On the Heterogeneous Eects of Sanctions on Trade

Gabriel Felbermayr Constantinos Syropoulos

Kiel Institute & Kiel University Drexel University
Erdal Yalcin Yoto V. Yotov*
Konstanz University Drexel University
of Applied Sciences ifo Institute

July 28, 2022

Abstract
With the help of a new, comprehensive sanctions database and the latest developments
in the structural gravity literature, we estimate the eects of economic sanctions on
international trade. We demonstrate that the average eects of sanctions may hide
signicant heterogeneity depending on the type of sanctions considered, their duration,
objectives and sender types. We also zoom in on the case of Iran. We nd that the
eects are signicant but also widely heterogeneous across sanctioning countries and
dependent on the direction of trade, even within the European Union. We complement
the aggregate analysis with estimates for 170 sectors, which show that sanctions have
been eective in decreasing bilateral trade at the sectoral level, however, the eects
vary signicantly across sectors and across complete vs. partial trade sanctions.

JEL Classication Codes: F1, F13, F14, F5, F51, H5, N4.
Keywords: Sanctions, Eectiveness of Sanctions, Structural Gravity.
* Contact information: FelbermayrKiel Institute, Kiel University. E-mail: felbermayr@ifw-kiel.de;
SyropoulosSchool of Economics, Drexel University; CESifo. E-mail: c.syropoulos@drexel.edu; Yalcin
Konstanz University of Applied Sciences; CESifo. E-mail: erdal.yalcin@htwg-konstanz.de; YotovSchool
of Economics, Drexel University; Center for International Economics, ifo Institute; CESifo. E-mail:
yotov@drexel.edu.
Acknowledgements: We are grateful to Rod Ludema for his insightful discussion of our
paper during the inaugural Scientic Workshop on Sanctions at Drexel University. We
also appreciate the useful comments and suggestions by Hana Attia, Richard Barnett, Ti-
bor Besedes, Albrecht Bohne, Edward Cartwright, Andrea Charron, Anne-Celia Disdier,
Jonathan Eaton, Lionel Fontagne, Annika Havlik, Friedrich Heinemann, James Harrigan,
Gary Hufbauer, Krisztina Kis-Katos, Udo Kreickemeier, Mario Larch, Tuan Luong, Ahmed
Mahmud, Thierry Mayer, Alejandro Molnar, Inmaculada Martinez-Zarzoso, Anna Miro-
manova, Daniel Mirza, Juan Mora-Sanguinetti, Cli Morgan, Volker Nitsch, Clara Portela,
Gerald Schneider, Holger Strulik, Farid Toubal, Alessia Lo Turco, Hylke Vandenbussche,
Joschka Wanner, Zheng Wang and seminar participants at De Montfort University, Drexel
University, the European Trade Study Group 2018, Katholieke Universiteit Leuven, Leib-
niz University Hannover, the Midwest Meetings in International Trade in Vanderbilt, Paris
School of Economics, the Organisation for Economic Co-operation and Development, Tech-
nische Universitat Darmstadt, University of Göttingen, ZEW Mannheim, University of Ho-
henheim, and the 2018 Annual Conference of the German Economic Association.
 At the beginning of the 21st century, the same as a century earlier, economic
sanctions remain an important yet controversial foreign policy tool.

(Hufbauer et al., 2007)

1 Introduction
From the Age of Pericles in Athens of ancient Greece, to the times of Napoleon and Thomas

Jeerson in the nineteenth century, and World War I, economic sanctions have served as a

prominent and purposive tool of coercive statecraft (Drezner, 1999). They have been im-

posed unilaterally or coalitionally and, especially in the post-Cold War era, with increasing

frequency.
1 Prior to WW II, economics sanctions were inextricably linked to military endeav-

ors, mostly taking the form of trade sanctions and economic blockades under the protective

umbrella of the sanctioning states' naval forces.

2 In more recent times, however, the arsenal

of weaponry in economic sanctions has been enriched to include boycotts, restrictions on

trade in arms, nancial sanctions, travel bans, and the withholding of military assistance,

among others.
3

The growing popularity of economic sanctions among policymakers raises a fundamental

question: Do they work? The conventional wisdom (e.g., Pape (1997)) is that ...Economic

1 Notable examples of such sanctions include: the Megarian decree of Athenians in 435 B.C.; the US's
sanctions at the time of Jeerson against France and Great Britain during their engagement in the Napoleonic
Wars; the sanctions against Italy by the League of Nations in 1935; the Anti-Apartheid Act against South
Africa in 1986; the US sanctions against Cuba; the United Nations nancial and trade sanctions against
Iraq prior to the overthrowing of Saddam Hussein in 2003; the recurrent sanctions on North Korea; the
comprehensive and multi-round economic sanctions by EU, the US and other nations against Russia; the
imposition of sanctions, under the US leadership, against Iran in 1987, their expansion in 1995 and 2006,
and their controversial renewal by the US in 2018; the sanctions against Venezuela under Hugo Chavez and
Nicolás Maduro; and, of course, the recent sanctions on Russia. Remarkably, and as emphasized later on in
our discussion of the content of the Global Sanctions Data Base (GSDB), the number of recorded sanctions
during 1950-2019 was 1101. What's more, their number over the last 25 years of this period more than
doubled.
2 In some cases, sanctions served as precursors of armed conict. As Findlay and O'Rourke (2009) amply
document in their impressive discourse of international trade in the second millennium, trade and warfare
were inseparable.
3 The increased reliance on these instruments nowadayswhich could be attributed to the liberal order
promulgated by the GATT/WTO and the dramatic reduction in transaction costs due to technological
advancesmay have improved the sanctioning states' ability to interfere in the sanctioned states' national
aairs. See Ahn and Ludema (2018) for an insightful analysis of targeted sanctions on Russia and Besedes
et al. (2018) for a valuable analysis of nancial sanctions on German economic activity.

1
Sanctions Do Not Work. Hufbauer et al. (2007), one of the most comprehensive contributions

to the related empirical literature, report that 34% of the cases they examined could be

viewed as successful. However, after subjecting this nding into further scrutiny, Pape

(1998) contends that the success rate was only 4%.

4 In contrast, more recently, Kirilakha

et al. (2021) provide evidence that the average success rate of sanctions is higher (around

40%). Why do analysts contend that sanctions do not work?

The evocative title of Morgan and Schwebach (1997), Fools Suer Gladly... may be in

the vicinity of truth. Hufbauer et al. (2007) and Drezner (1999) outline a number of reasons

that could explain the seemingly limited success of sanctions. First, the types of sanctions

used may be inadequate for the specic objective(s) considered. Second, the imposition

of sanctions may prompt vociferous opposition in the target country by uniting citizens and

domestic interests in rallies behind the ag. Third, powerful allies of the sanctioned country

may intervene (as black knights) to counteract the damaging eects of sanctions. Fourth,

uneven sharing of the costs of sanctions among the sender's allies and business interests may

impair unity in multilateral relationships thereby undermining their eectiveness. Fifth,

policy leaders may choose to deploy sanctions because they perceive them as a less damaging

substitute for military interventions. Last, but not least, while there is a broad consensus on

the presence of a positive link between the economic damage sanctions inict on targets and

their political success, the meta-analysis of Demena et al. (2021) suggests that the empirical

evidence for such links is not robust.

5

Motivated by the above insights, our objective/contribution in this paper is threefold.

First, we aim to demonstrate the usefulness of a recently created database, the Global

Sanctions Data Base (GSDB), cf. Felbermayr et al. (2020b) and Kirilakha et al. (2021), in

4 Also see Pape (1997), Baldwin (1985; 1999), Drezner (1999) and Kaempfer and Lowenberg (2007) for
insightful reviews of various contrasting views in this literature.
5 Morgan and Schwebach (1997) propose a theoretical framework and an empirical test that connects the
higher the cost of economic sanctions to the target to a higher the probability that sanctions will succeed. In
addition to emphasizing the above costs, careful theoretical contributions to this problem (e.g., Eaton and
Engers (1992; 1999); Drezner (1999)) also suggest that the lower the cost of sanctions to the sender(s), the
more likely that sanctions will succeed.

2
assessments of the impact of economic sanctions.
6 Second, we oer quantitative measures

of the costs of sanctions. Focusing on trade as the primary channel of economic pain on the

target and capitalizing on the latest developments in the structural gravity literature, we

estimate of eects of sanctions on aggregate and disaggregated trade for 170 sectors within

the broad categories of Agriculture, Mining & Energy, Manufacturing and Services. Third,

we show that the impact of sanctions is signicant and widely heterogeneous across the

various dimensions covered by the GSDB. Our methods and the wide heterogeneity in our

sanction estimates open opportunities for an analysis of the determinants of the eectiveness

of sanctions.

Taking advantage of the rich dimensionality of the GSDB and following the latest rec-

ommendations for gravity estimations, we obtain a series of partial equilibrium estimates of

the impact of sanctions on international trade. The broad conclusion from this analysis is

that average estimates of the eects of sanctions mask the presence of signicant heterogene-

ity across various dimensions. For example, while we do not obtain statistically signicant

estimates of the average impact of economic sanctions on trade, we obtain a negative and

statistically signicant estimate of the eects of nancial sanctions. Moreover, our estimates

reveal that complete trade sanctions reduce trade ows between sanctioned and sanctioning

countries signicantly (i.e., by more than 70 percent). In contrast, partial trade sanctions

do not aect overall trade signicantly. This nding suggests that even if partial sanctions

are eective in reducing trade in some sectors, their impact may be oset by a possible

increase in trade in other sectors. We also nd that the impact of sanctions is heterogeneous

across sanction cases (e.g., the sanction on Iran vs. all other trade sanctions), their duration,

objectives, and sender types (e.g., unilateral vs. multilateral and US vs. UN).

To deepen our understanding of the heterogeneous eects of trade sanctions, we zoom

6 See Section 2 for a. brief description of the GSDB.

3
in on the impact of the sanctions on Iran.
7 We obtain a negative, sizable, and statistically

signicant average estimate of the impact of the sanctions on Iran's trade and nd that the

impact of sanctions has been signicantly stronger on Iran's imports. We also observe that

the eects of sanctions vary across country pairs within the Iran sanctions (e.g., USA-Iran

vs. China-Iran). Our estimates suggest that the impact of sanctions also varies within

pairs depending on the direction of trade ows (e.g., Japan-Iran vs. Iran-Japan). Next,

using the EU case, we identify the eects of deeper provisions within the Iran sanctions and

nd that deeper provisions have been eective. However, their impact has also been very

heterogeneous across EU members and dependent upon the direction of trade. Finally, we

obtain some estimates that are not statistically signicant (e.g., Turkey-Iran and China-

Iran), reecting sanctions waivers, and even some positive estimates (e.g., United Arab

Emirates-Iran), reecting trade diversion eects for Iran's major source of imports.

We complement the aggregate-data analysis with estimates at the sectoral level. To

obtain these results we employ data from the International Trade and Production Database

for Estimations (ITPD-E) of the United States International Trade Commission (USITC),

cf. Borchert et al. (2022).

8 The ITPD-E enabled us to estimate the eects of complete and

partial trade sanctions for 170 sectors within the broad categories of Agriculture, Mining

& Energy, Manufacturing and Services. Our main ndings suggest that complete trade

sanctions have been eective in reducing trade at the sectoral level.

9 The average eects of

partial sanctions across the four broad sectors in our sample are smaller but all negative,

with the largest impact in services trade.

Our work complements and extends a series of studies that quantify the impact of sanc-

7 Our focus on this country is motivated by two factors. First, it is one of the most perplexing and widely
discussed cases in recent history. Second, the sanctions on Iran are multi-dimensional, varying in terms of
senders (e.g., UN vs. US vs. EU sanctions), sanction types (e.g., military goods vs. all goods vs. travel vs
nance vs. individuals), and time (e.g., EU sanctions were rst imposed in 2006 and reached a peak in terms
of stringency in 2012). We exploit this multi-dimensionality within the GSDB and identify the heterogeneous
eects of these sanctions in Section 2.2.
8 We oer a description of the ITPD-E in Section 2.
9 The strongest eects of complete trade sanctions are for Mining & Energy, followed by Agriculture,
and Manufacturing. Though these sanctions did not impact services trade signicantly, some partial trade
sanctions may have had signicant eects.

4
tions. For example, several papers have already used the gravity model to obtain partial

equilibrium estimates of the eects of sanctions (e.g., Hufbauer and Oegg (2003), Caruso

(2003), Yang et al. (2004), and Afesorgbor (2018)). We oer three contributions in relation

to these studies. First, we implement the latest developments in the estimation structural

gravity literature. With the help of the PPML estimator, we obtain country-pair and di-

rectional estimates of the impact of trade sanctions in a setting with asymmetric pair xed

eects and importer-time and exporter-time xed eects. Second, we capitalize on the rich

dimensionality of the GSDB to obtain novel estimates of the impact of trade sanctions.

Our study of the Iran sanctions complements several papers that have also examined

these sanctions (including, for example, Haidar (2017), which uses disaggregated customs

data to study export deection in Iran and Draca et al. (2017), which studies the incidence

of sanctions on Iranian rms) and papers that consider other sanction cases (e.g., Crozet

and Hinz (2017) and Miromanova (2019), which focus on Russia). We view these studies,

which use rm-level analysis, and our top down approach as complementary both in terms

of methods and ndings. From a broader, methodological perspective, and motivated by the

political science literature that views `large-N' studies and case studies as complementary

but mutually exclusive, the analysis of the sanctions on Iran demonstrate how our large-N

data and methods also allow us zoom in on the heterogeneous sanction eects for a specic

country.

We view the quantitative exercise as limited in scope and are keenly aware that it alone

cannot fully address the question of the eectiveness of sanctions.

