This article was downloaded by: [Universiti Putra Malaysia]

On: 19 July 2012, At: 20:46

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Applied Economics Letters

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/rael20

An application of the Ricardian trade model with trade

costs
Andrew J. Cassey

School of Economic Sciences, Washington State University, PO Box 646210, Pullman, WA,
99164, USA
Version of record first published: 22 Nov 2011

To cite this article: Andrew J. Cassey (2012): An application of the Ricardian trade model with trade costs, Applied
Economics Letters, 19:13, 1227-1230
To link to this article: http://dx.doi.org/10.1080/13504851.2011.617871

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.

Applied Economics Letters, 2012, 19, 12271230

An application of the Ricardian trade

model with trade costs
Andrew J. Cassey

Downloaded by [Universiti Putra Malaysia] at 20:46 19 July 2012

School of Economic Sciences, Washington State University, PO Box 646210,

Pullman, WA 99164, USA
E-mail: cassey@wsu.edu

Deardorffs (2004) broad definition of technology in Ricardian trade

models is useful for extending the explanatory power of comparative
advantage to account for a fact on firm level exporter clustering
unexplained under the standard definition.
Keywords: international trade; Ricardian; technology; agglomeration
JEL Classification: F10; B12

I. Introduction

for this fact if trade costs are included in technology

but not otherwise.1

The Ricardian trade theory is a cornerstone of economics. Its lesson on the efficiency of allocating
resources according to comparative advantage is
taught from high school to graduate courses.
Principle classes illustrate this lesson using a tworegion, two-goods model. Regions differ in the technology for turning labour, the unique input, into two
goods, which are traded without cost. Jones (1961),
among others, generalized the original Ricardian theory to the multi-region, multi-goods case. Shiozawa
(2007) generalized the theory further, to include intermediate goods and technology choice.
Deardorff (2004) extended the Ricardian model by
including trade costs in the technology of producing
and delivering a good from one country to another. I
use Deardorffs extension to account for a recently
documented fact: agglomeration of exporters by destination of shipments. Using Russian customs data,
Cassey and Schmeiser (2010) showed that exporting
firms are physically clustered by the destination of
shipments in addition to clustering around ports.
This fact is also documented by Koenig (2009) using
other data. I show the Ricardian model can account

II. Facts on Exporter Agglomeration

Glejser et al. (1980) showed exporting firms are physically clustered around ports but Cassey and
Schmeiser (2010) went further and showed that in
addition, exporters are clustered by the destination
of their shipments. Exporting firms in the same region
are more likely to send shipments to the same country
than exporting firms in another region identical in
size, distance and industrial composition.
The evidence is presented in Table 1. The table
contains the Morans (1950) I statistic for spatial
correlation:
PP
N
I PP
i

wij

wij Xi  Xc Xj  Xd

P

Xi  Xc 2

where N is the number of regions indexed by i and j, Xc

is the number of exporters shipping to country c, Xc is

Cassey and Schmeiser (2010) accounted for the agglomeration of exporters by destination by positing an externality in
shipping costs. An externality is consistent with the Ricardian theory given here, but it is just one story possible under a
Ricardian framework. Koenig (2009) used a logit estimator to document this fact, but does not account for it.
Applied Economics Letters ISSN 13504851 print/ISSN 14664291 online # 2012 Taylor & Francis
http://www.tandfonline.com
http://dx.doi.org/10.1080/13504851.2011.617871

