International Trade, Quality Sorting and Trade Costs:

The Case of Cognac∗

Antoine Bouet†, Charlotte Emlinger‡, Viola Lamani§

Abstract

This paper tests empirically the validity of the Alchian-Allen effect, using an original

dataset of French Cognac exports by quality designation. More specifically, we estimate

econometrically the impact of distance and tariffs on the quality mix of Cognac im-

porters. The definition of quality, based on the minimum time in oak of the youngest

eau-de-vie used in creating the blend, is subject to regulations and is constant and

objective, which makes the case of Cognac particularly relevant to analyze the impact

of different trade costs on the quality mix. Our econometric estimates confirm the

Alchian-Allen effect by showing the positive effect of distance on the share of high

quality (XO) in total imports of Cognac. Using detailed information on tariffs, we

show that ad valorem and per-unit duties have opposite effects on Cognac imports by

quality.

Keywords: Quality mix, Luxury product, distance, tariffs

JEL: F10, F13, F14.

∗Very preliminary draft. Please do not quote.†University of Bordeaux and International Food Policy Research Institute (IFPRI); Corresponding author:

a.bouet@cgiar.org‡Centre d’Etudes Prospectives et d’Informations Internationales (CEPII)§University of Bordeaux

1

1 Introduction

Transportation costs increase the relative demand for higher quality goods. First highlighted

by Alchian and Allen in 1964, this point gave rise to several theoretical analysis (Borcherding

and Silberberg 1978, Bauman 2004, Baldwin and Harrigan 2011, Sorensen 2014, Irarraza-

bal et al. 2015) but to seldom empirical verifications. Among them, most studies use unit

prices to proxy the quality of the product, as Hummels and Skiba (2004), except Lawson

and Raymer (2006) and Nesbit (2007) who rely on the different grades of gasoline to test

the Alchian-Allen theorem on the gasoline market.

In this paper, we test empirically the validity of the Alchian-Allen effect, using an origi-

nal dataset of French Cognac exports by quality designation. More specifically, we estimate

econometrically the impact of trade costs (distance and tariffs), on the quality mix of Cognac

importers. Focusing on Cognac to discuss the Alchian-Allen conjecture is particularly rel-

evant. Indeed, the definition of Cognac quality is based on the minimum time in oak of

the youngest eau-de-vie used in creating the blend. This ageing process and indications

are subject to regulations and are closely supervised by the Cognac inter-profession (the

Bureau National Interprofessionnel du Cognac BNIC). As a consequence, the definition of

quality for this product is objective and constant, contrary to other empirical analyses of

the quality mix in the literature which use the difference observed in the unit values for a

given category (see for example Hummels and Skiba 2004, Schott 2004) or on ratings from

experts or guidebooks (see for example Crozet, Head and Mayer 2012, for champagne).

Our work contributes to the literature on trade costs and their impact on product qual-

ity. As said above, despite distinguished theoretical studies (Verhoogen 2008, Hallak and

Sivadasan 2009), empirical work on the subject remains relatively scarce. Papers studying

the Alchian-Allen effect estimate the impact of transportation costs (usually distance) on

unit values, using either country-level data (Schott 2004, 2008, Hummels and Klenow 2005,

Baldwin and Harrigan 2011) or firm-level data (Bastos and Silva 2010, Manova and Zhang

2011, Martin 2012). Hummels and Skiba (2004) empirically show that the Alchian-Allen

assumption is mainly driven by per-unit shipping costs, that is theoretically confirmed by

Irarrazabal et al. 2015 with a Melitz-type model with additive transport costs. In this pa-

per, we estimate the impact of distance and also of the ad valorem and per-unit component

of tariffs on the quality mix of Cognac imports.

Our paper also relates to the literature on luxury goods. Indeed, as other spirits, Cognac

is a luxury product, based on Fogarty’s (2010) estimations of income-elasticity of demand

for spirits. Moreover, important Cognac Houses, e.g. Martell and Remy Martin, belong to

the famous list of the Comite Colbert, an association of 75 French luxury brands, founded

in 1954 by Jean-Jacques Guerlain to promote the concept of luxury. Among the different

papers on trade of luxury goods, Fontagne and Hatte (2013) estimate a gravity equation

of high-end products and find that exports of these products are less sensitive to distance

and more sensitive to Gross Domestic Product (GDP) than other products. For their part,

Martin and Mayneris (2015) conclude that high-end variety exporters are mainly charac-

terized by more distant markets on average, and that this goes hand-in-hand with a wider

geographic diversification of their exports. Our paper investigates the impact of trade costs

on the quality mix for a given luxury product, Cognac.

