レジュメ そ 3 （ 第 3 ） • “ Optimal Unemployment Benefit in the Harris-Todaro model — A Note on Calvo and Quibria ” ・・・・・・ • “ International Differences in Consumer Preferences and Trade — Evidence from Multicountry, Multiproduct Data ” ・・・・・・Oleksandr Movshuk • “ International Labor Standards and Their Harmonization ” ・・・・・・Laixun Zhao 学 第 47 2005 6 11 （ ） 援センター
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
報告レジュメ集その3
（午前の部第 3分科会）
•“ Optimal Unemployment Benefit in the Harris-Todaro model
— A Note on Calvo and Quibria”・・・・・・斎藤宗之
•“ International Differences in Consumer Preferences and Trade
— Evidence from Multicountry, Multiproduct Data”・・・・・・Oleksandr Movshuk
•“ International Labor Standards and Their Harmonization”・・・・・・Laixun Zhao
日本国際経済学会 第 47回関西支部総会
2005年 6月 11日（土曜）沖縄産業支援センター
Optim
alunem
plo
ym
ent
benefit
inth
e
Harr
is-T
odaro
model:
anote
on
Calv
oand
Quib
ria
大阪大学経済学研究科博士課程斉藤宗之
2005年
6月
11日
1.論文の目的
•標準的なハリス・トダロモデルにおける外生的な都市部門の高賃金を、独占的
な労働組合の賃金に関する効用最大化により内生的に賃金を決定する状況をモ
デル化
•そのもとで失業給付金を考慮し、最適失業失業給付金水準が存在するかを考察
1
1-1
.背景
•標準的なハリス・トダロモデルにける都市部門の高賃金の理由は、なにも最低
賃金法だけではない
•標準的なハリス・トダロモデルにおいて失業給付金の問題はあまり考察されて
いない
2
2.先行文献
•Calv
o(1
978)：標準的なハリス・トダロモデルに都市の労働組合を導入し、最
適な雇用・賃金補助金が存在しないことを示した
•Q
uib
ria(1
988)：
Calv
o(1
978)のモデルにインフォーマル・セクター、移動
条件に期待効用、組合の効用関数を変形し、最適な雇用補助金の存在を示した
•Chau(2
001)：
Quib
ria(1
988)の問題点を指摘
•Yabuuchi(1996)：標準的なハリス・トダロモデルに失業給付金を集めて移
転する政府の活動を導入
3
レジュメ集３－1 ページ
3.主要結論
•移動条件が期待賃金であろうと、期待効用であろうと最適失業給付金水準が存
在する
•Calv
o(1
978)の労働組合の効用関数を変形
4
4.モデル
•二部門モデル：都市部門と農村部門
•小国開放経済
•都市には労働組合が存在
•労働組合が組合の効用を最大化するように賃金を決定し、その賃金を所与とし
て都市の企業が利潤を最大化するように労働雇用量を決定（独占的労働組合）
•労働者の部門間の移動条件は期待賃金の場合と期待効用（危険回避的）の場合
5
•都市と農村部門の生産関数をそれぞれ、
Xu
=F(L
u)=
La u,
0<
a<
1(1
)
Xr
=G(L
r)(2
)
Xu：都市部門の生産量、
Xr：農村部門の生産量、
Lu：都市の労働雇用量、
Lr：
農村の労働雇用量
•都市部門の利潤最大化の一階の条件（
wu：都市部門の賃金率）
wu=
aL
a−1
⇒L
d u=
⎛ ⎜ ⎝a wu
⎞ ⎟ ⎠1
1−
a(3
)
•農村部門の利潤最大化の一階の条件（
wr：農村部門の賃金率）
G′ (
Lr)
=w
r(4
)
6
•各労働者の効用は所得の関数として定義
U′ >
0,
U′′
0,
and
U(0
)=
0(5
)
•労働組合の効用関数（
b：失業給付金）
UT(L
u,w
u,b
)=
Lu[U
(wu)−
U(b
)](6
)
•移動条件（
L∗ u：都市の失業者数、
Nu=
Lu+
L∗ u、
λ=
Lu/N
u）
U(w
r)=
λU(w
u)+
(1−
λ)U
(b)
(7)
•労働賦存条件
L=
Lr+
Nu=
Lr+
Lu λ
(8)
7
レジュメ集３－2 ページ
4-1
.独占的労働組合
•労働組合が企業の労働需要関数を制約として組合の効用を最大にするように賃
金率を設定
max
wu
UT(L
u,w
u,b
)=
Lu[U
(wu)−
U(b
)],
s.t.
