Lanka for currency shocks.
There are also some studies that point out that the cyclicality of remittances cannot be generalised since they are dependent on various country characteristics, for example Sayan (2006) find that remittances act different in each country in the event of GDP decreases. Further, Ruiz and Vargas-Silva (2014) argue that the dynamic nature of remittances should not be overlooked and that the cyclicality can change over time and are country dependent as well.
2.2.3 Remittances and other capital flows
The capital flows considered in this section are FDI, ODA, and FPI. First, the stabilisation effects of the flows in question are examined. Most empirical research finds that remittances are more stable compared to other flows, where remittances, ODA, FDI, and FPI are respectively most stable (Sirkeci et al., 2012; Bouoiyour et al., 2016; Gammeltoft, 2002). Constantinescu and Schiff (2014) dispute this result and find that ODA is more stable than remittances, nonetheless remittances are more stable than FDI. FPI is regarded as the most volatile flow due to its highly reversible nature. This explains their cyclicality, FDI and FPI have been found to be pro-cyclical in developing countries (De et al., 2019; Singer, 2008; Kaminsky et al., 2005), in contrast to remittances which research proves mostly to be counter-cyclical or a-cyclical, as seen in section 2.2.2.
2.3
Sanctions and Remittances
The channel through which the negative effects of sanctions can be mitigated by remittances is hardly addressed in the current empirical literature, despite remittances currently being one of the largest foreign capital flows. Figure 1 hypothesises what this relationship could be and whether it is positive or negative. As can be seen in the figure this relationship can either be direct or indirect, through income.
Figure 1: Hypothesised relationship between sanctions and remittances
Sanctions Remittances Income
−
− +
Increases in foreign capital flows are driven by underlying motives; be it the altruism or profit-seeking motive. This indicates that capital flows through an indirect channel. It is therefore that the only possible hypothesised direct relationship between sanctions and remittances would be that the sanction directly hinders sending remittances; so a negative effect is anticipated. As Spadoni (2010) showed in Cuba, restrictions following sanctions could be easily surpassed by transferring remittances through informal mechanisms, such as mules, or by delivering them in person. As these are unregistered forms of transferring remittances and these will not show up on the balance sheets of the country we expect the direct relationship between sanctions and remittances to be negative. Bear in mind that even though official records of remittances decrease this does not necessarily mean that the absolute amount of remittances actually decrease.
As for the possible indirect channel we first turn to the convincing evidence in section 2.2.2 that remittances are a counter-cyclical flow. Also, as sanctions are an exogenous shock, this means that they have a negative effect on the economy, and compared to other exogenous shocks, such as natural disasters and oil price shocks, we see that remittances tend to react counter-cyclically in such events. Therefore, remittances have the potential to help mitigate adverse effects or act as a stabiliser and therefore increase following a sanction.
The specific income channel presented in Figure 1 is explained as following: as GDP decreases after sanction imposition (Neuenkirch and Neumeier, 2015; Hufbauer et al., 2007) families are more prone to sending money back home to help compensate for losses in prosperity. This brings the following hypothesis:
Hypothesis: Sanctions lead to increased remittances
3
Methodology
3.1
Ordinary Least Squares (OLS)
In order to test the research question whether remittances are able to mitigate the externalities of sanctions we start with a pooled OLS estimation over all observations given,
RMit= α0+ β0Sit+ β1Xit+ uit (1)
Where RMitis the dependent variable that measures the remittances to GDP ratio for country i at year t,
α0 is the constant, β0 is the independent coefficient that measures the effect that economic sanctions have
on remittances, Sit is the dummy variable where 1 implies country i is subject to a sanction at year t and 0
otherwise, Xit is the vector of control variables, and uit is the error term.
The control variables are important determinants of the inflow of remittances and are chosen by drawing upon existing empirical research about the (uncontested) determinants of remittances and based on data availability. The control variables included in the regression are dependent on whether the direct or indirect relationship is being examined (see Figure 1). For the direct relationship the selected control variables are economic development, net migration, trade openness, and capital account openness. For the indirect relationship where income is at play, economic development is excluded as a control variable.
First, economic development is included since most researchers find that the level of income in the home country is related to the volume of remittances sent. Most studies find that this relation is negative (El-Sakka and McNabb, 1999; Aydas et al., 2005; Chami et al., 2005), meaning that the amount of remittances increases when the GDP decreases. Another obvious determinant of remittances is the level of migration (Gupta, 2005; Fajnzylber and Lopez, 2008); when a higher share of the population is migrated there is a higher possibility of receiving remittances. Following the excepted hypothesis and the empirical literature the coefficients for economic development and net migration are expected to be negative, implying that an increase in the relative amount of remittances is associated with a decrease in economic development and net migration. Furthermore, Cooray and Mallick (2013) find that capital account openness and trade openness of the host country are the most important factors for remittances inflows, where more openness leads to more remittances, because it makes transfers easier and cheaper. The coefficients for trade openness and capital account openness are expected to be positive, meaning that a higher degree of openness is associated with a higher amount of remittances.
However pooled OLS assumes that there is no heterogeneity, and considering the panel data set which is used over time and different countries, there must be heterogeneity. Therefore we should consider an alternative that accounts for individual heterogeneity; this can be either the fixed effects or random effects model. After running the Hausman test (unreported), it is concluded that fixed effects should be used rather than random effects.
Fixed effects can be country effects, time effects, or both. Country fixed effects are variables that affect RMitcross-sectionally but do not vary over time and are included to capture unique country characteristics.
