Singapore Management University
Institutional Knowledge at Singapore Management University
Research Collection Lee Kong Chian School Of
Business
Lee Kong Chian School of Business
10-1986
Determinants of Bilateral Trade Flows
Rajendra Kumar SRIVASTAVA
Singapore Management University, rajs@smu.edu.sg
Robert T. Green
Follow this and additional works at: http://ink.library.smu.edu.sg/lkcsb_research
Part of the Business Commons
Citation
SRIVASTAVA, Rajendra Kumar and Green, Robert T.. Determinants of Bilateral Trade Flows. (1986). The Journal of Business. , 59(4) ,
623. Research Collection Lee Kong Chian School Of Business.
Available at: http://ink.library.smu.edu.sg/lkcsb_research/4130
This Journal Article is brought to you for free and open access by the Lee Kong Chian School of Business at Institutional Knowledge at Singapore
Management University. It has been accepted for inclusion in Research Collection Lee Kong Chian School Of Business by an authorized administrator
of Institutional Knowledge at Singapore Management University. For more information, please email libIR@smu.edu.sg.
Rajendra K. Srivastava
Robert T. Green
University of Texas at Austin
Determinants of Bilateral
Trade Flows*
What factors determine the level of trade that
occurs between nations?Why does any given nation trade more with some nations than with
others?Are bilateraltradeflows of manufactured
items explained by the same factors as are raw
materials?These questions have received little
attentionin internationaltrade research. Considerable efforts have been made to understandthe
composition and the volume of nations' exports
and imports (see, e.g., Warner and Kreinin
1983). Yet the question, Why does country X
trade more with country Y than with country Z?
remains relatively less researched.
The study reportedin this paper representsan
attempt to understandthe relative level of trade
flows that occur between nations and their trading partners.It examines the bilateraltrade flows
between 45 exporting nations and 82 importing
nations in an effort to provide insights on the
questions posed above. The effects of multiple
determinantsof trade suggested by past literature are investigated in this study: distance,
product category, political instability, cultural
similarity,colonial past, membershipin an economic union, and standard demographic variables such as GDP and population.
Whatfactors influence
the level of trade flows
that occur between nations? Past studies
have answeredthis
question primarilyin
terms of distance and
economic size. The
present study considers
additionalvariablesas
potentialdeterminants
of trade relationsbetween nations. Also,
trade flows are analyzed by the type of
productsbeing exported and imported.
The findingsshow that
several additionalvariables are significantdeterminantsof the level
of bilateraltrade relations between nations.
In addition,it is found
that models of trade
flows are more appropriate for manufactured
goods than they are for
basic commodities.
* The authors wish to extend their appreciation to Tammy
B. Cornwell for assistance provided during all stages of this
project. The authors also acknowledge the support of the ICC
Institute and the Graduate School of Business of the University of Texas at Austin.
(Journal of Business, 1986,vol. 59, no. 4, pt. 1)
? 1986by The Universityof Chicago.All rightsreserved.
0021-9398/86/5904-0006$0
1.50
623
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
624
Journal of Business
The results of the study suggest relationsbetween the level of bilateral trade flows and the foregoing factors that have not been demonstrated previously. For instance, these factors have greater explanatory power for relative tradeflows of manufacturedgoods thanof such
basic productsas food and raw materials.In addition,politicalinstability is found to have a significantinfluenceon exports but little effect on
imports. Culturalsimilarityappearsto have a more pronouncedeffect
on relative trade flows than does shared membershipin an economic
union. Further, the GDP of the exporting country is found to be a
powerfulexplanatoryvariablein the relativeintensityof bilateraltrade
relations.The precedingfindingsare coupled with others that replicate
the findings of past research concerning such relations as those between distance and trade intensity.
The Determinantsof Trade:Past Research
Early studies on the determinantsof trade focused primarilyon the
relationbetween distance and trade. In general, it has been well established that distance is a strong determinantof the intensity of trade
flows that occur between nations;nationsthat are geographicallyproximate will tend to trade relatively more than will nations that are further apart (see, e.g., Beckerman 1956;Ullman 1956;Smith 1964;Linneman 1969;Yeats 1969).There has been some controversyabout the
precise nature of the relation between trade and distance, that is,
whetherit follows the classic gravitymodel or a variationof the gravity
model (see, e.g., Alcaly 1967; Black 1971). Nevertheless, a strong
relation has been found between distance and trade flows among
nations.
Other studies have been more multivariatein nature. Perhaps the
most widely quoted study on the determinantsof tradewas conducted
by Linneman(1966).Linnemanappliedan econometricmodel to study
the factors that determinedthe tradeflows between 80 nationsin 1959.
The independentvariables in the model were GNP, population, distance, and a preferentialtrade factor. The preferentialtrade factor
refers to a dummy variable denoting whether the nation was in the
British,French, or Belgianor Portugesesphereof influence.Linneman
ran his regressionon both exports and imports(separately).He found
that all the variables had a statistically significantrelation with the
volume of importsand the volume of exports flowingbetween the pairs
of nations. The variables with the greatest explanatorypower were
GNP (of both the importingand the exportingnations)and the distance
between the two nations. The other variables, while statistically
significant,were of more limited explanatoryvalue.