Lu=
Ld u=
⎛ ⎜ ⎝a wu
⎞ ⎟ ⎠1
1−
a
•組合の効用最大化の一階の条件
dU
T
dw
u=
Ld uU′ (
wu)+
[U(w
u)−
U(b
)]dL
d u
dw
u=
0(9
)
8
4-2
.期待賃金のケース
•移動条件が期待賃金のケースにおける組合の効用を最大化する賃金率は
wu=
b a(1
0)
•移動条件式と労働賦存条件式から
wrと
λが決定
λ=
a
(1−
a)
wr−
b
b(1
1)
λ=
Lu(b
)
L−
g(w
r)=
⎛ ⎜ ⎝b a2
⎞ ⎟ ⎠−1
1−
a
L−
g(w
r)(1
2)
9
wr
λ 1
Lu(b
)
L wr
λ=
Lu(b
)
L−
g(w
r)
λ=
a
(1−
a)b
wr−
1
−1
w∗ r
λ∗
10
•失業が存在しない、つまり
λ=
1とする
bと
wrの存在
•(1
1)式と
(12)式において
λ=
1と置けば、次式を得る
b=
aw
r(1
3)
b=
a2
[L−
g(w
r)](
1−
a)
(14)
この二本の方程式から
bと
wrが決定され、一意の最適失業給付水準が存在する
11
レジュメ集３－3 ページ
b
wr
a
a2
L(1
−a)
wr
w∗ r=
wc
b∗
12
4-3
.期待効用のケース
•個人の効用を
U=
lnw
uと特定化
•移動条件が期待効用のケースにおける組合の効用を最大化する賃金率は
wu=
[exp(1
−a)]
b(1
5)
•移動条件式と労働賦存条件式から
wrと
λが決定
λ=
1
(1−
a)ln
wr b
(16)
λ=
( be(1
−a) /
a) −
1
1−
a
L−
g(w
r)(1
7)
13
wr
λ
1
Lu(b
)
Lw
r
λ=
Lu(b
)
L−
g(w
r)
λ=
1
(1−
a)ln
⎛ ⎝wr b
⎞ ⎠
w∗ r
λ∗
−ln
b
(1−
a)
b1
14
•失業が存在しない、つまり
λ=
1とする
bと
wrの存在
•(1
6)式と
(17)式において
λ=
1と置けば、次式を得る
b=
a
{e[L
−g(w
r)]}
(1−
a),
(18)
b=
wre
−(1
−a) .
(19)
この二本の方程式から
bと
wrが決定され、一意の最適失業給付水準が存在する
15
レジュメ集３－4 ページ
b
wr
e−(1
−a)
a
(eL)(
1−
a)
wr
w∗ r=
wc
b∗
16
5.結論とその理由
•移動条件が期待効用でも期待賃金でも最適失業給付金水準が存在
•移動条件をどのように設定するかが問題ではない
•Calv
o(1
978)モデルの労働組合の効用関数では、都市部門における代替的賃
金が農村賃金であったが、そうではなく失業給付金に変形
17
6.今後の課題など
•雇用されている労働者の生産性は失業給付水準に依存していない
•ここでは明示的に政府の活動を導入してはいない。したがって、明示的に政府
の活動を導入した場合における最適な失業給付金が存在するのか
•インフォーマル・セクターが存在し、それと同時に失業が存在するケースにお
ける最適失業給付金が存在するのか
18
レジュメ集３－5 ページ
Inte
rnat
iona
l Diff
eren
ces i
n C
onsu
mer
Pre
fere
nces
and
Tra
de:
Evi
denc
e fr
om M
ultic
ount
ry, M
ultip
rodu
ct D
ata*
Ole
ksan
dr M
ovsh
uk
Dep
artm
ent o
f Eco
nom
ics,
Toya
ma
Uni
vers
ity,
3190
Gof
uku,
Toy
ama
930-
8555
, Jap
an
E-m
ail:
mov
shuk
@ec
o.to
yam
a-u.
ac.jp
May
200
5
(Com
men
ts w
elco
me)
Abs
trac
t
In th
is p
aper
, I e
xam
ine
a co
mm
on n
egle
ct o
f int
erna
tiona
l tra
de th
eorie
s of
dem
and-
side
di
ffere
nces
, fo
cusi
ng i
n pa
rticu
lar
on t
heir
assu
mpt
ion
of i
dent
ical
and
hom
othe
tic
pref
eren
ces
(IH
P) a
cros
s co
untri
es.
Usi
ng d
ata
from
the
lat
est
roun
d of
Int
erna
tiona
l C
ompa
rison
Pro
ject
for
199
6, w
hich
con
tain
s da
ta f
or 2
4 ca
tego
ries
of c
onsu
mpt
ion
in
115
coun
tries
, the
pap
er e
xam
ines
the
plau
sibi
lity
of I
HP,
spe
cific
ally
the
impl
icat
ion
of
iden
tical
con
sum
ptio
n sh
ares
acr
oss
coun
tries
. A
pply
ing
a ro
bust
tes
t fo
r m
ultiv
aria
te
outli
ers
to i
nter
natio
nal
cons
umpt
ion
shar
es, t
he p
aper
fou
nd a
wid
e di
vers
ity i
n th
ese
shar
es.
