Pitfalls of location choice analysis : the finished goods producer versus the
intermediate goods producer
著者 Hayakawa Kazunobu, Matsuura Toshiyuki
権利 Copyrights 日本貿易振興機構（ジェトロ）アジア
経済研究所 / Institute of Developing
Economies, Japan External Trade Organization (IDE‑JETRO) http://www.ide.go.jp
journal or
publication title
IDE Discussion Paper
volume 178
year 2008‑11‑01
URL http://hdl.handle.net/2344/796
IDE Discussion Papers are preliminary materials circulated to stimulate discussions and critical comments
Keywords: demand linkages; cost linkages; location choice JEL classification: F23; H32; R34
* Corresponding author: Kazunobu Hayakawa, Economic Integration Studies Group, Inter-Disciplinary Studies Center, Institute of Developing Economies, 3-2-2 Wakaba, Mihama-ku, Chiba-shi, Chiba 261-8545 Japan. Phone: 81-43-299-9754; Fax:
81-43-299-9763. E-mail: [email protected]
IDE DISCUSSION PAPER No. 178
Pitfalls of Location Choice Analysis:
The Finished Goods Producer versus the Intermediate Goods Producer
Kazunobu HAYAKAWA*
Toshiyuki MATSUURA
November 2008
Abstract: In literature related to firm location choice, estimation equations are
derived from the model of finished goods producers, but producer types are
generally not considered. Research presented in this paper shows that the use of
equations derived from such models against intermediate goods producers results
in several problems.
The Institute of Developing Economies (IDE) is a semigovernmental, nonpartisan, nonprofit research institute, founded in 1958. The Institute merged with the Japan External Trade Organization (JETRO) on July 1, 1998.
The Institute conducts basic and comprehensive studies on economic and related affairs in all developing countries and regions, including Asia, the Middle East, Africa, Latin America, Oceania, and Eastern Europe.
The views expressed in this publication are those of the author(s). Publication does not imply endorsement by the Institute of Developing Economies of any of the views expressed within.
I NSTITUTE OF D EVELOPING E CONOMIES (IDE), JETRO 3-2-2, W AKABA , M IHAMA - KU , C HIBA - SHI
C HIBA 261-8545, JAPAN
©2008 by Institute of Developing Economies, JETRO
No part of this publication may be reproduced without the prior permission of the
IDE-JETRO.
Finished Goods Produer versus the
Intermediate Goods Produer
Kazunobu HAYAKAWA y
Inter-Disiplinary StudiesCenter,InstituteofDevelopingEonomies,
3-2-2 Wakaba, Mihama-ku,Chiba-shi,Chiba261-8545 Japan
Toshiyuki MATSUURA z
TheInstituteof Eonomi Researh,HitotsubashiUniversity,
2-1Naka,Kunitahi,Tokyo 186-8603 Japan
Abstrat
In literature related to rm loation hoie, estimationequations
arederivedfromthemodelofnishedgoodsproduers,butproduer
typesaregenerallynotonsideredintheseletionofsample. Researh
presented inthispaper shows that theuse of equationsderived from
suh models against intermediate goods produers results in several
problems.
Keywords: demand linkages;ost linkages;loationhoie
JEL Classiation: F23; H32;R34
Corresponding author: Kazunobu Hayakawa, Eonomi Integration Studies Group,
Inter-Disiplinary Studies Center, Institute of Developing Eonomies, 3-2-2 Wakaba,
Mihama-ku, Chiba-shi, Chiba 261-8545Japan. Phone: 81-43-299-9754;Fax: 81-43-299-
9763. E-mail: kazunobuhayakawaide.go.jp.
y
We would like to thank Masahisa Fujita, Kyoji Fukao, Toshitaka Gokan, Nobuaki
Hamaguhi,KenItakura,TomohiroMahikita,and Kazuki Yokoyamafortheir valuable
ommentsandsuggestions.
z
E-mail: matsuuraier.hit-u.a.jp
1
A large number of studies have inluded empirial investigations of the role
of agglomeration benets suh asdemand linkages and ost linkages in rm
loation hoie by using various kinds of proxy variables. Of reent inter-
est are: Castellani and Zanfei (2004), Head and Mayer (2004), and Basile,
Castellani, and Zanfei (2008). In these studies, GDP or market potential
(sum of distane-weighted GDP) introdued by Harris(1954)has been used
asaproxy fordemand linkages,andtotal produtionvalues,value-added, or
the number ofrms ineah industryhavebeen used asproxiesfor ostlink-
ages. These studies have onsistently shown oeÆients to be signiantly
positive.
Head and Mayer(2004) examinedthe validity ofready-made proxy vari-
ables fordemand linkagesomparedwith marketpotentialmeasuresdiretly
derived fromthe new eonomi geographymodel(also known asKrugman's
model). Suh measures take into aount the extent of ompetitionand are
onstrutedbyusingestimatorsofimportingountrydummyvariablesinthe
well-known gravity equation. However, they nd that the \theory doesn't
pay" inthe sensethat Harrismarketpotentialoutperforms Krugman'smar-
ketpotentialinboth magnitudeofoeÆientandtofthe estimatedmodel.