Time fixed effects are variables that affect RMit that vary over time but are constant cross-sectionally and
RMit= α0+ β0Sit+ β1Xit+ µt+ λt+ vit (2)
This extension of Equation 1 decomposes its error term (uit) into a country fixed effect (µt), a time fixed
effect (λt), and an error term (vit).
In both regressions represented by equation 1 and 2 the sign of the main independent variable Sit is
expected to be significantly positive. This is anticipated following the hypothesis that sanctions lead to an increase in the remittance to GDP ratio.
3.2
Synthetic Control Method (SCM)
Even though the greatest complication of the pooled OLS, which is heterogeneity, is resolved using fixed effects instead, there are still some inevitable shortcomings when estimating the research question via OLS. First, most studies researching the effect of sanctions on capital flows are aggregate studies (Lektzian and Biglaiser, 2014, 2013; Early and Jadoon, 2016; Shin et al., 2016). Even though these studies are relevant in the sense that they provide aggregate evidence, their analysis raises some questions of reliability. In aggregating the results important nuances for time or between countries or even regions for that matter are easily overlooked. As explained in section 2.2.2 the way remittances behave in the event of a shock can be very country or time dependent, and aggregate estimations therefore do not capture all the individual variation. For instance, mitigating sanctions can take place soon after the sanction is imposed or even in anticipation of the sanction. It is therefore important to look at the impact of a sanction for individual cases. There are also drawbacks related to endogeneity issues when estimating via the OLS estimation method; first not all heterogeneity issues might be resolved and there still might be unmeasured attributes that are not controlled for. In many remittance research an Instrumental Variable (IV) approach has been adopted to circumvent these measurement issues. However, it still seems very challenging to find an appropriate instrument for this method and endogeneity is still a considerable issue.
An alternative interesting way to estimate the individual impact of a macro-policy intervention is to use a counterfactual impact evaluation method, this method implies you have a treatment group (the one that receives the intervention) and a control group (the unit(s) you are going compare the treatment to). This is done by constructing a so called ’counterfactual’, i.e. a unit that replicates the treated case as though it was not subject to the intervention. There are several counterfactual impact evaluation methods, such as the difference-in-difference (DID) approach, propensity score matching (PSM), or the synthetic control method (SCM). The last approach, by Abadie and Gardeazabal (2003) and extended by Abadie et al. (2010), will be adopted in this research. This method is different from other counterfactual impact evaluation methods in that it uses a data-driven method to construct the counterfactual; i.e. the treatment unit does not have a single control unit that comes closest to the treated unit but rather uses a composition of weighted averages of units (from the ’donor pool’) as the control unit. This synthetically constructed control unit exhibits similar dynamics in the covariates as the actual target before the sanction is imposed.
3.2.1 Formalisation
The potential outcomes are YitN and YitI. YitN is the outcome in the case of no sanction for countries, this is the counterfactual of interest. YI
it is the outcome of country (i = 1) exposed to the sanction at year
t = T0+ 1, ...T . T0 is the pre-treatment period, T0+ 1 is the year the treatment starts, and T is the last
year of the treatment period.
The outcome of interest for this research is the causal effect of the sanction on remittances for the treated unit in the treatment period shown by (α1,T 0+1, ..., α1,T), for t > T0.
α1t = YitI − Y N
it = Y1t− YitN (3)
Where Y1t is the observed outcome for unit 1 at year t.
The counterfactual of interest is constructed by the weighted averages of possible controls, shown by the vector W = (w2, ..., wJ +1)0, with wj ≥ 0 and all the weights add up to 1. Here j are the selected countries
from the donor pool.
X1is a (k x 1) vector of pre-intervention characteristics for the treated unit, and X0 is a (k x J) matrix
with the same variables for the units in the donor pool.
In order to reproduce the synthetic control unit we find the combination of untreated units that best resembles the treated unit before the intervention in terms of the values of k relevant covariates. The covariates used in the analysis will be the same as the control variables for the general OLS estimation, as well as the lagged value of the dependent variable. The lagged value is included an extra control variable to minimise kX1− X0W k and to account for unobserved heterogeneity, following Mitrut (2014). Then the
vector W∗= (w∗
2, ..., w∗J +1)0 is chosen to minimize kX1− X0W k.
This gives synthetic control estimator for the treatment period t
ˆ α1t= Y1t− J +1 X j=2 w∗jYjt (4)
So Y1t is the outcome variable of the treated country and Yjt is the value of the outcome of unaffected
countries for units j = 2, ..., J + 1 at year t.
After this the root mean squared prediction error (RMSPE) can be constructed which gives the goodness of fit for synthetic controls to the actual targets.
RMSPE = 1 T − T0 T X t=T0+1 Y1t− J +1 X j=2 w∗jYjt 2 1/2 (5)
Please see the paper by Abadie et al. (2010) for more technical details. 3.2.2 Significance Testing
In order to test significance placebo experiments are used where the sanction is reassigned to the states initially in the donor pool. In order to do this the SCM is applied iteratively to each country in the donor pool, consequently the a distribution of placebo effects is obtained. Next you have to compare the gap of the initial treatment state ( ˆα1t) to the distribution of all the placebo gaps ( ˆαP L1t ). Subsequently the p-value
p-value = PJ +1 j=2I( ˆα P L(j) 1t < ˆα1t) J (6)
In doing these placebo estimations the rank of ( ˆα1t) can be found in the total list of alle experiments
ˆ αP L
1t . A high rank indicates an unusual large positive or negative gap and confirms the hypothesis at stake.