1227

A. J. Cassey

1228
Table 1. Morans I for location of firms exporting to select countries

Canada
China
Germany
Great Britain
Japan
Poland
United States
Ukraine

Canada

China

Germany

Great Britain

Japan

Poland

United States

Ukraine

0.0345*
0.0089
0.0249
0.0215
0.0221
0.0146
0.0189
0.0258

-0.0012
0.0023
-0.0472*
-0.5160*
-0.0164
-0.0514*
-0.0367*
-0.0426*

0.0243
-0.0448*
0.0324*
0.0677**
0.0319
0.0581*
0.0413
0.0593*

0.0313
-0.0458*
0.0610**
0.0886***
0.0533*
0.0370
0.0571*
0.0801**

0.0300
-0.0147
0.0248
0.0676**
0.0227
0.0446
0.0525**
0.0626**

0.0196
-0.0450*
0.0547*
0.0277
0.0337
0.0115
0.0307
0.0520

0.0194
-0.0375*
0.0406
0.0783**
0.0593**
0.0450
0.0369*
0.0773**

0.0340
-0.0408*
0.0637**
0.0916***
0.0677**
0.0658**
0.0751**
0.0220

Downloaded by [Universiti Putra Malaysia] at 20:46 19 July 2012

Source: Authors calculation using data from Cassey and Schmeiser (2010).
Notes: Observations on the diagonal are the Morans I for the location of exporters shipping to that country. Off-diagonals are
the bivariate Morans I representing the degree of spatial correlation between Xc and Xd . The statistic has been normalized with
respect to regional Gross Domestic Product to account for the geographic lumpiness of Russian economy activity. Columns are
the c countries and the rows are the d countries. Bold observations are the bivariate statistic where both countries have a
statistically significant univariate Morans I.
*, ** and *** Indicate p-values less than 0.1, 0.05 and 0.01, respectively.

the mean across regions, Xd is the number of exporters

shipping to country d and wij is a spatial weight. (I use
an inverse Euclidean distance weight so that regions
whose centroids are closer are weighted more heavily.)
The statistic ranges from -1 to 1, with positives indicating clustering. Under no spatial correlation,
EI 1=N  1.
The Morans I is univariate or bivariate. For the
univariate version, the statistic differences the number
of exporters in each region i shipping to country c
from the average number of exporters also shipping
to c and compares it to the difference between the
number of exporters in another region j shipping to
country c from the average. A statistic different from
E(I ) is evidence of spatial autocorrelation, clustering
if positive. The bivariate version is similar except cd,
so the statistic differences the number of exporters in
each region i shipping to country c from the average
number of exporters also shipping to c and compares
it to the difference between the number of exporters in
another region j shipping to country d from the average also shipping to d. Because of its construction with
respect to the other country d, the bivariate Morans I
is not symmetric. A statistic different from E(I ) suggests a spatial pattern, clustering if postitive, between
exporters shipping to country c and those shipping to
country d.
The data to calculate the statistic are the number of
manufacturing firms exporting from each of 89
Russian regions to eight countries and are from
Cassey and Schmeiser (2010). The disadvantage of
these data is that exporter location is known only to
the regional level.
Observations on the diagonal of Table 1 are
the univariate Morans I for the location of exporters shipping to that country. Stars indicate statistically significant autocorrelation or clustering.

Off-diagonals are the bivariate Morans I indicating

the strength of spatial correlation between exporters
shipping to the two countries. If Russian exporters are
clustered around two destinations, such as Canada
and the United States (indicated by stars on the diagonal), the lack of off-diagonal stars indicates the pattern of clustering is not the same for exporters
shipping to those countries. If Russian exporters are
not clustered around either of the two destinations,
such as China and Poland (indicated by no stars on the
diagonal), off-diagonal stars indicate the randomness
in exporter location cannot be distinguished between
exporters shipping to China and exporters shipping to
Poland. (In this case, the negative sign indicates dispersion.) If there is a diagonal star for one country but
not the other and the off-diagonal is starred, such as
Japan and the United States, this indicates spatial
correlation in that exporters to Japan tend to cluster
about the exporters to the United States though
exporters to Japan do not exhibit spatial
autocorrelation.
Table 1 shows that Russian exporting firms cluster
by destination for four of the eight countries in the
sample: Canada, Germany, Great Britain and the
United States. Furthermore, the pattern of clustering
is not the same for seven of twelve possibilities.
Therefore, there is evidence of firm-level clustering
by destination of shipments.

III. A Ricardian Model and an Example

In a two-region, two-goods, zero trade costs Ricardian
model, if acg is the exogenous unit labour requirement
in region c to produce good g, then region 1 produces
and exports good 1 if and only if

Application of Ricardian model with trade costs

a11
a21

a12
a22

In Deardorffs (2004) model with trade costs from

shipping good g from region c to country d, tcgd , region
1 exports good 1 to country 1 if and only if