The contribution of this article is threefold. First, we show that distance has a positive effect

2

on the share of high quality (XO) in total imports of Cognac. By doing so, we confirm the

Alchian-Allen effect using an objective measure of quality. Second, we estimate the effect

of custom duties on the quality mix of Cognac imports, distinguishing the ad valorem from

the per-unit component of tariffs, and show that their impact on the shares of high and

low quality Cognac is different. Third, we confirm that Cognac is on average (whatever the

quality) a luxury product, as its imports increase with the price of the product, highlighting

a Veblen effect.

The paper is organized as follows. In a first section, we present our original database of

Cognac exports and give some details on the different quality designations. In a second

section, we perform an econometric estimation on the shares of high and low quality Cognac

exports to different countries. Section three proposes different robustness checks of these

estimations, using a longer period of time, or including domestic taxes in the model. We

finally end our paper by some concluding remarks.

2 Data and stylized facts

2.1 Cognac Quality Designations

The definition of the different qualities in Cognac depends on the production process. More

precisely, the classification in a given designation is based on the minimum time in oak of the

youngest eau-de-vie used in creating the blend. Cognac VS (for Very Special) is the lowest

quality of Cognac: the minimum time in oak legally required by the youngest eau-de-vie in

creating a VS cognac is two years. Cognac VSOP (for Very Superior Old Pale) is the medium

quality of Cognac: the minimum time in oak legally required by the youngest eau-de-vie in

creating a VSOP cognac is four years. Cognac XO (for Extra Old) is the highest quality of

Cognac: the minimum time in oak legally required by the youngest eau-de-vie in creating

a XO cognac is six years (see figure 1). The ageing process and indications are adopted by

all the ‘Maisons de Cognac’. They are subject to regulations and are closely supervised by

BNIC. The quality of the different designations is recognized by consumers, as Cognac VS

prices range from 25 to 45 euros per bottle, Cognac VSOP from 32 to 57 euros per bottle

and finally Cognac XO from 45 to 94 euros per bottle in 20151.

2.2 An original dataset

Raw data of Cognac shipments by year, destination and quality designation were provided

by the Bureau National Interprofessionnel du Cognac (BNIC).2 Different issues were encoun-

tered when compiling these raw data, mainly due to the creation/separation of countries, or

currency changeover (details may be requested from the authors).

1All these figures are from Sud-Ouest - April 11, 2015. Some bottles of Cognac are much more expensive.

The price of a bottle of Cognac Hennessy Paradis was 924 euros on February 2015 on the website LaCog-

natheque. On the same website you could buy at the same date a bottle of Richard Hennessy for 2780 euros.

Paradis is made of eau-de-vie of at least 25 years (40 years in the case of a bottle of Richard Hennessy2We are grateful to the Bureau National Interprofessionnel du Cognac for helping us from the start

with the constitution of this database, as well as for the constant attention and the interesting suggestions

addressed to our work. In particular we thank Janine Bretagne, Christakis Christodoulou, Stephane Feuillet,

Sebastien Freulon, Lionel Lalague and Catherine Lepage.

3

Figure 2 shows the evolution of the quality mix of French Cognac exports to all destinations

for 1992-2013. The shares of VS, VSOP and XO slightly change during the period but the

relative importance of each designation remains the same. The share of VS in total exports

goes from 43% in 2013 to 57% in 2003, the share of high quality XO from 9% in 2003 to

15% in 2011.

Figures 3 and 4 show respectively the share of the lowest (VS) and highest (XO) quality of

Cognac in countries’ imports in 2013. The quality mix appears to vary among countries.

European and American countries exhibit a higher share of VS than Asian countries which

present the highest share of XO (see details for the main markets in descriptive statistics of

VS and XO share tables 1 and 2 in the Appendix).