4
Data
4.1
Variables
In order to test the relationship between sanctions and remittances an unbalanced panel data set was created for the period 1980-2012. This lower-bound for the period is chosen since the remittances data, of which the source is World Bank, is only available from 1980 onward. The calculation of remittances was ”based on data from IMF Balance of Payments Statistics database and data releases from central banks, national statistical agencies, and World Bank country desks” (The World Bank, 2019). The main source, the IMF Balance of Payments, calculates remittances based on two components; personal transfers and compensation of employees. Even though the used source is the most reliable source on remittances data one must bear in mind that these calculations do not account for informal transfers of money. For example, a migrant giving their relatives cash money would not be recorded on any national account. Freund and Spatafora (2005) estimate that these informal remittances could account for 35% to 75% of official remittances to developing countries. Seeing as this is a very large share this should be kept in mind when looking at remittances data. In the analysis the main dependent variable is the remittances to GDP ratio as this provides for a better comparison than the absolute amount of remittances. The absolute GDP data is sourced from the World Bank too. However, one must bear in mind that the change in the ratio is therefore not only dependent on remittances, but the ratio can decrease due to a decrease in GDP, whilst the absolute amount of remittances might have been stable.
The main independent variable is the dichotomous sanction variable, meaning that the variable takes a value of 1 if a sanction is imposed, and 0 otherwise. The data comes from the TIES (Threat and Imposition of Sanctions) data set, by Morgan et al. (2014). This database is an update from a data set by Morgan et al. (2009) and covers 1412 cases with a time span from 1945-2005. Even though the last possible reported year of a sanction imposition is 2005, the data shows their duration up to 2012; hence the upper-bound year of the panel data set is 2012. The TIES database is very detailed with many variables that can be used to understand the nature of the sanction, such as the (primary) sender(s), target, sanction type, target cost, and the target settlement nature. The primary sender is the initiator of the sanction, the target is the country sanctioned, and the sanction type shows what type of sanction is implemented. Further, the target cost captures the costs of sanctions on the target state, where 1, 2, and 3 indicate a minor, major, and severe cost respectively. The target settlement nature captures how the sanction was settled, taking the effect the episode had on both the sender and the target into consideration.
Bank, 2019). The higher this share is, the more ’open’ a country is. The data on capital account openness is sourced by the Chinn-Ito Index (Chinn and Ito, 2010), and give a value between 0 and 1, where the scale goes from not open (0) to most open (1).
4.2
OLS sample
The OLS regression is based on final data sheet used for the SCM estimation. This means that the data has already been modified and trimmed. The steps taken can be found in section 4.3. The subsequential descriptive statistics for the data set can be found in Table 1 below. Here we see that not all observations of the remittances to GDP ratio are available, as only about 80% is available. This result is partly to blame by the absence of all GDP data, seen by economic development. In both cases the maximum amount of observations would have been 3,333. Also, the GDP per capita is not that high on average in the developing countries. Further, at an average of 19% of all observations a sanction was imposed. Lastly, we see that the developing countries are on average not that open, as capital account openness is only reported at 30%. These statistics are all in line with expectations of developing countries.
Table 1: Descriptive statistics for the OLS estimation
(1) (2) (3) (4) (5) Variables N Mean SD Min Max Remittance/GDP Ratio 2,682 0.04 0.06 0.00 0.50 Sanction 3,333 0.19 0.39 0.00 1.00 Economic Development 3,114 1,840.36 2,041.96 94.56 15,434.57 Net Migration 3,333 -0.00 0.01 -0.05 0.05 Trade Openness 2,911 69.94 36.19 0.17 375.38 Capital Account Openness 3,333 0.30 0.30 0.00 1.00
Table 2 shows the pairwise correlation matrix, which shows the correlation between all the variables. As expected, all control variables of remittances show significant correlation with the remittances. Though the relation between remittances and sanctions shows significance, it is very low. Therefore, in the regression the result is expected to be negative too, however this can change due to the inclusion of control variables. We also see that the rest of the variables have a low correlation too.
Table 2: Pairwise correlation matrix Of all the variables used in the estimation, * p < 0.5
Variables (1) (2) (3) (4) (5) (6) (1) Remittances / GDP 1 (2) Sanction -0.1366* 1 (3) Economic Development -0.0785* 0.0523* 1 (4) Migration -0.1356* 0.0151 0.1704* 1 (5) Trade Openness 0.2578* -0.1566* 0.1490* 0.0242 1 (6) Capital Account Openness 0.2014* -0.0098 0.2056* 0.0713* 0.1964* 1
Figure 2: The absolute amount of remittances (in billions) in developing countries per region
In the Appendix 7.1 Figure 7 provides insights in sanctioning practices. The amount of sanctions used in the regression are 107; one country can have multiple sanctions over the sample period. The US and EU alone are the primary senders in more than 75% of the cases, with 64 and 18 imposed sanctions respectively. Data on the amount of sanctions per region and per type are given over three decades: from 1980-1989 (I), from 1990-1999 (II), and from 2000-2005 (III). Most sanctions took place in the last decade (III), amounting to 60, these were evenly distributed over the regions of which most were trade or financial sanctions. However, compared to the decade before the amount of sanctions in Europe & Central Asia and sub-Saharan Africa increased a lot. As for the types of sanctions, compared to the decade before, financial sanctions have increased significantly; the amount was seven times higher. Interestingly no asset sanctions have been imposed exclusively, if they were imposed this was done so in combination with other sanctions (”Multiple” in Figure 7b).
