The Effects of Embargoes on International

Trade: Evidence From Russia ∗

Anna Miromanova ∗∗
University of Oregon
February 27, 2018
Comments are welcome

Abstract
This paper investigates the impact of the Russian food embargo, imposed in the
aftermath of the Crimean conflict of 2014, on Russian aggregate trade. I focus on
two effects the food embargo had on Russian trade: the main treatment effect (i.e.
the impact on trade of embargoed goods with the sanctioning countries) and the sub-
stitution effect (i.e. Russian imports of embargoed goods from the non-sanctioning
countries). Using monthly UN COMTRADE data and a gravity equation framework,
modified to accommodate the exogenous shock of embargo, I estimate a difference-in-
differences model of Russian trade flows. The main findings of the paper suggest a total
decrease in Russian imports of embargoed products from the sanctioning countries of
more than 80%. The estimated substitution effect points to an increase in imports
from non-sanctioning countries of about 40%. I also conduct an analysis of product
and country heterogeneity in an attempt to further separate the embargo effects by
product and country groups. The main findings suggest that the embargo’s impact
on consumption goods is significantly stronger than its impact on intermediate goods.
The OECD countries bear larger losses than ideologically close countries. Additionally,
I estimate the short and long run impacts of the embargo. The estimated effects are
stronger in the short run, due to the unexpected nature of the trade shock.

JEL: F14, F51, F55, R11

Keywords: International trade; Barriers to trade; Trade embargo; Sanctions; International
conflict; Import substitution

∗
I thank Anca Cristea, Woan Foong Wong, Bruce Blonigen, and the participants of Trade group and Micro
group at the University of Oregon for many helpful comments. All errors are my own.
∗∗
Contact information: amiroman@uoregon.edu; 419 Prince Lucien Campbell Hall, 1415 Kincaid St.,
Eugene, OR, 97401
1 Introduction

Trade policies, such as sanctions and embargoes have always been popular instruments
of foreign policy (United Nations Security Council (2017)). These tools are used to induce
behavioral changes in non-complying agents (countries) without military actions, which tend
to be extremely costly and involve human casualties. Understanding their impacts on trade,
welfare, growth and other economic outcomes of both sending and target countries is crucial
for a successful conflict resolution.
One of the most recent example of an enactment of these foreign policies is the economic
sanctions that were imposed on Russia in the aftermath of the Crimean conflict in 2014. The
political conflict escalated quickly due to several unpopular decisions made by the Russian
government and resulted in the imposition of economic and financial sanctions by the ma-
jority of the OECD and several European countries on Russian individuals and businesses.
The sanctions have been in place since March of 2014, and policy analysts predict no change
in the upcoming years. In August of 2014 Russia retaliated to the sanctions by imposing em-
bargo on a number of agricultural products from the sanctioning countries, including dairy
produce, meat, fish, fruits and vegetables.
This is the first incident in modern history when sanctions target a large market player
like Russia (unlike when sanctions were imposed on smaller countries like Iran, North Korea,
or South Africa). It is also the first time that the target country (Russia) retaliates by
imposing an embargo on the sending countries. These events provide an experimental set-
up, which allows for the analysis of the impacts of sanctions and embargoes on trade and
welfare1 . As preliminary evidence suggests, the embargo triggered a decrease in aggregate
Russian trade flows, as illustrated in Figure 1. In this paper, I concentrate on estimating
1
Even though the terms “sanctions” and “embargoes” are used interchangeably in the literature, it is im-
portant to distinguish between them for the purposes of this paper. The term “sanctions” refers to the
restrictive financial and economic measures imposed on Russia by several European countries, the US,
Canada and Australia in March 2014. The goal of sanctions is to coerce Russia into giving up control over
Crimean peninsula. The term “embargo” refers to the Russian embargo imposed in August of 2014 on the
sanctioning countries as a retaliation. Russia’s objective is to persuade the sanctioning countries to lift the
sanctions.

1
the impacts of the Russian food embargo on Russia’s aggregate trade flows.
Using monthly data on bilateral product-specific trade flows and the exogenous shock of
the embargo, I utilize a difference-in-differences estimation strategy to evaluate the impact
of the embargo on aggregate Russian trade. I concentrate on estimating the change in trade
flows of embargoed products from the sanctioning countries, and the substitution effect
created by the embargo - substitution towards the non-sanctioning countries in product
categories embargoed by Russia. Separation of the embargo impact into treatment and
substitution effects is a novel approach that allows for reduction of omitted variable bias2 . I
analyze the effects of the embargo from several perspectives. First, I conduct an analysis of
product and country heterogeneity in an attempt to separate the embargo effects by product
and country groups. I further explore which product class - consumption or intermediate
goods - was impacted the most by the imposition of the embargo. I also inspect which
countries experience the largest decrease in their trade with Russia due to the embargo.
Finally, I examine the dynamic effects of the embargo by distinguishing between short versus
long run effects.
The main findings of the paper suggest a total decrease in Russian imports of the embar-
goed products from the sanctioning countries of more than 80%. The estimated substitution
effect points to an increase in imports from non-sanctioning countries by about 40%. The
imports of non-embargoed goods from sanctioning countries also experience a decline, al-
though of a smaller magnitude that is not statistically significant. These results are robust
to a number of different tests - country and product heterogeneity, dynamics and falsification
exercise.
Switching focus to the export side, I do not find significant effects for Russian exports,
except for a decrease in exports of non-embargoed goods to the sanctioning countries. I
attribute this decrease to political motivations of the firms in sanctioning countries, who may
voluntarily refuse to continue business relationships with firms in Russia after the imposition
2
When estimating the embargo’s effect on trade flows, omitting the substitution effect leads to downward
bias in the main coefficient of interest - main treatment effect.

2
of the embargo. My findings are in agreement with the economic literature on the impacts
of trade policies on bilateral trade flows.
My research pertains to studies on the effectiveness of trade policies such as economic
sanctions (Eaton & Engers (1992), Eaton & Engers (1999), Bapat et al. (2013), Kaempfer
& Lowenberg (1988), Hufbauer & Schott (1985), van Bergeijk & van Marrewijk (1995)),
embargoes (Coulibaly (2009), Kuehnapfel (2015), Teegen et al. (2008), Irwin (2005)) and
boycotts3 (Heilmann (2016), Michaels & Zhi (2016), Chavis & Leslie (2009), Pandya &
Venkatesan (2016), Ashenfelter et al. (2007)). Papers on these topics explore a wide variety
of subjects, including theoretical modeling of sanctions, applied calculations of the amount
of lost trade due to sanctions and embargoes, case studies, etc. The intuitive conclusion that
embargoes, sanctions or boycotts hurt trade relations between countries, decreasing exports
and imports, is confirmed by some authors (Heilmann (2016), Michaels & Zhi (2016), Chavis
& Leslie (2009), Neuenkirch & Neumeier (2015)), while other papers find effects of trade bans
to be insignificant (Hufbauer & Schott (1985), Hufbauer et al. (1997)).
Currently there exists little research on the Russian sanctions and embargo, mostly due
to these events being fairly recent. The existing research can be divided in two groups: the
first group of studies concentrates on the effect of the sanctions and embargo on bilateral
trade flows between Russia and the sending countries, while the second group concentrates
on the macroeconomic effects of these trade policies.
From the first group, the most relevant study to my work is the working paper by Crozet &
Hinz (2016), which exploits French firm-level export data to study how the sending countries
were affected by the sanctions and the embargo. They find a large decrease in firms’ export
participation and the value exported to Russia. My results are in agreement with this
study, however I provide more robust estimates due to a higher data frequency (monthly vs.
annual) and a novel approach that separates the substitution effect from the main impact
3
Typically, boycotts are carried out by consumers against a certain brand or brands originating in the
embargoed country and are not officially enforced by the governments. Embargoes and sanctions, on the
other hand, are enforced by the imposing countries: they decide on what actions are included in the sanctions
packet and for how long they will be carried out.

3
of the embargo. Uzun & Loginova (2016) estimate that most sanctioning countries saw no
decline in the aggregate export trade flows, which they explain by the fact that the European
countries increase their exports to other countries to compensate for losses in the Russian
market. Boulanger et al. (2016) use a specific factors computable general equilibrium (CGE)
model to simulate the short-run impact of the Russian embargo. The results indicate that
Russia bears the highest income loss, while the EU recovers part of its lost trade through
expansion of exports to other markets. The studies of macroeconomic effects of the sanctions
and the embargo include Dreger et al. (2016),Kholodilin & Netsunajev (2016), Gurvich &
Prilepskiy (2015), Hinz & Monastyrenko (2016), Ahn & Ludema (2017). My work contributes
to the literature in the first group of studies by analyzing the effects of the Russian embargo
on Russian trade flows. To my knowledge, this is the only paper to examine this topic from
the perspective of Russia, bridging the existing gap in the economic literature.
Recent economic sanctions imposed on Russia and Russian retaliation embargo created
a significant disturbance (at least short run) in world trade flows4 . The goal of this paper is
to quantify these effects. The paper is organized as follows. Section 2 provides background
information on the imposition of the embargo; section 3 outlines the empirical model while
section 4 describes data sources used in the empirical analysis. Sections 5 and 6 describe
main empirical results for import and export trade flows, respectively, including robustness
checks. Section 7 concludes.

2 Sanctions Background

The imposition of economic and political sanctions against Russia was triggered by a
complicated political situation between Russia and Ukraine that evolved over the Russian
support of the separatist movements in Ukraine, and the Russian annexation of Ukrainian
Crimea.

4
Russia is one of the largest countries in the world, accounting for 1.77% of the world’s merchandise exports
and for 1.18% of the world’s merchandise imports in 2014 (World Trade Organization (2017)).

4
 In 2013, the President of Ukraine, Viktor Yanukovych, had a choice of signing an asso-
ciation agreement with the European Union or a gas loan bailout from Russia. Yanukovich
decided not to sign the pending EU agreement and instead chose to deepen ties with Russia,
which triggered protests, especially among young pro-EU Ukrainians. After a sharp esca-
lation of violence against protesters, the foreign ministers of France, Germany and Poland
brokered a deal between Yanukovych and Ukraine’s main opposition leaders on February
21st 2014 (Veebel & Markus (2015)). Yanukovych, however, fled to Russia, leaving the
opposition in control. This division of the country allowed Russia to start its annexation
of Crimea (formerly an autonomous republic in the Soviet Union, which was annexed to
Ukraine in 1954).
Russia’s actions sparked the biggest crisis in relations between Russia and the West
since the Cold War; as the crisis deepened, western governments, led by the European
Union and the United States, imposed economic sanctions against Russia. There were 4
tiers (or waves) of sanctions. The first and second tiers, imposed on March 17th 2014
(active till present), included asset freezes in foreign banks, visa bans for the individuals,
and interruption of any cooperation with the businesses who supported the annexation of
Crimea, both in Russia and Ukraine. These sanctioning practices are typical procedures
standardized by international organizations and are the first steps, which are taken when
a sanctioning process is initiated (United Nations Security Council (2017)). The third tier
of sanctions was imposed following the shooting down of flight MH17 on July 17th 2014
over the territories controlled by a pro-Russian separatist military group. This round of
measures expanded the previous sanctions by extending the group of individuals under asset
freezes and travel bans, and included restrictions on lending to Russian state banks, an arms
embargo, an export ban on oil technology and services that could be used for Arctic or deep-
sea drilling or shale oil projects, and an export ban on dual-use goods equipment, such as
specialist computers or heavy engineering vehicles that could be used for military purposes.
The fourth tier of sanctions refer to the Russian retaliation through a food embargo. On the