10 Our motivation in

pursuing this task is twofold. First, we believe that any attempts to bridge the gap between

scholarship and policy making are doomed to remain unsatisfactory (or fail) if the costs that

senders impose on targets through trade-related channels and the burden of their actions

to themselves are not appraised. We think the analysis in this paper can rectify, or at

10 This is so because the problem we are dealing with is multi-dimensional in nature and has component
parts that are notoriously dicult to quantify and assess. As noted earlier, these diculties are at least in
part related to the potentially conicting objectives of politicians and policymakers as well as the nuanced
strategic policy interactions in the global economy.

5
least mitigate, this problem from the perspective of trade costs. Second, while we recognize

the importance of earlier eorts to quantity these eects, we believe that the fact that our

methodology has solid theoretical and general equilibrium foundations positions it well to

capture both the heterogeneous eects of the various types of sanctions considered and the

distributional eects on the burden of costs among senders.

The rest of the paper is organized as follows. Section 2 describes our data and sources

with a special focus on the GSDB. It also oers some descriptive motivational evidence on

the relationship between sanctions and international trade. Section 3 briey reviews the

structural gravity theory and species our econometric model. Section 4 presents a series of

results that demonstrate that the eects of sanctions are very heterogeneous across various

dimensions. Section 5 concludes. A supplementary appendix includes additional support

materials and demonstrates how the partial estimates of the impact of sanctions on trade can

be translated into general equilibrium welfare eects within the same theoretical framework.

2 Data: Description and Sources

To perform our empirical analysis we rely on the new edition of the Global Sanctions Data

Base (GSDB), cf. Kirilakha et al. (2021), which we describe in Subsection 2.1. Subection

2.2 focuses on the sanctions on Iran, which we explore in detail in our empirical analysis.

Finally, Subsection 2.3 describes the additional data and sources we use.

2.1 The Global Sanctions Database

To quantify the impact of sanctions on international trade, we will utilize the new edition

of the Global Sanctions Data Base (GSDB).

11 The GSDB covers all publicly traceable mul-

tilateral, plurilateral, and purely bilateral sanctions over the period 1950-2019. In light of

our objectives, we begin this section with a brief description of the main dimensions of the

11 The GSDB is freely available and can be requested by e-mail at GSDB@drexel.edu.

6
GSDB. Then we take a close look at trade sanctions because these sanctions play a leading

role in our quantitative analysis. For a comprehensive and full data description interested

researchers and readers are referred to Felbermayr et al. (2020b) and Kirilakha et al. (2021).

The GSDB includes 1,101 sanction cases. A case refers to an ocial sanction policy that

is initiated by an individual country, a group of countries, or an international organization

against another nation and that is identied on the basis of an ocial decision (e.g., an

executive order, a parliamentary decision, or a resolution). Therefore, cases are related to

single sanctioning countries (e.g., the USA), groups of countries (e.g., the EU) or interna-

tional organizations (e.g., the UN). Sanctions are classied along three primary dimensions:

type, objective, and success. First, sanctions that are classied by type fall into one of

the following six categories: trade, nancial activity, arms, military assistance, travel, and

other sanctions. Trade sanctions represent national or international legal actions that aim

to prevent the movement of goods over a dened period. In comparison to existing promi-

nent sanction datasets (such as Hufbauer et al. (2007) and Morgan et al. (2014)) the GSDB,

not only identies the existence of trade sanctions between two nations, but also species

the direction of aected trade ows (i.e., exports and/or imports), and the type of trade

restrictions considered (partial or complete sanctions). In particular, the GSDB identies

whether imports from or exports to a target are banned totally or partially. At this stage the

database does not permit quantitative assessments of the banned share of trade.
12 Sanctions

in TIES include taris, export controls, embargoes, import bans, travel bans, freezing assets,

cutting foreign aid, and blockades. As a result the number of sanctions that are included in

TIES amount to over 1400 cases. Structurally, the GSDB is closer to Hufbauer et al. (2007)

because neither database includes classical trade policy interventions, such as taris and/or

anti-dumping measures, as sanctions. In fact all sanctions listed in Hufbauer et al. (2007)

12 The most widely used database released in 1983 and updated several times over the past years is Hufbauer
et al. (2007). This database focuses on approximately 200 identied cases of trade and nancial sanctions.
First published in 2006, Morgan et al. (2014) oer the Threat and Imposition of Economic Sanctions database
(TIES) and cover not only imposed sanctions but also sanction threats. Although the TIES also has an
economic and nancial focus, it diers from Hufbauer et al. (2007) and from the GSDB in that it treats
defensive trade measures, like anti-dumping duties, as economic sanctions.

7
are also included in the GSDB. Additionally, the GSDB includes information on the politi-

cal objectives of senders and documents the extent to which listed sanctions are considered

successful in achieving senders' dened objectives.

Fig. 1 depicts the evolution of all identied sanctions between 1950 and 2015. For each

year, panel (a) of Fig. 1 reports the number of newly imposed sanctions and the cases that

were initiated in previous years. During the 1950-1978, 1992-1995, and 2010-present time

periods, the number of sanctions increased gradually or abruptly. In contrast, during the

1978-1992 and 1994-2006 periods, the number of sanctions remained the same or declined.

During 1950-1978, we observe a gradual and steady rise in the use of sanctions. This period

is characterized by civil wars, and political and territorial conicts (e.g., the civil war in

Nigeria (1967), the Cuba crisis (1960), and South Africa (1963)) which shape the evolution

of sanctions. Between 1992 and 1995 political turmoil in African countries (e.g., Algeria

(1992), Angola (1993), Burundi (1995), Cameroon (1992), Congo (Zaire) (1992), Kenya

(1992), Libya (1992), Nigeria (1993), Togo (1993)) and the war in former Yugoslavia led to an

acceleration in the adoption of sanctions. In the subsequent period, until 2010, both the total

number of sanctions in force and the newly initiated cases amounted to an annual number

of around 150 observed cases. However, a strong and steady increase of initiated sanctions

policies is observed from 2010 until the end of the sample period (which also continues

today). The signicant rise of sanctions during the last decade is special in that: (i) it has

occurred at a time during which new civil unrest (e.g., Egypt, Syria, Yemen, Venezuela,

Ukraine, Sudan, or Myanmar) has appeared; and (ii) the US administration under president

Trump increasingly relied on sanctions to implement its foreign policy objectives (e.g., the

US sanctions against Iran, North-Korea, China and Venezuela). Overall, the number of

sanctions has been increasing steadily over the last seven decades. We view this as evidence

of the rising popularity of sanctions as a tool of coercive diplomacy. As a consequence, one

8
 Figure 1: Number of Sanctions over Time

(a) New vs. Existing Cases (b) Frequency by Type

Panel (a): number of sanctions in force inherited from last year, and number of total (inherited plus new) sanctions in force per
year. Panel (b): number of sanctions by type (trade sanctions, arms sanctions, military assistance sanctions, nancial sanctions,
travel sanctions, and other sanctions), stacked.

may also suspect that the economic impact of sanctions will have been increasing as well.
13

Panel (b) of Fig. 1 displays the evolution of the number of sanctions by type.
14 Several

ndings stand out: First, trade sanctions are the main type of sanctions implemented be-

tween 1950 and the late 70s. During this period, all other types of sanctions played a minor

role. Second, over the years two specic policy measures have been applied with increasing

frequency: nancial and arms sanctions. Travel bans, restrictions on military assistance and

other sanctions have also been imposed at an increasing rate. In contrast, the number of

trade sanctions remained constant over the years which, in combination with the increasing

number of other sanctions, suggests that the number of trade sanctions is relatively smaller.

This raises the question of whether trade sanctions, which have been treated with particular

13 An alternative explanation for the changing sanction patterns over time is to link them to the various
stages of the cold war. In the early cold war (i.e., 1950-1975) period, sanctions increased continuously over
time. Thereafter, in the late cold war (i.e., 1975-1990) period, the number of sanctions continued to rise.
However, their volatility rose as well. While there was a strong rise in sanctions use during some years, in
other years there was a substantial (drop due to a signicant rise in the number of lifted sanctions). In the post
cold war (1990-2005) period, there was a strong slow-down in new sanction policies, reecting the positive
developments in global economic integration and cooperation such as the foundation of the World Trade
Organization (WTO). Finally, the last (i.e., post-2005) period, which is characterized by a dramatic rise in
imposed sanctions, starts with the terrorist attacks on the US in 9/11 and continues until today. The latter
development has been accompanied by a declining political convergence within international organizations
such as the WTO or the UN.
14 The exact number of countries that have been sanctioned with a specic type of sanction can be found
in Felbermayr et al. (2020a).

9
care in the GSDB, are still an important and eective policy tool.

In the GSDB, trade sanctions are broadly dened as limitations of trade ows. The GSDB

distinguishes between several types of trade sanctions depending on coverage, direction, and

participating countries. First, it is possible that only exports to a target state or imports from

such a state are temporarily banned. It is also possible that both exports and imports are

not banned during the sanctions period. Accordingly, depending on the direction of banned

trade ows, the GSDB distinguishes between sanctions on exports from the sender to the

target, sanctions on imports from the target to the sender, and sanctions that simultaneously

apply to both the exports and the imports between the two sides. Second, trade sanctions

may apply only to specic goods (partial trade sanctions) or to exports and/or imports as

a whole (complete trade sanctions). Thus, the GSDB distinguishes between partial and

complete sanctions within each of the three dimensions, depending on the direction of

banned trade ows.

15 Finally, an important aspect of sanctions is the scope of participating

countries that may vary from one country imposing sanctions unilaterally, to sanctions that

are imposed multilaterally by, for example, all members of the United Nations.
16 To perform

the structural empirical analysis, we capitalize on the variation across each of the three

dimensions of trade sanctions in the GSDB.

Fig. 2 tracks partial and complete trade sanctions over time. In the early 1950s, all coun-

tries that imposed sanctions on imports restricted imports to the full extent. In subsequent

years, however, an increasing number of sanctioners restricted imports only partially. In

2019, the share of all countries applying import sanctions stood at around 65%. In contrast,

as illustrated in panel (b) of Fig. 2, over the past 70 years, countries have been less eager to

15 Fora range of cases, the GSDB additionally includes detailed trade ban information (e.g., export controls
of small aviation, helicopter, aviation parts and electronics, or export restrictions of high-tech products).
The partial character of this type of trade sanctions is very heterogeneous as the product ranges dier
substantially.
16 A prominent example of a complete unilateral sanction is the full trade sanction policy imposed by the
United States on Cuba that was introduced by President John F. Kennedy in February 1962. The UN
sanction on Iran based on the 1996 resolution is an example for a (partial) sanction policy that was imposed
by all UN member states. The various dimensions of trade sanctions in the GSDB are explained in detail in
Felbermayr et al. (2020a)

10
 Figure 2: Partial versus Complete Trade Sanctions

a) Partial or Complete Import Sanction b) Partial or Complete Export Sanction

100% 100%

90% 90%
80% 80%
70% 70%
60% 60%
50% 50%
40% 40%
30% 30%
20% 20%
10% 10%
0% 0%
1950
1953
1956
1959
1962
1965
1968
1971
1974
1977
1980
1983
1986
1989
1992
1995
1998
2001
2004
2007
2010
2013
2016
2019

1950
1953
1956
1959
1962
1965
1968
1971
1974
1977
1980
1983
1986
1989
1992
1995
1998
2001
2004
2007
2010
2013
2016
2019
Partial Import Sanction Complete Import Sanction Partial Export Sanction Complete Export Sanction

Partial and complete import (Panel (a)) and export (Panel (b)) sanctions over time (1950 to 2019).

impose restrictions on all exports. Between 1950 and 1990, around 70% of countries sanc-

tioning exports imposed partial restrictions. In the following period of 20 years, almost half

of all export restricting countries applied complete export sanctions, whereas in recent years

two thirds of the countries participating in export sanctions applied only partial sanctions.

These patterns illustrate the importance of identifying the diering extent of trade sanctions

as partial vs. complete and export vs. import sanctions. In the econometric analysis below

we demonstrate that the eectiveness of sanctions varies across these dimensions.

Who imposes sanctions on whom? Fig. 3 oers two radial dendrograms by major regions.

Arrows starting in a specic region and pointing to a particular region indicate the number

of imposed trade sanctions. Panel (a) of Fig. 3 illustrates the sanction activities between

regions for the year 2015. Countries from North-Western Europe (NW Europe) imposed the

largest number of trade sanctions in Africa (brown arrow). At the same time, however, not

a single state from Africa imposed a trade ban against a North-Western European state.

Interestingly, some regions are barely sanctioned by other regions while at the same time

others have been confronted with sanctions from almost every listed region. For example,

East and South Asia have been sanctioned by almost all regions in 2015. Panel (b) of Fig.

3 oers a radial dendrogram that illustrates the sanction activities between regions for the

11
 Figure 3: Bilateral Structure of Sanctions

(a) 2015 (b) 1950

Note: These two radial dendrograms display sanctions between dierent regions in the world for the years 1950 and 2015.
Regions are classied according to the UN Geoscheme. For details see the Appendix.

year 1950. For the sake of comparability, we hold the regional classication of countries

constant to that in 2015. The gure illustrates a much smaller variety and number of

sanctions policies among dierent regions. The biggest share of arrows indicate that trade

sanctions took place between members of the Eastern and Western blocks at that time.

The GSDB also illustrates that the USA and the EU countries have been the most active

nations in imposing trade sanctions against other states, followed by North-African nations

and Canada (cf. Felbermayr et al. (2020a))

The descriptive analysis of trade sanctions reveals that they are used by many countries

in an intricate network. Overall the GSDB suggests that trade sanctions are relevant.

2.2 The Sanctions against Iran

As discussed in the introduction, we nd the sanction on Iran especially interesting (i)

because it is one of the most prominent and most discussed cases in recent history, (ii)

because it is multi-dimensional, and (iii) because it has been imposed by almost all countries

12
in the world. These features allow us to obtain a large set of pair-specic estimates and to

study their heterogeneity across countries. This section oers further details on the nature

and evolution of the sanctions on Iran.