4.3
SCM sample
For the SCM estimation several steps had to be taken before the data was fit to use in the estimation. First all developed countries were excluded, based on The World Bank (2019) classification. Even though developing countries are the aim of this study, one must note that the data quality is generally lower than in developed countries. Next, all countries with either no remittance data or remittance data starting only 2000 were removed. The year 2000 was set as a lower-bound because the TIES database only covers cases that were implemented until 2005. Since we need at least 5 years of pre-sanction data to construct a fitting control unit for the SCM, remittance data only starting in 2000 will not suffice and is therefore excluded. Then the most extreme outliers (for remittance to GDP ratio) were dropped, namely Tonga and Lesotho, to prevent biased estimates and so increase the accuracy of the prediction method.
implies that the target country had some form of economic damage by the sanction. Of these cases those were removed that only had limited pre-sanction (< 5 years) or post-intervention (less than up until the year the sanction ended plus 3 years) remittance data. Also those with no covariate data at all for at least one variable were removed, as this prevents a pre-sanction synthetic control to be constructed. Appendix 7.2 refers to these cases as ” ”Not enough data”.
After excluding the cases that were not eligible as a result of these restrictions 26 cases remained (”Included” in Appendix 7.2). Of these 26 cases only cases with a good pre-treatment fit of the synthetic control unit were used for the analysis, which resulted in a total of 9 cases (”Included*” in appendix 7.2). Table 3 shows some important data of these 9 cases. The pre-sanction period is selected on basis of data availability or in the case of a previous sanction starting 3 years after the previous sanction ended. The sanction period is the duration of the sanction and for most countries this coincides with the year the sanction was imposed. For Mali, Niger, and Bolivia we see that this is not the case, this means that the years before the imposition only a threat was in place. Only two primary imposers are present in these cases, namely the US and the EU. This was expected as these two imposed more than three quarter of all sanctions in the aggregate sample. As for the sanction type mainly financial sanctions can be observed, in the form of the termination of foreign aid. As for the type of trade sanction, these are mainly import restrictions, which the most common type of trade sanction according to Hufbauer et al. (2007). The only asset sanction is in combination with an import sanction and a non-economic sanction. Seeing as asset sanctions have high costs for the sender country we would expect this type to be combined with other types.
Table 3: Case information for the 9 selected cases used in the individual estimation
Country Pre-sanction period Sanction period Sanction imposition Primary imposer
Sanction type Type category
Latin America
Bolivia 1984-1999 2000-2009 2004 US Termination Foreign Aid Financial Costa Rica 1980-1991 1992-1997 1992 US Import restriction Trade Dominican
Republic
1980-1991 1992-2001 1992 EU Import restriction Trade Honduras 1980-1991 1992-2001 1992 EU Import restriction Trade
Sub-Saharan Africa
Guinea 1994-2001 2002-2006 2002 EU Termination Foreign Aid Financial Mali 1980-1999 2000-2009 2003 US Termination Foreign Aid Financial Niger 1980-1999 2000-2009 2003 US Termination Foreign Aid Financial Nigeria 1980-1992 1993-1999 1993 US Import restriction, Asset
Freeze, Travel Ban
Other Togo 1980-1991 1992-2007 1992 EU Termination Foreign Aid Financial
5
Results
5.1
OLS Estimation
The results for the OLS estimations are shown in table 4 below. Column 1 reports the pooled OLS, column 2 includes country fixed effects, and column 3 includes country and time fixed effects. All estimations include robust standard errors.
Table 4: Results of the OLS Estimations
Dependent variable: remittances to GDP ratio
(1) (2) (3) Variables Pooled OLS Fixed Effects Fixed Effects Sanction -0.013*** -0.001 -0.003 (0.002) (0.003) (0.003) Ln(Economic Development) -0.000*** -0.000 -0.000*** (0.000) (0.000) (0.000) Net Migration -0.561*** 0.007 -0.090 (0.152) (0.452) (0.457) Trade Openness 0.000*** 0.000** 0.000 (0.000) (0.000) (0.000) Capital Account Openness 0.040*** 0.032*** 0.029***
(0.004) (0.011) (0.011) Constant 0.012*** 0.003 0.022* (0.002) (0.011) (0.011) Observations 2,493 2,493 2,493 R-squared 0.142 0.089 0.174 Number of COW 97 97
Country FE YES YES
Year FE YES
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
When examining the table, we see that all coefficients of the pooled OLS estimation are significant at a 1% level, while the fixed effect estimates yield less significant results and vary in their significance on the control variables. However, as explained in the methodology section, the model with country and time fixed effects (column 3) is the preferred specification of the OLS estimations. This can be also be justified by looking at the R-squared of each model, where column 3 has the highest R-squared. This indicates that the model has the best overall fit and can best explain the proportion of variance in the remittances to GDP ratio. All things considered, the reported R-squared is still relatively low as only 17.4% can be explained by the country and time fixed effects model.
causing remitters to send less remittances to their home country. However, in the fixed effects estimations the coefficients are far smaller and lose their significance. The considerable stability in the flows of remittances is likely to contribute to this result. Another relevant factor is that these data include all sanctions, many of which have too little impact to trigger a rise in the remittances to GDP ratio. We therefore conclude that the OLS estimations provided no indication that sanctions have a notable effect on the remittances to GDP ratio.
As for the control variables, more trade openness and capital account openness are associated with a higher remittances to GDP ratio, which is an expected result. Capital account openness is the only variable that remains significant in all models. The log of economic development is significant and negative in column 1 and 3, which suggests that more economic development leads to a lower remittances to GDP ratio. This result is also in line with the expectations. The pooled OLS yields a significant negative effect for the net migration, which is the anticipated result that more migration leads to more remittances sent back to the home country. However, the fixed effects estimations do no show significant effects. We should be cautious when drawing conclusions from this control variable as the data is only available as a 5-year average, leading to important nuances to be lost for the OLS estimation.