5
6th of August, 2014 Russia imposed a one-year ban on imports of a long list of agricultural
products from the sanctioning countries, including fruit, vegetables, flowers, fish, meat and
cheese. The embargo has been extended multiple times, and currently is in place until
the end of 20185 . Certain goods, like wine, spirits and dry goods (for example, pasta) are
not embargoed. The Russian government also allows the embargoed products to enter the
country if their intended use is as intermediate goods in the production of special diet foods
(eg. baby food, food for diabetic patients).
Prior to the Crimean conflict, Russia, being one of the leading suppliers of oil and gas
in the world economy, experienced a significant drop in growth due to the rapid decrease in
world oil prices at the end of 2013 - beginning of 2014. The oil price shock was caused by
a significant decrease in the oil demand by the United States. 40% of Russian total exports
and more than 50% of the budget revenues depend on oil and gas, while other industries
are not as developed. This makes the Russian economy especially vulnerable to volatile
oil prices, causing the exchange rate between the Russian ruble and foreign currencies to
increase significantly after the oil price shock. The unanticipated imposition of sanctions in
the spring of 2014 contributed to further increases in the exchange rates, which in turn led
to lower real disposable income, a collapse in government revenues, lower public spending
and increasing inflation. The Russian economy was in deep recession in 2015, characterized
by a negative GDP growth rate (-2.8%), and inflation of almost 15%6 . In 2016, the economy
performed better due to substitution of the trade flows away from embargoed countries and
due to several measures taken by the Russian Central Bank. However, the negative growth
persisted (-0.2% in 2016, according to Russian Statistics Committee).
The unanticipated economic sanctions and the Russian food embargo had a devastating
effect on the Russian economy, causing a deficit of production inputs, which in turn led to a
rapid inflation. Several Russian food markets significantly depend on imports. For example
about 50% of all dairy products or fruits and vegetables are imported, and almost 74% of
5
Decisions on sanctions’ extensions are decided collectively by the sanctioning countries every 6 months.
6
According to World Bank Indicators

6
pork is imported7 . Due to the imposition of the embargo, Russian producers had to increase
domestic production, and retail businesses had to search for new trade partners. Overall this
led to a significant decrease in the variety of goods. Domestically produced goods often suffer
from a lower quality relative to imported goods. According to the opinion poll, conducted
by the Russian Research Holding Romir, almost 33% of respondents remarked on the lower
quality of dairy products in December of 20158 . This could be due to the substitution of
high quality imported inputs with domestic inputs of lower quality (substitution of cow milk
with powdered milk).
It is difficult to say how much longer the sanctions and the embargo will stay in place
and what repercussions they will have for the world economy and global trade, but they have
already had a significant negative impact on Russian trade and economy, as I show in this
paper and several other studies clearly show (Dreger et al. (2016),Kholodilin & Netsunajev
(2016), Crozet & Hinz (2016)). Quantifying the effects of sanctions and embargo are a
difficult task that can provide helpful estimates of the costs of these economic measures,
which in turn allow for calculations of changes in welfare.

3 Empirical Model

In this section I describe the estimating equation used to analyze the embargo’s effect on
trade flows. The motivation for the choice of this particular specification is presented further.
I estimate a difference-in-differences OLS model to analyze the changes in (logged) Russian
trade flows (exports and imports) caused by the imposition of the Russian embargo. I also
compare the standard OLS coefficients to estimates obtained from estimation techniques
that account for zero trade flows (e.g. PPML).
The estimation equation is derived from the standard gravity model, which has been
widely used to estimate the responsiveness of trade flows to various economic factors. I
7
According to statistics provided by Central European Financial Observer https://financialobserver.
eu/cse-and-cis/russia/the-embargo-has-transformed-the-russian-food-market/
8
Detailed poll description is provided at http://romir.ru/studies/711_1443646800/

7
follow Anderson & Van Wincoop (2003) and Feyrer (2009) in setting up the basic gravity
relationship:
 1−σ
yit yjt τijt
T radeijt = (1)
yωt Pit Pjt

where T radeijt is bilateral trade between countries i and j at time t; yit , yjt and yωt denote
the incomes of the exporter country, the importer country, respectively the world at time t.
τijt stands for bilateral resistance term; Pit and Pjt denote the country-specific multilateral
resistance terms at time t. Taking logs of both sides of the equation (1), this can be re-written
as:

ln(T radeijt ) = ln(yit ) + ln(yjt ) − ln(yωt ) + (1 − σ) [ln(τijt ) − ln(Pit ) − ln(Pjt )] (2)

I transform equation (2) to arrive at my main estimation equation. Since the model
applies to Russia’s trade only, and Russia is always a trading partner, I drop subscript “i”to
simplify notation. The data allow me to add product dimension, which is denoted by the
subscript “k”.
The central hypothesis of my work is that the embargo that was imposed by Russia on
the sanctioning counties in August 2014 led to a significant decrease in Russian trade with
those countries in embargoed products, and an increase in trade with the non-sanctioning
countries. My identification strategy relies on the assumption that the imposition of the
embargo was an exogenous shock to the bilateral trade flows. This is a plausible assumption
because the imposition of the embargo was triggered by the shooting down of the flight
MH-17, which was an event that was not possible to predict.
I model Russia’s embargo as a bilateral trade friction (i.e., part of τijt ). To define the
embargo treatment variable of interest, it is worth noting that there are four groups of
country-product interactions in the data: sanctioning country trading in embargoed, re-
spectively in non-embargoed products, and non-sanctioning country trading in embargoed,
respectively in non-embargoed goods. In my model, I include dummy variables for each of

8
the first three groups to account for any changes in trade that takes place in groups that are
somewhat affected by the sanctions and the embargo. The omitted (i.e. reference) group
consists of the non-sanctioning countries trading in non-embargoed goods.
In order to account for multilateral resistance terms Pit , Pjt , and world income yωt , I
include country - year fixed effects, where I denote years by “y” subscript (since the data
is available monthly, I preserve the “t” index for the month-year periods). Log GDP of
the partner country is included to model yjt in equation (2). Using “j”, “t”, “y”, and
“k”subscripts for partner country, month-year time period, year, and product, respectively,
the main estimating equation takes the following form:

ln(T radejkt ) = α + γjk + γt + γjy + γky + β1 × emb treatmentjkt +

(3)
β2 × emb substitutionjkt + β3 × emb spilloverjkt + ln(GDPjt ) + ln(T radecostjt ) + ε

The country-product fixed effects (γjk ) account for the time-invariant determinants of
bilateral trade such as bilateral distance, common borders, and common language. The time
period fixed effects (γt ) account for any Russian macroeconomic factors that might affect
trade (e.g. inflation, currency movements, etc). It is important to account for these factors
because of a dramatic devaluation of the Russian ruble due to the oil shock that took place
in the beginning of 2014. The country-year fixed effects (γjy ) allow me to control for the
multilateral resistance terms, while the product-year fixed effects allow me to control for
product-specific trends. To deal with a large number of fixed effects, I use an estimator
described in Correia (2016). ln(T radecostjt ) is a time-varying trade cost variable in logs,
which is constructed in the following fashion:

ln(T radecostjt ) = ln(distancej ) × ln(oilpricet )

where distancej is the distance between Russia and the partner country j and oilpricet is

9
the price of barrel of oil at time t. This trade cost variable is a proxy for bilateral shipping
cost.
I will next describe the treatment dummy and cost variables in further detail.

1. emb treatmentjkt captures the main effect of the embargo on trade in embargoed products
with the sanctioning countries. This variable reflects whether the Russian food embargo
targets a particular product k imported from country j at time t. It is constructed as
follows:
emb treatmentjkt = Dcountry sanction × Dproduct embargo × Dtime

where Dcountry sanction is a dummy that takes value of 1 if the partner country is sanctioning
Russia, and 0 otherwise; Dproduct embargo is a dummy that takes value 1 if the product is
embargoed, and 0 otherwise; Dtime is a dummy that takes value of 1 in all the months
following August 2014 - the period when the embargo was imposed. My hypothesis is
that the coefficient β1 on embt reatment is negative, because the embargo was aimed to
reduce the embargoed goods’ flows in order to coerce the sanctioning countries to lift the
sanctions.

2. emb substitutionjkt measures the substitution effect of the embargo. This dummy variable
reflects how the trade flows in embargoed good k with the non-sanctioning country j at
time t were affected by the embargo. It is constructed as follows:

emb substitutionjkt = Dcountry nonsanction × Dproduct embargo × Dtime

where Dcountry nonsanction is a dummy that takes the value of 1 if the partner country is
not sanctioning Russia and 0 otherwise; Dproduct embargo and Dtime are as described above.
The coefficient β2 for the substitution effect should have a positive sign because I expect
Russia to import more embargoed products from the non-sanctioning countries.

3. emb spilloverjkt is a dummy variable that takes a value of 1 if a sanctioning country j

10
 trades in a non-embargoed good k at time t and 0 otherwise. The justification to include
this variable is the possibility that sanctioning countries voluntarily reduce trade with
Russia in other product categories as well. This variable is constructed as:

emb spilloverjkt = Dcountry sanction × Dproduct nonembargo × Dtime

where Dcountry sanction and Dtime are as defined above and Dproduct nonembargo is equal to 1
if a product is not embargoed, and 0 otherwise. My prior is that the coefficient β3 on
emb spillover will be negative, but not necessarily significant. The sanctioning countries
might stop exporting non-embargoed products as a result of the embargo due to either
political or logistical reasons. For instance, it could be the case that the firms in sanction-
ing countries that export non-embargoed goods into Russia decide to boycott Russia to
retaliate against the embargo. Alternatively, there may be economies of scale in shipping
and the firms that export embargoed goods might also export non-embargoed goods. If
the relative size of the latter in total exports is small, then the incentive to only trade in
non-embargoed goods with Russia is small, potentially interrupting trade altogether.

4 Data

To estimate the regression model described above, I use several data sources, including
the UN COMTRADE, CEPII Gravity and World Bank Global Economic Monitor databases.
The UN COMTRADE data are available at product (HS-4 digit) and monthly level over the
period January 2010 to December 2016. I analyze the impact of the Russian embargo on
both its exports and imports. The dataset on Russia’s imports is assembled from foreign
countries’ reported exports to Russia, while the Russian export dataset is assembled from
countries’ reported imports from Russia9 .

9
This approach is chosen because in the COMTRADE database Russia starts reporting trade at monthly
level starting in January 2012, while other countries’ reports are available earlier. In order to increase
the number of observations, I proceed using other countries’ reports. In UN COMTRADE, imports are

11
 There are 119 countries in my sample, of which 38 are sanctioning countries. There are
1240 HS-4 digit product codes, of which 48 are embargoed. I provide the list of sanctioning
countries and embargoed products in the Appendix, tables 1 and 2, respectively. Both
lists are compiled from the Russian laws and the Russian President’s Executive Orders,
which contain the detailed description of the countries and products to be embargoed. The
embargoed products are listed at the HS-4 digit level, which dictates the use of HS-4 digit
level COMTRADE data.
The CEPII (Centre d’Etudes Prospectives et d’Informations Internationales) Gravity
database provides information on the geographic distance between Russia and its trading
partners. It also provides the information on common language and common border, on trade
agreements, and other controls typically used as regressors in the gravity equation. Data
on the partner countries’ GDP are taken from the World Bank Global Economic Monitor
database, and are recorded on a quarterly basis. I use quarterly GDP values to construct a
measure of the average monthly GDP level of a country. Monthly time series data on the oil
price are provided by St. Louis FRED. I use the price of BRENT oil in US dollars for the
calculation of trade costs.
After combining all these data sources I obtain a panel dataset of Russia’s import and
export trade flows by country, by product and by month of transaction. The import sample
contains 1,587,216 observations. Table 1 reports the summary statistics.
It is worth pointing out that most of the sanctioning countries are members of EU,
which trade with Russia extensively due to the geographical proximity and long established
historical connections. This is among the main reasons why the trade share of the sanctioning
countries is 71.5% versus 28.5% for the non-sanctioning countries. The embargoed products
consist of different food groups, including fruit and vegetables, meat and dairy products.
The share of the embargoed products in Russia’s imports is 10.6% of the sample; on average
the embargoed products are cheaper than the non-embargoed products (log of embargoed

recorded cif (cost insurance and freight) while exports are fob (free on board).