Over the past decades the UN Security Council passed several resolutions that imposed

increasingly tighter sanctions on Iran. The triggering event of more recent UN sanctions

against Iran has been the report by the International Atomic Energy Agency (IAEA) re-

garding Iran's nuclear activities, especially the country's uranium enrichment activities. The

rst UN resolution imposing economic sanctions on Iran came into force when the county

rejected the Security Council's request to suspend uranium enrichment activities. The rst

economic sanctions against Iran commenced in July 2006 based on the UN Security Council

resolution 1696. These sanctions were extended and continuously tightened in subsequent

years. Starting with trade sanctions on goods that could be used in Iran's nuclear and ballis-

tic missile program, sanctions were extended by increasing the number of goods whose trade

was banned and by imposing nancial sanctions and travel bans for individual people.

Table 1: UN Security Council Resolutions imposing sanctions on Iran - since 2006

Resolution 1696 from July 2006.

Resolution 1737 from December 2006.
Resolution 1747 from March 2007.
Resolution 1803 from March 2008.
Resolution 1835 from September 2008.
Resolution 1929 from June 2010.
Resolution 1984 from June 2011.
Resolution 2049 from June 2012.
Resolution 2105 from June 2013.
Resolution 2159 from June 2014.

The US and the EU imposed their own sanctions on Iran on top of the UN resolutions

listed in Table 1. In 1979 the US imposed import sanctions after Iranian students stormed

the US embassy in Tehran. Moreover, in 1984 the US imposed additional sanctions on Iran

because of the country's purported support for acts of international terrorism. In 1995 the US

formulated an executive order to prohibit its bilateral trade with Iran. European sanctions

reached their peak in January 2012 when the EU introduced an oil embargo and froze the

13
assets of Iran's central bank. In addition to imposing sanctions on foreign trade, the US

imposed sanctions on nancial services, and banned the provision of insurance by insurers

in member states to Iran. Finally, all Iranian banks that were identied as institutions in

breach of US or EU sanctions were disconnected from the SWIFT, the global payment system

that connects banks. Table 2 summarizes all sanctions against Iran by other countries which

are listed in the Global Sanctions Data Base. In the empirical analysis each of these listed

sanctions is accounted for as a separate sanction case against Iran.

Table 2: Sanctions against Iran by further countries

Country Initiated in Type of Sanctions

United States 1979 Parital Import Sanctions, Financial Sanctions
United States 1980 Extenstion of Import Sanctions
United States 1984 Partial Export, Arms and Financial Sanctions
United States 1987 Complete Import Sanctions
United States 1995 Complete Export Sanctions
United States 1996 Partial Export Sanctions
United States 2017 Travel Restrictions
United States 2019 Financial Sanctions
Canada 2010 Partial Export, Financial, and Arms Sanctions
Canada 2011 Extension Partial Export Sanctions
Canada 2012 Extension of Existing Sanctions
Canada 2013 Complete Import and Export Sanctions
Canada 2016 Partial Export Sanctions, Financial Sanctions
EU 2007 Arms Sancations
EU 2011 Financial Sanctions and Travel Restrictions
EU 2012 Partial Export Sanctions
EU 2012 Partial Import Sanctions
Japan 2006 Partial Import Sanctions
Australia 2008 Patial Export/Import Sanctions, Travel Restrictions,
Financial, Arms, and Military Sanctions
Switzerland 2011 Partial Export/Import Sanctions, Financial and Travel Restrictions
Switzerland 2016 Partial Export Sanctions, Travel Restrictions
South Korea 2010 Partial Import Sanctions
South Korea 1018 Partial Import Sanctions
Note: This table lists all non-UN sanctions against Iran in the period 1950-2019.

2.3 Other Data and Sources

In addition to the data on sanctions, we employ three dierent sources to construct the

trade datasets for our analysis. To match the time coverage of the GSDB, we construct an

aggregate trade ows dataset for the period 1950-2019 from the Direction of Trade Statis-

14
tics (DOTS) of the International Monetary Fund (IMF), which we complement with data

from the United Nations Comtrade Database.

17 To obtain maximum number of trade ow

observations, we use a mirroring procedure. Specically, consistent with theory, we use CIF

imports as the baseline data, which we complement with data on FOB exports. We also

replace all missing trade values with zeroes.

18

We complement the aggregate analysis with sectoral estimates, which are obtained with

data from the second release of the USITC's International Trade and Production Database

for Estimation (ITPD-E-R02), cf. Borchert et al. (2022).

19 The ITPD-E has three main

advantages for our purposes. First, similar to our aggregate data, it covers all countries in

the world. Second, consistent with gravity theory, it includes not only international but also

domestic trade ows. Finally, and most important for our purposes, it includes 170 industries,

which cover the complete economy within four broad sectors, including Agriculture, Mining

& Energy, Manufacturing, and Services.

20 A downside of the TPD-E relative to our aggregate

dataset is its shorter time coverage (i.e., 1986-2019 for Agriculture, 1988-2019 for Mining &

Energy and for Manufacturing, and 2000-2019 for Services).

Finally, we rely on the Dynamic Gravity Database of the USITC, cf. Gurevich and

Herman (2018), for data on some of the additional gravity covariates we use in our specica-

tions (i.e., regional trade agreements (RTAs), membership in the World Trade Organization

(WTO), and membership in the European Union (EU)). Due to the fact that we use pair

xed eects in all of our econometric specications, we do not need data on the standard

time-invariant gravity variables (such as bilateral distance, common ocial language, colonial

ties, contiguous borders, etc.)

17 Detailed information about the DOTS database can be found at https://data.imf.org/?sk=9D6028D4-

F14A-464C-A2F2-59B2CD424B85, and for the Comtrade Database at https://comtrade.un.org/.
18 We demonstrate that our main results are very similar when the analysis is based solely on positive trade
ow data.
19 Details on the ITPD-E database can be found at https://www.usitc.gov/data/gravity/itpde.htm.
20 The complete list of industries appears in Table 5.

15
3 Quantifying the Impact of Sanctions with Gravity
To quantify the heterogeneous impact of sanctions on trade we rely on the structural gravity

equation  the workhorse model of international trade. This section oers a brief review of

the gravity model and, guided by theory, sets up our econometric specication. As famously

demonstrated by Arkolakis et al. (2012), the following gravity equation, which will motivate

our empirical specication, can be obtained from a series of alternative micro foundations:

 1−σ
Yi,t Ej,t tij,t
Xij,t = , (1)
Yt Pj,t Πi,t

Here, at each point of time t, Xij,t denotes trade ows from exporter i to destination j ; Ej,t
is the total expenditure in importer j ; Yi,t is the value of total production in exporter i ; Yt
is the value of world output; tij,t denotes bilateral trade frictions between partners i and j;
σ >1 is the elasticity of substitution among goods from dierent countries. Equation (1)

intuitively links bilateral exports to market size (the rst term on the right-hand side) and

trade frictions (the second term on the right-hand side). The numerator of the trade cost term

includes bilateral trade frictions (tij,t ), which we model explicitly below as a function of the

possible impact of sanctions among other variables that have been recognized as signicant

determinants of trade ows in the existing literature. Finally, coined by Anderson and

van Wincoop (2003), Pj,t and Πi,t are the multilateral resistance terms (MRs, inward and

outward, respectively).

Guided by theory and capitalizing on the developments in the empirical gravity literature,

as summarized by Yotov et al. (2016), we set an econometric gravity model that corresponds

to the structural gravity equation (1). The following equation will enable us to obtain

partial equilibrium estimates of the eects of sanctions on international trade across various

dimensions:

Xij,t = exp[SAN CTij,t α + GRAVij,t β + µij + πi,t + χj,t ] + ij,t . (2)

16
As noted earlier, Xij,t denotes nominal trade ows from exporter i to importer j at time

t. Most importantly for our purposes, SAN CTij,t is the vector of sanction variables that

take central stage in our analysis. To demonstrate the importance of properly capturing

the heterogeneous impact of sanctions on Iran's trade, in the estimation analysis we start

with a baseline of sanction variables and then gradually decompose these variables across

the various dimensions of interest to us. We motivate and describe each step of decomposing

vector SAN CTij,t in detail in Section 4 below.

Specication (2) includes three sets of xed eects, which are standardly used in the

gravity literature. Specically, πi,t denotes the set of time-varying exporting-country dum-

mies, which control for the outward multilateral resistances of Anderson and van Wincoop

(2003) as well as for any other observable and unobservable exporter-specic factors that

may inuence bilateral trade. Similarly, χj,t encompasses the set of time-varying destination-
country dummy variables that account for the inward multilateral resistances as well as for

any other observable and unobservable importer-specic characteristics that may inuence

international trade.

The third set of xed eects, denoted by µij , are country-pair xed eects, which serve

two main purposes. First, they comprehensively account for all time-invariant bilateral trade

costs by absorbing any observable and unobservable time-invariant bilateral determinants of

trade costs. Second, as demonstrated by Baier and Bergstrand (2007), the pair xed eects

absorb most of the linkages between the potentially endogenous bilateral policy variables

(e.g., RTAs, sanctions etc., and the error term ij,t ).21 We also note that, as in Baier et al.

(2016), we employ directional bilateral xed eects. The benets are: (i) the directional xed

eects allow us to control for the presence of asymmetric trade costs; and (ii) this treatment

21 Another factor that mitigates potential endogeneity concerns with respect to sanctions is that, by de-
nition, sanctions are usually imposed in response to actions/inactions that are specic to the target country.
Therefore, the use of exporter-time and importer-time xed eects in our econometric specication com-
pletely controls for any such target-specic linkages. The downside of using country-time xed eects is, of
course, that they also absorb any multilateral eects of sanctions. Therefore, our estimates of the bilateral
impact of sanctions should be interpreted as lower bounds. We will revisit this point in the next section
when we present our estimates.

17
is consistent with the fact that many of the sanctions in our database are directional.

The last term in equation (2), GRAVij,t , is a vector that includes some standard time-

varying gravity covariates. Specically, we control for the presence of regional trade agree-

ments (RTAs) with a dummy variables (RT Aij,t ), which takes the value of one if there is an

RTA between countries i and j at time t, and zero otherwise. We also account for member-

ships in the WTO and membership in the EU with dummy variables (W T Oij,t and EUij,t ),
which take the value of one if an exporter and an importer are jointly members of the WTO

and the EU, respectively, at time t, and zero otherwise. Finally, we note that the pair xed

eects (µij ) in our econometric model will fully absorb the eects of all time-invariant grav-

ity covariates (such as the logarithm of bilateral distance and indicator variables for colonial

relationships, common language, etc.) which are standardly used in gravity estimations.

We employ panel data to estimate all of our specications. In addition to improving

eciency, the panel data allow for the use of country-pair xed eects and enable us to

study the impact of deeper sanction provisions (e.g., the EU sanctions on Iran) and to study

the evolution of sanction eects over time within the same specication. Following recent

recommendations from Egger et al. (2022), we employ consecutive year data for every year in

our sample.
22 Following the recommendations of Santos Silva and Tenreyro (2006), we rely

on the Poisson Pseudo Maximum Likelihood (PPML) estimator to obtain our results. The

benets of using PPML are: (i) this estimator handles successfully the heteroskedasticity in

trade data which renders OLS estimates inconsistent; and (ii) due to its multiplicative form,

the PPML estimator enables us to take advantage of the information contained in the zero

trade ows. Finally, we cluster the standard errors by country pair in all specications.

22 Consistent with the critique of Cheng and Wall (2005) that `[f]ixed-eects estimation is sometimes
criticized when applied to data pooled over consecutive years on the grounds that dependent and independent
variables cannot fully adjust in a single year's time' (footnote 8, p. 52), many papers have used interval trade
data to estimate gravity models. For example, Baier and Bergstrand (2007) use 5-year intervals, while Olivero
and Yotov (2012) experiment with 3-year and 5-year interval data. Most recently, Egger et al. (2022) argue
against the use of interval data and show that gravity models with three-way xed eects, such as the model
we employ, deliver similar estimates of the common estimates of FTAs.

18
4 Estimation Results and Analysis
This section explores the heterogeneous eects of sanctions across various dimensions. In

Subsection 4.1 we capitalize on the rich dimensionality of the GSDB to study the eects of

sanctions by type, objectives, duration, and over time. Subsection 4.2 zooms in on the impact

of the sanctions on a specic target: Iran. Finally, Subsection 4.3 explores the heterogeneous

eects of sanctions across sectors.

4.1 On the Heterogeneous Eects of Sanctions

Table 3 reports estimates of the heterogeneous eects of sanctions on trade across vari-

ous dimensions. All estimates are based on equation (2), i.e., using the PPML estimator,

exporter-time xed eects, importer-time xed eects, directional pair xed eects, and af-

ter controlling for RTAs, WTO and EU memberships. All standard errors are clustered by

country pair. The dierences between the alternative specications in Table 3 are in the

denition and composition of the sanction covariates.

The results in column (1) of Table 3 are obtained with a single indicator variable for

the presence of a sanction between a pair of countries at any point of time; that is, the key

covariate takes the value of one if there is sanction of any type between two countries, and

zero otherwise. Without going into details, we note that the estimates of the standard gravity

controls (RTA, EU, and WTO) are comparable to those form the literature (e.g., we obtain

positive and statistically signicant estimates of the impact of RTAs and EU membership).

The estimate on WTO is negative and marginally statistically signicant, which is consistent

with the results from Rose (2004).

23 Most important for the current purposes, we obtain a

negative and statistically signicant, but also very small estimate of the eects of sanctions

on trade. A possible explanation for this result is that the single sanction dummy may hide

23 Larch et al. (2019) provide a solution to Rose's puzzle that the WTO does not promote trade.
Specically,
they show that when gravity is estimated after including theory-consistent domestic trade ows, the eects
of the WTO on trade are actually large, positive, and signicant. We conrm this result in Section 4.3,
where we obtain sectoral gravity estimates based on a dataset that also includes domestic trade ows.

19
signicant heterogeneity across various dimensions. This result motivates our main objective

to explore the heterogeneity of the impact of sanctions in order to improve our specication

in pursuit of estimating the true sanction eects.