2.3 Quality mix and trade costs

We saw in the previous sub-section that the quality mix of Cognac imports differs from

country to country. High quality (XO) accounts for more than a third of Cognac imports

in some countries (China, Hong Kong), while the lowest quality (VS) represents the largest

share of imports in USA, UK or Ireland. In this section, we analyze the quality mix of dif-

ferent markets, according to different measures of trade costs with France. We suspect that

trade costs, especially distance, partly explain the distribution of quality among imports of

Cognac.

Figure 5 plots the share of VS and XO designations in imports of various countries in 2013,

according to their distance to France. It shows a decreasing (respectively increasing) rela-

tionship between the VS (respectively XO) share in a country’s imports and its distance

to France. This stylized fact seems to confirm the Alchian-Allen conjecture, stating that

transport costs increase the relative demand for more expensive/higher quality goods.

A similar pattern can be observed in figure 6 which plots the nominal exchange rate of

countries’ currencies vis-a-vis the Euro in 2013 with the share of VS and XO in Cognac im-

ports.3 The appreciation of the Euro relative to the destination country’s national currency

seems to be in favor of high quality Cognac, as the share of XO increases with the exchange

rate. On the other side, the share of VS in imports appears to be negatively affected by

the exchange rate. These statistics might be an evidence of a Veblen effect in the case of

the highest quality of Cognac. The demand for this product increases with its price, due to

conspicuous consumption.

Finally, we look at the relationship between the share of VS and XO in total imports of each

country and custom duties. In the case of Cognac, two types of custom instruments are

applied: ad valorem duties, that are defined in percentage terms, and per-unit or specific

duties, defined in monetary units by units of volume. We use information on ad-valorem

and specific duties on an annual basis from the International Trade Center.4 Our database

covers 18 years (1996 to 2013) and 193 importing countries. Figures 7 and 8 show the share

of VS and XO according to respectively ad valorem and per-unit duties. If ad valorem duties

seem to have a negative impact on the share of VS, there is no clear relationship between

per-unit duties and the shares of XO and VS, or between the share of XO and ad valorem

3Annual averages of daily observations are from fxtop.com. Whenever fxtop.com data were missing, we

resorted to annual averages of midpoint daily rates from oanda.com4Special thanks to Xavier Pichot from the International Trade Centre

4

duties. The following section proposes a more in-depth analysis of the relationship between

the Cognac trade quality mix and trade costs, using an econometric model.

3 Empirics

3.1 Empirical strategy and data

Our empirical analysis assesses the determinants of the quality mix of French exports of

Cognac. We study in particular to what extent trade costs impact the shares of high and

low quality Cognac in total exports. We expect the share of XO in total exports (i.e. high

quality) to be positively correlated with distance, which would confirm the Alchian-Allen

conjecture.

We estimate separately the determinants of the share of VS (low quality) and XO (high

quality) in total Cognac exports in volume. As we distinguish three different qualities of

Cognac, the two equations are not complementary, the reference being each time the two

other qualities. The equation we estimate for k = V S and k = XO is the following:

V oljkt∑k V oljkt

= β1 lnGDPjt + β2 lnGDP/capjt + β3 ln distj + β4 lnExRatejt + β5Landlockedj

+ β6 lnAdV aloremjt + β7 lnPer − unitjt + νt + uj,t (1)

With V oljkt the French export of quality k to country j the year t. GDPjt and

GDP/capjt are the GDP and the GDP per Capita of country j at year t, distj the dis-

tance of France to country j, ExRatejt the nominal exchange rate of country j’s currency

vis-a-vis the Euro at year t and Landlockedj a dummy variable equal to one whether the

country j is landlocked, and zero otherwise. AdV aloremjt and Per − unitjt are the ad

valorem and per-unit component of the tariff.5 Finally, time fixed effects are included.

We use trade and tariff data previously presented in section 1. Data on each country j’s

GDP and GDP per capita are in constant 2005 US dollars and were retrieved from the World

Bank World Development Indicators Databank. Nominal exchange rates are annual aver-

ages of daily observations from fxtop.com and we use CEPII’s weighted bilateral distance

measure and landlocked variable.

Equation 1 is estimated using OLS for the period 1996 − 2013, as tariffs data are not avail-

able before 1996. However we proceed to the same estimations without the duties variables

on a 47-year-long database on Cognac exports as robustness check (period 1967 − 2013, see

section 4).