This paper inludes onsideration of dierent aspets on demand and
osts linkages from Head and Mayer (2004) suh as type of produers (for
example, nished or intermediate goods produers). Despite the fat that
estimationequationsarederivedfromthemodelofnishedgoodsproduers,
produer typesare generallynot onsidered in the seletionof sample. This
paper shows that the use of equations derived from suh models against
intermediategoodsproduersresults inseveral problems. Setion2 inludes
derivation of prot funtions of nished and intermediate goods produers
from the standard model found in the literature related to loation hoie.
Problems enountered in analyzing loation hoie without distinguishing
produer typesare disussed in Setion3.
2 Loation Choie Model
Prot funtionsof nishedand intermediategoodsproduersfromthe stan-
dard model are presented in this setion. In the nished goods setor, the
well-employed model found in loation hoie studies suh as that of Head
2
lated by following the standard new eonomi geography model (Krugman
and Venables, 1995)is inorporated.
2.1 Finished Goods Produers
Arepresentativeonsumerineahregionisassumedtohaveatwo-tierutility
funtion. The upper tier is a Cobb-Douglas funtion of the utility derived
from onsumption of nished goods. Speially, the following utility fun-
tion of the onsumer inregion r is applied:
U
r
= H
Y
h=1
C h
r
h
; H
X
h=1
h
=1;
where C h
r
isthe aggregate onsumptionof nished goodsh in region r.
Expenditurealloationisformalizedinnishedgoodsonsistingofmulti-
ple varietiesomittingthe subsript representing the name of nishedgoods.
The onsumer has the following preferene speied as aonstant elastiity
of substitution (CES) funtion overvarieties:
C
r
= 2
4 R
X
i=1 Z
Nr
0 x
r;i (j)
t
r;i
! 1
dj 3
5
1
where R , N
r
, and x
r;i
(j) are the number of ountries, the number (mass) of
nishedvarieties,andthedemandofregionrfornishedvarietiesjprodued
in region i, respetively. Transations in nished goods between regions r
and s are modeled as faing Samuelsonian ieberg osts, t
r;s
( 1). t
r;s
= 1
if r = s. is the elastiity of substitution between nished varieties and is
assumed tobe greater thanunity. The utility maximizationyields:
x
r;i
= t 1
r;i p
i P
1
i Y
r
(1)
where p
i
and P
r
denote the prie of the variety produed in region i and
the prieindex inregionr,respetively(variety notationj isdropped where
larity permits). Y
r
is total expenditure inregion r.
The market struture in the nished goods setor is assumed to be a
Chamberlinian monopolisti ompetition. The nished goods produer of
eahregionisaombinationofaompositeindexaggregatedarossvarieties
3
This is based on a Cobb-Douglas model. The omposite beomes a part of
the ost funtionfor eah produer through aCES aggregatoras follows:
C(x
r )=w
1
r G
r x
r +F
r
; G
r
=
"
Z
M
r
0 q
r (j)
1
dj
# 1
1
;
wherew
r
denotestheprieindexforprimaryfatorsthatisemployedbyeah
nished goodsproduer inthe produtionof total outputx
r (=
P
i x
r;i ). G
r
is the prie index for intermediate goods, and F F
r
represents xed osts.
is a linkage parameter between nished and intermediate goods. M
r , q
r (j),
and are respetivelythe number(mass)of intermediatevarietiesprodued
in region r, the prie of j-th varieties produed in region r, and the elas-
tiity of substitution between intermediate goods, respetively. Elastiity is
again assumed to be greater than unity. Note that for easy omparison of
the prot funtion between nished and intermediate goods produers, the
intermediategoodsmarketis assumed tobesegmented; transation osts of
intermediate goods aross regions are prohibitively high. Eah rm maxi-
mizes its protwith respet to quantity in orderto derive produerpries:
p
r
=
1
w 1
r G
r
: (2)
Using(1) and(2), aprotfuntionofanishedgoodsproduerinregion
r may bederived asfollows:
F
r
=
( 1) 1
w
(1 )(1 )
r
G (1 )
r
"
R
X
i=1 t
1
i;r P
1
i Y
i
#
F
r :
The seond braket of the RHS, P
R
i=1 t
1
i;r P
1
i Y
i
, will hereafter be alled
\market potential" and denoted by MP
r
. The prot funtion may thus be
rewritten as follows:
F
r
=
( 1) 1
w
(1 )(1 )
r
G (1 )
r
MP
r F
F
r
: (3)
2.2 Intermediate Goods Produers
In the ase of loation hoie of intermediate goods produers, the prot
funtion (3) qualitatively hanges. Considering the prodution tehnology
4
termediate goods are produed not only with primary fators but also with
intermediate goods themselves. As in the nished goods produer, the in-
termediate goods produer of eah region ombines a omposite index ag-
gregated aross varieties of intermediate inputs and primary fators using a
Cobb-Douglas model. The ompositebeomesapartofthe ostfuntionfor
eah produerthrough aCES aggregator:
C(z
r )=w
1
r G
r z
r +F
I
r
;
where z
r
denotes total output of anintermediatevariety produed inregion
r,and isalinkage parameteramong intermediategoods. F I
denotes xed
osts. Then the prot funtion is given by:
I
r
=
( 1) 1
w
(1 )(1 )
r
G (1 )
r
"
R
X
i=1
1
i;r G
1
i
(X
i +Z
i )
#
F I
r
;
where denotes ieberg osts. X
i
is equal to N
i p
i P
r x
r;i . Z
i
is equal to
M
i q
i z
i
. In this ase, the omposition of demand linkages beomes omlex.