By means of a robustness test the same aggregate OLS estimations were performed by once excluding economic development as a control variable and once excluding trade openness and capital account openness. The results are shown in Appendix 7.3 in Tables 6 and ??, respectively. This is done to take the indirect relationship presented in Figure 1 into account, as explained in section ??. For the main variable of interest we see no change in the coefficient when excluding economic development and after excluding trade- and capital account openness a slight decrease is experienced. Overall, the robustness checks show no change in sign in all the models and therefore no indirect link can be established for the income channel.
5.2
SCM Estimation
According to the OLS estimations, sanctions do not lead to an increase in the remittances to GDP ratio. However, as seen in section 2.2.2 remittances act differently in each country and have a dynamic nature, therefore this section explores this notion further by presenting and discussing the 9 selected cases which have experienced an imposed sanction following the SCM estimation. These entail 5 Sub-Saharan African and 4 Latin American countries. As explained in the data section the individual cases are selected when they when they were imposed (so no threats) and experienced target costs. However, this section also tries to take the broader environment of the target country into consideration when interpreting the results and looks at previously imposed sanctions (without target costs) and threats, as well as the macroeconomic situation in the target country.
In Figures 3 and 4 the year the sanction started can be seen in each figure by the vertical dashed line (also referred to as the ’treatment’) The period before this line is the so called pre-treatment period and for each case the RMSPE is minimised over the entire pre-treatment period. The year the sanction ended is represented by the blue vertical line and everything to the right of this line is post-sanction period. This is included to see what happened after the sanction.
the donor pool to construct the synthetic control can be found in this section as well, for the experiments presented in this section (experiment A) as well as the results with a limited donor pool as shown in section 5.3.2 (experiment B).
5.2.1 Latin America
In Latin America a similar pattern for every nation is observed, namely an increase in remittances after the sanction starts; this confirms the general hypothesis. Empirical studies, such as the study by Fajnzylber and Lopez (2008), underpins this point and finds that remittances behave counter-cyclically and increase sharply following macroeconomic (policy) shocks in Latin America. Also, Beaton et al. (2017) finds that remittances facilitate consumption smoothing.
A reason for this increase in the Latin American countries could be that they all have a medium settlement nature and target costs, indicating that the sanction has led to some degree of damage in the target country. This would explain the increase in remittances for they are counter-cyclical. Possible other reasons for the significant increases in remittances can be natural disasters. In Bolivia the absolute peak of the remittance ratio occurs in 2007, which is the same year the government declared a state of national emergency due to heavy rainfall. Also, for the Dominican Republic the greatest gap in the sanction period was in 1999, coincidentally this is the same year as a devastating hurricane hit Dominican Republic. So it is unsure if the result in Bolivia and Dominican Republic can be ascribed exclusively to the sanction imposition as literature points to counter-cyclicality of remittances in the event of natural disasters.
In Bolivia and Costa Rica clear evidence of anticipation effect can be seen, as the increase of the remittance to GDP ratio starts before the sanction is imposed, this is in line with the expectations. However, as the impact of sanctions is greatest at the beginning of the sanctions one would expect the amount of remittances to GDP to peak at the beginning and decrease over the coarse of the sanction. This is not the case in Latin America, where sanctions continue to increase during the sanction. In Honduras and Dominican Republic the amount even continued to increase after the sanction ended. This can either mean that that the sanction has inflicted a lot of damage on the target economy, meaning that remittances are still needed after the sanction has ended, or can mean that the increase is attributed to other effects.
Figure 3: SCM estimation - Latin America
(a) Bolivia (b) Costa Rica
(c) Dominican Republic (d) Honduras
In Costa Rica, the Dominican Republic, and Honduras the sanctions under examination start in 1992. However in all three countries an additional sanction was imposed in 1995, which coincides with the year the sanctions started to outperform their synthetic control units for these countries. Therefore, it seems that these 1995 sanctions provide more evidence for the increases in remittances to GDP ratio in these countries than the one they experienced in 1992. Also the sanctions are still ongoing which explains why the remittances to GDP ratio is still much higher in the Dominican Republic and Honduras than their relatively stable counterfactual. In Costa Rica another sanction has been imposed in 2002 and ending in 2009, which evidently is the year in which the ratio starts to decrease.
5.2.2 Sub-Saharan Africa
No region commonalities can be found when looking at the effects of a sanction on the remittances to GDP ratio in sub-Saharan Africa, as seen in in Figure 4. No particular trends can be observed as the effects following a sanction differ in their timing and direction in each country. This is not entirely unexpected as the empirical results are ambiguous with some providing evidence for the consumption smoothing effect of remittances (Nnyanzi, 2013), and others advocating a-cyclicality (Nwaogu and Ryan, 2015). Naud´e and Bezuidenhout (2014) also found that remittances are unresponsive in the event of outbreaks of conflict in sub-Saharan Africa.
in remittances after the sanction was imposed is observed. Interestingly, the year that the ratio returned to its equilibrium rate was in 2006, which coincides with the year that the sanction ended. In Nigeria the gap increased by almost 16 times in the year of the sanction, showing a slight anticipation effect, and remained large up until one year after the sanction ended. The economic costs following the sanction were severe for Nigeria seeing as the sanction consisted out of three different types: an important restriction, an asset freeze, and a travel ban. Also for Nigeria there is short period of stability after which a steep increase is observed again. This can be explained by the additional sanction imposed in 1995. After the 1993 sanction ended the remittances to GDP ratio decreases once again and almost returns to its equilibrium a few years after. However, after 2003 an enormous surge in remittances is observed again which coincides with an additional sanction imposed in 2003. In Togo the initial effect is lower than its counterfactual and the remittances only surpass its synthetic control unit after 2000, this is the same year a threat was imposed on Togo. Additionally, in 2005, another sanction was imposed on Togo. After the sanction ended in 2007 the ratios decreased in the two years hereafter, however remained far higher than that of the synthetic control. However, why the effect of the sanction was very lagged is not clear.