3.2 Results

Table 1 and 2 report the results of our estimation of equation 1, with respectively k = V S

and k = XO. In both tables, column 1 reports our baseline estimate of equation 1 without

custom duties.

The distance estimated coefficient is significant and negative in the VS estimation (table 1

5we take the power of the duties, i.e. (1+ ad valorem) and (1+ per-unit)

5

) while its coefficient is positive in the XO estimations (table 2). The dummy variable land-

locked follows the same pattern. This result confirms the Alchian-Allen conjecture, stating

that countries export higher quality varieties to remote destinations, due to transportation

costs. These results are robust to the inclusion of ad valorem and per-unit duties in column

2 and to the use of the population of country j the year t instead of the GDP to control

for the size of the destination country in column 3. Column 4 performs another robustness

check of the baseline estimate by excluding countries with a high level of re-exportation

(Hong Kong, Netherlands, Panama, Singapore and United Arab Emirates).6 The distance

coefficient is slightly reduced, but remains negative and significant for the VS and positive

and significant for the XO.

Similarly, the estimated coefficient for the exchange rate is significant and negative for the

VS estimation and positive for the XO estimation. The appreciation of the Euro increases

the share of XO in Cognac imports, at the expense of VS, whose share is reduced. This

result suggests a Veblen effect in the case of Cognac’s highest quality. The demand for this

product increases with its price, due to conspicuous consumption.

Column 2 adds the ad valorem and per-unit components of custom duties. The ad valorem

duties have a negative and significant impact on the share of VS in Cognac imports of coun-

tries, while per-unit duties positively impact the share of VS (table 1). Again, these results

remain unchanged with the inclusion of the population in column 3 and when we restrict

the sample in column 4. On the contrary, the share of XO in imports is not impacted by

tariffs, as neither the ad valorem nor the per-unit component of tariffs have a significant

coefficient in table 2. As a consequence, tariffs seem to mainly impact the relative share of

VS and VSOP, as the XO market share is not impacted by duties.

The GDP of country j appears to have a negative impact on the share of VS and a positive

impact on the share of XO. We obtain similar results when we use the population of country

j instead of its GDP to control for the size of the importing country. This result means

that bigger countries tend to import a higher share of high quality Cognac, while small

countries import more low quality in proportion. The results of the GDP per capita are

counter-intuitive in both tables, as the variable exhibits a positive impact on both the VS

and XO shares. This result suggests that rich countries have a higher proportion of high

quality and low quality of Cognac, at the expense of middle quality. This result is robust to

the different specifications.

4 Robustness checks

4.1 Estimations on trade volumes

As a robustness check, we perform the same equation as equation 1 but on the volume of

VS and XO imports of country j instead of their share in total Cognac imports. We use the

following equation:

6list provided by the BNIC

6

lnV oljkt = β1 lnGDPjt + β2 lnGDP/capjt + β3 ln distj + β4 lnExRatejt + β5Landlockedj

+ β6 lnAdV aloremjt + β7 lnPer − unitjt + νt + uj,t (2)

Results for VS and XO Cognac imports are presented respectively in tables 3 and 4.

4.2 Estimation with long period

4.3 The impact of local taxes

5 Conclusion

Using a unique dataset of annual Cognac shipments by volume, destination and quality

designation, we test empirically the validity of Alchian and Allen’s (1964) thesis accord-

ing to which per-unit transportation costs increase the relative demand for higher-quality

goods. We find that the share of XO (higher-quality Cognac) in total imports increases with

distance, and the opposite effect for the share of VS imports (lower-quality Cognac). Our

results, robust to different specifications, seem to validate the Alchian-Allen effect meaning

that for all else equal, higher-quality varieties are shipped to remote destinations, due to

transportation costs. We also estimate the impact of policy-related costs: ad-valorem and

per-unit custom duties. Our results suggest that while ad-valorem (per-unit) duties have

a negative (positive) impact on the VS share of imports, they have no significant impact

on the higher-quality share of Cognac imports. Further research is needed especially in

understanding the relationship between the Cognac quality mix and custom duties and the

possible existence of a Veblen effect. We also plan to test the robustness of our results in

line with the Alchian-Allen conjencture using other specifications.

References

Alchian, A. A. and Allen, W. (1964). University Economics: Elements of Inquiry. Belmont,

CA: Wadsworth Publishing Company.