Table 3: The heterogeneous eects of sanctions

(1) (2) (3) (4) (5) (6) (7)

ALL TYPE TRADE POSTV DUR1 DUR2 MULTI
ANY_SANCT -0.039
(0.017)∗
RTA 0.084
(0.028)∗∗
EU 0.428
(0.035)∗∗
WTO -0.110
(0.065)+
TRADE_SANCT -0.038
(0.027)
ARMS_SANCT 0.002 0.012 0.016 0.012 0.012 0.012
(0.041) (0.040) (0.039) (0.040) (0.040) (0.040)
MLTRY_SANCT 0.013 0.012 0.018 0.009 0.009 0.012
(0.029) (0.029) (0.029) (0.029) (0.029) (0.029)
FINCE_SANCT -0.094 -0.093 -0.092 -0.092 -0.092 -0.093
(0.035)∗∗ (0.035)∗∗ (0.034)∗∗ (0.035)∗∗ (0.035)∗∗ (0.035)∗∗
TRAVL_SANCT -0.080 -0.083 -0.080 -0.082 -0.082 -0.083
(0.055) (0.055) (0.055) (0.055) (0.055) (0.055)
OTHER_SANCT 0.052 0.094 0.090 0.084 0.084 0.094
(0.040) (0.040)∗ (0.039)∗ (0.040)∗ (0.040)∗ (0.040)∗
PARTL_SANCT -0.029 -0.042 -0.031 -0.031 -0.029
(0.027) (0.026) (0.027) (0.027) (0.027)
COMPL_SANCT -1.281 -1.202 -0.250 -0.280
(0.209)∗∗ (0.220)∗∗ (0.120)∗ (0.155)+
SANCT_DURTN -0.072 -0.067
(0.009)∗∗ (0.023)∗∗
SANCT_DURTN_2 -0.000
(0.000)
MULTILAT -1.285
(0.209)∗∗
UNILAT -1.265
(0.280)∗∗
N 2077421 2077421 2077421 1190935 2077421 2077421 2077421
Notes: This table reports estimates of the heterogeneous eects of sanctions on trade. The dependent variable is trade in
levels. All estimates are obtained with the PPML estimator and exporter-time, importer-time and directional country-pair
xed eects. In addition, we control for the presence of RTAs, WTO membership, and EU membership. Column (1) imposes
a common eect of all sanctions regardless of their type. Column (2) allows for heterogeneous sanction eects by type.
Column (3) further distinguishes between the eects of complete vs. partial trade sanctions. Column (4) reproduces the
estimates using positive trade ows only. Columns (5) and (6) explore the links between sanction duration and trade. Finally,
column (7) allows for dierential eects of multilateral vs. unilateral sanctions. Standard errors are clustered by country
pair. + p < 0.10, ∗ p < .05, ∗∗ p < .01. See text for further details.

To obtain the results in column (2) of Table 3, we take advantage of an important dimen-

sion of the GSDB, its distinction among the following six types of sanctions: Trade, Arms,

Military Assistance, Financial, Travel, and Other. According to the estimates in column (2),

only Financial sanctions (F IN CE _SAN CT ) have a statistically signicant (and negative)

impact on bilateral trade ows. The link between trade and nancial constraints is a natural

explanation for this result. We do note, however, that the estimate is relatively small. A

20
possible explanation for this and for the insignicant estimates on other types of sanctions

(except trade sanctions) is that, especially in recent years, economic sanctions have become

more targeted (`smart'), with few spillover eects into non-targeted activities. On a related

note, such eects (e.g., the country-specic eects of nancial sanctions) are fully controlled

for by the country-specic xed eects in our econometric model.

The most surprising result from column (2) of Table 3 is the insignicant estimate of

the eects of Trade sanctions (T RADE _SAN CT ). We oer an explanation in column (3),

where we distinguish between the eects of complete trade sanctions, which are imposed on

all trade ows between the sender and the target, vs. partial trade sanctions, which only

apply to specic sectors and activities.

24 The estimates in column (3), which we view as

our main results, reveal that complete trade sanctions have led to a signicant decrease in

trade. Specically, our estimate implies that, on average, bilateral trade between senders

and targets of complete trade sanctions has fallen by more than 72%. Using the level of

trade at the time of the imposition of the sanctions, this implies a decrease in trade by about

37 billion. Costas: What are the units behind the 37 billion and the gures in the following

footnote?
25 The insignicant estimate on partial trade sanctions on aggregate trade is not

surprising, as they may be targeting very specic activities and sectors. We oer evidence

for this hypothesis in Section 4.3. Finally, the estimates in column (4) reveal that the results

are very similar when we estimate the model with data on positive trade ows only.

The main takeaway from Table 3 is that, as expected, complete trade sanctions have been

most eective in impeding trade. Based on this, and to keep the analysis and exposition

manageable, for the rest of this section we focus on the heterogeneous eects of complete

trade sanctions. We do this by capitalizing on the additional information from the GSDB

and guided by existing theories of the impact of sanctions. We start with an analysis of the

24 Unfortunately, the GSDB does not identify the specic sectors and activities that are targeted by partial
trade sanctions. However, based on the sectoral analysis that we perform in Section 4.3, we show that the
eects of partial trade sanctions are fairly heterogeneous across sectors, which may indicate their targets.
25 If all trade sanctions were as eective as complete trade sanctions, they would have reduced trade by
1.5 trillion. For comparison, based on our estimates of the eects of RTAs, we nd that the corresponding
partial eect of RTAs is an increase of trade by about 216 billion.

21
eects of sanctions across three time dimensions. In our rst specication, we follow Egger

et al. (2022) and Dai et al. (2021), who study the evolution over time of the eects of FTAs

and sanctions, respectively. Specically, we introduce leads and lags for the complete trade

sanctions in our sample. An additional advantage of this specication from an econometric

perspective is that it enable us to test for pre-existing trends, which may contaminate our

results.

The estimates in panel (a) of Fig. 4 are obtained from the specication in column (3) of

Table 3 after replacing the single variable for complete trade sanctions with 16 new variables

that include ve sanction leads (which are designed to capture pre-trends or anticipation

eects), a contemporaneous eect at the time when the sanction is imposed, and ten sanction

lags (which are intended to capture the evolution of the phasing-in eects of sanctions). Two

primary messages stand out from Fig. 4. First, we do not nd evidence for pre-trends

or anticipation eects. Second, we see that the eects of complete trade sanctions are

immediate, strong, and persistent over time.

A possible interpretation of the phasing-in sanction estimates in Fig. 4 is that they evolve

non-monotonically, reaching a pick around 5 years after implementation. Such interpreta-

tion is consistent with the histogram of sanctions duration in panel (b) of Fig. 4, which

shows that the most common sanctions are those lasting 5 years. We study the relationship

between sanctions duration and trade more formally in columns (5) and (6) of Table 3. Both

specications are the same as our main specication from column (3). The dierence is that

we have introduced interactions between sanction duration and complete trade sanctions (in

column (5)) and between the square of sanction duration and complete trade sanctions (in

column (6)). The main message from column (5) is that longer sanctions have stronger neg-

ative eects on trade.

26 However, column (6) does not oer any evidence for non-monotonic

sanction eects depending on duration.

In our next experiment, we explore a dierent time dimension by allowing the eects of

26 The nding that longer sanctions have stronger eects on trade is consistent with the results from Dai
et al. (2021), who obtain separate estimates of the eects of short vs. long sanctions at 5-year cuto.

22
 Figure 4: The heterogeneous eects of sanctions: Over time and by objective

(a) Pre-trends and phasing-in (b) Sanctions duration

.5

2
Sanction Gravity Estimates

1.5
0

Density
-.5

1
-1

.5
-1.5

0
t-5 t-4 t-3 t-2 t-1 t t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+9 t+9 t+10 1 2 3 4 5 6 7 8 9 10
Time Years of sanction duration

(c) By period (d) By objective

1
-.5

Sanction Gravity Estimates

0
Sanction Gravity Estimates
-1

-1
-1.5

-2 -3
-2

cy

ar

ct

es
ht
ng

ng
fli
W
ra

tiv
ig

on
ha

ha
oc

ec
d

lC
En
C

C
em

an

bj
e

y
ia

O
-2.5

um

lic
D

im

or

le
Po
eg

rit
H

tip
r
R

Te

ul
1950-1974 1975-1989 1990-2004 2005-2019

M
Time Objectives

Note: The panels of this gure display estimates of the heterogeneous eects of sanctions across various dimensions. All
estimates are obtained from the main specication in column (3) of Table 3 after replacing the single variable for complete trade
sanctions with alternative covariates. The results in panel (a) allow for pre-trend and phasing-in sanction eects. Panel (b)
depicts the density of sanctions with duration of 10 years or less. The estimates in panels (c) and (d) allow for heterogeneous
eects across four time periods and depending on the sanction objectives, respectively. See text for further details.

sanctions to vary depending on the period in which they were imposed. To this end, we

follow Morgan et al. (2022) who use the GSDB to identify four `sanction eras,' including

the periods 1950-1974, 1975-1989, 1990-2005 and 2005-2019. Our estimates are shown in

panel (c) of Fig. 4. The impact of complete trade sanctions is not statistically dierent

across the four sanction eras. However, we do nd the evolution of the sanction eects in

terms of magnitude intuitive. Consistent with the descriptive evidence that the use of trade

sanctions has decreased over time (Felbermayr et al. (2020b)), we observe a decrease in their

eects across the rst three eras. Most recently, however, the negative impact of complete

23
trade sanctions is stronger. A possible explanation for this result is that the line between

economic sanctions and trade policy has become blurred in recent years, e.g., during the

Trump presidency (Hufbauer and Jung (2020)).

Motivated by ndings from the political science literature, e.g., Demena et al. (2021), that

multilateral sanctions are more successful in reaching their political objectives, in our next

experiment we distinguish between the eects of unilateral vs. multilateral sanctions. To

perform this analysis, we use the GSDB to construct two indicator variables: M U LT ILAT
takes a value of one if a sanction is imposed by an international organization (e.g., UN, EU,

the Arab Union, etc.) or jointly by a group of countries; and U N ILAT is equal to one if a

sanction is imposed by a single sender (e.g., USA, Canada, Japan, etc.)

The estimates of the eects of multilateral vs. unilateral sanctions from column (7)

of Table 3, which are very similar to each other, have two implications. First, given the

signicantly larger number of senders in the multilateral sanctions, our results imply that

multilateral sanctions cause signicantly more damage to the target states due to lower

trade. In turn, the higher economic cost may increase the probability for political success.

Second, given that many countries that participate in multilateral sanctioning either receive

exemptions or do not implement the sanctions completely,

27 the equal estimates we obtain

between the two types of sanctions imply that the impact of multilateral sanctions for some

specic senders is larger than the average eect of unilateral sanctions.

We conclude the analysis in this section by taking advantage of another key dimension

of the GSDB. Specically, we allow for dierential eects of complete trade sanctions based

on their objectives. The GSDB distinguishes between the following nine sanction objectives:

promote democracy, defend human rights, regime change, end war, prevent war, resolve

territorial conict, ght terrorism, change policy, and other. Moreover, (i) some sanction

cases may include more than one (up to 4) objectives, and (ii) some target states may be

subject to more than one sanction at the same time. Therefore, in addition to the original

27 We oer evidence for both cases in subsequent sections.

24
nine categories from the GSDB, we construct an additional variable for `Multiple objectives'.

Our ndings appear in panel (d) of Fig. 4. The gure does not include estimates for

sanctions with objectives to ght terrorism, to prevent war, or other objectives. The reason

is that there were no complete trade sanctions in the GSDB with such objectives. Our

estimates reveal some signicant dierences between the impact of complete trade sanctions

with dierent objectives. Specically, sanctions that target regime change and sanctions with

multiple objectives have the strongest eects on trade. We nd the latter result intuitive.

We also nd that sanctions in support of democracy and human rights are among those

with the weakest impact on trade. This result is in contrast to the descriptive evidence from

Morgan et al. (2022) that these two objectives are the ones that are most often successfully

achieved by sanctions. This raises the question for the link between the economic impact of

sanctions and sanction success, which, in our view, is not well established in the literature

and deserves more attention.

4.2 The Eects of the Sanctions on Iran

Our objective in this section is twofold. First, from a methodological perspective and mo-

tivated by the political science literature that views `large-N' studies and case studies as

complementary but mutually exclusive, we wish to demonstrate how our large-N data and

methods also allow us zoom in on the heterogeneous sanction eects for a specic country.
28

Second from a policy perspective, we select and provide new estimates and analysis of the

most widely studied sanctions: the sanctions on Iran. Consistent with our approach in the

previous section, we base the specications in this section on our main specication from

column (3) of Table 3. Then, we gradually allow for heterogeneous eects of the sanctions

on Iran across various dimensions. Due to the large number of sanction estimates, we only

report results for complete trade sanctions and for the sanctions on Iran.

28 Motivated by our analysis and capitalizing on the methods we use, Kirilakha (2022) obtains sanction-
specic estimates, pair-specic estimates within sanctions, and even directional pair-specic estimates for
each complete sanction in the GSDB. Then, she uses the distribution of these estimates to study the deter-
minants of sanction eectiveness and success.

25
 Our estimates appear in Table 4. We start with a model, in column (1) of Table 4, which

isolates the eects of the sanctions on Iran from all other sanctions. To ease interpretation,

we subtract the dummy variable for the sanctions on Iran from all other sanctions variables

in our specication.
29 Thus, all estimates in column (1) should be interpreted in levels rather

than as deviations from each other. The main results from column (1) are that the sanctions

on Iran have hurt Iran's trade; however, the average impact of these sanctions is almost twice

smaller as compared to the average eect of complete trade sanctions.