In explainingthe variance that was not accounted for by the variables of his model, Linnemanidentifiedseveralfactors. One factorwas
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Determinants of Bilateral Trade Flows
625
the commoditycompositionof tradebetween pairsof nations, which is
to say that the volume of trade between any pair of nations will be
affected by the nature of the productionin the two nations. Two nations that produceprimarilycorn will not tradewith each other. Other
factors not accounted for in his model were seen to be politicalforces
(internaland external), the "psychic distance" that exists between
pairs of countries (see Beckerman 1956), and the existence of economic alliances between nations.
It was the Linnemanstudy that providedthe basic take-offpoint for
the research reported in this paper. The authors have attempted to
extend the findingsof the Linnemanstudy in numerousrespects. First,
the authors have incorporatedseveral of the factors that could affect
trade flows but that were omitted from the Linneman study. These
factors include political instability, membershipin specific economic
unions, and such culturalfactors as religionand language.Second, the
present study controls for variationin the size of nations' economies.
A substantialand significantfactor in explainingthe flow of trade between nations is GNP, a fact that has been well established by past
research. On the other hand, the association between GNP and trade
flows is obvious (there will of course be a tendency for largenationsto
trade with large nations because they are large nations), and it could
obscure other trade determinantsthat are simply overwhelmedby the
effect of economic size. As explained below, the dependent variable
has been constructed in such a way that it controls for the size of the
economy and, by so doing, provides a measure of the intensity of
bilateraltrade relations between nations.
A third way in which the present study represents an extension of
the Linnemanstudy is that it provides independentmeasuresfor individual product categories as well as for total trade between nations.
The inclusion of individual product categories in the analysis overcomes one of the acknowledgedweaknesses of the Linnemananalysis,
for example, that it was unableto account for the commoditycomposition of tradebetween nations. The inclusionof productcategoriesalso
extends the model by determiningwhether the identifiedfactors are
better at explainingtrade flows in some categories than in others.
Method
Trade Data
The trade data used in this study were obtainedfrom the United Nations ExternalTradeExtractData for 1977.These dataconsist of trade
that has occurredbetween all pairs of reportingnationsfor all product
categories. The trade flows are disaggregatedto the 5-digitlevel of the
StandardInternationalTrade Classification(SITC) system.
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Journal of Business
626
Because of the large amount of data availableand the limited computerstorage/processingcapacities, this study has been restrictedto 45
exporting countries and 82 importingcountries (see table Al). This
study analyzes total trade flowingfrom the exportingto the importing
countries as well as trade at the 1-digitlevel of the SITC code. The
"miscellaneous" product category (SITC 9) was not employed in the
study, resultingin nine product categories. The exportingand importing countries were selected on the basis of their trade volume and to
achieve representationfrom all geographicareas. A few majorcountries (e.g., China, the Soviet Union, and Easternbloc nations)are not
included in the list of exporters because of a lack of comprehensive
data. The chosen set of countries, however, accounts for the overwhelming majority of world trade: the 45 exporting countries accounted for more than 80% of world exports in 1977, while the 82
importingcountries accounted for more than 95%of world imports.
The Dependent Variable
For the dependent variablein this study, a Trade Intensity Index was
constructedto reflectthe strengthof traderelationsbetween all pairsof
nations within each of the nine SITC product categories. The index
numberis the ratio of the actual volume of trade between two nations
to the expected volume of tradebetween the two nations. For example,
if the United States accounts for 10%of worldfood exports and if India
accounts for 8% of world food imports, then the expected volume of
U.S. food exports to India would be 0.8% (.10 x .08) of world food
exports. If the actual volume is 1.2%,the index of U.S. food exports to
India would be 1.5; if the actual volume is 0.4%, then the index is 0.5.
Indices of greater than 1.0 therefore reflect stronger-than-expected
traderelationsbetween nations, whereas indices of less than 1.0 reflect
weaker-than-expectedrelations.
The values of the TradeIntensity Index are restrictedto a minimum
value of zero, while their maximumvalues are unrestrained.Accordingly, the distributionof these values is positively skewed. It is therefore necessary to use a logarithmic transformationto reduce the
skewness since regressionanalysis assumes a normallydistributeddependentvariable.The dependentvariableused in the regressionanalysis was the logarithmof one plus 10 times the Trade Intensity Index.
This transformationdoes not interfere with the interpretationof the
results because the scaling is monotonic. The logarithmof the Trade
Intensity Index is identical to the "mutual information"ratio developed by Theil (1967, p. 358), but the procedurefor the prediction of
bilateraltrade is different. Theil's approachfocused on the prediction
of period-t flows based on trade during a prior period. This paper
focuses on distance and the character of relations between pairs of
countries. Theil's objective was to develop forecasts of trade flows.
The objective here is to attemptto understandand explaintradeflows.
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Determinants of Bilateral Trade Flows
627
Table 1 presents a samplingof the Trade Intensity Indices (without
the logarithmictransformation)for selected nations. For the United
States, for instance, it can be seen that the indices are relatively high
for WesternHemispherenations and for other nations with which the
United States has special relations, such as Japan, Israel, and Egypt.