12
good’s price is 0.968 versus 2.433 for the non-embargoed good). This is plausible considering
that only non-durable food items were embargoed, while the expensive, durable goods were
not affected by the embargo. On average, Russia imports 81 distinct HS-4 digit product
codes from the sanctioning countries and 48 from the non-sanctioning countries.

5 Results: Imports Sample

In this section I present the estimation results for the main regression equation (3). I
then analyze the effect’s dynamics and conduct several robustness exercises. I exploit the
variation in Russian import and export trade flows created by the exogenous shock of the
Russian embargo imposed in August of 2014 to identify its effect. The imposition of the
embargo was completely unanticipated by both Russian and foreign firms, thus the shock
was truly exogenous. I make an important identifying assumption that no other shock
happened at the same time targeting the same set of countries and commodities as the
embargo, and that my estimations are picking up solely the effect of the embargo. I first
analyze the embargo’s impact on Russian imports, since the embargo was imposed on the
imported goods, and I expect to see a much greater effect of the embargo on imports than
on exports. I briefly describe the effects on exports in the end of this section.

5.1 Main Specification

Table 2 provides the motivation for choosing the main estimating equation (3) as my
benchmark specification. Each column contains different sets of fixed effects together with
the three treatment dummy variables of interest; I compare the effects of the embargo across
these specifications. In all specifications, I account for country - product fixed effects; the
reference group for all of the specification consists of non-sanctioning countries trading in
non-embargoed products. In all of the specifications, the main coefficients of interest are:
1) β1 , which measures the impact of the embargo on the embargoed goods’ flows sourced

13
from the sanctioning countries; and 2) the substitution effect (emb substitutionjkt ), ex-
pressed through β2 , which measures Russia’s import substitution in embargoed products
from sanctioning to non-sanctioning countries. I also consider the effect of the embargo on
the sanctioning countries’ trade in non-embargoed goods (emb spilloverjkt ), β3 .
Column 1 of table 2 provides the estimation results for the least restrictive version of
the main estimating equation, in which I control for country-product and time period (i.e.,
month-year) fixed effects. The main results are in line with my hypothesis - the embargo
had a significant negative impact on the embargoed products’ flows from the sanctioning
countries; Russia substituted toward non-sanctioning countries in its imports of embargoed
goods. I also find evidence of a significant fall in Russian imports of non-embargoed goods
from the sanctioning countries. This specification, however, does not control for the multilat-
eral resistance terms (Anderson & Van Wincoop (2003)); I add country-year fixed effects in
the next specification to account for them. The results are presented in column 2. The main
treatment and substitution effects remain significant at 1 percent level and the magnitude
of both coefficients is similar to the results in column 1. The significance and magnitude of
the spillover effect goes down. The spillover effect is of a smaller magnitude in all of the
specifications, so adding additional controls decreases this effect.
The third column of table 2 presents the results from an even more restrictive model,
that includes product-year fixed effects in addition to country-product and country-year
fixed effects. In this specification, I control not only for multilateral resistance terms, but
also for product-specific time trends. The results are consistent with those in the other
specifications: the main effect is negative and significant, and of similar magnitude as in
previous two columns. The substitution effect is positive and significant. The non-embargo
effect loses significance in comparison to previous specifications. This could be due to the
fact that the embargo had a relatively smaller impact on this group of country-product
interactions, and adding both country-year and product-year trends removes the variation
needed to identify this effect.

14
 The last column of table 4 presents the results of the most restrictive specification,
in which I control for country-product, country-period and product-period fixed effects.
This specification does not allow me to estimate the spillover effect. This is likely due
to the fact that most sanctioning countries trade in non-embargoed goods (417,245 out of
528,756 observations of non-embargo product trade flows are attributed to the sanctioning
countries), so the product-period trends are highly collinear with the non-embargo dummy,
which makes it impossible to estimate this effect. We see the same pattern in column 3, (in
which I control for product-year trends) that the significance of the spillover effect decreases.
Country-period fixed effects in column 4 account for the log of partner country’s GDP and
for the trade cost, which vary by country and time period, so I exclude them from the model
to avoid multicollinearity.
The two effects of interest - the main treatment and substitution effects - have the
predicted signs in all of the specifications, and are of a similar significance and magnitude.
This provides confirmation that the embargo indeed had a significant impact on Russian
import flows. I choose as my benchmark model the specification in equation (3), which
accounts for country - product, period, country-year and product-year fixed effects. Its
estimation results are presented in column 3 of table 2. This model is the most conservative
specification, which allows me to estimate all of the effects of interest and account for most
of the unobserved variation. Thus, I proceed by interpreting the results of my benchmark
specification.
The embargo imposed by the Russian government on a number of products from the
sanctioning countries had a significant negative impact on trade with those countries. Im-
ports decrease on average by 82.5% (i.e., e−1.742 − 1 = −0.825 ). The substitution effect
is an increase of 57% (i.e., e0.452 − 1 = 0.57) in Russian imports of embargoed goods from
non-sanctioning countries. We cannot compare these effects directly because the mean trade
flows of the two country-product categories are different: the sanctioning countries’ average
logged trade flow in embargoed goods is 11.48, while for the non-sanctioning countries this

15
value is 10.66. Because non-sanctioning countries have smaller trade flows in embargoed
goods, the substitution effect does not fully mitigate the shock on embargoed goods’ trade
flows caused by the embargo.
The embargo had a negative effect on trade in non-embargoed goods with the sanction-
ing countries as well; however, this effect is insignificant under my benchmark specification.
Although Russia did not explicitly restrict the imports of the non-embargoed goods from
the sanctioning countries, trade within that group fell as well, but this fall is insignificant
relative to the two main effects. This fall could be due to several factors. Some firms in the
sanctioning countries might have made a decision to boycott the Russian market to ”pay
back” for the firms trading in the embargoed products. Another reason could be that these
firms export both embargoed and non-embargoed products to Russia, but the share of em-
bargoed goods in their exports is higher, so once the exports of the embargoed goods were
restricted, exports of the non-embargoed products ceased to be profitable. Imposition of the
embargo was a large unexpected shock to both Russian and foreign trading partners, which
led to a significant negative drop in Russian imports of the embargoed goods despite the
presence of a significant substitution effect.
Figure 2 illustrates the effect of the embargo on import trade flows in the embargoed
products from the sanctioning countries. The significant drop in Russian imports happens in
August 2014, precisely when the embargo was imposed by Russia and persists all the way to
the end of the sample. This estimate, as discussed previously, does not separate the effect of
the embargo from the substitution effect, which must be accounted for. The full impact of the
embargo, comprised of the main treatment and substitution effects, is depicted in Figure 3.
This figure demonstrates a significant drop in the embargoed commodities’ imports from the
sanctioning countries (main effect) and a smaller increase in Russian imports of embargoed
goods from the non-sanctioning countries in August 2014, the month when Russian embargo
was imposed. One might notice a negative trend in July 2014, a month preceding the
imposition of the embargo. This decline is barely significant at the 10% significance level and

16
could be due to the anticipatory effects of the market because the sanctions were intensified
in July 2014 after the shooting down of the flight MH-17. However, while analyzing news
reports from July 2014, I did not find any evidence or statements by the Russian government
about the upcoming imposition of the embargo. This decline might be indicative of the fact
that some firms in the sanctioning countries wanted to boycott Russian government’s actions
in Ukraine which resulted in civilian casualties. This explanation is, however, speculative10 .
It is worth pointing out that my main dataset includes the observations on HS-4 level
product codes, aggregated from HS-6 level. The HS-6 level has further disaggregations
that are not reported in COMTRADE (HS-6 is the highest level of disaggregation reported).
These further HS classification levels might include the products that Russian government did
not embargo specifically, but are reported as embargoed due to the Russian laws containing
only information on HS-4 level. This may explain why trade does not go to zero in all
embargoed HS-4 product categories. Unfortunately, my data does not allow me to distinguish
between those observations and thus, the effects I find are the upper bound estimate. Further
disaggregation of the goods in the dataset might provide lower estimates.

5.2 Product and Country Heterogeneity

In this section I explore which product class - consumption or intermediate - was im-
pacted the most by the imposition of the embargo. I also look at how exclusion of oil and gas
products from the dataset affects my estimates - in other words, I check whether assuming
monopolistic competition biases the estimates. I explain further in this section why it is
important to check whether this assumption holds. I conclude this section by inspecting
which countries lose the most from the embargo. These results are presented in tables 3

10
To ensure that this negative trend does not impact my estimation results significantly, I generate a new time
indicator and expand the embargo’s time to include July 2014. I then generate the treatment dummies as
described in section 3 and re-estimate the benchmark model. These results are presented in the appendix
table A3. The main effects carry the expected signs and are highly significant. The main effect has a lower
magnitude (a decrease of 72% versus 82% in the benchmark specification), while the substitution effect is
higher that in the benchmark model. These differences can be interpreted as an evidence of the presence
of an upward bias when including July in the treatment time variable.