Figure 4: SCM estimation - Sub-Saharan Africa
(a) Guinea (b) Mali
(c) Niger (d) Nigeria
(e) Togo
5.3
SCM robustness tests
5.3.1 Individual OLS Estimation
no significance is found.
When the results of the SCM estimations for Latin America and sub-Saharan Africa are compared to those found in the individual OLS estimation we find some differences. The outcomes of Bolivia, Costa Rica, Guinea, Mali, Nigeria, and Togo are in line with the individual OLS estimation results. However, for Honduras, Dominican Republic, and Niger the sign is opposite. However, the results for Niger were only significant in the robustness check excluding trade- and capital account openness as control variables.
All in all, for all these cases the observed effect of the individual OLS estimation is marginal at best, whilst they are quite significant when looking at the SCM. The contrasting results found when comparing both estimation methods highlight an important caveat of the OLS estimation; it is important to look at sanction case-by-case. This is important as only then one can see how the remittances develop during the sanction and if there are anticipation effects and are thus placed in perspective over time.
5.3.2 Limiting the donor pool
By means of a first robustness test the donor pool was limited to only countries from the same region, the results from this test can be found in Appendix 7.3 in Figure 8. This test was done to address interpolation bias, since countries in the same region have similar characteristics. However, due to data availability, this restriction can limit the donor pool significantly which can bias the results in turn as well. Also, at least 10 countries (J = 10) are needed to be able to derive any conclusions from the experiments since the p-value is derived from a ranking and at least 10% significance level is required (J1 = 101 = 0.1). We see that, for example, in Latin America many sanctions were imposed in the same period in the region. This limited the donor pool to the extent that hardly any experiment B could successfully be applied in this region. For the other countries experiment B could be conducted for, the amount of countries in the donor pool was smaller than 10 (J < 10), meaning that no meaningful interpretation can be derived from the results. Nevertheless, the test is performed as a robustness check to see if the results are similar. Figure 8 shows the countries for which the experiment could be replicated with a limited donor pool. For Mali, Niger, Nigeria, and Togo the results are similar to those in section 5.2.2. For Mali, Niger, and Nigeria this can be explained by the weights assigned to countries in the donor pool for experiment A and for experiment B. For these countries the weights in experiment A are already almost solely sub-Saharan African countries (for exact details please refer to Appendix 7.4). The pre-treatment fit after experiment B is very bad for Bolivia so no conclusions can be derived from these.
5.3.3 Placebo tests
Another robustness check and the way to derive significance from the Synthetic Control Method can be done by conducting placebo tests as explained in section 3.2.2. This section only reports the two cases in which significance was found in experiment A due to limited space. The placebo tests for all other experiment A cases can be provided upon request. Figures 5 and 6 show the distribution of the unit of interest ( ˆα1t) in
orange and all other experiments in grey ( ˆαP L
1t ). In both cases we see that the gap is not unusually large for
in the placebo tests that includes all countries in the donor pool (in Figure 5a and 6a).
the gap only became significant just before the sanction, this indicates that that indeed the sanction has led to an unusually large decrease in the remittances to GDP ratio in Mali after the imposition of a sanction and provides evidence of an anticipation effect. Unfortunately, this significant result shows exactly the opposite of what we expected to happen and the hypothesis.
Figure 5: Mali
(a) MSPE All (b) MSPE < 5
Also after correcting for a maximum of five times the MSPE we find a significant result in Honduras, as shown in Figure 6b. Contrary to Mali, in Honduras the sanction has led to a small, but significant increase in the remittances to GDP ratio, which is in line with the hypothesis. However, the significant result is lagged, starting from 1999 up until the end of the of the sample period. Seeing as the result continues to outperform the donor pool after the sanction has ended in 2002 and bearing in mind that another sanction has been imposed on Honduras in 1995 and is still ongoing, this can mean that the sanctions still have effect on the economy and remittances are still needed to mitigate the negative effects of these sanctions.
Figure 6: Honduras
6
Conclusion
6.1
Discussion
This study looked at the effect of sanctions on a not yet before researched capital flow: remittances. Several contributions to the literature have been made through two estimation methods. Via its macro-empirical OLS analysis of 101 developing countries, from 1980-2012, the findings suggest that economic sanctions have no effect on remittances. This point towards the stable nature of remittances, which is in line with the literature (for example, De et al. (2019)). However, the SCM results show strong evidence in support of the hypothesis; where in almost all the cases the remittances to GDP increased. This is in line with the counter-cyclicality of remittances found in the literature. Additionally, this provides a preliminary answer to the research question, in that remittances can possibly mitigate the negative effects associated with sanctions, thereby contributing to the literature regarding the mitigation of sanctions shock through third-parties. However, further research is needed to answer the research question with certainty. Nonetheless, the SCM results were not robust, as significance is found in only two cases, of which one found an opposite sign as compared to the hypothesis.