Conversely, U.S. trade is less than expected with many European
nations, Cuba, and Kenya. In the case of India it can be seen that that
nation has relatively weak trade relations with much of Europe, althoughit has strong trade ties with the United Kingdom.Many of the
countries with which India has strong trade relations are located in
Asia and East Africa. A scanning of the indices associated with the
other nations provides furtherindicationsof the natureof the dependent variableused in this study.
The Independent Variables
A summarystatement of the independentvariables employed in this
study, as well as their sources and the way they are measured, is
presentedin table A2. These variableshave been classifiedin ways that
are consistent with the past literature.First, there are the demographic
variables-GDP and population-which relate to the size and stage of
economic development of exporting and importingcountries. These
factors are included in the study despite the controllingeffect of the
dependentvariableto determinewhether the size of an economy has
an independentinfluenceon traderelationsaside from the obvious one
associated with relative buying power. The second group of independent variables comprises the political variables-political instability,
membershipin a particulareconomic union, and colonial heritage.The
thirdgroup comprises the culturalvariables-religion and languageand the final variableis geographicdistance.
Statistical Analysis
The strengthof association between the dependentvariable(tradeintensity) and the predictors (the determinantsof trade) was examined
via a linear-in-logsregression model. Ten separate regression equations were estimated,one for total tradeand one for tradein each of the
1-digitSITC categories. The linear-in-logsregression model was used
because interactionsand multiplicativerelations were anticipatedbetween the independentvariables.For example, if the geodesic distance
between two tradingcountries is large, the other predictors, such as
culture, would have little effect.
Since a linear-in-logsmodel was used, all dummyvariablepredictor
variableswere coded on a 1-2 scale rather than on a 0-1 scale. For
example, if the pair of countries had a colonial relationor belonged to
the same economic union, the correspondingdummy variable was
coded as 2 (1 otherwise). Because the correlationbetween language
and religion was leading to unstable coefficients, these two cultural
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
628
Journal of Business
Ci
ci
0
cl
X
CZ
cl
W)
Cl
M 'D
0 o 0
-.4
Cl
-4
s..
.0
u
0
0
u u >
cl
M cj
0M
cl
u
aN
cq cl
cl Cl
+-j
'n
4-4
-.4
cl
CZ
IZ
cl C'I
.
+-j
v)
W)
Cl
--4
cl
-4
ci
tocl
;--4
4
7
-.4
cl
cl
Cl
144
;--4
--4
-6
0
I--,
C!
CZ
Cl
'C
C,
cl
u
u
u
aN
>1
-4
cl
M
cl cq
0
00
.,t
0
ci
cq
>1
CZ
5
4
u
C'd -"
>0
C,
C,d
C'j C13
t
C,3
C13
--4
X
00
ci
m
o
Cd
-W
CZ C13
"Cl C,3 >
ci
4-4
Cld >
-.4
cn
C'd
1-4
0
-.4
0
llc
--4
..4
A)
Cld
cli en
o
C's
O
14 :
v. CS
u
-4
o
M
m
+1
ct
cl
0
0
0
7:$ Q
w
I'C
JZ
P.
u
=
00
C's
cl
cl
.0
CIO--, E
--4
(.q
C13
0
O
I..
C's
cq =
C13 ',-"
Cd
C,3 C,3
;-4
--!
- -4
cld
m
o ci
7
ci
>,
-CC33
C,3
C13
C)
Wm
v.
004
"M
um
C13
C13
00
a-, r-
A)
cid
0
llc
C,3
0-4
O
Cla
u
ci
C,3
>, CI
r4
C,3
X
C,3
;-4
u
cl
;-4
cl
P4
cl
u
O
0
cld
ci
4)
;-4
cl
4)
;-4
t
0
;-4
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Determinants of Bilateral Trade Flows
629
variableswere combined. If two countries had both languageand religion in common, the dummy variable was coded as 2 (1 otherwise).
Finally, the political instability of exporting and importingcountries
was also coded as dummyvariables(1 if there was no irregulartransfer
of power, 2 if there was one or more irregulartransfers).
Summaries of the resulting regression equations are provided in
tables 2-3. Because of the large number of data points involved (45
exportingcountries x 82 importingcountries = 3,690) it is advisable
that stringent significance levels be applied. Thus, although all relations significantat a. < .05 or better are reported,only those significant
at ot< .005 are likely to contributesubstantiallyto explainedvariance.
Table 2 presents the standardizedregressioncoefficients. The relative
importanceof the predictor variables is proportionalto the square of
these coefficients. Thus, for total trade, the GDP of the exporting
country is approximatelyfour times as importantas colonial status.
Table3 presents the unstandardizedregressioncoefficientsfor only the
significantpredictors. Because the regression model estimated was
linearin logs, the relationbetween tradeintensity(TI) and the determinants of trade can be expressed as (for "total trade," e.g.)
TI
= k(GDPe)201(popi).079(LR).297(cs)'.828
(D)'.496(INSTe)-321
where k is a constant, e and i are subscripts denoting exporting and
importingcountries, CS = colonial status, POP = population, D =
distance, INST = political instability, and LR = commonalityof language and religion.
Findings
The standardizedregression coefficients and r2 values presented in
table 2 depict the associations between each of the independentvariables and the dependentvariablefor the individualproductcategories.