The magnitudeof intermediate aswell as nishedgoodsprodution isposi-
tivelyrelatedtotheprotofplantsproduingintermediategoods. Assuming
prohibitivelyhigh ie-berg osts, the following isobtained:
I
r
=
( 1) 1
w
(1 )(1 )
r
G
( 1)(1 )
r
(X
i +Z
i ) F
I
r
: (4)
3 Pitfalls
Mostpreviousstudieshaveexpliitlyorimpliitlyonsidered loationhoie
of nished goods produers and estimated their prot funtion as in (3) by
using onditionallogits. Inprotfuntion(3), F F
r
isassumedtobeidential
for tratabilityaross regionsas seen in Head and Mayer (2004). As mono-
toni transformations leave orderingof prot unhanged, the nished goods
produerhoosestheregioninwhihthefollowinglog-funtionismaximized:
ln
r
= V
r +"
r
= (1 )(1 )lnw
r
+(1 )lnG
r
+lnMP
r +"
r
(5)
where "
r
denotes unobservable regional harateristis.
5
sample, the use of equation (5) for intermediate goods produers results in
severalproblems: First,thepowers ofw
r
andG
r
infuntion(4)are dierent
from those in funtion (3). In partiular, while the power of G
r
is positive
in funtion (4), it is negative in funtion (3). This asymmetry implies that
the magnitude of its oeÆient may suer from a serious aggregation bias
when equation(5) is appliedtointermediategoodsproduers. Seond, from
a qualitativepoint of view, unlike itsrole infuntion (3), G
r
infuntion (4)
playsarole inapturing apartofdemandlinkages ratherthanostlinkages.
The small G
r
in funtion (4) implies bad aess to input markets rather
than existene of many ompetitors. This results in lower operating prot.
Third, onsidering funtion (4), the demand omponent X
r +Z
r
is not
log-linearlyrelated toMP
r
. This leads toanerrors-in-variable problemand
results in inonsisteny of estimators. This emerges in estimating equation
(5) forloationhoie of intermediategoodsproduers.
A more appropriate proedurewould be toseparately estimateequation
(5) for nished goods produers and the equation based on(4) for interme-
diate goods produers. In both ases, wage data are usually available, but
thereis alimitationofdata relatedtothe prieindex forintermediategoods
G. Intheliterature,thevariablereetingmagnitudeofagglomeration(total
prodution values in eah industry) is often used. From a theoretial point
of view, the prie index for intermediate goods is low in regions with suh
large agglomerations,so this proxy is somewhat plausible. Variablesrelated
to demand linkages (MP
r
or X
r +Z
r
) are also troublesome. In the ase
of nishedgoodsproduers, (sum ofdistane-weighted) GrossDomesti Ex-
penditure beomesa good proxy for MP sine onsumers of nished goods
are all people living in the region. On the other hand, in the ase of inter-
mediate goodsproduers, the total produtionvalues of nished goods and
those of intermediate goods are good proxies though non-linear estimation
tehniques are neessary in estimating the prot funtion. In addition, the
use of a diret measure suh as the expenditure on intermediate goods in
the region maybebetter thanthose variables. Ineither ase,those data are
diÆult to obtain, and the input-output table seems to be the only soure.
If the sample overs many regions,suh data are likely tobe unavailable.
Insum, the estimation ofequation (5) forthe loationhoieof interme-
diate goods produers yields an errors-in-variable problem, and this makes
estimators inonsistent. In addition,the magnitude of some oeÆients suf-
fersfromaggregationbiasanddiÆultyininterpretation. Tostritlyanalyze
6
data is neessary inluding total prodution values of nished goods and
intermediategoods.
Referenes
[1℄ Basile,R., Castellani,D., andA.Zanfei, 2008,Loationhoiesof multi-
nationalrms in Europe: the role of EU ohesion poliy, Journal of In-
ternationalEonomis 74, 328-340.
[2℄ Castellani,D.andA. Zanfei,2004,Choosinginternationallinkagestrate-
giesintheeletronisindustry: theroleofmultinationalexperiene,Jour-
nalof EonomiBehavior and Organization53, 447-475.
[3℄ Harris,C., 1954, Themarketasafator inthe loalizationofindustryin
the United States, Annals of the Assoiation of Amerian Geographers
64,315-348.
[4℄ Head, K. and T. Mayer, 2004, Market potential and the loation of
Japanese investment in the European Union, Review of Eonomis and
Statistis86,959-972.
[5℄ Krugman,P.R.and A.J.Venables, 1995, Globalizationand theinequal-
ity of nations,Quarterly Journal of Eonomis 110, 857-880.
7