Anderson, J.E. and van Wincoop, E. (2004). Trade Costs. Journal of Economic Literature,

42(3), 691-751.

Baldwin, R. and Harrigan, J. (2011). Zeros, Quality, and Space: Trade Theory and Trade

Evidence. American Economic Journal: Microeconomics, 3(2), 60-88.

Bastos, P. and Silva, J. (2010). The quality of a firm’s exports: Where you export to matters.

Journal of International Economics, 82(2): 99-111.

Bauman, Y. (2004). Shipping the good apples out: A new perspective. Economic Inquiry,

42(3), 534-536.

Borcherding, T. E., and Silberberg, E. (1978). Shipping the Good Apples Out: The Alchian

and Allen Theorem Reconsidered. Journal of Political Economy, 86(1), 131-138.

7

Crozet, M., Head, K. and Mayer, Th. (2012). Quality Sorting and Trade: Firm-level Evidence

for French Wine. Review of Economic Studies, 79(2), 609-644.

Fogarty, J. (2010). The demand for beer, wine and spirits: A survey of the literature. Journal

of Economic Surveys, 24(3), 428-478.

Fontagne, L. and Hatte, S. (2013). European high-end products in international competition.

Paris School of Economics Working Paper No. 37.

Hallak, J. C. and Sivadasan, J. (2009). Firms’ Exporting Behavior under Quality Con-

straints, NBER Working Papers No. 14928.

Hummels, D. and Klenow, P.J. (2005). The Variety and Quality of a Nation’s Exports.

American Economic Review, 95(3): 704-723.

Hummels, D. and Skiba, A. (2004). Shipping the Good Apples Out? An Empirical Confirma-

tion of the Alchian-Allen Conjecture. Journal of Political Economy, 112(6), 1384-1402.

Irarrazabal, A., Moxnes, A. and Opromolla, L. D. (2015). The tip of the Iceberg: A Quan-

titative Framework for Estimating Trade Costs. Review of Economics and Statistics,

97(4), 777-792.

Jorgensen, J. G. and Schrder P. J. H. (2005). Welfare-ranking Ad Valorem and Specific

Tariffs under Monopolistic Competition. Canadian Journal of Economics, 38(1), 228-

241.

Lawson, R. and Raymer, L. (2006). Testing the Alchian-Allen Theorem: A study of Con-

sumer Behavior in the Gasoline Market. Economics Bulletin, 35(4), 1-6.

Manova, K. and Zhang, Z. (2009). Export Prices across Firms and Destinations. The

Quaterly Journal of Economics, 127(1), 379-436.

Martin, J. (2012). Markups, quality, and transport costs. European Economic Review, 56(4),

777-791.

Martin, J., and Mayneris, F. (2015). High-end variety exporters defying gravity: Micro facts

and aggregate implications. Journal of International Economics, 96(1), 55-71.

Nesbit, T. (2007). Excise Taxation and Product Quality: The Gasoline Market. Economic

Issues, 12(2), 1-14.

Schott, P. K. (2004). Across-product Versus Within-product Specialization in International

Trade. The Quarterly Journal of Economics, 119(2), 646-678.

Schott P. K. (2008). The relative Sophistication of Chinese Xxports. Economic Policy,

23(53), 5-49.

Sorensen, A. (2014). Additive versus multiplicative trade costs and the gains from trade

liberalizations. Canadian Journal of Economics, 47(3), 1032-1046.

Verhoogen, E. A. (2008). Trade, quality upgrading, and wage inequality in the Mexican

manufacturing sector. The Quarterly Journal of Economics, 123(2), 489-530.