The first columnof numbersrepresentsthe regressionfindingsfor total
tradebetween pairs of countries. Here it can be seen that the independent variablesexplain 30.95%of the varianceand that six of the independentvariablesare significantlyrelatedto the intensity of total trade
flows between nations. The r2 compares favorablywith that obtained
by Linnemandespite the fact that the dependentvariablecontrols for
much of the effect of economic size, the primarycontributorin the
Linnemanmodel. The size of the r2is undoubtedlyaffected by the use
of more explanatoryvariablesin the present study and by the natureof
the dependent variable.
An examinationof the product-specificfindingsshows that there are
considerabledifferences in explained variance across product catego-
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Journal of Business
630
00
$
Y
*
*
*
*
*
*
*
*
*
*
*
*
cO U
*
*
*
*
*
*
*
*
*
*
*
-
*
*
o
Q)
t
m
*
*
c
00
or-
00
*
c
*
*
r*
*
4r-'I
*
?
*
Q
V
Oe
Cd
Fj
C
*
*
*
U-
U)
*
*
*
U
*
*
*
*
*
*
o
<,
E
*
*
*
*
*
*
*
*
*
*
*
*
v)
-co
00
m
0
i
U
@
*II
.
M
*
)
fflo
^
-
r-e
c*
CQ 1
'U
U)
W
'm
u
U
'Um
H N
o
*
m
?
*
*t
*m
I
I
~~~~~~SII
I
I I
*
*
*
*
*
*
*
v)
11- 11T
*
*
CN v)
*
r-
v~~~~I
*
**
*
*
*
v- C l
Cl
~
*
a
*
*
*
CN
C)
*
*
*
00
0~~~~
e
CD
m
**
*
**
*
*
.O
r-
. .O .
N
00 C$
*
*
0
el
n
e^
*
~
,0
'U
o oooo
*?
~
U
m
r-
v) OOOi
*
Cl
*
0 0)
Cl0
CNaN a
*
*
*
*
*
* *
Q
C0
~~*
O X
.N
Ez
*
*
00C
*
'o
Ez
**
Om
X
*
*
*
*
*
*
*
o
o 0
N 0* ?
?
? ? N? ^
0f
*
~~~~~~~~~~~~~~~
r
-3
? t 8 Cxg C ? 90= 1 4 :L?2
vY
.~~~~~~~~~~~~~
0
~~~
u~~~~~
8?
r
C-C
*w~~~~~~~~~~~~~~~~~~~~~~~~7
MM
ao
00Nell
*
c
o~~~
*
*
ell
0
m
XmtN
*
*
*
H c
U
Q
ao
*
0
0
*
~~~~*
el v
'IC
CQ4C4 tNO0
0O
*
~I
@
o
l-'N
o
*
00 t
*
-
?
I
3-
^ 00
ii
~~*
*N
>
U
00 'IC
*
*
00 00
4.r-
3
b
=
d
U~~Er.>
,
D
-
^
4~
'
.Cst
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
~~~~~~~~~~~~~~~~~~~~~
*
631
Determinants of Bilateral Trade Flows
0)
*
*
*
*
**
oo
-
;
CQ
Q
0-
00, c 0)
=OX
0
*
*
*
CQ
OT m ?
.
~~~*
-
m
V
0
U)
**
*
*
0
Y
*
.2
>
I
*
*
*
t ?t
*
) o-
*
.*
. *
*
. *"It
*
1.-)e
*
o*
C-
ZO
^
a
*
*?
00~.0
)
*~.*
*
*
0
0
I
m0
00
*
*
*
*t***
CQ4C
. *
0
**
f
C'E
*
m
c
m
0U
*
*
*
*
*N
d0
00
*
N
I
*.
0 00
El
.
*
*
* *
*
*
*1
o
^N
*
*
*
, \ 4.- )
s
.
oo O
CN 00
\
*
*
m
I
~~~~~000~~~~~~~~~~~~0
_
*
*
)
*
>
*
.*
*
*
0
C-
o
e
rU)0lU
*
.w~~~~~~~~~~~~~~~~~~~~~~
OR - i
O
3 ~iO. CQ
O
\ ..
m
v
C.'
*
Un
o
*
*
elO v) *1 b
0)
v
*
*
*
?
o~~~~~~~~~
S.Y CQ
I
*
*
*
*
c
7:I)
*
*
'
*
*
'
*
*.*~
*
*
*
*
~~~~~~~~~
00
.'
C
*
*
*
.*
*
*
00~~~~~~~~~~~~~~~~~~~~~~~0C4C
c')
r
\
X
c)
0
t
?
_
00
O
a)
C
0
w
O0
U
*
*
~~~~*
*
*
*
*
*
*
0
Ez ?
00
g
*
0
._-
*N
tI cd'
*
*
*
*
*
*
*
*
*
*
.*
* 1-
~~~
Ez Q $
ul ~ ~~->o
V
*
*
*
C.')l- m ffi!El
*
*t
*
*
*
*
*
.*
W
. *
C.')
*
*
r~~~~~~~~~~~~~~~~~~~~~~~~~~~Cs-'
0
.