17
and 4. I first analyze product heterogeneity results, which are presented in table 3, then I
describe the country heterogeneity results and present them in table 4.
Column 1 of table 3 reproduces the results of the benchmark specification (table 2
column 3) to enable comparison. The dataset used to derive the results in column 2 consists
of observations on the HS-4 level product code disaggregated by Broad Economic Categories
classification (BEC). I begin with the dataset with HS-6 digit product code level and merge
it with the BEC conversion dataset available from the UN COMTRADE. Each BEC code
belongs to one of the three categories - consumption, intermediate or capital goods. As a
result, I create an HS-6 digit code level product dataset, where each observation belongs
to one of the three BEC categories. I then proceed by aggregating the dataset to the HS-
4 digit product code by country and product. Each observation in this dataset is a HS-4
digit product code with BEC classification attached. In some cases there can be multiple
BEC classifications within a single HS-4 code. I estimate my benchmark specification using
this new dataset; the results are recorded in column 2 in order to enable the comparison of
heterogeneity analysis.
The results of the benchmark specification obtained from both datasets (i.e., main
sample and BEC-classification sample) are of similar magnitude (82.5% decline in column 1
versus 80% in column 2). The statistical difference between these estimates is insignificant,
so I proceed to use the estimates in column 2 for comparison of the heterogeneity results.
Because no capital goods were embargoed, I can estimate the embargo effect for consumption
and intermediate goods only. I analyze the embargo effect on the intermediate goods, which
makes the consumption goods my reference group in this specification. To analyze the
magnitude of the embargo effects on different groups of goods, I include an interaction term
between the main treatment effect (emb treatmentjkt ) and an indicator variable for the
intermediate goods. The same procedure is done for the other two treatment variables -
substitution and non-embargo effects. The results are recorded in column 3 of table 3.
I proceed to describe the embargo’s effects on consumption and intermediate goods for

18
each country-product group. The embargo’s impact on consumption goods is equal to βi ,
i = 1, 2, 3 or eβi − 1 if translated into a percentage change. The effect on intermediate
goods is the summation of the main coefficient and the coefficient on the interaction term:
βi + βi × Dintermediate .
The embargo’s impact on consumption goods imported from sanctioning countries is the
largest of all three effects, -1.708 or a decline of 82%. This is not surprising, considering that
the imports of embargoed goods (mostly food products) from the sanctioning countries were
completely banned in August 2014, and the largest share of the embargoed goods (78%) are
classified as consumption class. Imports of intermediate goods experienced a smaller drop as
a result of the embargo of 46% (e(−1.708+1.094) − 1 = −0.459). This effect if significant at the
10% significance level11 . Smaller drop in intermediate goods’ imports may be attributed to a
smaller share of those goods in the sample. Another reason for the smaller effect could be due
to the fact that some embargoed goods from the sanctioning countries were allowed to cross
the Russian border if they were intended to be used for production of special diet foods (eg.
baby food, food for diabetic patients). Another reason for a smaller effect on intermediate
goods might be due to the fact that the embargo could be less strongly enforced in vital
intermediate goods that are key inputs for domestic production. My data do not allow me
to disentangle the intermediate embargoed goods that are used for those specific types of
production from all the other products, which could lead to the estimates being upward
biased.
Embargoed consumption goods experience a smaller substitution effect towards the non-
sanctioning countries compared to intermediate goods from those countries - an increase of
54%. The substitution effect for the intermediate goods is 0.432 + 0.071 = 0.503, or a 65%
increase. This effect is statistically significant only at the 10% level. This makes intuitive
sense because the Russian government emphasized the importance of using domestically-
produced inputs in the production process, so I would not expect the Russian firms to

11
Established using an F-test; H0 : βi + γi × interaction term.

19
increase intermediate goods’ imports significantly and instead try to use domestic inputs.
The impact of the embargo on non-embargoed goods from the sanctioning countries (spillover
effect) is insignificant.
Next, I investigate whether the results so far are sensitive to the inclusion of oil and
gas products. Russia is one of the largest oil and gas exporters in the world (though not an
importer). I rely on an assumption that the prices in all goods’ markets is set according to
the same mechanism (monopolistic competition). However, oil markets do not follow the mo-
nopolistic competition assumption because there are few producers with large market power
(oligopoly), so the prices of these goods are set differently, which impacts their supply and
demand. I test whether including oil and gas products in the dataset might be affecting the
estimates. To test this assumption, I drop the HS product codes for oil and gas products from
the main sample; the results are presented in column 4. The results from both estimations
- with and without oil and gas products in the sample - are very close in magnitude for all
effects - β1 , β2 and β3 are very similar in magnitude and significance (columns 1 and 4). The
results hold for performing this procedure with the dataset containing BEC-classification.
Thus, including oil and gas products in the sample has no impact on the estimates.
When analyzing country heterogeneity, I concentrate on separating the embargo effects
into the impact on the countries that protested Russia’s annexation of Crimea and those
that supported it (or at least acknowledged the Referendum). I compare the impact of the
embargo on the OECD countries (most of these countries are imposing sanctions on Russia)
to the effects on a group of ideologically close countries, which includes China, several of the
former Soviet republics (except Ukraine and the countries of the Baltic region), Afghanistan,
Venezuela and several other communist states. These results are presented in table 4.
Column 1 of table 4 records the results of the benchmark specification, provided for
comparison. I create two interaction terms for each of the three treatment variables to esti-
mate these effects. The first interaction is with an indicator variable for whether the partner
country is a member of OECD (these countries protested Crimean annexation); these re-

20
sults are recorded in column 2. The second interaction is with the indicator for ideologically
close countries described above; these results are reported in column 3. There is no overlap
between sanctioning countries and ideologically close countries, thus the interaction of the
two country-product groups containing the sanctioning countries with the indicator for ideo-
logically close countries produces a 0. Thus the main treatment and spillover effects cannot
be estimated for the ideologically close countries. I compare the effect of the embargo on
the two groups of countries - OECD and ideologically close countries - to the countries that
belong to neither group (control group).
I concentrate on the results in column 4, which pulls together all the three effects and
the 2 interaction terms for each effect. The results in columns 2 - 4 do not differ in a major
way in coefficient significance or magnitude. The impact of embargo on the countries that
are in neither group is large and significant, a 95% decrease in trade (e(−3.147) − 1). The
OECD countries see a slightly smaller decline in trade of 76% (e(−3.027+1.599) − 1). This effect
is highly significant.
The substitution effect for the control group is positive and statistically significant. It
translates approximately to a 47% (i.e. e(0.387) − 1) increase in the control group’s imports.
The ideologically close countries see a greater increase in their exports to Russia as a result
of the embargo, but the difference effect is imprecisely estimated. This substitution pattern
is expected, as Russia would naturally gravitate to the countries that support (or at least
do not object to) its political ambitions. The insignificance of the substitution effect of
the ideologically close countries could be explained by the fact that the size of this group
is relatively small in comparison to the size of the control group, which translates to a
more restricted sample of import goods available. As a result, trade with the ideologically
close countries increased, but this increase is insignificant in comparison with the increase in
imports of the countries in the control group. The joint significance of the ideology impact
is, however, highly significant. The substitution effect for the OECD countries is smaller
than for the control group but again the difference is statistically insignificant. The spillover

21
effect for the sanctioning countries is negative and insignificant for both the control and the
OECD groups.

5.3 Dynamics

Table 5 presents the results of the dynamics exercise, in which I disentangle the short
and long term impacts of the embargo on Russian imports. To do so I create an indicator
variable for short and long periods. The short run spans the first five months of the embargo
(August 2014 till December 2014), while the long run includes all the other months following
the embargo (January 2015 till December 2016). I construct an interaction term with the
long-term indicator variable for each treatment effect, and compare the long and short term
effects (my control group consists of the short term effects). My benchmark model fails to
estimate the substitution effect in the long run. This could happen if the product-year fixed
effects absorb most of the variation necessary for the identification of this coefficient, i.e. if
for some of the non-sanctioning countries trade flows in embargoed products (substitution
effect) are observed in the short but not in the long run. This is plausible given that the
bulk of the substitution towards the non-sanctioning countries happened shortly after the
embargo in order to compensate the losses from the decrease in trade with the sanctioning
countries. In order to estimate the long run of the substitution effect, I use a less restrictive
model, in which I remove the product-year fixed effects from the benchmark model. The
results in table 5 are separated into the results derived using the benchmark model (columns
1 and 2) and the less restrictive model (columns 3 and 4). Columns 1 and 2 present the
results of the benchmark model for comparison purposes and the dynamic exercise based on
the benchmark model, respectively. Columns 3 and 4 provide the results of a less restrictive
model (no product-year fixed effects), and the results of the dynamic effects estimated using
the less restrictive model, respectively.
Across both specifications, the main embargo effect has a greater magnitude and sig-
nificance in the short run - Russian imports from sanctioning countries decrease by 81% in

22
column 2 (benchmark model) and by 80% in column 4 (less restrictive specification). The
long run impacts across both specification are smaller and individually insignificant. This is
feasible since the embargo was an exogenous shock which was imposed unexpectedly for all
economic agents. Thus, the largest portion of the decline in the imports of the embargoed
products from the sanctioning countries happened in the first 5 months after the embargo
was imposed. The overall long-run impact of the embargo on Russian imports estimated in
column 2 is −1.661 + 0.307 = −1.354 or a 74% (71 % in column 4) decline. This shows that
in the long run, the imports decrease less than in the short run. This may indicate either a
relaxation of embargo or an increase in domestic production.
I concentrate on the results in column 4 when analyzing the substitution effect, in order
to be able to estimate its long run dynamics. The strongest impact of the substitution effect
also happens in the short run, 0.328 and highly significant, or a 38.8% increase in imports
of the embargoed goods from non-sanctioning countries, while the long run coefficient is
statistically insignificant. This could serve as the evidence that in the long run domestic
production makes searching for the new foreign trade partners less needed, and thus decreases
the significance of the long run interaction term. The F-test for the joint significance of the
treatment and substitution effect reject the null hypothesis at 1%. I interpret these results
as follows. Both, the main treatment and substitution effects are the strongest in the short
run, as evidenced by the individual coefficients, β1 and β2 , which have the expected signs and
high statistical significance. No large changes in trade flows happen for either effect (main
treatment or substitution) in the long run, as evidenced by the low statistical significance of
the interaction coefficient. However, the negative impacts of both effects persist in the long
run, which is shown by the high joint significance of the long-run effects.
The same pattern emerges when I analyze the embargo effects on imports of non-
embargoed goods with the sanctioning countries. The short-run spillover effect is significant
at 5%, although its magnitude is much lower than the magnitudes of the other two effects - I
observe a drop of only 8%. In the long run I see an insignificant increase in Russian imports

23
of non-embargoed goods’ trade flows from the sanctioning countries. This short run negative
impact could be the result of a significant disturbance in the market, caused by the embargo.
The firms in sanctioning countries trading in non-embargoed goods might have decreased
their exports to Russia in the short run in an anticipation of Russia expanding the embargo
to include other, goods. However, in the long run there was no change to the embargo’s
scope, and these firms might have resumed their exports. These results are inconsistent with
my hypothesis that the fall in trade in non-embargoed products recorded in column 1 and 3
(even though it might be insignificant) with the sanctioning countries is most likely due to
political reasons. If this were true, I would expect the non-embargo effect to be stronger in
the long run, which is not what I observe.
To enhance the analysis of the dynamic effects of the Russian embargo, I analyze whether
the “smart” sanctions had any impact on Russian imports. The “smart” sanctions refer
to the restrictions imposed on certain Russian individuals and firms, who supported the
Crimean annexation. The first round of the “smart” sanctions was imposed immediately
after the annexation of Crimea in March of 2014. These sanctions include a ban on all
commercial activities with firms in Russia and Ukraine that supported Crimean annexation.
I disentangle the impact of the embargo from the impact of the “smart”sanctions and present
the results in table 6.
Column 1 of table 6 presents the benchmark model’s results for the imports’ sample. I
create a new time indicator for the “smart” sanctions; it takes the value of 1 for periods
from March 2014 until August 2014, when the embargo was imposed. I then generate
treatment variables by interacting the time dummy with the indicators for the sanctioning
or non-sanctioning countries and embargoed or non-embargoed product (same procedure as
described in section 4). This approach is similar to the one presented in Crozet & Hinz (2016)
- they disentangle the impacts of the Russian sanctions and embargo on the sanctioning
countries’ exports to Russia. I re-estimate the benchmark model including the old and new
treatment variables; the results are presented in column 2 of table 6.

24
 I find a significant drop in Russian imports of embargoed goods from the sanctioning
countries following the imposition of the “smart”sanctions, however this effect is of a much
smaller magnitude than the effect of the embargo (-0.216 vs -1.831). This finding suggests
that the sanctions did have a negative impact on Russian imports from the sanctioning
countries. This finding is in line with the study by Crozet & Hinz (2016). The sanctions
target a small number of firms, so their impact, relative to the size of the embargo, which
targets multiple countries and products, is not expected to be large. I do not find significant
impact of sanctions on either substitution or the spillover coefficients.