In column (2) of Table 4 we separate the indicator variable for the Iran sanctions into two

dummies, one for Iranian exports to the sanctioning countries (SAN CT _IRAN _EXP ) and

another for Iranian imports from the sanctioning countries (SAN CT _IRAN _IM P ).
30 The

main result from column (2) is that the eects of the sanctions on Iran's exports and imports

are asymmetric. Specically, we obtain a negative and statistically signicant estimate of

the eects of sanctions on Iran's imports, which is twice larger than the corresponding

insignicant eect on Iran's exports. Our explanation for this result is that many countries

did not impose sanctions on Iran's main export item (i.e., oil), and we conrm this hypothesis

later in this section, when we study the heterogeneous eects of the EU sanctions on Iran.

In column (3) of Table 4 we allow for directional pair-specic sanction eects. Specically,

we isolate the eects on Iranian exports and imports with a series of individual countries

and regions (e.g., the US vs. the EU). Our choice of sanctioning countries is based on

the following three criteria: (i) Identifying the major exporters to Iran (e.g., United Arab

Emirates and Turkey) and the main destinations for Iranian exports (e.g., Japan and China)

at the beginning of the sanctioning period. (ii) Selecting countries that imposed individual

sanctions on Iran (e.g., US, Canada, and Australia). (iii) Isolating the impact of sanctions

29 We remind the reader that all specications include the full set of covariates from column (3) of Table
3. The complete sets of gravity estimates are available upon request.
30 We also obtained estimates of the eects of the Iran sanctions by type. We do not report these results
due to space limitations. However, they are similar to our main ndings (i.e., we nd that the eects of the
Iran sanctions on trade are driven by trade sanctions). The main dierence with our main results is that
we do not obtain a signicant estimate of the eects of the nancial sanctions on Iran. Our explanation
is that the eects of the nancial sanctions on Iran have been multilateral and thus controlled for by the
country-time xed eects in our econometric model.

26
 Table 4: The Eects of Sanctions on Iran

(1) (2) (3) (4)

ALL EXPIMP CNTRY EU2012
SANCT_IRAN -0.451
(0.236)+
COMPL_SANCT -1.019 -1.019 -1.006 -1.006
(0.144)∗∗ (0.144)∗∗ (0.143)∗∗ (0.143)∗∗
SANCT_IRAN_EXP -0.357 -0.127 -0.127
(0.304) (0.359) (0.359)
SANCT_IRAN_IMP -0.715 -0.047 -0.048
(0.180)∗∗ (0.261) (0.262)
SANCT_IRN_USA -2.663 -2.660
(0.227)∗∗ (0.227)∗∗
SANCT_USA_IRN -2.367 -2.368
(0.170)∗∗ (0.170)∗∗
SANCT_IRN_CAN -2.691 -2.687
(0.334)∗∗ (0.335)∗∗
SANCT_CAN_IRN -1.025 -1.025
(0.205)∗∗ (0.206)∗∗
SANCT_IRN_AUS -0.896 -0.894
(0.379)∗ (0.380)∗
SANCT_AUS_IRN -1.662 -1.661
(0.242)∗∗ (0.243)∗∗
SANCT_IRN_CHN 0.352 0.344
(0.357) (0.357)
SANCT_CHN_IRN 0.059 0.052
(0.229) (0.231)
SANCT_IRN_TUR 0.376 0.377
(0.367) (0.368)
SANCT_TUR_IRN 0.089 0.082
(0.229) (0.230)
SANCT_IRN_JPN 0.180 0.186
(0.382) (0.384)
SANCT_JPN_IRN -0.781 -0.774
(0.227)∗∗ (0.228)∗∗
SANCT_IRN_ARE -0.377 -0.389
(0.367) (0.366)
SANCT_ARE_IRN 0.621 0.614
(0.236)∗∗ (0.236)∗∗
SANCT_IRN_RUS -0.308 -0.311
(0.358) (0.359)
SANCT_RUS_IRN -0.801 -0.806
(0.234)∗∗ (0.235)∗∗
SANCT_IRN_IND 0.227 0.222
(0.356) (0.356)
SANCT_IND_IRN 0.004 0.000
(0.220) (0.221)
SANCT_IRN_EU -0.415 -0.133
(0.404) (0.415)
SANCT_EU_IRN -0.445 -0.222
(0.231)+ (0.254)
SANCT_IRN_EU_2011 -1.364
(0.425)∗∗
SANCT_EU_IRN_2011 -0.824
(0.260)∗∗
N 2077421 2077421 2077421 2077421
Notes: This table reports estimates of the heterogeneous eects of trade sanctions on
Iran. The dependent variable is trade in levels. All estimates are obtained with the
PPML estimator and exporter-time, importer-time and directional country-pair xed
eects. In addition, we control for the presence of RTAs, trade sanctions on exports
and imports, complete trade sanctions, and all other sanctions. Column (1) obtains a
single average estimate of the eects of the sanctions on Iran. Column (2) decomposes
the impact of of the Iranian sanctions on the exports vs. imports of Iran. Column (3)
further allows for dierential eects across country pairs and in each direction of trade
ows. Finally, column (4) captures the impact of the stricter EU sanctions on Iran
post 2011. Standard errors are clustered by country pair. + p < 0.10, ∗ p < .05, ∗∗
p < .01. See text for further details.

27
whose stringency changed over time (e.g., the stringency of the EU sanctions on Iran reached

a peak in 2012). Once again, to ease interpretation, we subtracted the directional pair-

specic sanctions dummies from the corresponding sanction indicators for Iranian exports

and imports.

Most of the estimates on the country-specic sanction eects in column (3) are negative

but very heterogeneous across countries. We also obtain some positive estimates. The

estimates on the sanctions of the US, Canada, and Australia are the largest, and they are

all negative and statistically signicant in both directions of trade. The explanation is that

these countries imposed individual sanctions on Iran. The estimates for Japan and Russia

are negative and statistically signicant but only for their exports to Iran. The estimates

for Iran's trade with China and Turkey are not statistically signicant, possibly because

these countries were granted sanction waivers.

31 The single positive and signicant estimate

we obtain is on exports from the United Arab Emirates, suggesting the presence of `trade

creation' eects for Iran's largest source. Finally, the estimates of the eects of the EU

sanctions are both negative, but they are relatively small and only the estimate on EU

exports is marginally statistically signicant. We oer a possible explanation next.

The estimates from column (4) of Table 4 allow for additional (presumably stronger)

eects of the EU sanctions on Iran. The motivation for this specication is that the EU sanc-

tions reached a peak in 2012 with the imposition of a series of new provisions and additional

sanctions on Iran. To capture these eects we introduce two new indicator variables for the

EU sanctions on Iranian exports to the EU post 2011 (SAN CT _IRN _EU _2012) and for

the EU sanctions on Iranian imports from the EU post 2011 (SAN CT _EU _IRN _2012),

which we subtracted from the original dummies for the EU sanctions. Consistent with our

expectations, the estimates from column (4) indicate that, on average, the impact of the EU

sanctions on Iran more than doubled during the period post-2011, decreasing EU exports

31 For
some sanctions on Iran (e.g., the 2017-2019 US sanctions), eight countries received a waiver allowing
them to temporarily continue buying Iranian Oil. These countries include China, India, Greece, Italy,
Taiwan, Japan, Turkey and South Kore.

28
to and imports from Iran by 56% and 74%, respectively. The natural explanation for the

stronger impact on Iran's exports is that the new EU sanction provisions covered oil.

We conclude the analysis in this section by going a step deeper in studying the heteroge-

neous eects of sanctions. Specically, we zoom in on the impact of the 2012 EU sanctions

on Iran by allowing for directional and pair-specic eects; that is, we allow for dierential

eects of the EU sanction on Iran for exports from Germany to Iran vs. imports of Bulgaria

from Iran.
32 Our new specication includes all variables from column (4) of Table 4 but we

replace the two 2012 EU covariates with a series of directional and pair-specic dummies for

the impact of the sanctions on Iran for each EU member. Due to the large number of addi-

tional estimates, instead of a table format, we use a graphical presentation. The left panel of

Fig. 5 presents estimates of the eects on EU exports to Iran, while the right panel reports

results for the corresponding eects on Iran's exports to the EU countries. The countries on

the x-axis of each panel of Fig. 5 are ordered based on the size of the gravity estimates of

the 2012 eects of Iran's sanction.

Three salient ndings emerge from the left panel of Fig. 5. First, all estimates are

negative and all but ve of them are statistically signicant. This means that the 2012 EU

sanction on Iran has been eective in reducing Iran's exports to the EU. More importantly,

the graph captures very wide heterogeneity in terms of the magnitude of the sanction eects.

In terms of impact, the total eects of the sanction on Iran's exports to the EU are very

large and very heterogenous, ranging from complete elimination of trade, e.g., decreasing

it by 99.4 percent (for Finland) to being small and statistically insignicant (for Slovakia,

Lithuania, and Greece).

32 EU members decide on sanctions policies within the framework of the Common Foreign and Security
Policy (CFSP) as set out in Article 21 of the European Union Treaty. Restrictive measures such as sanctions
are either implemented at EU or at national levels. Measures such as arms embargoes or restrictions on
admission are implemented directly by EU members, which are legally bound to act in conformity with CFSP
Council Decisions. Further measures interrupting or reducing, in part or completely, economic relations with
a third country, such as freezing funds and economic resources, are implemented by means of a regulation,
adopted by the Council. (A detailed explanation of the procedure can be found in the EU's guidelines
on restrictive measures.) Because all EU regulations are binding for member states one could argue that
sanctions should have uniform eects for EU member states.

29
 Figure 5: On the Heterogeneous Impact of the EU Sanctions on Iran

(a) Iran's exports to the EU (b) Iran's imports from the EU

2
0

Estimate of the sanction effect

Estimate of the sanction effect

1
-2

-1 0
-4

-2 -3
-6

H LT
PON
BG L
CR
G E
BEC
FL
G IN
ESR
AUP
PRT
SV T
D N
FRU
A
SWA
SVE
N K
D D
LUK
R X
U
H L
RV

LVP
ESA
LT T
U
LUN
H X
RV
SW L
N E
CD
PRE
G T
BR

LVK
H A
FRN
ESA
ES T
POP
D L
MU
C T
BG P
R R
U
AUA
SV T
BEN
G L
LTC
SVU
K

IR
IT
Z

Y
U

O
L
IR

IT
Z

Y
FI

C
D

EU Members EU Members

Gravity Estimate CI min95/max95 Gravity Estimate CI min95/max95

Note: These gures present estimates of the impact of the post-2011 EU sanctions on Iran's trade with EU countries. The
left panel reports estimates of the eects on Iran's exports to Iran, while the right panel includes results for the corresponding
eects on Iran's imports from the EU countries. The countries on the x-axis of each panel are ordered based on the size of the
gravity estimates of the sanction estimates. See text for further details.

The left panel of Fig. 5 reports results for the impact of the sanction on Iran's imports

from the EU countries. Our estimates reveal that the sanction has been eective in limiting

Iran's imports from the EU but, consistent with our ndings from column (4) of Table 4, not

as eective as limiting Iran's exports. Once again, our explanation for this is that the 2012

EU sanctions targeted Iran's key export, oil. Turning to the heterogeneity of our estimates,

we note that 23 of the 28 estimates are negative, and 18 of them are statistically signicant.

Once again, we see wide heterogeneity in the estimates of the sanction eects, which, in

terms of volume eects, range from decreasing trade by more than 85 percent (for Malta,

Hungary, Poland, and Bulgaria) to actually being positive and statistically signicant (for

Lithuania, Estonia, and Latvia).

We draw six main conclusions on the basis of our analysis in this section. First, our

results demonstrate that the impact of sanctions is pretty heterogeneous across sanctions

(e.g., the sanction on Iran vs. all other trade sanctions). Second, we nd that the eects of

sanctions vary signicantly across country pairs within sanctions (e.g., USA-Iran vs. China-

Iran). Third, our estimates reveal that the impact of sanctions also varies within pairs

depending on the direction of trade ows (e.g., Japan-Iran vs. Iran-Japan). Fourth, we

30
nd that the additional provisions in the EU sanctions on Iran have been very eective on

average. However, fth, we also document very heterogeneous eects of sanctions even within

the European Union. Finally, we also obtain several estimates that are not statistically

signicant (e.g., Turkey-Iran and China-Iran), reecting sanction waivers, and even some

positive estimates (e.g., United Arab Emirates-Iran and Lithuania-Iran), reecting trade-

creation eects.

4.3 The Sectoral Eects of Economic Sanctions

We conclude the empirical analysis by investigating the eects of sanctions across sectors.

To this end, we capitalize on the separability property of the gravity model, whose empirical

implication is that specication (2) applies to individual sectors at any level of aggregation,

and we utilize the new release of the International Trade and Production Database for esti-

mation (ITPD-E-R02), which enables us to obtain estimates of the eects of sanctions for 170

industries, covering the complete economy within four broad sectors, including Agriculture,

Mining & Energy, Manufacturing, and Services.

33 By oering detailed estimates for such a

large number of sectors that cover the whole economy we complement existing studies that

focus on individual sectors, e.g., Hall and Sey (2021)for tourism, Hall and Sey (2021) for

banking and nance, Larch et al. (2021) for agriculture, and Larch et al. (2022) for mining.

Table 5: Sectoral Sanction Eects, ITPD-E-R02

ID Industry Description Broad Sector Complete Partial

Estim. Std.Err. Estim. Std.Err.
1 Wheat Agriculture -1.070 0.700 -0.110 0.100
2 Rice (raw) Agriculture -0.620 0.650 0.140 0.120
3 Corn Agriculture 0.100 0.480 0.0500 0.170
4 Other cereals Agriculture -0.760 0.570 -0.0200 0.110
5 Cereal products Agriculture -4.070 0.960 -1.250 0.280
6 Soybeans Agriculture -2.160 0.280 -0.790 0.160
7 Other oilseeds (excluding peanuts) Agriculture -0.980 0.380 0.110 0.0800
8 Animal feed ingredients and pet foods Agriculture -1.270 0.460 0.120 0.0900
9 Raw and rened sugar and sugar crops Agriculture 0.760 1.450 -1.070 0.580
10 Other sweeteners Agriculture -0.910 0.700 -0.330 0.290
11 Pulses and legumes, dried, preserved Agriculture -0.770 0.320 -0.100 0.0800
12 Fresh fruit Agriculture -1.430 0.220 -0.360 0.0900
13 Fresh vegetables Agriculture 0.0700 0.240 -0.150 0.100
14 Prepared fruits and fruit juices Agriculture -0.500 0.330 0.0600 0.130
15 Prepared vegetables Agriculture -4.310 1.780 2.810 3.910
16 Nuts Agriculture -1.380 0.830 -0.0400 0.0900
17 Live Cattle Agriculture -1.300 0.690 -0.320 0.320
18 Live Swine Agriculture -3.340 1.610 -0.0900 0.230
19 Eggs Agriculture 0.890 0.320 0.340 0.120
20 Other meats, livestock products, and live animals Agriculture -1.470 0.470 -0.360 0.100
Continued on next page
33 The complete list of industries appears in Table 5.