NN
*m
\*0
m
?
- -
. ?-d
g
/'
. *
0*~~~~~~~~~~~3~
I
*N
c
?C)) .( C'
'
*N
0
t
fi
EE
--
*NN
-
~~~~~~~~~~~~~~~~o
El ??
w
E
C'
1-I
U)
0*)g
*
*
^ 8 61 t
to
C >
^~~~~~~~~~~~~~~~~~~~o
o
O^N
r
ON
.
.
0
.*
g t
jE
u
O
>
-Q
n
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
0)
.
632
Journal of Business
ries. The most strikingdifferenceoccurs between manufacturedgoods
(SITC5, 6, 7, and 8) and the nonmanufacturedproductcategories. The
independentvariables used in the present study explain substantially
greateramounts of variance for manufacturedgoods than they do for
food, beverages and tobacco, raw materials, fuels, and animal and
vegetable fats. The differences here are pronounced, with all the r2
values in the manufacturedgoods categories being greater than 0.30.
The nonmanufacturedproduct categories all have values less than
0.30, and one is even less than 0.20. While some of the independent
variables are significantacross all product categories, there are large
variations in their levels of contribution to the explained variance
across product categories. The following discussion will consider the
individualgroups of independentvariablesand will discuss their effect
within the differentproduct categories.
Distance
Distance is the single most importantdeterminantof trade intensity
amongnationsof the variablesemployedin the regression.This finding
applies to all the product categories as well as to the total trade category. The importanceof distance had been anticipated,given previous
findings.In fact, the correlationcoefficientbetween tradeand distance
is considerablyhigherin the present study than had been found in the
Linnemanstudy. This difference seems primarilydue to the natureof
the dependent variable. Since Linnemanused the absolute volume of
tradebetween nations as the dependentvariable,it would be expected
that the large trade flows that occur between highly developed but
distant nations (e.g., the United States and Germany)would override
the smallerbut higher-intensitytrade that occurs between neighboring
nations (e.g., Switzerlandand Germany).
Althoughdistance has the largest standardizedregressioncoefficient
within each of the product categories, the size of the regression
coefficient varies across the product categories. In general, distance
accounts for approximately50%of the total explainedvariationin the
individualproduct categories. There are some notable exceptions to
this rule, however. With regardto raw materials,distanceaccountsfor
almost two-thirdsof the explainedvariance. Conversely,for two categories of manufactures(SITC 7 and 8) distance accounts for only 40%
or less of the explained variance.
The productcategory findingsfor distance appearto portraystatistically one of the classic relations definedin location theory: industries
thatproduceproductscharacterizedby low value-to-bulkratiostend to
be located closer to their markets than do industrieswhose products
have high value-to-bulkratios (see, e.g., Hoover 1948; Isard 1956).
Root (1984, p. 232) noted the need to integrate location and trade
theories. The findingspresentedhere have statisticallydemonstrateda
link that exists between the two areas.
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Determinants of Bilateral Trade Flows
633
Demographic Variables
As noted above, demographicvariablesrelatingto the populationand
economic size of importersand exporterswere includedin the analysis
because of their significancein past research.Theirinclusion, despite a
dependent variable that controls for size, was intended to determine
whether size (economic, population, or both) has an independentinfluence on the intensity of bilateraltrade flows.
With regard to total trade flows, only the GDP of the exporting
country(a supplyfactor)and the populationof the importingcountry(a
demandfactor) have a significantrelationwith the dependentvariable.
These findingsmay be influencedby the correlationsthat exist between
the GDP and populationof the exporter(0.32) and the GDP and populationof the importer(0.33). These correlations,however, are not large
enough to cause any majorestimationproblems.
The findings within product categories provide more revealing insights into bilateral trade relations. The significanceof the exporting
country's GDP exists across each of the nine product categories. Of
equal importance,the relation between trade intensity and exporter's
GDP exists without any consistent relation to the population of the
exportingcountry. This findingmay reflect the marketpower exerted
by large economies and the correspondinglack of market power of
small economies.1 Nations with large economies tend to have large
firmsthat diversify their exports across a numberof importingnations.
Smaller nations may tend to be dependent on a limited number of
marketsfor their exports. While these smallernations may have very
high trade intensities with a limited numberof countries, with the vast
majorityof nations their trade relations (and thus intensities) are almost nonexistent. The foregoing explanation is not intended to be
definitive. The rationalefor the relationbetween exporter's GDP and
trade intensity deserves additionalstudy by trade researchers.
The findingswithinproductcategoriesalso uncover some interesting
relations between importer's GDP and trade intensity that had been
masked in the aggregate analysis. For all nonmanufacturedproduct
categories except SITC 3 (oil and other energy-relatedproducts)there
is a significantpositive relationbetween tradeintensity and importer's
GDP. For the two categories of manufacturedgoods that contain the
largest concentration of technology-intensivegoods (SITC 5 and 7),
there is a negative relation between trade intensity and importer's
GDP.
1. The strongrelationbetween tradeintensity and the GDP of the exportingcountry
was unanticipatedand led the authorsto conductsubsequenttests to determinewhether
this findingwas due primarilyto the effect of the United States on traderelations.The
United States was removedfrom the analysis, and the tests were rerun.Even with the
United States excluded, the relation between trade intensity and exporter GDP remained,thus supportingthe generalnatureof the observed relation.