8

Figure 1: Definition of Cognac Qualities

Figure 2: Quality mix of Cognac exports

40%

50%

60%

70%

80%

90%

100%

0%

10%

20%

30%

VS VSOP XO

9

Figure 3: Share of VS in Cognac’s imports in 2013

[0,20](20,40](40,60](60,80](80,100]No data

Figure 4: Share of XO in Cognac’s imports in 2013

[0,20](20,40](40,60](60,80](80,100]No data

Figure 5: Share in imports according to distance

050

100

6 7 8 9 10

2013

95% CI Fitted valuesVS Share

ln Distance

Graphs by YEAR

(a) VS

020

4060

6 7 8 9 10

2013

95% CI Fitted valuesXO Share

ln Distance

Graphs by YEAR

(b) XO

10

Figure 6: Share in imports according to the exchange rate

050

100

0 5 10

2013

95% CI Fitted valuesVS Share

ln Exchange rate

Graphs by YEAR

(a) VS

020

4060

0 5 10

2013

95% CI Fitted valuesXO Share

ln Exchange rate

Graphs by YEAR

(b) XO

Figure 7: Share in imports according to the ad valorem duties

-50

050

100

0 .5 1

2013

95% CI Fitted valuesVS Share

ln Ad valorem duty

Graphs by YEAR

(a) VS

020

4060

0 .5 1

2013

95% CI Fitted valuesXO Share

ln Ad valorem duty

Graphs by YEAR

(b) XO

Figure 8: Share in imports according to the per-unit duties

050

100

0 2 4 6 8

2013

95% CI Fitted valuesVS Share

ln Per Unit duty

Graphs by YEAR

(a) VS

020

4060

0 2 4 6 8

2013

95% CI Fitted valuesXO Share

ln Per Unit duty

Graphs by YEAR

(b) XO

11

Table 1: Market share of VS

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

GDPjt -3.94*** -3.98*** -4.00***

(0.25) (0.29) (0.29)

GDP per capitajt 2.67*** 2.09*** -1.89*** 2.75***

(0.40) (0.52) (0.49) (0.52)

Distancej -11.66*** -11.80*** -11.81*** -9.95***

(0.61) (0.73) (0.73) (0.76)

Exchange ratesjt -0.76*** -0.97*** -0.97*** -1.20***

(0.19) (0.23) (0.23) (0.23)

Landlockedj -7.84*** -9.04*** -9.06*** -8.22***

(1.42) (1.71) (1.71) (1.70)

Ad valorem Dutyjt -3.57** -3.57** -4.58***

(1.54) (1.54) (1.53)

Per-Unit Dutyjt 1.06*** 1.07*** 0.72***

(0.25) (0.25) (0.25)

Populationjt -4.01***

(0.29)

N 2,487 1,832 1,832 1,754

r2 0.239 0.246 0.247 0.244

Notes: All variables in logarithms. Year fixed effects included.

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

Table 2: Market share of XO

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

GDPjt 0.91*** 0.91*** 0.88***

(0.14) (0.16) (0.16)

GDP per capitajt 0.74*** 0.79*** 1.71*** 0.49*

(0.23) (0.29) (0.27) (0.29)

Distancej 5.84*** 5.09*** 5.09*** 4.61***

(0.35) (0.41) (0.41) (0.43)

Exchange ratesjt 0.70*** 1.00*** 1.00*** 0.98***

(0.11) (0.13) (0.13) (0.13)

Landlockedj 4.91*** 4.68*** 4.69*** 4.40***

(0.80) (0.95) (0.95) (0.96)

Ad valorem Dutyjt 0.59 0.59 0.76

(0.86) (0.86) (0.87)

Per-Unit Dutyjt -0.02 -0.02 0.05

(0.14) (0.14) (0.14)

Populationjt 0.92***

(0.16)

N 2,487 1,832 1,832 1,754

r2 0.143 0.144 0.144 0.133

Notes: All variables in logarithms. Year fixed effects included.

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

12

Table 3: Volume of VS exports

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

GDPjt 0.50*** 0.51*** 0.51***

(0.02) (0.02) (0.02)

GDP per capitajt 0.65*** 0.60*** 1.10*** 0.57***

(0.03) (0.04) (0.04) (0.04)

Distancej -0.82*** -0.95*** -0.95*** -1.03***

(0.05) (0.06) (0.06) (0.06)

Exchange ratesjt -0.08*** -0.09*** -0.09*** -0.08***

(0.02) (0.02) (0.02) (0.02)

Landlockedj -0.64*** -0.89*** -0.89*** -0.92***

(0.11) (0.14) (0.14) (0.14)

Ad valorem Dutyjt -0.34*** -0.34*** -0.31**

(0.13) (0.13) (0.13)

Per-Unit Dutyjt 0.15*** 0.15*** 0.16***

(0.02) (0.02) (0.02)

Populationjt 0.51***

(0.02)

N 2,487 1,832 1,832 1,754

r2 0.596 0.602 0.601 0.594

Notes: All variables in logarithms. Year fixed effects included.