5.4 Robustbess checks

The previous estimates from the heterogeneity and dynamics analysis of the embargo’s
effects are of a similar significance and magnitude. This provides evidence of the stability and
robustness of the results. I conduct two additional robustness checks, including a falsification
exercise and addressing the issue of zero trade flows, described in detail in Santos Silva &
Tenreyro (2006). The results for the falsification exercise and zero trade flows exercise are
presented in tables 7 and 8, respectively. First I describe the procedure and the results
for the falsification exercise; then I analyze the issue of the zero trade flows in the gravity
equation set-up.

5.4.1 Falsification exercise

The aim of the falsification exercise is to provide evidence that the strong negative im-
pact found in the previous estimations is indeed due to the imposition of embargo and not
other unaccounted exogenous shocks and processes happening in the economy that started
before the imposition of the embargo. The methodology of the falsification exercise is as fol-
lows. First, I limit the sample to the period until August 2014 and create a false embargo time
indicator by changing the treatment time to the periods from January 2012 till July 2014. I
then use this counterfactual dummy to create three false embargo treatment variables (false

25
treatment, substitution and spillover variables) by interacting it with the respective dum-
mies for sanctioning or non-sanctioning country and embargoed or non-embargoed products,
respectively. Finally, I estimate my benchmark specification using the new counterfactual
variables in place of the old treatment, substitution and spillover effects. The results of the
falsification exercise are presented in table 7.
The first thing to notice when analyzing the results in table 7 is the fact that the coun-
terfactual main treatment effect is statistically insignificant and of a significantly different
magnitude than in my benchmark model with the original treatment variables (provided in
column 1 of table 7 for comparison). This provides confirmation that no other exogenous
negative shock affected Russian imports prior to the imposition of the embargo. However,
the benchmark model, which includes product-year fixed effects, proves to be too restrictive
for the estimation of either counterfactual substitution or spillover effects. In the truncated
dataset, there are 13,589 observations for the counterfactual substitution country-product
group12 . The total number of fixed effects for this group is 3030 for the pre-counterfactual
embargo period and 2363 for the post-counterfactual embargo, so there must be not enough
variation to estimate the counterfactual substitution effect. Prior to the embargo, the non-
sanctioning countries played a significantly smaller role in Russian trade. During the pre-
embargo period, Russia traded in a larger number of goods with the sanctioning countries
than the non-sanctioning countries: there are 1230 distinct HS-4 digit product codes in the
pre-embargo period for the sanctioning countries against 1210 HS-4 codes traded with the
non-sanctioning countries in the same period. After the imposition of the embargo, the
number of distinct HS-4 codes for both types of countries fell to 1210 and 1197 for the
sanctioning and non-sanctioning countries, respectively. The average logged trade flow of
both types of countries in embargoed goods contracts; however, the fall experienced by the
sanctioning countries is much larger in magnitude (a decrease from 11.73 to 10.26) than for
the non-sanctioning countries (a decrease from 11.97 to 11.53). This provides proof for the
12
This number of observations is split in the following fashion: 8316 observations are in the pre-counterfactual
embargo period (2010-2012) and 5025 are in the post-counterfactual embargo period (2012-2014).

26
hypothesis that the embargo led to an increase in the substitution effect in the post-embargo
period, however there is no substitution effect to estimate in the pre-embargo period.
When I relax the restrictions by excluding the product-year fixed effects (the results
are presented in column 3), I am able to estimate the false substitution effect in addition
to the false main treatment effect. Because both false effects that I am able to estimate
(main treatment and substitution) are insignificant, it is safe to assume that the spillover
effect will also not be statistically significantly different from zero. I relax the model further
by removing the country-year fixed effects; these results are presented in column 4 of table
7. This provides enough variation for the estimation of all three effects of interest. As
expected, neither of the counterfactual coefficients is of similar magnitude as in the original
model; all estimated coefficients of interest are statistically insignificant. However, omitting
the country-year fixed effects introduces bias into the estimates, because country-year fixed
effects account for the multilateral resistance terms. To conclude, the falsification exercise
confirms that the significant drop in bilateral trade between Russia and the sanctioning
partner countries that I estimate is indeed attributed to the imposition of the embargo.

5.4.2 Zero trade flows

When estimating log-linearized models, an issue of omitting the observations with a

zero value arises quite often, because it is impossible to take a logarithm of a zero. The
standard estimating procedures, like OLS, omit these observations, introducing bias into
the estimates. Gravity model estimations are especially prone to this pitfall due to the
underlying data-generating process - few countries trade with every single country in the
world, generating a large number of zeros in the trade flows matrix for the country pairs
that do not trade with each other.
My imports sample faces a similar issue - the imposition of the Russian food embargo
leads to a truncation of the sample by omitting the zero trade flows in embargoed goods with
the sanctioning countries. There are several standard procedures in the economic literature

27
used to address this issue, including Heckman correction model, linear transformation of the
dependent variable and others. I proceed by using the Poisson Pseudo-Maximum Likelihood
(PPML) estimator, suggested by Santos Silva & Tenreyro (2006), to ensure that my estimates
are robust to this issue.
I expand my imports sample by generating a matrix of all possible partner country,
product and time period interactions. I then populate this matrix with the values of import
trade flows from the original sample, and the missing observations are assigned the value
of zero. Next, I generate the dummy variables for the three main effects, as described in
section 3. In the new model, the dependent variable is the value of an import trade flow
from country j in product k at time t to Russia in level; the regressors include the dummy
variables for the three main effects (treatment, substitution and spillover). I estimate two
specifications of the new model, a less restrictive one, with a smaller set of fixed effects
(country-product and period) and a more restrictive one with country-product, period and
country-year fixed effects. They are reported in columns 2 and 4, respectively, of Table 8.
Due to a large number of fixed effects, I am unable to estimate the model with the full set
of fixed effects, proposed in the estimating equation (3), and I omit the product-year fixed
effects. I provide the results of the benchmark model with the corresponding sets of fixed
effects in columns 1 and 3 of table 8 for comparison.
The first thing to note is that the main treatment effect of the embargo is negative, highly
significant and of a similar magnitude across all specifications, which provides confirmation
that my estimates are robust to the issue of the zero trade flows. I concentrate on interpreting
the results in columns 3 and 4, because these specifications include the most restrictive set
of fixed effects and account for the multilateral resistance term, represented by the country
- year fixed effects.
OLS estimation uses only 40% of the sample (1,580,711 out of 3,996,636 observations)
and over-estimates all three effects of the embargo, with substitution and spillover effects
being significantly over-estimated. Santos Silva & Tenreyro (2006) also find that OLS has

28
the tendency to over-estimate the coefficients. Under the PPML estimator, the substitution
and spillover effects lose significance. The main treatment effect is -1.606 under OLS and
-1.514 under PPML, corresponding to 80% and 78% decrease in imports of embargoed goods
from the sanctioning countries, respectively. Thus, under both estimators, the decrease in
trade in that country-product group is of similar magnitude and significance.
The disparity between the significance levels of the substitution and spillover effects
from the OLS and PPML estimations is quite large, which implies that these results should
be interpreted with caution. The OLS could significantly overstate the substitution effect if
the positive trade flow data under-represents the amount of times when the switch in trade
to non-sanctioning partners is actually zero. More analysis of these effects is required in
the future work. To conclude, even though the OLS estimates significantly overstate the
substitution and spillover effects of the embargo and more work needs to be carried out to
produce reliable estimates, the main effect of the embargo is confirmed to be negative, large
and statistically significantly different from zero.

6 Results: Exports Sample

Next, I estimate the results of the embargo on Russian exports. The summary statistics
for the exports sample are presented in panel B of table 1. There are more partner-countries
in the exports sample than in the import sample - 129, 38 of which are sanctioning countries.
There are 1365 distinct HS-4 digit level product codes in the sample; 48 HS-4 codes are em-
bargoed. Trade flows with sanctioning and non-sanctioning countries follow the same pattern
in the exports sample as in the import sample - Russia trades with the sanctioning countries
more than with the non-sanctioning countries; the trade share of sanctioning countries in
both samples is about 70%. However, the imbalance between embargoed and non-embargoed
goods is heavily skewed towards the non-embargoed goods in the exports sample - 99% of
the exports sample is comprised by the non-embargoed goods. Embargoed goods in the

29
exports sample are on average cheaper than non-embargoed goods. Russia primarily exports
oil, gas, mineral resources and timber, which are non-embargoed goods and are typically
more expensive than embargoed goods, most of which are agricultural commodities. On av-
erage, Russia exports 81 distinct HS-4 digit product codes to the sanctioning countries and
45 to the non-sanctioning countries. Because most of the goods in the exports sample are
non-embargoed goods, and also because the embargo targets the imported goods from the
sanctioning countries, I do not expect to find significant embargo effects on Russian exports.
The estimation results for the exports sample are presented in table 9. I follow the
same protocol for exports as I do for imports and compare the results of the main estimat-
ing equation across different specifications, ranging from the least (column 1) to the most
(column 4) restrictive. As expected, the main treatment and substitution effects for Russian
exports are insignificant and are of much smaller magnitude than the respective effects for
the imports sample. This result is in line with theory - only imported goods were embargoed,
while none of Russian exports were embargoed by the Russian government, and the share of
the embargoed goods in the sample is only 1%.
In columns 2 and 3 of table 9, I observe a significant negative impact of the embargo
on Russian exports of the non-embargoed goods to the sanctioning countries (the spillover
effect). This result is somewhat surprising, since I do not expect the embargo to hurt Russian
exports as its aim was to hurt the sanctioning countries. My estimates might be picking up
the effects of the “smart”sanctions on Russian exports. The “smart”sanctions are described
in the dynamics section of this paper. Because the exports’ sample consists of predominantly
non-embargoed goods, my results might be picking up the effects of these measures on
Russian trade. The “smart”sanctions are analyzed in detail by Ahn & Ludema (2017).
Unfortunately, I do not have the firm level data to disentangle this effect between the firms
that are targeted by these sanctioning measures and firms that are not. In my sample trade
flows are aggregated by product and country, which enables me to estimate the sanctions’
or embargo’s effect on aggregate but not firm level Russian trade flows. The spillover effect

30
could also be due to the politically motivated decision of the firms in the sanctioning countries
to stop importing from Russia to boycott its decision to impose the embargo. I provide
analysis of the dynamics of the three effects of interest in order to disentangle the impact of
the embargo on Russian exports from the impact of the “smart”sanctions. These results are
presented in table 10.
Column 1 presents the benchmark model’s results for the exports’ sample. To check
whether my results are picking up the effects of the “smart”sanctions, I create a new time
indicator for the “smart sanctions”; it takes the value of 1 for periods from March 2014
until August 2014, when the embargo was imposed. I then generate treatment variables
by interacting the time dummy with the indicators for the sanctioning or non-sanctioning
countries and embargoed or non-embargoed product (same procedure as described in section
3). I re-estimate the benchmark model including the old and new treatment variables. These
results are presented in column 2 of table 10.
I find a significant drop in Russian exports of embargoed goods to the sanctioning
countries immediately following the imposition of the “smart”sanctions, but see no significant
effect on Russian exports in this group after the embargo was imposed. I believe, these results
confirm that the “smart”sanctions had a significant negative impact on certain sanctioned
firms and this effect persisted after the embargo was imposed, which is reflected in the
insignificant main treatment coefficient. It could also be the case that after the initial shock
of “smart”sanctions dissipated, trade resumed and followed its initial trend. This hypothesis
is supported by the fact that the substitution effect is insignificant after both shocks -
imposition of the “smart”sanctions and the embargo.
The spillover effect is significant only after the embargo was imposed and is insignificant
after the imposition of the “smart”sanctions. Thus, Russian exports of non-embargoed goods
to the sanctioning countries experiences a 16% (e−0.172 − 1 = −0.158) drop. I believe this
decrease could be attributed to either political or logistical motivations of the partner firms.
It could be the case that firms in the sanctioning countries stopped importing from Russia

31
to boycott the imposition of the embargo. This decline could also be motivated by the input
- output relationship; if the embargo negatively impacted the input goods, which Russian
firms import from the sanctioning countries for the production of non-embargoed outputs
that are then exported to the sanctioning countries, this could explain the highly significant
negative embargo’s effect on this group. This story is, however, less likely than the politically
motivated decline, because the previous analysis of heterogeneity shows that the intermediate
goods experience a significantly smaller impact than consumption goods due to the embargo.
If the export trade flows dropped due to the lack of inputs, the exports should drop with
all countries, not just the sanctioning. However, this question presents an important avenue
for the future research - using a firm level data and the input-output relations for Russian
industries it would be possible to test this hypothesis.
I find that the embargo has a negative impact on Russian exports of non-embargoed
goods to the sanctioning countries (spillover effect), while the main treatment and substi-
tution effects are insignificant. I explain this by the political motivations of the trading
partners - foreign firms trading in non-embargoed goods might have stopped importing from
Russia to boycott the embargo which hurts primarily the agricultural producers.