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
634
Journal of Business
The positive relationbetween tradeintensitiesof basic productsand
importer'sGDP may to a large extent reflect the fact that large economies tend to be more diverse and industrialized,thereforerepresenting
marketsfor diverse sources of raw materialsto supporttheir industry.
Trade intensities would therefore be higher across a broader cross
section of tradingpartners.
The negative relation between importerGDP and trade intensity of
sophisticatedmanufacturesoccurs in conjunctionwith significantpositive relationsbetween populationand tradeintensitiesfor these goods.
These findings appear to depict statistically a classic trade relation
between relatively rich and poor nations, that is, sophisticatedmanufacturedgoods flowingfrom more developed nationsto less developed
nations. The importantdifference between the findingshere and the
traditionalexplanationjust presented is that these results may portray
the less developed nations as being engagedin more diversifiedbuying
across industrializedcountries for their sophisticated manufactures.
Given the nature of the dependent variable, the significantrelations
suggest that the less developed nations draw their imports of these
products from -a more diverse set of sources than do nations with
relatively high per capita GDP. This relation is also one that would
benefit from more in-depth study.
For three product categories there are not any significantrelations
between trade intensity and importer-relateddemographics.Intensity
ratios for SITC 3 are unrelatedto importerdemographics.This finding
has face validity since energy productsare importedby nations across
the full range of GDP and population. In addition, the findings here
suggest that overall trade intensitiesare unrelatedto these demographics. Similarly, demographicsdo not explain the intensity of flows of
goods from SITC 6 and 8, two categories of manufacturedgoods that
contain primarilylow-technology goods that are importedacross the
entire spectrumof nations.
Cultural Similarity
Culturalsimilaritybetween nations has been proposed by Beckerman
(1956)as an importantdeterminantof tradeflows between nations. The
findingsof the present study tend to confirmthis postulate, although
there is considerable variation among product categories. There is a
significantrelationbetween culturalsimilarityand the intensitylevel of
total trade among nations, although the magnitudeof the regression
coefficient is relatively small when compared to those of the other
independentvariables.Likewise there are significantrelationsbetween
cultural similarity and trade intensity for food and for tobacco and
beverages, but the magnitudes of the coefficients are low. The
strongest associations with cultural similarity are found within the
manufacturedproductscategories. For all the manufacturedgoods cat-
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Determinants of Bilateral Trade Flows
635
egories the relation is both significantand relatively substantial:only
distance and the GDP of the exporting country rank substantially
above culturalsimilarityin terms of the amountof variationaccounted
for with respect to trade intensities for manufacturedgoods. Some
qualificationtherefore seems to be needed when expressing the relation between trade and culturalsimilarity:it is most pronouncedwith
respect to trade in manufacturedgoods.
Political Variables
The political variables with significantregression coefficients are the
political instability of the exporting nation and the previous colonial
relationsexisting between the exportingand the importingnations. As
anticipated,low levels of instabilityfor the exportingnationare associated with hightradeintensityindices. This relationis shown to exist for
total trade and for all product categories except fuels and beverages
and tobacco. Stable nationstend to be the higher-levelexporterswhen
bilateraltrade relations are examined. Conversely, there is very little
effect of the instabilityof the importingnationon the intensityof trade,
this variable being an insignificantdeterminantof trade in all but a
couple of instances.
The former colonial relations between countries is found to be a
significantdeterminantof the intensity of total trade as well as of trade
in each of the product categories. Despite the attainmentof independence, the ties of former colonial times seem to exert an influenceon
bilateral trade relations. Such a finding is not surprisingsince businesses of the former colonists are generally well representedin the
previous colonies and, in many cases, since close political ties still
exist between nations and their former colonies.
Membershipin the same economic union does not have a significant
relationwith total trade flows, althoughthere are significantrelations
shown for several of the productcategories. The relativelyweak effect
of membershipin the same economic union may be explainedat least
partiallyby the relatively large effect of distance on the determination
of trade flows. Since most nations in an economic union are in close
proximityto each other, a large part of the associated tradeties would
have been incorporatedin the distance factor.
Limitations
It would be a naturalreactionto wonderwhy the present study has not
accountedfor a greaterportion of the variancein trade flows. Indeed,
the authors had wondered the same thing. Several possible explanations could be posited with regard to the variance that has been left
unexplained. For one, not all the potential determinantsof trade are
accounted for in the model. Nations vary in terms of the restrictions
placed on trade, a factor that could influence trade flows. There are
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
636
Journal of Business
also such factors as the historical relations between nations, which
have not been quantifiedto the extent that they can be incorporated
into a formalanalysis. Many nationsthat are geographicallyproximate
are in fact hostile to one another (e.g., India and Pakistan, Iraq and
Iran),and this factor would undoubtedlycontributeto the level of trade
between these countries.