31
ID Industry Description Broad Sector Complete Partial
Estim. Std.Err. Estim. Std.Err.
21 Cocoa and cocoa products Agriculture -3.080 0.990 -0.540 0.320
22 Beverages, nec Agriculture -2.380 0.710 0.260 0.0800
23 Cotton Agriculture -1.490 0.380 -0.0500 0.130
24 Tobacco leaves and cigarettes Agriculture -0.680 0.260 -0.150 0.0800
25 Spices Agriculture -1.100 0.450 -0.0100 0.0700
26 Other agricultural products, nec Agriculture -0.970 0.230 -0.320 0.0900
27 Forestry Agriculture -0.570 0.210 0.200 0.100
28 Fishing Agriculture -1.260 0.360 -0.400 0.130
29 Mining of hard coal Mining and Energy -3.300 0.850 -0.200 0.0900
30 Mining of lignite Mining and Energy -4.150 1.830 -1.110 0.500
31 Extraction crude petroleum and natural gas Mining and Energy -0.270 0.250 -0.0800 0.130
32 Mining of iron ores Mining and Energy -10.79 1.460 0.450 0.200
33 Other mining and quarring Mining and Energy -0.840 0.460 -0.130 0.130
34 Electricity production, collection, and distribution Mining and Energy -0.140 0.330 0.120 0.220
35 Gas production and distribution Mining and Energy 0.190 1.390 0.460 0.600
36 Processing/preserving of meat Manufacturing -0.240 0.440 -0.390 0.100
37 Processing/preserving of sh Manufacturing -1.460 0.420 -0.0400 0.0900
38 Processing/preserving of fruit and vegetables Manufacturing 0.180 0.190 -0.0100 0.0400
39 Vegetable and animal oils and fats Manufacturing -0.640 0.220 0.0300 0.0900
40 Dairy products Manufacturing -0.530 0.270 -0.260 0.0700
41 Grain mill products Manufacturing -0.710 0.330 0.0400 0.0700
42 Starches and starch products Manufacturing -0.780 0.220 0.190 0.100
43 Prepared animal feeds Manufacturing -0.190 0.320 -0.0100 0.0800
44 Bakery products Manufacturing -0.350 0.210 -0.0200 0.0500
45 Sugar Manufacturing -1.020 0.810 0.480 0.140
46 Cocoa chocolate and sugar confectionery Manufacturing -0.120 0.170 0.0100 0.0600
47 Macaroni noodles and similar products Manufacturing -0.590 0.320 0.140 0.0900
48 Other food products n.e.c. Manufacturing -0.810 0.320 0.0400 0.0400
49 Distilling rectifying and blending of spirits Manufacturing -1.550 0.640 -0.240 0.0900
50 Wines Manufacturing -1.600 0.360 -0.140 0.0400
51 Malt liquors and malt Manufacturing -0.330 0.230 -0.130 0.100
52 Soft drinks; mineral waters Manufacturing -1.560 0.550 -0.0300 0.0800
53 Tobacco products Manufacturing -0.270 0.490 0.380 0.160
54 Textile bre preparation; textile weaving Manufacturing -0.650 0.260 -0.110 0.0500
55 Made-up textile articles except apparel Manufacturing -0.0800 0.210 -0.0100 0.0600
56 Carpets and rugs Manufacturing -0.280 0.300 -0.0700 0.0900
57 Cordage rope twine and netting Manufacturing -1.330 0.360 -0.0500 0.0600
58 Other textiles n.e.c. Manufacturing -0.560 0.170 0.0500 0.0700
59 Knitted and crocheted fabrics and articles Manufacturing 0.230 0.380 0.130 0.0900
60 Wearing apparel except fur apparel Manufacturing -0.540 0.480 0.150 0.0800
61 Dressing and dyeing of fur; processing of fur Manufacturing -1.390 0.380 -0.270 0.150
62 Tanning and dressing of leather Manufacturing -0.560 0.310 -0.160 0.0900
63 Luggage handbags etc.; saddlery and harness Manufacturing -0.270 0.350 -0.120 0.0500
64 Footwear Manufacturing -0.580 0.260 -0.160 0.0900
65 Sawmilling and planing of wood Manufacturing -1.080 0.260 -0.0400 0.100
66 Veneer sheets plywood particle board etc. Manufacturing -0.460 0.130 -0.130 0.0900
67 Builders' carpentry and joinery Manufacturing 0.0500 0.240 -0.220 0.0700
68 Wooden containers Manufacturing -1.750 0.200 0.0200 0.0800
69 Other wood products; articles of cork/straw Manufacturing -0.320 0.290 -0.0600 0.100
70 Pulp paper and paperboard Manufacturing -0.300 0.150 0.0300 0.0800
71 Corrugated paper and paperboard Manufacturing -0.530 0.180 -0.0800 0.0600
72 Other articles of paper and paperboard Manufacturing -0.560 0.110 -0.110 0.0500
73 Publishing of books and other publications Manufacturing -0.330 0.130 -0.0200 0.0400
74 Publishing of newspapers journals etc. Manufacturing -0.0800 0.510 0 0.0900
75 Publishing of recorded media Manufacturing 0.180 0.210 -0.0900 0.110
76 Other publishing Manufacturing -0.970 0.230 -0.160 0.120
77 Printing Manufacturing -1.020 0.230 -0.0600 0.0900
78 Service activities related to printing Manufacturing -0.500 0.360 0.120 0.180
79 Coke oven products Manufacturing -1.910 0.360 0.220 0.170
80 Rened petroleum products Manufacturing -0.360 0.390 -0.260 0.100
81 Processing of nuclear fuel Manufacturing -1.990 0.890 0.0100 0.130
82 Basic chemicals except fertilizers Manufacturing -0.520 0.210 0.0500 0.0600
83 Fertilizers and nitrogen compounds Manufacturing -0.890 0.370 0.100 0.0500
84 Plastics in primary forms; synthetic rubber Manufacturing -0.400 0.210 0.0100 0.0400
85 Pesticides and other agro-chemical products Manufacturing -0.390 0.200 -0.0800 0.0500
86 Paints varnishes printing ink and mastics Manufacturing -0.0400 0.200 -0.0300 0.0500
87 Pharmaceuticals medicinal chemicals etc. Manufacturing -0.340 0.150 -0.110 0.0800
88 Soap cleaning and cosmetic preparations Manufacturing -0.720 0.150 -0.0700 0.0300
89 Other chemical products n.e.c. Manufacturing -0.570 0.140 -0.0600 0.0400
90 Man-made bres Manufacturing 0.760 0.570 -0.160 0.0700
91 Rubber tyres and tubes Manufacturing -0.610 0.240 -0.0600 0.140
92 Other rubber products Manufacturing -1.070 0.290 -0.0200 0.0400
93 Plastic products Manufacturing -0.340 0.120 -0.0400 0.0400
94 Glass and glass products Manufacturing -0.310 0.120 -0.0300 0.0600
95 Pottery china and earthenware Manufacturing -1.090 0.250 -0.0600 0.0400
96 Refractory ceramic products Manufacturing -0.180 0.340 -0.0500 0.0900
97 Struct.non-refractory clay; ceramic products Manufacturing -0.560 0.230 0.170 0.100
98 Cement lime and plaster Manufacturing -1.320 0.470 -0.260 0.110
99 Articles of concrete cement and plaster Manufacturing -0.320 0.250 -0.300 0.100
100 Cutting shaping and nishing of stone Manufacturing -0.930 0.370 0.0300 0.0600
101 Other non-metallic mineral products n.e.c. Manufacturing -0.120 0.140 -0.200 0.0500
102 Basic iron and steel Manufacturing -0.800 0.140 -0.0400 0.0700
103 Basic precious and non-ferrous metals Manufacturing -0.140 0.290 -0.140 0.0700
104 Structural metal products Manufacturing 0 0.140 -0.0600 0.0600
105 Tanks reservoirs and containers of metal Manufacturing -0.560 0.190 0.0300 0.0700
106 Steam generators Manufacturing -1.360 0.330 -0.290 0.130
107 Cutlery hand tools and general hardware Manufacturing -0.660 0.120 -0.150 0.0800
108 Other fabricated metal products n.e.c. Manufacturing -0.240 0.0900 0.0800 0.0700
109 Engines and turbines (not for transport equipment) Manufacturing -1.420 0.610 0.0800 0.0700
110 Pumps compressors taps and valves Manufacturing -0.630 0.290 -0.250 0.0600
Continued on next page

32
 ID Industry Description Broad Sector Complete Partial
Estim. Std.Err. Estim. Std.Err.
111 Bearings gears gearing and driving elements Manufacturing -0.470 0.230 0.0900 0.0600
112 Ovens furnaces and furnace burners Manufacturing -0.190 0.240 -0.140 0.0800
113 Lifting and handling equipment Manufacturing -0.210 0.380 -0.0200 0.0600
114 Other general purpose machinery Manufacturing -0.660 0.180 -0.200 0.0500
115 Agricultural and forestry machinery Manufacturing 0.0100 0.190 0.0400 0.0400
116 Machine tools Manufacturing 0 0.160 -0.120 0.0600
117 Machinery for metallurgy Manufacturing -0.790 0.380 0.120 0.180
118 Machinery for mining and construction Manufacturing -2.070 0.560 -0.140 0.0400
119 Food/beverage/tobacco processing machinery Manufacturing -0.0300 0.300 -0.0600 0.0500
120 Machinery for textile apparel and leather Manufacturing -0.780 0.270 -0.220 0.110
121 Weapons and ammunition Manufacturing -0.660 0.590 -0.130 0.110
122 Other special purpose machinery Manufacturing -0.620 0.280 0.0200 0.0900
123 Domestic appliances n.e.c. Manufacturing -0.580 0.150 -0.120 0.0700
124 Oce accounting and computing machinery Manufacturing -0.590 0.160 -0.0600 0.0800
125 Electric motors generators and transformers Manufacturing -1.200 0.250 -0.140 0.0600
126 Electricity distribution and control apparatus Manufacturing -0.660 0.140 -0.110 0.0400
127 Insulated wire and cable Manufacturing -0.520 0.390 0.0900 0.0600
128 Accumulators primary cells and batteries Manufacturing -0.350 0.220 -0.350 0.100
129 Lighting equipment and electric lamps Manufacturing 0.410 0.210 -0.280 0.0800
130 Other electrical equipment n.e.c. Manufacturing -0.480 0.120 0.0800 0.100
131 Electronic valves tubes etc. Manufacturing -0.530 0.200 -0.110 0.120
132 TV/radio transmitters; line comm. apparatus Manufacturing -0.310 0.200 -0.180 0.0800
133 TV and radio receivers and associated goods Manufacturing -0.580 0.260 -0.160 0.0500
134 Medical surgical and orthopaedic equipment Manufacturing -0.220 0.120 -0.0700 0.0400
135 Measuring/testing/navigating appliances etc. Manufacturing -0.210 0.210 -0.100 0.0400
136 Optical instruments and photographic equipment Manufacturing -0.280 0.270 0.0400 0.0500
137 Watches and clocks Manufacturing 0.680 0.230 -0.170 0.0900
138 Motor vehicles Manufacturing -0.280 0.220 0.0900 0.0800
139 Automobile bodies trailers and semi-trailers Manufacturing 0.540 0.190 0.0300 0.0700
140 Parts/accessories for automobiles Manufacturing 0.130 0.270 0.160 0.0700
141 Building and repairing of ships Manufacturing 0.220 0.560 -0.190 0.150
142 Building/repairing of pleasure/sport. boats Manufacturing -0.540 0.340 -0.0700 0.0700
143 Railway/tramway locomotives and rolling stock Manufacturing -1.250 0.320 0.120 0.0900
144 Aircraft and spacecraft Manufacturing 0.140 0.400 0.190 0.0600
145 Motorcycles Manufacturing -0.660 0.250 0.0600 0.0900
146 Bicycles and invalid carriages Manufacturing -0.400 0.290 0.0600 0.0900
147 Other transport equipment n.e.c. Manufacturing -0.620 0.330 -0.190 0.0700
148 Furniture Manufacturing -0.200 0.210 -0.0700 0.0600
149 Jewellery and related articles Manufacturing -2.290 0.850 0 0.0800
150 Musical instruments Manufacturing -0.470 0.350 0.0200 0.0500
151 Sports goods Manufacturing -0.770 0.170 0.0200 0.0800
152 Games and toys Manufacturing 0.190 0.400 -0.0100 0.110
153 Other manufacturing n.e.c. Manufacturing -0.450 0.250 -0.0400 0.0500
154 Manufacturing services on physical inputs Services 0 0 -0.480 0.380
155 Maintenance and repair services n.i.e. Services 0 0 -0.0400 0.160
156 Transport Services 1.520 0.560 -0.170 0.100
157 Travel Services -0.280 0.180 -0.130 0.0700
158 Construction Services 1.200 1.190 -0.100 0.250
159 Insurance and pension services Services 1.240 0.200 0.170 0.250
160 Financial services Services -0.320 0.670 -0.0400 0.110
161 Charges for use of intellectual property Services -0.630 0.570 0.0900 0.120
162 Telecom, computer, information services Services -0.220 0.140 -0.240 0.170
163 Other business services Services -0.500 0.470 -0.180 0.0800
164 Heritage and recreational services Services 0 0 -1.560 0.210
165 Health services Services 0 0 0.0700 0.140
166 Education services Services -1.570 0.260 0.0700 0.110
167 Government goods and services n.i.e. Services 1.450 0.460 0.100 0.100
168 Services not allocated Services 0 0 -0.170 0.330
169 Trade-related services Services 0 0 0.540 0.160
170 Other personal services Services 0 0 -1.300 0.280
Notes: This table reports estimates of the eects of `complete' and `partial' trade sanctions for each of the 170 industries in ITPD-
E-R02. Column (1) lists ITPD-E-R02 ID codes for each of the industries. The corresponding industry descriptions are in column
(2). Column (3) includes the broad sectoral descriptions. Columns (4) and (5) report the estimates of the eects of `complete' trade
sanctions and their corresponding standard errors, respectively. Finally, columns (6) and (7) report the estimates of the eects of
`partial trade sanctions' and their corresponding standard errors, respectively. All estimates are obtained with the PPML estimator
from the same econometric model (2). See text for further details.