Downloaded by [Universiti Putra Malaysia] at 20:46 19 July 2012

a11 t111
a21 t211

a12 t122
a22 t222

The data in Section II show that firms in different

Russian regions export to Canada and the United
States despite that all Russian firms face the same
importer tariffs, language and cultural barriers, and
that these countries are roughly the same distance and
buy the same products from Russia. The standard
Ricardian model cannot account for this because it
predicts whichever region has the lowest relative
labour cost will export the good to both countries.
Now consider a model with two regions with access
to the only resource, labour. They produce the same
good, but they do not consume it. Instead there is
potential sales to Canada and the United States.
There is a technology that converts labour into exports
deliverable to the United States and there is a different
technology that converts labour into exports deliverable to Canada. Table 2 shows the units of labour
required to deliver one unit of exports to each country.
The standard Ricardian definition of technology is
the transformation of labour into goods traded without cost. But the definition in this example is the
transformation of labour into a good deliverable to a
specific destination. It may take more labour for
region 1 to produce and deliver exports to the United
States than it does to Canada. This expanded interpretation of technology incorporates variable trade
costs such as physical distance and fixed trade costs
such as getting permits.
By reducing the number of goods from two to one,
Equation 3 simplifies to a1 t11 =a1 t12 
a2 t21 =a2 t22 . A change of notation of
ac tcd xcd reduces this example to Equation 2
where x replaces a and d replaces g.
In Table 2, the cost of region 1 delivering a unit of
exports to Canada is two exports to the United States
Table 2. An example

Region 1
Region 2

Canada

United States

10
6

5
6

Note: Units of labour required to deliver one export from

each region to each country.

1229
whereas region 2 can deliver a unit of exports to
Canada for one US export. Therefore, this theory
predicts that region 1 will produce exports to be delivered in the United States and region 2 will produce
exports to be delivered in Canada. This theory generates specialization of destinations by region: the firms
in region 1 specialize in exporting manufactured goods
to the United States instead of Canada because they
have a relative cost advantage in producing and delivering goods there.
Why does technology for producing and delivering
exports differ across regions? Why does technology
differ in Ricardos example of England and Portugal
producing wine and cloth? These are the assumptions
of the model.
The replacement of goods with countries in this
example from the standard Ricardian story yields an
interpretation of goods differentiated by place of consumption. This contrasts with Armington (1969)
because in that work exports are unique goods to the
producing country whereas here exports are unique
goods to the consuming country. British tea is tea
that is consumed in Britain, no matter where it is
made, rather than tea produced in Britain and consumed anywhere. The point, however, is this uncomfortable interpretation from redefining goods is not
necessary. Rather a broader interpretation of technology is acceptable and more digestible.

IV. Conclusion
David Ricardo described his model of trade using a
thought experiment. The theory has developed since
then, and though both theoretical and empirical challenges confront it, the Ricardian model has been enormously influential and persistent. Increasingly, the
international trade literature has been using customslevel data to identify new facts about international
trade. Cassey and Schmeiser (2010) used customs
data from Russia to show that exporters agglomerate
around the destination of exported shipments. I present more evidence of this exporter agglomeration by
destination.
My contribution is to show that Deardorffs (2004)
broader theoretical interpretation of technology can
account for exporter clustering by destination.
Technology transforms labour into a commodity
delivered to a destination and thus regions specialize
if they have sufficiently low delivery costs relative to
others. This is useful to extend the idea of the comparative advantage to account for facts that at first
seem inconsistent with standard Ricardian theory.

1230
Acknowledgements
Thanks to Jeremy Sage for ArcGIS and Geoda help,
Katherine Schmeiser for the data and Julian Diaz and
Mark Gibson for comments. Partial support for this
work by the Agricultural Research Center Project
#0540 at Washington State University.
References

Downloaded by [Universiti Putra Malaysia] at 20:46 19 July 2012

Armington, P. (1969) A theory of demand for products

distinguished by place of production, International
Monetary Fund Staff Papers, 16, 15978.
Cassey, A. and Schmeiser, K. (2010) The agglomeration
of exporters by destination, Forum for Research
on Empirical International Trade Working Paper
No. 233.
Deardorff, A. (2004) Local comparative advantage: trade
costs and the pattern of trade, University of Michigan

A. J. Cassey
Research Seminar in International Economics Discussion Paper No. 500, University of Michigan, Ann
Arbor, MI.
Glejser, H., Jacquemin, A. and Petit, J. (1980) Exports in
an imperfect competition framework: an analysis of
1,446 exporters, Quarterly Journal of Economics, 94,
50724.
Jones, R. (1961) Comparative advantage and the theory of
tariffs: a multi-country, multi-commodity model,
Review of Economic Studies, 28, 16175.
Koenig, P. (2009) Agglomeration and the export decisions
of French firms, Journal of Urban Economics, 66,
18695.
Moran, P. (1950) Notes on continuous stochastic phenomena, Biometrika, 37, 1723.
Shiozawa, Y. (2007) A new construction of Ricardian trade
theory a many-country, many-commodity case with
intermediate goods and choice of production techniques, Evolutionary and Institutional Economics Review,
3, 14187.