A furtherlimitationof the present study that undoubtedlycontributed to the unexplainedvariance concerns the natureof the measures
used for some of the independent variables. This limitation applies
particularlyto culturalsimilarityand political instability.The measure
of culturalsimilarity-similarity of languageand religion- was necessarily aggregateand did not consider the fact that nations may have
severalmajorreligions(e.g., the United States) or morethanone major
language(e.g., Belgium).If culturalsimilarityhad been able to be more
finely measured,it is likely that this variablewould have accountedfor
a greaterportion of variance and thereby increased the total variance
explainedby the model.
Likewise, the measure of political instability used in the model is
subject to limitations.The measure's primarylimitationis that it measured instabilityfor a period that ended 10 years prior to the year for
which the trade statistics have been analyzed. Some nations that were
unstableduringthe earlierperiodhad attainedrelative stabilityby 1977
and vice versa. Despite this limitation, however, instabilityhas been
depicted as a significantdeterminantof trade. Like culturalsimilarity,
it is probablethat a better measureof instabilitywould add to the total
variationexplained by the model.
Discussion
The purpose of this study has been to extend and refinecurrentunderstanding of the determinantsof trade between nations. The primary
reference point for this work has been the study of Linneman(1966)
since he had performedthe most extensive analysis of the subject to
date. The main differencesbetween the present study and past studies
are found in the numberof variables employed, the measure of trade
relationsbetween nations, and the product-specificnatureof the analysis. The present study incorporatedmore independentvariables into
the analysis-including measures of culturalsimilarity,political instability, and membershipin particulareconomic unions-that had not
been formally included in past studies. In addition, the magnitudeof
trade flows between nations in the present study is measuredby the
relative strengthof trade between nations ratherthan by the absolute
volume of the tradeflow. In essence, this study has analyzedthe determinantsof the relative strengthof trade ties between nations, whereas
most past studies have been concerned with the determinantsof the
volume of trade between nations. Finally, whereas past studies have
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Determinants of Bilateral Trade Flows
637
been concerned with aggregate trade between nations, the present
study has examined the determinantsof aggregate trade as well as
trade in specific product categories.
There are several findings emanatingfrom this study that extend
currentknowledge of determinantsof trade. First, the importanceof
the individualdeterminantswill depend in part on the type of product
being exported. In general, the independentvariablesemployedin this
study are better for explainingtrade flows of manufacturedgoods than
of nonmanufactureditems. The two variablesthat appearto be responsible for the greaterexplanatorypower for manufacturesare the political instability of the exporter and, especially, cultural similarity.
These two variablesaccounted for considerablymore varianceamong
manufacturedgoods than they did amongother products. One possible
explanationof this findingis that it is likely that a greater numberof
sources will exist for manufacturesthan for basic products,which tend
to be available only from the limited numberof nations in which they
are found. Thus the political stability and degree of culturalsimilarity
of sources of supply of basic products are less of a consideration.
A second finding of the present study that extends current understandingof the determinantsof trade relates to the effect of the demographicvariables of GDP and population. Since past studies had employed total volume of trade as their dependent variable, they had
naturally found a strong association between trade flows and such
economic size variables as GDP. The present study, by partiallycontrollingfor the economic size of the countries via the Trade Intensity
Index, has found that the demandvariablesexert an influenceon trade
flows that is out of proportionto nations' relative economic size in the
internationalcommunity. This findingis consistent across all product
categories and deserves future research to explore the underlying
causes of the relation between economic size and trade intensity.
In the manner noted above, the present study has contributedin
substantiveways to the understandingof trade relations between nations. In his study Linnemanhad found that the majorcontributionto
explainingthe determinantsof trade had been made by GDP and by
distance. The other variablesused in the study-population, preferential tradewith the United Kingdom,France, or Portugaland Belgiumwere found to be statisticallysignificantbut not importantenoughto be
consideredfundamentalforces in tradeflow determination(see Linneman 1966, p. 88). The present study, however, has identifiedseveral
forces that go beyond simple GDP or distance in determiningthe trade
ties that exist between nations, especially in the area of manufactured
goods. In most cases distance accounted for only half (or less) of the
explainedvariationin the intensity of trade flows between nations;the
additionof other factors such as culturalsimilarityis necessary for a
more comprehensive understandingof the strength of trade ties that
exist between nations.
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Journal of Business
638
Conclusions
The intensity of bilateral trade relations between nations is a multidimensionalphenomenon. The present study has demonstratedstatisticallyhow several factors are significantlyrelatedto the intensityof
tradeflows between nations. Furtherimprovementsin explainingtrade
flows could be made throughrefinementsin the measures of the individual variables (e.g., political instability and cultural similarity)and
throughthe incorporationof additionalvariables (e.g., the extent of
traderestrictionsand the natureof political relationsbetween pairs of
nations). The extension of past findings provided by this study has
contributedto current understandingof bilateral trade relations between nationsand can serve as a takeoffpoint for furtherstudies in the
area.