7 Conclusion

Interest to the topic of barriers to trade such as trade sanctions and embargoes in
the economic literature has increased significantly in the past decades. In this paper I
use difference-in-differences estimation of the log-linearized gravity model to quantify the
effects of the Russian embargo imposed on the imports from the sanctioning countries in the
aftermath of the Crimean conflict. I argue that the embargo was an exogenous shock that
impacted both Russian and foreign firms, which led to a significant decrease in embargoed
goods’ trade with sanctioning countries. I also find a significant substitution effect, which
indicates that Russia substituted its imports of embargoed goods away from the sanctioning

32
towards the non-sanctioning countries. The relative magnitudes of these effects are such that
the substitution effect does not mitigate the main negative impact. The sanctioning countries
also experience a drop in their imports of non-embargoed goods from Russia, mostly due to
political or logistical reasons.
I conduct an analysis of product and country heterogeneity in an attempt to further
separate the embargo’s effects by product and country groups. As expected, the embargo’s
impact on consumption goods is significantly stronger than its impact on intermediate goods,
because the embargo mostly targets consumption goods. The OECD countries bear larger
losses than ideologically close countries; substitution effect towards ideologically close coun-
tries is weaker than for all other non-sanctioning countries. Analysis of the embargo’s effects’
dynamics show that all of the effects are strongest in the short run, due to the unexpected
nature of the shock.
I find no significant effect of the embargo on Russian embargoed goods’ exports to the
sanctioning countries. However, I do find a significant negative impact on the exports of the
non-embargoed goods to the sanctioning countries (spillover effect). I explain this by the
political motivations of the trading partners - foreign firms trading in non-embargoed goods
might have stopped importing from Russia to boycott the embargo which hurts primarily
the agricultural producers.
Understanding the effects of sanctions and embargoes is an important task, because
they have always been popular tools of foreign policy used to induce behavioral changes
in non-complying agents without military actions. Military actions are very costly and
involve human casualties, while sanctions target certain economic and financial sectors of
the non-complying country and may induce the desired change in behavior without military
involvement. Therefore, it’s is crucial to analyze the effectiveness of these episodes, their
impact on both target’s and sender’s economies.
The incident of Russian sanctions and embargo presents a good opportunity to analyze
the impacts of these economic policies on world’s trade flows, economy and welfare because

33
Russia is a large market player and it is the first time that a large market player is sanctioned.
My findings are in line with several other studies; I, too, find significant negative impacts on
Russian trade flows. There are several possible options for the future research of this topic,
crucial for our understanding of the impacts of the trade policies. Analyzing the embargo’s
effects on Russian firms, including the exit - entry and input - output angles, will provide
important insights in how these policies affect firms. The unintended consequences of the
embargo might be welfare losses born by the firms in the sending countries due to loss of the
trade partners and access to markets.

34
References

Ahn, D. P., & Ludema, R. D. (2017). The Sword and the Shield: The Economics of Targeted
Sanctions.

Anderson, J. E., & Van Wincoop, E. (2003). Gravity with Gravitas : A Solution to the
Border Puzzle. The American Economic Review , 93 (1), 170–192.

Ashenfelter, O., Ciccarella, S., & Shatz, H. J. (2007). French Wine and the U.S Boycott:
Does Politics Really Affect Trade?

Bapat, N. A., Heinrich, T., Kobayashi, Y., & Morgan, T. C. (2013). Determinants of
Sanctions Effectiveness: Sensitivity Analysis Using New Data. International Interactions,
39 (1), 79–98.

Boulanger, P., Dudu, H., Ferrari, E., & Philippidis, G. (2016). Russian Roulette at the
Trade Table: A Specific Factors CGE Analysis of an Agri-food Import Ban. Journal of
Agricultural Economics, 67 (2), 272–291.

Chavis, L., & Leslie, P. (2009). Consumer boycotts: The impact of the Iraq war on french
wine sales in the U.S. Quantitative Marketing and Economics, 7 (1), 37–67.

Correia, S. (2016). Linear models with high-dimensional fixed effects: An efficient and
feasible estimator. Tech. rep. Working Paper.

Coulibaly, B. (2009). Effects of financial autarky and integration: The case of the South
Africa embargo. Journal of International Money and Finance, 28 (3), 454–478.
URL http://dx.doi.org/10.1016/j.jimonfin.2008.11.003

Crozet, M., & Hinz, J. (2016). Collateral damage: The impact of the Russia sanctions on
sanctioning countries’ exports. CEPII Working Paper , (p. 59).
URL http://www.cepii.fr/PDF{_}PUB/wp/2016/wp2016-16.pdf

35
Dreger, C., Kholodilin, K. A., Ulbricht, D., & Fidrmuc, J. (2016). Between the hammer and
the anvil: The impact of economic sanctions and oil prices on Russia’s ruble. Journal of
Comparative Economics, 44 (2), 295–308.

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

Eaton, J., & Engers, M. (1999). Sanctions : Some Simple Analytics. The American Economic
Review: Papers and Proceedings of the One Hundred Eleventh Annual Meeting of the
American Economic Association ( May , 1999 ), 89 (2).

Feyrer, J. (2009). Distance, Trade, and Income - the 1967 to 1975 Closing of the Suez Canal
as a Natural Experiment.

Gurvich, E., & Prilepskiy, I. (2015). The impact of financial sanctions on the Russian
economy. Russian Journal of Economics, 1 (4), 359–385.
URL http://www.sciencedirect.com/science/article/pii/S2405473916000039

Heilmann, K. (2016). Does political conflict hurt trade? Evidence from consumer boycotts.
Journal of International Economics, 99 , 179–191.

Hinz, J., & Monastyrenko, E. (2016). Bearing the cost of politics : Consumer prices in
Russia . (July).

Hufbauer, G. C., Elliott, K. A., Cyrus, T., & Winston, E. (1997). US Economic Sanctions:
Their Impact on Trade, Jobs, and Wages.
URL https://ideas.repec.org/p/iie/wpaper/wpsp-2.html

Hufbauer, G. C., & Schott, J. J. (1985). Economic Sanctions Reconsidered: History and
Current Policy. Institute for International Economics.

Irwin, D. A. (2005). Welfare Effects of Autarky: Evidence from the Jeffersonian Embargo
of 18071809. Review of International Economics, 13 (March 1809), 631–645.

36
Kaempfer, W., & Lowenberg, A. (1988). The Theory of International Economic Sanctions :
A Public Choice Approach. The American Economic Review , 78 (4), 786–793.

Kholodilin, K. A., & Netsunajev, A. (2016). Crimea and Punishment: The Impact of
Sanctions on Russian and European Economies. SSRN Electronic Journal .
URL https://www.ssrn.com/abstract=2768622

Kuehnapfel, E. (2015). Welfare effects of embargoes - How does the embargo of the United
States affect economic and social development in the Republic of Cuba?
URL https://www.researchgate.net/publication/275154919{_}Welfare{_}effects{_}of{_}emba

Michaels, G., & Zhi, X. (2016). Freedom Fries. American Economic Journal: Applied
Economics, 2 (3), 256–281.

Neuenkirch, M., & Neumeier, F. (2015). The impact of UN and US economic sanctions on
GDP growth. European Journal of Political Economy, 40 , 110–125.
URL http://dx.doi.org/10.1016/j.ejpoleco.2015.09.001

Pandya, S. S., & Venkatesan, R. (2016). French Roast: Consumer Response to International
Conflict - Evidence from Supermarket Scanner Data. Review of Economics and Statistics,
98 (1), 42–56.
URL http://www.mitpressjournals.org/doi/pdf/10.1162/REST{_}a{_}00526

Santos Silva, J., & Tenreyro, S. (2006). The log of gravity. The Review of Economics and
Statistics, November 2006, 88 , 88(4)(November), 641–658.

Teegen, H., Forrer, J., Askari, H., & Yang, J. (2008). Economic and Strategic Impacts of
U.S. Economic Sanctions on Cuba.

United Nations Security Council (2017). Sanctions. https://www.un.org/sc/suborg/en/

sanctions/information. Online; accessed 12 November 2017.

37
Uzun, V., & Loginova, D. (2016). Russian food embargo : minor losses. Russian Economic
Developments, 1 (9), 32 – 37.

van Bergeijk, P. A., & van Marrewijk, C. (1995). Why do sanctions need time to work?
Adjustment, learning and anticipation. Economic Modelling, 12 (2), 75–86.

Veebel, V., & Markus, R. (2015). At the Dawn of a New Era of Sanctions: Russian-Ukrainian
Crisis and Sanctions. Orbis, 60 (1), 128–139.

World Trade Organization (2017). Trade Profiles. http://stat.wto.org/CountryProfile/

WSDBCountryPFReporter.aspx?Language=E. Online; accessed 01 November 2017.

38
 Table 1: Summary statistics

Panel A: Imports
Sanctioning country Non-sanctioning country Embargoed product Non-embargoed product
Log trade flows 10.68 10.71 11.61 10.65
(2.868) (2.93) (2.935) (2.874)

Log GDP of the partner country 10.55 11.71

(1.642) (5.363)

Log price 0.968 2.433

(1.384) (1.968)

Trade shares 0.715 0.285 0.106 0.894

Number of sanctioning countries in sample = 119: 38 81

Number of embargoed products in sample = 1240: 48 1192

Average number of products countries trade in 81 42

Observations (total = 1587225) 1244435 342790 58231 1528994

Panel B: Exports
Sanctioning country Non-sanctioning country Embargoed product Non-embargoed product
Log trade flows 10.04 10.09 10.55 10.05
(3.388) (3.342) (2.944) (3.376)

Log GDP of the partner country 11.45 10.03

(1.724) (2.139)

Log price 1.380 1.795

(1.346) (2.228)

Trade shares 0.714 0.286 0.006 0.994

Number of sanctioning countries in sample = 129: 38 91

Number of embargoed products in sample = 1365 : 48 1317

Average number of products countries trade in 81 45

Observations (total = 1222879) 718122 504757 24279 1198389

Variable means; standard deviations in parentheses.

The table presents the sample summary statistics for the variables used to estimate specification (3) for imports sample (Panel A) and exports sample (Panel B).