This research uses two different estimation methods to be able to distinguish between the effects of sanctions between (through the OLS estimation) and within (through the SCM estimation) developing countries. By using this approach this research distinguishes itself from previous research on the effect of sanctions on capital flows, in accounting for the differences between regions and countries in their reaction to sanctions shocks by doing an individual analysis. Additionally, the SCM provides benefits as one can take other macro-economic situations, such as political unrest, into consideration as well as other (previous or simultaneous) sanctions or threats. Also, the SCM ensures we can look at the development of the remittances to GDP ratio over time, in this way anticipation effects, and short- and long-term effects can be accounted for.
The discrepancies in the estimation results underpin the importance of a case-by-case study due to the dynamic nature of remittances in developing countries; each country responds different in the event of a sanctions shock. The policy implication this has is that countries should think twice before implementing sanctions as a measure of coercion depending on the country, as remittances could possibly mitigate the negative effects.
6.2
Limitations
The methods used also present some limitations to the research. First, the OLS estimation uses a data set that was adjusted and trimmed to fit the SCM estimation. This might have biased the results, but was used for equivalence purposes. Also sanction threats were not included in the sanction dummy variable which might led to the underestimation of the effect of sanctions on remittances. Further, by performing an OLS estimation method important nuances are lost for time. For instance, that the effect of a sanction may differ depending on its duration or in the short- or long run. Also, if a country has been sanctioned before this might have lasting effects, even after the sanction ends. The OLS estimation does not take this into consideration.
Besides the OLS method limitations, the SCM estimation method has several drawbacks too. To start, the construction of the synthetic control can have limitations. In assigning certain weights to certain countries when constructing a counterfactual in some cases a lot of weight is assigned to a particular country. For instance, in the case of Nigeria only one country was used in the construction of the synthetic control, namely Ghana. Also, Bolivia’s synthetic counterfactual was constructed using 82% of Ghana, which is not a country in the same region. When applying such weights to a single country the novelty of the SCM is undermined. However, in the case of Bolivia the donor pool sample could not be limited to those in the same region because there were not enough eligible donors in the region to be able to test for significance. This can occur because other countries that also experienced the treatment (i.e. a sanction) are automatically excluded from the donor pool or due to missing data, which is not uncommon in developing countries. Also only data from the pre-intervention period is used to construct the synthetic control, meaning that in the period before the sanction they may have been equals, however a lot has changed in the past few decades, which can make comparing the treated country and its synthetic counterfactual flawed. Thus, estimates become more reliable when a longer pre-intervention period is available. Further, in hindsight selecting cases to be included in the SCM estimation by first filtering on cases that had target costs, hereby implying that mainly such cases were harmful for the target state, might not have been the best selection method. As seen in the SCM estimation threats or imposed sanctions without target costs could possibly lead to increases as well. One could argue that the selection of the 9 cases for the SCM method could be exposed to selection bias.
As for the results one must bear in mind that changes in remittances might have been caused by other factors rather than the sanction. Finally, for only 2 of the 9 cases significance was found in the robustness tests for the SCM estimation; a significant decrease for Mali and a significant increase for Honduras. This indicates that in the rest of the countries the increase or decrease was not unique in comparison to the donor pool.
6.3
Suggestions for further research
References
Abadie, A., Diamond, A., Hainmueller, J., 2010. Synthetic control methods for comparative case studies: Estimating the effect of California’s tobacco control program. Journal of the American statistical Association 105, 493–505.
Abadie, A., Gardeazabal, J., 2003. The economic costs of conflict: A case study of the Basque country. American economic review 93, 113–132.
Abdih, Y., Chami, R., Dagher, J., Montiel, P., 2012. Remittances and institutions: Are remittances a curse? World Development 40, 657–666.
Afesorgbor, S. K., 2019. The impact of economic sanctions on international trade: How do threatened sanctions compare with imposed sanctions? European Journal of Political Economy 56, 11–26.
Afesorgbor, S. K., Mahadevan, R., 2016. The impact of economic sanctions on income inequality of target states. World Development 83, 1–11.
Attzs, M., 2008. Natural disasters and remittances: Exploring the linkages between poverty, gender and disaster vulnerability in Caribbean SIDS. No. 2008.61, Research paper/UNU-WIDER.
Aydas, O. T., Metin-Ozcan, K., Neyapti, B., 2005. Determinants of workers’ remittances: the case of Turkey. Emerging Markets Finance and Trade 41, 53–69.
Bali, M., 2018. The impact of economic sanctions on Russia and its six greatest European trade partners . Bapat, N. A., Morgan, T. C., 2009. Multilateral versus unilateral sanctions reconsidered: A test using new
data. International Studies Quarterly 53, 1075–1094.
Barajas, A., Gapen, M. T., Chami, R., Montiel, P., Fullenkamp, C., 2009. Do workers’ remittances promote economic growth? No. 2009-2153, International Monetary Fund.
Barry, C. M., Kleinberg, K. B., 2015. Profiting from sanctions: Economic coercion and us foreign direct investment in third-party states. International Organization 69, 881–912.
Beaton, M. K., Cerovic, M. S., Galdamez, M., Hadzi-Vaskov, M., Loyola, F., Koczan, Z., Lissovolik, M. B., Martijn, M. J. K., Ustyugova, M. Y., 2017. Migration and remittances in Latin America and the Caribbean: engines of growth and macroeconomic stabilizers? International Monetary Fund.
Bettin, G., Presbitero, A. F., Spatafora, N. L., 2015. Remittances and vulnerability in developing countries. The World Bank Economic Review 31, 1–23.