Appendix
TABLE Al
South Africa
Libya*
Morocco*
Sudan
Tunisia*
Egypt*
Cameroon
Gabon*
Zaire
Ghana
Algeria*
Ivory Coast*
Kenya
Nigeria*
Zambia
Canada*
United States*
Argentina*
Brazil*
Chile
Colombia
Ecuador
Mexico*
Peru
Venezuela*
El Salvador
Guatemala
Countries Included in This Study
Cuba
Dominican Republic
Jamaica
Netherlands Antilles
Trinidad-Tobago
Panama
Israel*
Japan*
Iran*
Jordan
Kuwait*
Lebanon
Saudi Arabia*
Syria
Yemen
Turkey*
Bangladesh
Hong Kong*
India*
Indonesia*
South Korea*
Malaysia*
Pakistan
Philippines
Singapore*
Thailand*
Taiwan
China
Belgium*
Denmark*
France*
Germany*
Ireland*
Italy*
Netherlands*
United Kingdom*
Austria*
Finland*
Norway*
Portugal*
Sweden*
Switzerland*
Greece*
Spain*
Yugoslavia*
Bulgaria
Czechoslovakia
East Germany
Hungary
Poland
Romania
Soviet Union
Australia*
New Zealand
* Denotes countries that were used as both exporting and importing nations in this study.
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
Determinants of Bilateral Trade Flows
TABLE A2
639
Independent Variables Employed in This Study
1977 GDP (Business International Corp. 1981)
Population in 1975 (Business International Corp.
1981)
Political instability (Taylor
and Hudson 1971)
Economic union membership (Ball and McCulloch 1982)
Colonial heritage (Bjorklund 1970)
Religion (De Blij 1982)
Language (De Blij 1982)
Distance (National Geographic Society 1981)
US$; GDPs of both exporting and importing countries
were considered in the analysis
Measured in thousands of inhabitants for both exporter
and importer
Number of irregular political power transfers, 1958-67;
measures for both exporter and importer
European Economic Community, Association of
Southeast Asian Nations, European Free Trade Association, Economic Community of West African
States, Andean Common Market, Central American
Common Market, Council for Mutual Economic Assistance
Colonial power administering the nation in 1946
Predominant religion practiced in the country: mostly
Roman Catholic, mostly Protestant, mostly Eastern
rites, Sunni Moslem, Shiah Moslem, Hindu, Buddhism, Chinese religions, Shintoism, Judaism, traditional and shamanist religions
Predominant language spoken in the country: Germanic, Baltic, Slavic, Celtic, Romance, Greek,
Armenian, Indo-Iranian, Afro-Asiatic, Niger-Congo,
Saharan-Sudanic, Khoisan, Ural-Altaic, Sino-Tibetan, Japanese and Korean, and Malay/Polynesian
Geodesic distance: the shortest distance between
boundaries of two countries, coded by a logarithmic,
nine-point scale, with 1 = contiguous, 2 = less than
500 miles, 3 = 500-999 miles, 4 = 1,000-1,999
miles, . . ., and 9 = greater than 9,000 miles
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions
640
Journal of Business
References
Alcaly, R. E. 1967. Aggregation and gravity models: Some empirical evidence. Journal
of Regional Science 7 (Summer): 61-73.
Ball, D. A., and McCulloch, W. H. 1982. International Business: Introduction and
Essentials. Plano, Tex.: Business Publications.
Beckerman, W. 1956. Distance and the pattern of inter-European trade. Review of Economics and Statistics 38:31-40.
Bjorklund, 0. 1970. Historical Atlas of the World. New York: Barnes & Noble.
Black, W. R. 1971. The utility of the gravity model and estimates of its parameters in
commodity flow studies. Proceedings of the Association of American Geographers
3:28-32.
Business International Corp. 1981. Worldwide Economic Indicators Comparative Summary for 131 Countries. New York: Business International Corp.
De Blij, H. J. 1982. Human Geography, Culture, Society and Space. New York: Wiley.
Hoover, Edgar M. 1948. The Location of Economic Activity. New York: McGraw-Hill.
Isard, Walter. 1956. Location and Space-Economy. Cambridge, Mass.: MIT Press.
Linneman, Hans. 1966. An Economic Study of International Trade Flows. Amsterdam:
North-Holland.
Linneman, Hans. 1969. Trade flows and geographical distance, or the importance of
being neighbors. In H. C. Bos (ed.), Towards Balanced International Growth: Essays
Presented to J. Tinbergen. Amsterdam: North-Holland.
National Geographic Society. 1981. Atlas of the World. Washington, D.C.: National
Geographic Society.
Root, Franklin R. 1984. International Trade and Investment. 5th ed. Cincinnati: Southwestern.
Smith, R. H. T. 1964. Toward a measure of complementarity. Economic Geography 40
(January): 1-8.
Taylor, C. L., and Hudson, M. C. 1971. World Handbook of Political and Social Indicators II. New Haven, Conn.: Inter-University Consortium for Political Research.
Theil, Henri. 1967. Economics and Information Theory. Chicago: Rand McNally.
Ullman, E. L. 1956. The role of transportation and the bases for interaction. In William
L. Thomas (ed.), Man's Role in Changing the Face of the Earth. Chicago: University
of Chicago Press.
Warner, D., and Kreinin, M. E. 1983. Determinants of international trade flow. Review
of Economics and Statistics 65 (February): 96-104.
Yeats, M. H. 1969. A note concerning the development of a geographical model of
international trade. Geographic Analysis 1:399-404.
This content downloaded from 202.161.46.2 on Wed, 25 Jun 2014 00:18:21 AM
All use subject to JSTOR Terms and Conditions