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

Table 4: Volume of XO exports

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

GDPjt 0.81*** 0.84*** 0.84***

(0.02) (0.02) (0.02)

GDP per capitajt 0.60*** 0.53*** 1.37*** 0.43***

(0.03) (0.04) (0.04) (0.04)

Distancej 0.16*** -0.02 -0.02 -0.26***

(0.05) (0.06) (0.06) (0.06)

Exchange ratesjt -0.01 -0.00 -0.00 0.02

(0.02) (0.02) (0.02) (0.02)

Landlockedj 0.10 -0.15 -0.14 -0.27**

(0.12) (0.14) (0.14) (0.13)

Ad valorem Dutyjt -0.20 -0.20 -0.10

(0.13) (0.13) (0.12)

Per-Unit Dutyjt 0.12*** 0.12*** 0.16***

(0.02) (0.02) (0.02)

Populationjt 0.84***

(0.02)

N 2,487 1,832 1,832 1,754

r2 0.588 0.589 0.589 0.603

Notes: All variables in logarithms. Year fixed effects included.

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

13

6 Appendix

Table 1: Descriptive Statistics - share of VS in imports of main markets

Share in Share in Share of VS

total exports VS exports Nb. Obs. mean sd min max

Total - - 2842 45.22 27.57 0.00 99.57

USA 32.20 50.61 22 74.56 2.92 69.72 79.97

Singapore 17.62 3.52 22 1.95 2.15 0.32 8.89

China 12.66 2.16 22 8.30 16.41 0.00 71.15

UK 6.43 12.35 22 86.42 1.69 81.99 89.03

Germany 3.73 4.11 22 56.45 5.05 47.60 63.03

Hong Kong 2.93 0.26 22 1.75 1.07 0.64 3.99

Netherlands 2.27 2.33 22 60.59 8.93 45.54 74.46

Norway 1.73 1.62 22 51.47 9.65 35.19 66.01

Finland 1.42 1.93 22 60.82 3.52 53.90 67.71

Russia 1.33 1.62 22 37.00 11.62 8.24 55.01

Latvia 1.30 1.28 21 39.64 10.15 7.84 51.06

Malaysia 1.04 0.01 22 0.60 0.38 0.18 1.62

Nigeria 1.03 1.42 22 34.36 19.93 0.02 67.00

Ireland 0.93 2.07 22 97.96 0.60 96.68 98.96

Canada 0.89 1.35 22 65.40 4.88 51.57 73.35

Table 2: Descriptive Statistics - share of XO in imports of main markets

Share in Share in

total exports XO exports Nb. Obs. mean sd min max

Total - - 2842 13.64 14.47 0.01 95.01

USA 32.20 8.80 22 3.37 1.26 1.89 7.22

Singapore 17.62 27.54 22 27.48 6.36 18.98 39.08

China 12.66 28.24 22 34.94 15.69 10.16 73.42

UK 6.43 0.71 22 1.53 0.18 1.22 2.00

Germany 3.73 2.54 22 5.63 2.14 2.20 9.50

Hong Kong 2.93 9.93 22 38.80 9.19 17.93 53.78

Netherlands 2.27 1.11 22 4.73 1.87 1.89 7.92

Norway 1.73 2.75 22 18.12 7.93 2.13 26.05

Finland 1.42 1.11 22 11.21 4.44 6.16 23.68

Russia 1.33 0.94 22 12.25 4.56 2.43 22.23

Latvia 1.30 1.02 21 11.42 5.71 1.15 27.46

Malaysia 1.04 3.72 22 29.50 12.33 12.98 49.93

Nigeria 1.03 0.19 22 1.96 0.95 0.38 4.36

Ireland 0.93 0.01 22 0.25 0.08 0.11 0.44

Canada 0.89 0.43 22 6.68 1.21 3.66 8.46

14

Figure 1: Distribution

0.0

05.0

1.0

15.0

2.0

25D

ensi

ty

0 20 40 60 80 100VS Share

(a) Share of VS

0.0

2.0

4.0

6.0

8D

ensi

ty

0 20 40 60 80 100XO Share

(b) Share of XO

15