Bettin, G., Zazzaro, A., 2018. The impact of natural disasters on remittances to low-and middle-income countries. The Journal of Development Studies 54, 481–500.
Bouoiyour, J., Selmi, R., Miftah, A., 2016. What mitigates economic growth volatility in Morocco?: Remittances or FDI. Journal of Economic Integration pp. 65–102.
Chami, R., Fullenkamp, C., Jahjah, S., 2005. Are immigrant remittance flows a source of capital for development? IMF Staff papers 52, 55–81.
Chinn, M. D., Ito, H., 2010. The Chinn-Ito index. A de jure measure of financial openness. Online version http://web. pdx. edu/˜ ito/Chinn-Ito website. htm pp. 243–76.
Combes, J.-L., Ebeke, C. H., Etoundi, S. M. N., Yogo, T. U., 2014. Are remittances and foreign aid a hedge against food price shocks in developing countries? World Development 54, 81–98.
Constantinescu, I. C., Schiff, M., 2014. Remittances, FDI and ODA: stability, cyclicality and stabilising impact in developing countries. International Journal of Migration and Residential Mobility 1, 84–106. Cooray, A., Mallick, D., 2013. International business cycles and remittance flows. The BE Journal of
Macroeconomics 13, 515–547.
De, S., Islamaj, E., Kose, M. A., Reza Yousefi, S., 2019. Remittances over the business cycle: Theory and evidence. Economic Notes: Review of Banking, Finance and Monetary Economics 48, e12143.
Dizaji, S. F., Van Bergeijk, P. A., 2013. Potential early phase success and ultimate failure of economic sanctions: A VAR approach with an application to Iran. Journal of Peace Research 50, 721–736.
Durmaz, A., Akku¸s, ¨O., 2019. The effects of international sanctions on external debt. Gaziantep University Journal of Social Sciences 18.
Early, B. R., 2015. Busted sanctions: Explaining why economic sanctions fail. Stanford University Press. Early, B. R., Jadoon, A., 2016. Do sanctions always stigmatize? The effects of economic sanctions on foreign
aid. International Interactions 42, 217–243.
El-Sakka, M. I., McNabb, R., 1999. The macroeconomic determinants of emigrant remittances. World Development 27, 1493–1502.
Escrib`a-Folch, A., Wright, J., 2010. Dealing with tyranny: International sanctions and the survival of authoritarian rulers. International Studies Quarterly 54, 335–359.
Eyler, R., 2007. Economic sanctions: international policy and political economy at work. Springer.
Fajnzylber, P., Lopez, J. H., 2008. Remittances and development: lessons from Latin America. The World Bank.
Freund, C., Spatafora, N., 2005. Remittances: transaction costs, determinants, and informal flows. The World Bank.
Gammeltoft, P., 2002. Remittances and other financial flows to developing countries. International migration 40, 181–211.
Giuliano, P., Ruiz-Arranz, M., 2009. Remittances, financial development, and growth. Journal of Development Economics 90, 144–152.
Hatipoglu, E., Peksen, D., 2018. Economic sanctions and banking crises in target economies. Defence and Peace Economics 29, 171–189.
Hufbauer, G. C., Schott, J. J., Elliott, K. A., Oegg, B., 2007. Economic sanctions reconsidered. Peterson Institute for International Economics .
IMF, 2003. Fund Assistance for Countries Facing Exogenous Shocks. August 2003.
Jeong, J. M., 2019. Do sanction types affect the duration of economic sanctions? The case of foreign aid. International Political Science Review 40, 231–245.
Kaminsky, G. L., Reinhart, C. M., Vegh, C., 2005. When it rains, it pours; procyclical flows and macroeconomics policies.
Khodeir, A. N., 2015. Migration remittances inflows and macroeconomic shocks: The case of Egypt. International Journal of Economics and Financial Issues 5, 1001–1010.
Kholodilin, K. A., Netsunajev, A., 2016. Crimea and punishment: The impact of sanctions on Russian and European economies .
Lektzian, D., Biglaiser, G., 2013. Investment, opportunity, and risk: Do us sanctions deter or encourage global investment? International Studies Quarterly 57, 65–78.
Lektzian, D., Biglaiser, G., 2014. The effect of foreign direct investment on the use and success of US sanctions. Conflict Management and Peace Science 31, 70–93.
Lueth, E., Ruiz-Arranz, M., 2007. Are workers’ remittances a hedge against macroeconomic shocks? The case of Sri Lanka. International Monetary Fund.
McLean, E. V., Whang, T., 2010. Friends or foes? Major trading partners and the success of economic sanctions. International Studies Quarterly 54, 427–447.
Mirkina, I., 2018. FDI and sanctions: An empirical analysis of short-and long-run effects. European Journal of Political Economy 54, 198–225.
Mitrut, A., 2014. Remittances after natural disasters: Evidence from the 2004 Indian tsunami .
Morgan, T. C., Bapat, N., Kobayashi, Y., 2014. Threat and imposition of economic sanctions 1945–2005: Updating the TIES dataset. Conflict Management and Peace Science 31, 541–558.
Morgan, T. C., Bapat, N., Krustev, V., 2009. The threat and imposition of economic sanctions, 1971—2000. Conflict Management and Peace Science 26, 92–110.
Naud´e, W. A., Bezuidenhout, H., 2014. Migrant remittances provide resilience against disasters in Africa. Atlantic Economic Journal 42, 79–90.
Neuenkirch, M., Neumeier, F., 2015. The impact of UN and US economic sanctions on GDP growth. European Journal of Political Economy 40, 110–125.