1
 Table 2: Motivation for specification choice

Dependent variable: log trade

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

Sanctioning country × Embargoed product -1.735*** -1.606*** -1.742*** -1.675***

(Treatment effect) [0.329] [0.322] [0.216] [0.190]

Non-sanctioning coutry × Embargoed product 0.353*** 0.340*** 0.452*** 0.560**

(Substitution effect) [0.090] [0.100] [0.154] [0.252]

Sanctioning country × Non-embargoed product -0.174** -0.065* -0.052 -

(Spillover effect) [0.068] [0.034] [0.035] -

Log trade cost -0.027 -0.179*** -0.185*** -

[0.069] [0.054] [0.055] -

Log GDP of the partner country 1.009** 0.613 0.634 -

[0.447] [0.655] [0.657] -

FE Country × product; Country × product; Country × product; Country × product;

period country × year country × year product × period
period product × year country × period
period

Observations 1,587,216 1,587,216 1,587,216 1,587,216

R-squared 0.795 0.798 0.807 0.845
Robust standard errors in brackets; clustered at country level
*** p<0.01, ** p<0.05, * p<0.1

This table provides the motivation for choosing the main estimating equation (3) as the benchmark specification for the imports sample.
Each column contains a different set of fixed effects along with the three treatment dummy variables of interest.

1
 Table 3: Heterogeneity across products

Dependent variable: log trade

(1) (2) (3) (4)
Benchmark specification Benchmark specification Consumption vs Intermediate Benchmark specification:
(BEC dataset) (no oil)

Sanctioning country × Embargoed product -1.742*** -1.589*** -1.708*** -1.739***

(Treatment effect) [0.216] [0.212] [0.214] [0.215]

Treatment × Intermediate 1.094***

[0.329]

Non-sanctioning country × Embargoed product 0.452*** 0.446*** 0.432*** 0.453***

(Substitution effect) [0.154] [0.141] [0.156] [0.154]

Substitution × Intermediate 0.071

[0.338]

Sanctioning country × Non-embargoed product -0.052 -0.047 -0.046 -0.050

(Spillover effect) [0.035] [0.034] [0.037] [0.035]

Spillover × Intermediate -0.002

[0.023]

Log trade cost -0.185*** -0.190*** -0.190*** -0.185***

[0.055] [0.054] [0.054] [0.056]

Log GDP of the partner country 0.634 0.649 0.649 0.632

[0.657] [0.623] [0.623] [0.656]

FE Country × product; Country × product; Country × product; Country × product;

country × year; country × year; country × year; country × year;
product × year; product × year; product × year; product × year;
period period period period

Observations 1,587,216 1,801,167 1,801,167 1,581,942

R-squared 0.807 0.796 0.796 0.807
Robust standard errors in brackets; clustered at country level
*** p<0.01, ** p<0.05, * p<0.1

This table records the results of product heterogeneity analysis. The effect of the embargo on the consumption goods as by BEC classification is compared to
the effect on intermediate goods; BEC dataset refers to a sample aggregated from HS-6 digit product code level with a BEC classification assigned to each
observation to HS-4. The Intermediate variable refers to a dummy variable, which is equal to 1 if the product is an intermediate good. The reference group
is a consumption good. In the no oil sample (column 4) all oil and gas products are removed to test the monopolistic competition assumption.

1
 Table 4: OECD vs Ideologically close countries

Dependent variable: log trade

(1) (2) (3) (4)
Benchmark specification OECD effects Ideologically close effects Both effects
Sanctioning country × Embargoed product -1.742*** -3.032*** -1.738*** -3.027***
(Treatment effect) [0.216] [0.481] [0.216] [0.480]

Treatment ×OECD 1.600*** 1.599***

[0.583] [0.582]

Treatment × Ideology - -

Non-sanctioning country × Embargoed product 0.452*** 0.461*** 0.364** 0.387*

(Substitution effect) [0.154] [0.151] [0.184] [0.215]

Substitution × OECD -0.262 -0.187

[0.158] [0.226]

Substitution × Ideology 0.225 0.161

[0.228] [0.265]

Sanctioning country × Non-embargoed product -0.052 -0.045 -0.052 -0.046

(Spillover effect) [0.035] [0.101] [0.035] [0.101]

Spillover × OECD -0.010 -0.010

[0.100] [0.100]

Spillover × Ideology -

Log trade cost -0.185*** -0.185*** -0.188*** -0.187***

[0.055] [0.053 ] [0.055] [0.053]

Log GDP of the partner country 0.634 0.641 0.630 0.638

[0.657] [0.659] [0.658] [0.659]

FE Country × product; Country × product; Country × product; Country × product;

country × year; country × year; country × year; country × year;
product × year product × year; product × year; product × year;
period period period period

Observations 1,587,216 1,587,216 1,587,216 1,587,216

R-squared 0.801 0.807 0.807 0.807
Robust standard errors in brackets; cluster at country level
*** p<0.01, ** p<0.05, * p<0.1

This table presents results of the country heterogeneity analysis. I compare the impacts the embargo had on the OECD (sanctioning countries) to its
impacts on the ideologically close countries (communist regimes (China, North Korea, Venzuela), political allies and countries of the former Soviet block,
excluding Ukraine and countries of the Baltic region). The OECD variable refers to the dummy variable that takes the value of 1 if a country is a member OECD;
Ideology is an indicator for the ideologically close country.

1
 Table 5: Dynamics

Dependent variable: log trade

(1) (2) (3) (4)
Benchmark specification Dynamic effects Less restrictive model Dynamic effects:
less restrictive model

Sanctioning country × Embargoed product -1.742*** -1.661*** -1.606*** -1.510***

(Treatment effect) [0.216] [0.155] [0.322] [0.163]

Treatment × Long run 0.307 0.254

[0.356] [0.321]

Non-sanctioning country ×Embargoed product 0.452*** 0.269** 0.340*** 0.328***

(Substitution effect) [0.154] [0.108] [0.100] [0.115]

Substitution × Long run - 0.013

[0.111]

Sanctioning country × Non-embargoed product -0.052 -0.075** -0.065*** -0.078**

(Spillover effect) [0.035] [0.035] [0.034] [0.036]

Spillover × Long run 0.710 0.400

[0.536] [0.482]

Log trade cost -0.185*** -0.184*** -0.179*** -0.177***

[0.055] [0.055] [0.054] [0.054]

Log GDP of the partner country 0.634 0.624 0.613 0.605

[0.657] [0.653] [0.655] [0.648]

FE Country × product; Country × product; Country × product; Country × product;

country × year; country × year; country × year; country × year;
product × year; product × year; period period
period period

Observations 1,587,216 1,587,216 1,587,216 1,587,216

R-squared 0.807 0.807 0.798 0.798
Robust standard errors in brackets; clustered at country level
*** p<0.01, ** p<0.05, * p<0.1

In this table the results of the dynamics exercise are presented. I disentangle the short and long term impacts of the embargo on Russian imports.
The short run spans the first five months of the embargo (August 2014 till December 2014), while the long run includes all the other months following
the embargo (January 2015 till December 2016). Long run is an indicator variable for the long run. Columns 1 and 2 present results for a more
restrictive specification, while columns 3 and 4 omit product-year FE.

1
 Table 6: Smart sanctions vs Embargo for imports’ sample

Dependent variable: log trade

(1) (2)
Benchmark Dynamics

Smart sanctions: Treatment effect -0.216***

[0.060]

Embargo: Treatment effect -1.742*** -1.831***

[0.216] [0.224]

Smart sanctions: Substitution effect 0.023

[0.078]

Embargo: Substitution effect 0.452*** 0.405**

[0.154] [0.162]

Smart sanctions: Spillover effect -0.002

[0.037]

Embargo: Spillover effect -0.052 -0.055

[0.035] [0.043]

Log trade cost -0.185*** -0.186***

[0.055] [0.055]

Log GDP of the partner country 0.634 0.641

[0.657] [0.660]

FE Country × product; Country × product;

country × year; country × year;
product × year; product × year;
period period

Observations 1,587,216 1,587,216

R-squared 0.807 0.807
Robust standard errors in brackets; clustered at country level
*** p<0.01, ** p<0.05, * p<0.1

As part of the analysis of dynamic effects, I disentangle effects of the smart sanctions and the embargo on
Russian aggregate trade flows. The smart sanctions refer to the restrictions imposed on certain Russian
individuals and firms, who supported the Crimean annexation; they were imposed in March of 2014. I
generate a separate indicator variable for smart sanctions, which receives the value of 1 for time periods
from March till July 2014 and keep the original time dummy variable, which receives value of 1 in time periods
following August 2014. I then generate the three coefficients of interest (main, substitution and spillover effects)
as described in section 3, using the new time indicator for the smart sanctions. I include these variables
along with the original treatment variables in the main specification (3), and re-estimate the model.

1
 Table 7: Falsification exercise
(1) (2) (3) (4)
Benchmark specification Benchmark specification: Less restrictive model: Least restrictive model:
False treatment False treatment False treatment

Sanctioning country × Embargoed product -1.742***

(Treatment effect) [0.216]

Non-sanctioning coutry × Embargoed product 0.452***

(Substitution effect) [0.154]

Sanctioning country × Non-embargoed product -0.052

(Spillover effect) [0.035]

False treatment effect -0.121 -0.084 -0.067

[0.080] [0.055] [0.071]

False substitution effect - -0.034 -0.011

[0.074] [0.072]

False spillover effect - - 0.021

[0.046]

Log trade cost -0.185*** -0.334*** -0.330*** -0.255***

[0.055] [0.077] [0.076] [0.090]

Log GDP of the partner country 0.634 0.436 0.440 0.913***

[0.657] [0.455] [0.448] [0.324]

FE Country × product; Country × product; Country × product; Country × product;

country × year; country × year; country × year; period
product × year; product × year; period
period period

Observations 1,587,216 1,043,689 1,043,687 1,043,687

R-squared 0.807 0.828 0.822 0.820
Robust standard errors in brackets; clustered at country level
*** p<0.01, ** p<0.05, * p<0.1

Table 7 records the results of the falsification exercise. I limit the sample to the period until August 2014 and create a false embargo time indicator
by changing the treatment time to the periods from January 2012 till July 2014. I then use this counterfactual dummy to create three false embargo
treatment variables (false treatment, substitution and spillover variables) by interacting it with the respective dummies for sanctioning or
non-sanctioning country and embargoed or non-embargoed products, respectively. Finally, I estimate my benchmark specification using the new
counterfactual variables in place of the old treatment, substitution and spillover effects.

1
 Table 8: Zero trade flows

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

Benchmark PPML Benchmark PPML

Sanctioning country × Embargoed good -1.735*** -1.331*** -1.606*** -1.514 ***

(Treatment effect) [0.329] [0.287] [0.322] [.202 ]

Non-sanctioning country × Embargoed good 0.353*** 0.441** 0.340*** .066

(Substitution effect) [0.090] [0.217] [0.100] [.144]

Sanctioning country × Non-embargoed good -0.174** 0.192 -0.065* .005

(Spillover effect) [0.068] [0.243] [0.034] [.063]

Dependent variable Log trade Level trade Log trade Level trade

FE country × product country × product country × product country × product

period period country × year country × year
period period

Observations 1,580,711 3,996,636 1,580,711 3,760,224

Robust standard errors in brackets; clustered at country level
*** p<0.01, ** p<0.05, * p<0.1

Table 8 records the estimates obtained from using an alternative estimator - PPML - to estimate the effects of the
embargo on Russian trade flows in order to deal with the issue of zero trade flows. I expand my imports sample
by generating a matrix of all possible partner country, product and time period interactions. I then populate this matrix
with the values of import trade flows from the original sample, and the missing observations are assigned the value of zero.
I generate the dummy variables for the three main effects, as described in section 3, and include them in the model
in place of the original treatment variables. The dependent variable is the level of import trade flows.
Two specifications are estimated - the least restrictive one (includes country-product and period fixed effects and the more
more restrictive model, which accounts for the multilateral resistance term by including country-year fixed effects.
Results of the benchmark models with corresponding sets of fixed effects are provided in columns 1 and 3 for comparison.