Our results appear in Table 5 and can be visualized with the help of Fig. 6. These results

are obtained from our main specication from column (3) of Table 3 but for each of the 170

sectors in the ITPD-E. For brevity and clarity of exposition, we only report the estimates

for complete and partial trade sanctions. Based on these estimates, we draw the following

conclusions. Complete trade sanctions have been eective in reducing trade at the sectoral

level. This is supported by the fact that 86% of the estimates on complete trade sanctions

33
 Figure 6: Sectoral Sanction Estimates

A. Complete Sanctions B. Partial Sanctions

4

1
Sanction Gravity Estimate

Sanction Gravity Estimate

.5
0 2

-.5 0
-2 -4

-1
5 15 25 35 45 55 65 75 85 95 105 115 125 135 145 155 5 15 25 35 45 55 65 75 85 95 105 115 125 135 145 155
Sector ID: Ranked by Estimate Sector ID: Ranked by Estimate

Estimate Lower 95 CL Upper 95 CL Estimate Lower 95 CL Upper 95 CL

Notes: The two panels of this gure display the estimates from Table 5. Panel A plots the estimates (together with the
corresponding 95% condence limits (CL)) for `complete' trade sanctions. Panel B plots the estimates (together with the
corresponding 95% condence limits (CL)) for `partial' trade sanctions. See text for further details.

are negative and most of them are statistically signicant. We also nd that partial sanctions

have had signicant impact on some sectors as 65% percent of them are negative and 39 of

those are statistically signicant.

Turning to the variation across sectors, we nd that the strongest eects of complete

trade sanctions are for Mining & Energy, followed by Agriculture, and Manufacturing. In-

terestingly, the average eects of complete trade sanctions are positive for services, where

the only negative and signicant estimate we obtain is for `Education Services'. The aver-

age eects of partial sanctions are negative across broad sectors with the largest impact in

services trade.

Overall, the estimates in this section reveal that sanctions have decreased bilateral trade

between senders and targets at the sectoral level. Nonetheless, we see several opportuni-

ties for renement and further contribution. For example, we obtain about 4% of positive

and signicant estimates of complete trade sanctions and about 5% of positive estimates

of partial trade sanctions. While these are relatively small numbers, a closer look at the

individual sectors may generate interesting insights. Thus, for example, we note that half

of the positive and signicant estimates of complete trade sanctions are for service sectors.

34
Given the increased importance of both sanctions and trade in services, we believe a thor-

ough investigation of the impact of sanctions on services trade is an important direction for

future research.

5 Conclusion
In this study we illustrate that sanctions policies can be eective in the sense that they

signicantly reduce trade with the target. Importantly, we also show that the impact of trade

sanctions is very heterogeneous depending on whether sanctions are complete or partial, and

whether they are imposed on the exports or on the imports of the target country. They also

depend on their objectives and the time of their imposition. The specic sanctions on Iran

we have analyzed, illustrate that their direct/partial equilibrium eects on bilateral trade

dier signicantly across countries and depending on the direction of trade, even within

the European Union. We complement the aggregate partial sanction estimates with sectoral

analysis, which conrm that sanctions have been successful in decreasing trade at the sectoral

level and also reveal signicant heterogeneity in the eects of complete and partial sanctions

across a large number of industries.

From a methodological perspective, we demonstrate how the latest developments in the

quantitative trade literature helps analyze not only the average eects of sanctions but also

the specic sanction eects on individual sectors, countries, pairs of senders and targets, and

the direction of trade ows. Our research unveils many opportunities for further analysis

across dierent dimensions. An important message stemming from our analysis is that the

average estimates of the eects of sanctions may hide signicant heterogeneity, thus rendering

sanctions ineective while in fact they may have very signicant economic costs and impact.

Through their imposition of various restrictions on trade, nancial transactions and travel,

sanctioning countries aim to induce sanctioned countries to comply with policy requests by

raising the economic costs of noncompliance. According to this logic, the eectiveness of

35
sanctions policies crucially depends on the strength of their adverse eects on sanctioned

countries. Thus, by estimating the heterogeneous eects of sanctions on trade, we hope that

we have contributed to the proper quantication of the costs of economic sanctions and,

indirectly, to an improved evaluation of the sanctions success and eectiveness.

References
Afesorgbor, S., The Impact of Economic Sanctions on International Trade: How Do
Threatened Sanctions Compare with Imposed Sanctions?, European Journal of Political
Economy, 06 2018.

Ahn, P. Daniel and Rodney D. Ludema, The Sword and the Shield: The Economics
of Targeted Sanctions, Manuscript, 2018.

Aichele, Rahel and Inga Heiland, Where is the Value Added? Trade Liberalization and
Production Networks, CESifo Working Paper No. 6026, 2016.

Anderson, James E. and Eric van Wincoop, Gravity with Gravitas: A Solution to
the Border Puzzle, American Economic Review, 2003, 93 (1), 170192.

Arkolakis, Costas, Arnaud Costinot, and Andrés Rodríguez-Clare, New Trade

Models, Same Old Gains?, American Economic Review, 2012, 102 (1), 94130.

Baier, Scott L. and Jerey H. Bergstrand, Do Free Trade Agreements Actually In-
crease Members' International Trade?, Journal of International Economics, 2007, 71 (1),
7295.

Baldwin, D.A., Economic Statecraft, Princeton University Press, Princeton, 1985.

, The Sanctions Debate and the Logic of Choice, International Security, 1999, 24(3),
80107.

Besedes, Tibor, Stefan Goldbach, and Volker Nitsch, Cheap Talk? Financial Sanc-
tions and Non-Financial Activity, Manuscript, 2018.

Borchert, I., M Larch, S. Shikher, and Y. V. Yotov, The International Trade and
Production Database for Estimation - Release 2 (ITPD-E-R02), Working Paper 2022-07-
A, USITC 2022.

Caliendo, Lorenzo and Fernando Parro, Estimates of the Trade and Welfare Eects
of NAFTA, Review of Economic Studies, 2015, 82 (1), 144.

Caruso, R., The Impact of International Economic Sanctions on Trade - An empirical

Analysis, University Library of Munich, Germany, 2003.

36
Cheng, I-Hui and Howard J. Wall, Controlling for Heterogeneity in Gravity Models of
Trade and Integration, Federal Reserve Bank of St. Louis Review, 2005, 87 (1), 4963.

Crozet, M. and J. Hinz, Friendly re - the trade impact of the Russia sanctions and
counter-sanctions, Kiel Institute for the World Economy (IfW), 2017.

Dai, Mian, Gabriel Felbermayr, Aleksandra Kirilakha, Constantinos Syropoulos,

Erdal Yalcin, and Yoto V. Yotov, Timing the Impact of Sanctions on Trade, in the
Research Handbook on Economic Sanctions, Edited by Peter A.G. van Bergeijk, 2021.

Demena, B., A. Alemayehu, S. Reta, B Jativa, P. Kimararungu, and P.. van

Bergeijk, Publication bias of economic sanctions research: a meta-analysis of the im-
pact of trade linkage, duration and prior relations on sanctions success, in the Research
Handbook on Economic Sanctions, Edited by Peter A.G. van Bergeijk, 2021.

Draca, M., J. Garred, N.Warrinnier, and L. Stickland, On Target? The Incidence
of Sanctions Across Listed Firms in Iran, University of Warwick Working Paper, 2017.

Drezner, D.W., The Sanctions Paradox: Economic Statecraft and International Rela-
tions, Cambridge University Press, 1999.

Eaton, J. and M. Engers, Sanctions, Journal of Political Economy, 1992, 100(5), 899
928.

and , Sanctions: Some Simple Analytics, American Economic Review, 1999, 89(2),
404414.

Egger, Peter H., Mario Larch, and Yoto V. Yotov,  Gravity Estimations with Interval
Data: Revisiting the Impact of Free Trade Agreements, Economica, January 2022, 89
(353), 4461.

Felbermayr, G., Aleksandra Kirilakha, C. Syropoulos, E. Yalcin, and Y.V. Yotov,

The Global Sanctions Data Base, School of Economics Working Paper Series 2020-2,
LeBow College of Business, Drexel University., 2020.

Felbermayr, Gabriel, Aleksandra Kirilakha, Constantinos Syropoulos, Erdal Yal-

cin, and Yoto V. Yotov,  The global sanctions data base, European Economic Review,
2020, 129 (C).

Findlay, R. and K.H. O'Rourke, Power and Plenty: Trade, War, and the World Econ-
omy in the Second Millennium, Princeton University, 2009.

Gurevich, Tamara and Peter Herman, The Dynamic Gravity Dataset: 1948-2016,
2018. USITC Working Paper 2018-02-A.

Haidar, J., Sanctions and Export Deection: Evidence from Iran, Economic Policy, 2017,
32(90).

Hall, C. Michael and Siamak Sey, Tourism and sanctions, in the Research Handbook
on Economic Sanctions, Edited by Peter A.G. van Bergeijk, 2021.

37
Hufbauer, G. and B. Oegg, The Impact of Economic Sanctions on US trade: Andrew
Rose's Gravity Model, Peterson Institute for International Economics, 2003.

Hufbauer, G. C., J. J. Schott, K. A. Elliott, and B. Oegg, Economic Sanctions

Reconsidered, (3rd edition). Washington, DC: Peterson Institute for International Eco-
nomics., 2007.

Hufbauer, Gary Clyde and Euijin Jung, What's new in economic sanctions?, Euro-
pean economic review, 2020, 130, 103572.

Kaempfer, W.H. and A.D. Lowenberg, The Political Economy of Economic Sanctions,
Handbook of Defense Economics, edited by T. Sandler and K. Hartley, 2007, 2.

Kirilakha, Aleksandra, On the determinants of the success and eectiveness of sanctions,
Manuscript, Drexel University, 2022.

, Gabriel Felbermayr, Constantinos Syropoulos, Erdal Yalcin, and Yoto V.

Yotov,  The Global Sanctions Data Base: An Update to Include the Years of the Trump
Presidency, in the Research Handbook on Economic Sanctions, Edited by Peter A.G. van
Bergeijk, 2021.

Larch, Mario, Je Luckstead, and Yoto Yotov,  Economic Sanctions and Agricultural
Trade, School of Economics Working Paper Series 2021-16, Drexel University 2021.

©
, Josà -Antonio Monteiro, Roberta Piermartini, and Yoto Yotov,
Eects of GATT/WTO Membership on Trade: They are Positive and Large After All,
 On the

School of Economics Working Paper Series 2019-4, Drexel University May 2019.

, Serge Shikher, Constantinos Syropoulos, and Yoto V. Yotov,  Quantifying the

impact of economic sanctions on international trade in the energy and mining sectors,
Economic Inquiry, 2022, 60 (3), 10381063.

Miromanova, A., Quantifying the Trade Reducing Eect of Embargoes: Firm Level Evi-
dence from Russia, Unpublished Manuscript, 2019.

Morgan, T. Clifton, Constantinos Syropoulos, and Yoto V. Yotov, Economic sanc-

tions, Manuscript, 2022.

, Navin Bapat, and Yoshiharu Kobayashi, Threat and imposition of economic sanc-
tions 1945-2005: Updating the TIES dataset, Conict Management and Peace Science,
2014, 31 (5), 541558.

Morgan, T.C. and V. L. Schwebach, Fools Suer Gladly: The Use of Economic Sanc-
tions in International Crises, International Studies Quarterly, 1997, 41, 2750.

Olivero, MarÃa PÃa and Yoto V. Yotov, Dynamic Gravity: Endogenous Country Size
and Asset Accumulation, Canadian Journal of Economics, 2012, 45 (1), 6492.

Pape, R.A., Why Economic Sanctions Do Not Work, International Security, 1997, 22(2),
90136.

38
 , Why Economic Sanctions still Do Not Work, International Security, 1998, 21(1), 6677.

Rose, Andrew K., Do We Really Know That the WTO Increases Trade?, American
Economic Review, 2004, 94 (1), 98114.

Santos Silva, João M.C. and Silvana Tenreyro, The Log of Gravity, Review of Eco-
nomics and Statistics, 2006, 88 (4), 641658.

Yang, J., H. Askari, J. Forrer, and H. Teegen, US Economic Sanctions: An Empirical
Study, The International Trade Journal, 2004, 18 (1), 2362.

Yotov, Yoto V., Roberta Piermartini, Jose-Antonio Monteiro, and Mario Larch,
An Advanced Guide to Trade Policy Analysis: The Structural Gravity Model, Geneva:
UNCTAD and WTO, 2016.