1
 Table 9: Main specification results for the exports’ sample

Dependent variable: log trade

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

Sanctioning country × Embargoed product -0.022 -0.052 0.004 0.140

(Treatment effect) [0.098] [0.068] [0.087] [0.123]

Non-sanctioning coutry × Embargoed product 0.114 0.114 0.134 0.160

(Substitution effect ) [0.127] [0.118] [0.094] [0.150]

Sanctioning country × Non-embargoed product -0.101 -0.111*** -0.114*** -

(Spillover effect ) [0.085] [0.029] [0.029] -

Log of trade cost -0.064 -0.095** -0.096** -

[0.040] [0.042] [0.043] -

Log GDP of the partner country 1.974*** 1.238 1.251 -

[0.688] [0.982] [0.987] -

FE Country × product; Country × product; Country × product; Country × product;

period; country × year; country × year; product × period;
period product × year; country × period;
period

Observations 1,084,560 1,084,560 1,084,560 1,084,560

R-squared 0.799 0.802 0.808 0.827
Robust standard errors in brackets; clustered at country level
*** p<0.01, ** p<0.05, * p<0.1

This table provides the motivation for choosing the main estimating equation (3) as the benchmark specification for the exports sample.
Each column contains a different set of fixed effects along with the three treatment dummy variables of interest.

1
 Table 10: Smart sanctions vs Embargo for exports’ sample

Dependent variable: log trade

(1) (2)
Benchmark Dynamics

Smart sanctions: Treatment effect -0.277**

[0.126]

Embargo: Treatment effect 0.004 -0.107

[0.087] [0.122]

Smart sanctions: Substitution effect 0.083

[0.101]

Embargo: Substitution effect 0.134 0.130

[0.094] [0.107]

Smart sanctions: Spillover effect -0.078

[0.071]

Embargo: Spillover effect -0.114*** -0.172**

[0.029] [0.072]

Log of trade cost -0.096** -0.096**

[0.043] [0.043]

Log GDP of the partner country 1.251 1.244

[0.987] [0.982]

FE Country × product; Country × product;

country × year; country × year;
product × year; product × year;
period period

Observations 1,084,560 1,084,560

R-squared 0.808 0.808
Robust standard errors in brackets; clustered at country level
*** p<0.01, ** p<0.05, * p<0.1

This table presents the results of the analysis of dynamic effects of the embargo on Russian exports.
The procedure for the dynamic exercise is described in detail in section 5.3 and in the footnote for Table 6.

1
 ADDENDUM II: FIGURES

Figure 1 depicts the changes in logs of Russian aggregate import trade flows of embargoed
and non-embargoed goods introduced by the Russian food embargo, imposed in August of
2014. Data source is UN COMTRADE. Each good in the original imports sample receives
an indicator as an embargoed or non-embargoed good, depending on whether it is included
in the embargoed products list by the Russian government. I the plot the average log trade
for each good category against month-year time periods.

1
Figure 2 plots 18 leads and 18 lags of the main coefficient of interest, which measures the
effect of the embargo on Russian imports of embargoed goods from the sanctioning countries.
The 95% CI is represented by the grey area around the coefficient line. The aggregate trade
flows decrease significantly when the embargo is introduced in August of 2014. Trade flows
do not experience significant deviations from trend in 18 months preceding the embargo.
Note that this figure does not fully demonstrate the effect of the embargo because it does
not account for the substitution effect.

2
Figure 3 depicts the estimated impact of the embargo on Russian aggregate trade flows,
comprised by the main (solid line) and substitution effects (dashed line). The 95% CI are
represented by the grey area around the coefficient line. Aggregate trade flows respond
significantly after the imposition of the embargo in August of 2014: Russian imports of
embargoed goods from the sanctioning countries falls, and a substitution effect is observed
(imports of embargoed goods are redirected towards the non-sanctioning countries). The
substitution effect does not fully offset the negative main effect of the embargo.

3
Appendix

Table 1A: List of sanctioning countries

Country name COMTRADE code

Albania 8
Australia 36
Austria 40
Belgium 56
Bulgaria 100
Canada 124
Croatia 191
Cyprus 196
Czech Republic 203
Denmark 208
Estonia 233
Finland 246
France 251
Germany 276
Greece 300
Hungary 348
Iceland 352
Ireland 372
Italy 381
Latvia 428
Lichtenstein/Switzerland 757
Lithuania 440
Luxembourg 442
Malta 470
Montenegro 499
Netherlands 528
Norway 579
Poland 616
Portugal 620
Romania 642
Slovakia 703
Slovenia 705
Spain 724
Sweden 752
Turkey 792
Ukraine 804
United Kingdom 826
USA 842

1
 Table 2A: List of embargoed products
Sanctioned Description
products
0201 Meat and edible meat offal
0202 Meat of bovine animals, frozen.
0203 Meat of swine, fresh, chilled or frozen.
0207 Meat and edible offal, of the poultry of heading 01.05, fresh, chilled or
frozen.
0210 Meat and edible meat offal, salted, in brine, dried or smoked; edible flours
and meals of meat or meat offal.
0301 Live fish.
0302 Fish, fresh or chilled, excluding fish fillets and other fish meat of heading
03.04.
0303 Fish, frozen, excluding fish fillets and other fish meat of heading 03.04.
0304 Fish fillets and other fish meat (whether or not minced), fresh, chilled or
frozen.
0305 Fish, dried, salted or in brine; smoked fish, whether or not cooked before
or during the smoking process; flours, meals and pellets of fish, fit for
human consumption.
0306 Crustaceans, whether in shell or not, live, fresh, chilled, frozen, dried,
salted or in brine; smoked crustaceans, whether in shell or not, whether
or not cooked before or during the smoking process; crustaceans, in shell,
cooked by steaming or by boiling in water, whether or not chilled, frozen,
dried.
0307 Molluscs, whether in shell or not, live, fresh, chilled, frozen, dried, salted
or in brine; smoked molluscs, whether in shell or not, whether or not
cooked before or during the smoking process; flours, meals and pellets of
molluscs, fit for human consumption.
0308 Aquatic invertebrates other than crustaceans and molluscs, live, fresh,
chilled, frozen, dried, salted or in brine; smoked aquatic invertebrates other
than crustaceans and molluscs, whether or not cooked before or during the
smoking process; flours, meals and pellets of aquatic invertebrates other
than crustaceans and molluscs, fit for human consumption.
0401 Milk and cream, not concentrated nor containing added sugar or other
sweetening matter.
0402 Milk and cream, concentrated or containing added sugar or other sweet-
ening matter.
0403 Buttermilk, curdled milk and cream, yogurt, kephir and other fermented
or acidified milk and cream, whether or not concentrated or containing
added sugar or other sweetening matter or flavoured or containing added
fruit, nuts or cocoa.
0404 Whey, whether or not concentrated or containing added sugar or other
sweetening matter; products consisting of natural milk constituents,
whether or not containing added sugar or other sweetening matter, not
elsewhere specified or included.
0405 Butter and other fats and oils derived from milk; dairy spreads.

2
0406 Cheese and curd.
0701 Potatoes, fresh or chilled.
0702 Tomatoes, fresh or chilled.
0703 Onions, shallots, garlic, leeks and other alliaceous vegetables, fresh or
chilled.
0704 Cabbages, cauliflowers, kohlrabi, kale and similar edible brassicas, fresh
or chilled.
0705 Lettuce (Lactuca sativa) and chicory (Cichorium spp.), fresh or chilled.
0706 Carrots, turnips, salad beetroot, salsify, celeriac, radishes and similar ed-
ible roots, fresh or chilled.
0707 Cucumbers and gherkins, fresh or chilled.
0708 Leguminous vegetables, shelled or unshelled, fresh or chilled.
0709 Other vegetables, fresh or chilled.
0710 Vegetables (uncooked or cooked by steaming or boiling in water), frozen.
0711 Vegetables provisionally preserved (for example, by sulphur dioxide gas, in
brine, in sulphur water or in other preservative solutions), but unsuitable
in that state for immediate consumption.
0712 Dried vegetables, whole, cut, sliced, broken or in powder, but not further
prepared.
0713 Dried leguminous vegetables, shelled, whether or not skinned or split.
0714 Manioc, arrowroot, salep, Jerusalem artichokes, sweet potatoes and sim-
ilar roots and tubers with high starch or inulin content, fresh, chilled,
frozen or dried, whether or not sliced or in the form of pellets; sago pith.
0801 Coconuts, Brazil nuts and cashew nuts, fresh or dried, whether or not
shelled or peeled.
0802 Other nuts, fresh or dried, whether or not shelled or peeled.
0803 Bananas, including plantains, fresh or dried.
0804 Dates, figs, pineapples, avocados, guavas, mangoes and mangosteens, fresh
or dried.
0805 Citrus fruit, fresh or dried.
0806 Grapes, fresh or dried.
0807 Melons (including watermelons) and papaws (papayas), fresh.
0808 Apples, pears and quinces, fresh.
0809 Apricots, cherries, peaches (including nectarines), plums and sloes, fresh.
0810 Other fruit, fresh.
0811 Fruit and nuts, uncooked or cooked by steaming or boiling in water, frozen,
whether or not containing added sugar or other sweetening matter.
0813 Fruit, dried, other than that of headings 08.01 to 08.06; mixtures of nuts
or dried fruits of this Chapter.
1601 Sausages and similar products, of meat, meat offal or blood; food prepa-
rations based on these products.
1901 Other

3
2106 Other
2501 Salt (including table salt and denatured salt) and pure sodium chloride,
whether or not in aqueous solution or containing added anti-caking or
free-flowing agents; sea water.

4
 Table 3A: Embargo start time

Dependent variable: log trade

(1) (2)
Benchmark specification July trend

Sanctioning country x Embargoed product -1.742***

(Treatment effect) [0.216]

Non-sanctioning coutry x Embargoed product 0.452***

(Substitution effect) [0.154]

Sanctioning country x Non-embargoed product -0.052

(Non-embargo effect) [0.035]

Treatment effect -1.295***

[0.197]

Substitution effect 0.506***

[0.177]

Non-embargo effect -0.008

[0.037]

Log trade cost -0.185*** -0.194***

[0.055] [0.053]

Log GDP of the partner country 0.634 0.598

[0.657] [0.611]

FE Country × product; Country × product;

country × year; country × year;
product × year; product × year;
period period

Observations 1,587,216 1,587,216

R-squared 0.807 0.807
Robust standard errors in brackets; clustered at country level
*** p<0.01, ** p<0.05, * p<0.1

This table addresses the concern of a negative trend in Russian aggregate trade flows in July of 2014,
which is significant at 10% (for reference see Figure 3). To address this issue, I generate a new time indicator
and expand the embargo’s time to include July 2014. I then generate the treatment dummies as described in
section 3 and re-estimate the benchmark model with the new treatment dummies. These results are presented in
column 2. The estimation results of the original benchmark model are presented in column 1.