39
Appendix
Classication of regions based on UN geoscheme
Africa: Algeria, Angola, Benin, Botswana, Burkina Faso, Burundi, Cabo Verde, Cameroon, Central African
Republic, Chad, Comoros, Cote d'Ivoire, DR Congo, Djibouti, Egypt, Eritrea, Ethiopia, Equatorial Guinea,
Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Liberia, Libya, Madagascar, Malawi, Mali,
Mauritania, Mauritius, Mayotte, Morocco, Mozambique, Namibia, Niger, Nigeria, Reunion, Republic Congo,
Rwanda, Saint Helena, Ascension and Tristan da Cunha, Sao Tome and Principe, Senegal, Seychelles, Sierra
Leone, Somalia, South Sudan, South Africa, Sudan, Swaziland, Tanzania, Togo, Tunisia, Uganda, Western
Sahara, Zambia, Zimbabwe.

Northern America: Bermuda, Canada, Greenland, Saint Pierre and Miquelon, United States of America.

Central America: Belize, Costa Rica, Clipperton Island, El Salvador, Guatemala, Honduras, Mexico,
Nicaragua, Panama, Caribbean Anguilla, Antigua and Bermuda, Aruba, Bahamas ,Barbados, Bonaire, Sint
Eustatius and Saba, British Virgin Islands, Cayman Islands, Cuba, Curacao, Dominica, Dominican Re-
public, Grenada, Guadeloupe, Haiti, Jamaica, Marinique, Montserrat, Navassa Island, Puerto Rico, Saint-
Barthelemy, Saint Kitts and Nevis, Saint Lucia, Saint Martin, Saint Vincent and the Grenadines, Sint
Maarten, Trinidad and Tobago, Turks and Caicos Islands, United States Virgin Islands.

Southern America: Argentina, Bolivia, Bouvet Island, Brazil, Chile, Colombia, Ecuador, Falkland Islands,
French Guiana, Guayana, Paraguay, Peru, Suriname, Uruguay, Venezuela.

Northwestern Europe: Austria, Belgium, Bulgaria, Czech Republic, Denmark, Germany, Estonia, Faroe
Island, Finland, France, Germany (Federal Republic), Guernsey, Hungary, Iceland, Isle of Man, Jersey,
Latvia, Lichtenstein, Lithuania, Luxembourg, Monaco, Netherlands, Norway, Poland, Republic of Ireland,
Romania, Slovakia, Sweden, Switzerland, United Kingdom.

Southern Europe: Albania, Andorra, Bosnia and Herzegovina, Croatia, Gibraltar, Greece, Italy, Republic
of Macedonia, Malta, Montenegro, Portugal, San Marino, Serbia, Kosovo, Slovenia, Spain, Vatican.
Eastern Europe: Belarus, Republic of Moldova, Russian Federation, Ukraine.

Western Asia: Armenia, Azerbaijan, Bahrain, Cyprus, Georgia, Iraq, Israel, Jordan, Kuwait, Lebanon,
Oman, Qatar, Saudi Arabia, State of Palestine, Syria, Turkey, United Arab Emirates, Yemen.

Central Asia: Kazakhstan, Kyrgystan, Tajikistan, Turkmenistan, Uzbekistan.

Southern Asia: Afghanistan, Bangladesh, Bhutan, India, Iran, Maldives, Nepal, Pakistan, Sri Lanka.

Southeastern Asia: Brunei Darussalam, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines,
Singapore, Thailand, Timor-Leste, Vietnam.

Eastern Asia: China, Taiwan, Hong Kong, Japan, Macau, Mongolia, DPR Korea, Republic of Korea.

Oceania, Australia: Christmas Island, Cocos (Keeling) Island, New Zealand, Norfolk Island, Fiji, New
Caledonia, Papua New Guinea, Solomon Islands, Vanuatu, Guam, Kiribati, Marshall Islands, Micronesia,
Nauru, Northern Mariana Islands, Palau, American Samoa, Cook Islands, French Polynesia, Niue, Pitcairn
Islands, Samoa, Tokelau, Tonga, Tuvalu, Wallis and Futuna.

40
 Supplementary Web-Appendix
(not for publication)
for the article

On the Heterogeneous Eects of Sanctions on Trade

General Equilibrium Eects on Trade and Welfare

Using a new, comprehensive sanctions database and capitalizing on the latest developments
in the structural gravity literature, we estimate the eects of economic sanctions on inter-
national trade. In this supplementary appendix we provide a standard general equilibrium
analysis to illustrate how partial empirical estimates can be utilized to assess welfare eects
resulting from sanction policies.
We quantify the general equilibrium eects of the sanctions on Iran on trade, real GDP,
and sectoral value added. To this end, we rely on the gravity framework of Aichele and
Heiland (2016), who calibrate a multi-sector version of the gravity model presented in Section
3, which, following Caliendo and Parro (2015), also includes intermediate linkages for 130
countries and 57 sectors using the GTAP data.
34 The baseline year for the analysis is 2014
and the counterfactual experiment employs our pair-and-direction-specic estimates from
the previous section to simulate a hypothetical world in the absence of the sanctions on Iran.
Since our empirical estimates relate only to the goods market and do not vary across sectors,
in the counterfactual simulations, we assume a uniform shock on trade costs for all goods
sectors and a uniform trade elasticity of 3.5 (the average value across sectors from Aichele
and Heiland (2016). The sectoral trade cost shocks relate the counterfactual level of iceberg
τ̂i,j ≡ τi,j /τi,j .35
0 0
trade costs τi,j to their baseline levels τi,j such that
First we describe the eects on bilateral trade. The estimates in Table 1 are constructed
0
as the ratio of counterfactual to baseline expenditure shares π̂i,j ≡ πi,j /πi,j .
For brevity and clarity of exposition, we focus on four countries: Iran, Germany, USA
0
and China. For i = j = {IRN }, πIRN,IRN = 63.8 and πIRN,IRN = 68.1, implying π̂IRN,IRN =
0.94, which means that undoing the sanctions makes the country substantially more open to
international trade. The share of German expenditure that falls on Iranian goods would go
up from 0.01 to 0.19 percent, so that π̂ = 27, whereas the share of US expenditure falling
on Iranian goods would increase from virtually zero to 0.06 percent, an increase by a factor

34 The model assumes perfect competition. Trade is Ricardian in the sense that countries purchase only
the cheapest available variety of each good. A key feature of the model is the intersectoral linkages, both
intra- and internationally. Production requires labor and a composite of inputs drawn from other sectors.
Wages are equalized across sectors so that sectoral variation results only from reallocation eects. While the
values of imports and exports need not coincide at the country level, the trade balance normalized by GDP
is assumed constant. See Aichele and Heiland (2016) and Caliendo and Parro (2015) for additional details.
35 For the pairs involving Iran, the average trade cost shock is 0.84; that is, in the counterfactual, trade
costs are 16 percent lower than in the baseline. The median is 0.79, the minimum of 0.24 is for US imports
from Iran; the maximum of 1.22 is for Iranian imports from the United Arab Emirates.

1
Table 1: Counterfactual Analysis: Trade Eects (% of Baseline)

Exporter Importer π0 π π 0 /π
IRN CHN 0.16 0.38 0.42
IRN DEU 0.19 0.01 26.89
IRN USA 0.06 0.00 165.71
IRN IRN 63.84 68.05 0.94

DEU CHN 0.21 0.21 1.00

DEU IRN 0.71 0.08 8.64
DEU USA 0.28 0.28 1.00
DEU DEU 50.41 50.51 1.00

USA CHN 1.03 1.03 1.00

USA DEU 1.54 1.56 0.99
USA IRN 0.81 0.03 28.75
USA USA 67.57 67.73 1.00

CHN DEU 1.94 1.94 1.00

CHN IRN 3.61 4.79 0.75
CHN USA 2.29 2.29 1.00
CHN CHN 72.51 72.45 1.00

Notes: This table reports the results of coun-

0
terfactual simulations. π /π relates counter-
factual (no sanctions) to baseline (sanctions)
equilibrium.

2
of 176. These trade changes are dramatic, but the new levels of trade are still very low.
Similarly, Iran's expenditure on German and US products goes up as well, by a factor of
9 and by a factor of 29, respectively. The share of Iran's spending on Chinese goods goes
down from 4.8 to 3.6 percent, reecting the reduction in trade diversion due to sanctions.
This loss of trade is even more pronounced for Chinese spending on Iranian goods, where
the expenditure share falls from 0.38 to 0.16 percent. For country pairs not directly aected
by sanctions, such as USA-Germany or Germany-China, expenditure shares do not change
signicantly.
Next, we turn to the eects of undoing sanctions on real income, our welfare measure.
The results are visualized in Fig. 1. The biggest winner of terminating sanctions, not
surprisingly, is Iran. It's real per capita income is predicted to rise by about 4.2 percent.
This might seem small; but the overall gains from trade in this quantitative model are
usually estimated to be rather small; see Costinot and Rodriguez-Clare (2015). So, it may
not be overly surprising that the termination of sanctions does not imply higher gains for
Iran. Also note that services sectors are aected only indirectly, as demand for their output
in manufacturing sectors may fall. Moreover, the sanctions have led to considerable trade
diversion, especially to China, so that Iran's overall openness has not been reduced by much
either. The country with the next largest welfare gain from the removal of the sanctions on
Iran is Armenia, a neighboring country to Iran. This is intuitive, since Armenia has dicult
political and economic relations with its other neighbors such as Azerbaijan and Turkey. So,
it relies overproportionately on trade with Iran, and lower income in Iran worsens Armenia's
terms of trade.
Top 15 winners from undoing sanctions
Figure 1: Counterfactual Analysis: Welfare Eects (% of baseline)

4,5 0,45
4 0,4
3,5 0,35
3 0,3
2,5 0,25
2 0,2
1,5 0,15
1 0,1
0,5 0,05
0 0
MLT
ARM
IRN

MDA

LKA
MNG

CYP

GRC
KEN
MWI

OMN
KHM
GEO

ZAF
KGZ

Note: Percent changes in real per capita income resulting from an end of sanctions against Iran. Only countries with largest
eects are shown: Iran (IRN), Armenia (ARM), Moldova (MDA), Malta (MLT), Sri Lanka (LKA), Mongolia (MNG), Malawi
(MWI), Kyrgizstan (KGZ), Georgia (GEO), Kenia (KEN), South Africa (ZAF), Cyprus (CYP), Cambodia (KHM), Oman
(OMN), Greece (GRC).

Amongst the 15 countries most positively aected form the removal of the Iran sanction,
are small nations such as Malta, Cyprus and Greece, which provide shipping services to Iran's

3
oil industry, and other geographically close countries such as Georgia or the centralasian
countries Mongolia and Kyrgyzstan. Gains for third countries are typically small: they lie
between about 0.4 percent for Armenia and 0.07 percent for Greece. The Western initiators
of the sanctions against Iran and the UN Security Council members, such as the USA,
France, and Great Britain, are barely aected; gains from undoing the sanctions lie below
0.03 percent of GDP; the UK is even predicted to lose a tiny amount. In contrast, the
normalization of trade relationships between these countries and Iran hurts those who have
benetted from trade diversion. The largest losses are predicted to occur in Korea, Panama
and the Ukraine. Oil producing countries also tend to lose from undoing the sanctions as
the additional supply of Iranian oil drives down the world price of oil.
Finally, we turn to the eects on sectoral value added, which appear in Fig. 2. Even
though, by design, the trade cost shocks are uniform across sectors, the sectoral structure of
comparative advantage of Iran and its trading partners implies a rich pattern of changes in
their terms of trade which, in turn, aect sectoral value added. Not surprisingly, the model

Figure 2: Counterfactual Analysis: Sectoral Value Added Eects (% of baseline)

40 1
IRN CHN DEU
30 0.75

20 0.5

10 0.25

0 0

-10 -0.25

-20 -0.5

-30 -0.75

-40 -1
Crops nec

Metals nec

Manufactures nec

Trade

Forestry

Minerals nec
Raw milk

Electricity

Sea transport
Air transport

Electronic equipment

Construction

PubAdmin/Defence/Health/Education

Chemical, rubber, plastic prods

Cattle, sheep, goats, horses

Sugar cane, sugar beet

Gas
Petroleum, coal products
Gas manufacture, distribution
Leather products

Textiles
Wool, silk-worm cocoons

Fishing

Dwellings
Dairy products

Plant-based fibers

Beverages and tobacco products

Wood products

Meat: cattle, sheep, goats, horses

Ferrous metals
Mineral products nec

Water

Wheat

Communication

Metal products

Motor vehicles and parts

Oil seeds
Recreation and other services

Paper products, publishing

Sugar
Vegetable oils and fats
Transport equipment nec

Vegetables, fruit, nuts

Transport nec

Paddy rice

Animal products nec

Food products nec

Business services nec

Financial services nec

Meat products nec

Cereal grains nec

Machinery and equipment nec

Processed rice
Insurance

Wearing apparel

Coal

Note: Percent changes in sectoral value added resulting from an end of sanctions against Iran in Iran (IRN), China (CHN), and
Germany (DEU).

predicts that an end to the sanctions would benet Iran's oil and gas sectors most strongly.
Value added would rise most in this area, reaching almost 40 percent in the gas sector, where
substitution eects in the face of sanctions are dicult due to a rigid system of pipelines.
Sizable gains also occur in Iran's leather, textile, and transportation sectors. In the former
two, the country enjoys a comparative advantage relative to its trade partners; in the latter,
the eects are driven by additional demand for transportation services as trade picks up.
Some positive eects are predicted in agri-food, in particular in nuts production. Iran is an
important producer of almonds. But those gains lie below 10 percent.
In contrast, losses are concentrated in comparative disadvantage sectors. Again, the
impact on agri-food looms large. Arid Iran is a net importer of vegetables, sugar and
rice, so that ending the sanctions drives up imports even more in these sectors leading to
their shrinkage. For many services sectors, which are not directly aected by sanctions as

4
modeled in our exercise, value added eects are very close to zero. The transportation sector
mentioned above is an interesting exception. Sectoral value-added eects are smaller by
almost two orders of magnitude in the US, China or Germany. They do not exhibit any
clear sectoral pattern, but they tend to be positive for Germany and negative for China,
reecting trade creation and trade diversion, respectively.