A l
arge
num
ber
of c
ount
ries
exhi
bite
d ou
tlyin
g pa
ttern
s in
the
ir co
nsum
ptio
n sh
ares
, es
peci
ally
cou
ntrie
s w
ith t
otal
con
sum
ptio
n pe
r ca
pita
bel
ow r
ough
ly 5
,000
in
tern
atio
nal d
olla
rs.
Con
vers
ely,
the
sim
ilarit
y of
con
sum
ptio
n sh
ares
was
muc
h m
ore
evid
ent
for
of r
ich
coun
tries
tha
t ex
ceed
ed t
he c
utof
f va
lue.
In
the
sec
ond
part
of t
he
pape
r, I a
pply
the
sign
test
, ran
k te
st a
nd S
pear
man
rank
cor
rela
tion
test
s of
Bow
en e
t al.
(198
7) to
test
the
rela
tions
hip
betw
een
idio
sync
ratic
con
sum
ptio
n sh
ares
and
trad
e flo
ws.
Th
e pa
per
foun
d th
at u
nusu
ally
lar
ge d
eman
d ha
s th
e (e
xpec
ted)
pos
itive
im
pact
on
impo
rts, b
ut th
e (u
nexp
ecte
d) p
ositi
ve im
pact
on
expo
rts, a
ppar
ently
indi
catin
g th
e ‘h
ome
effe
ct’ f
rom
larg
e do
mes
tic m
arke
ts o
n ex
pert
com
petit
iven
ess.
How
ever
, the
impa
ct o
f de
man
d di
ffere
nces
on
trade
turn
ed o
ut h
ighl
y in
sign
ifica
nt in
the
subs
et o
f ric
h co
untri
es,
impl
ying
tha
t IH
P m
ay i
ndee
d be
a f
easi
ble
assu
mpt
ion
for
anal
yzin
g tra
de,
but
only
am
ong
rich
econ
omie
s.
JEL
clas
sific
atio
n nu
mbe
rs: F
11, D
12, C
21.
* I am
gra
tefu
l to
Kiy
oshi
Fuj
ikaw
a, K
eith
Mas
kus a
nd E
iichi
Tom
iura
for u
sefu
l com
men
ts o
n ea
rlier
dr
afts
of t
his p
aper
. The
usu
al d
iscl
aim
er a
pplie
s.
1
Intr
oduc
tion.
Even
thou
gh in
tern
atio
nal d
iffer
ence
s in
sup
ply
and
dem
and
can
conc
eiva
bly
be s
ourc
es
of c
ompa
rativ
e ad
vant
ages
, th
e co
nven
tiona
l tra
de t
heor
y ha
s no
t co
nsid
ered
dem
and
diff
eren
ces
as a
n im
porta
nt s
ourc
e of
trad
e.
A ty
pica
l exa
mpl
e is
the
Hec
ksch
er-O
hlin
-
Vane
k th
eory
of
inte
rnat
iona
l tra
de.
It
focu
ses
on i
nter
natio
nal
diffe
renc
es i
n fa
ctor
endo
wm
ent,
and
neut
raliz
es th
e im
pact
of d
eman
d by
ass
umin
g id
entic
al a
nd h
omot
hetic
pref
eren
ces
(IH
P) a
cros
s th
e w
orld
. T
his
assu
mpt
ion
impl
ies
that
rep
rese
ntat
ive
cons
umer
s in
all
tradi
ng c
ount
ries
have
a c
omm
on u
tility
func
tion,
and
that
inte
rnat
iona
l
inco
me
diffe
renc
es h
ave
no e
ffect
on
cons
umpt
ion
shar
es.
Tak
en t
oget
her
with
the
assu
mpt
ion
of fr
ee tr
ade
and
zero
tran
spor
tatio
n co
sts,
this
pat
tern
of p
refe
renc
es g
reat
ly
sim
plifi
es t
he a
bsor
ptio
n si
de o
f th
e H
ecks
cher
-Ohl
in-V
anek
the
ory,
with
exp
endi
ture
shar
es o
f tra
ded
good
s af
fect
ed b
y on
ly r
elat
ive
pric
es,
with
no
impa
ct f
rom
diff
eren
t
tast
es o
r inc
ome
leve
ls a
cros
s cou
ntrie
s.
The
feas
ibili
ty o
f ass
umin
g IH
P w
as re
peat
edly
que
stio
ned
in re
cent
yea
rs.
Afte
r
all,
the
assu
mpt
ion
is a
t odd
s w
ith th
e En
gel l
aw th
at li
nks
incr
easi
ng in
com
e le
vels
with
decr
easi
ng s
hare
of
food
exp
endi
ture
s. D
avis
and
Wei
nste
in (
2001
) ac
know
ledg
ed th
at
the
dem
and
side
of
the
Hec
ksch
er-O
hlin
-Van
ek m
odel
“cr
ies
out f
or m
ore
rese
arch
” (p
.
39),
whi
le L
eam
er (
2002
) po
inte
d ou
t th
at t
he “
assu
mpt
ion
that
bud
get
shar
es a
re
inde
pend
ent o
f in
com
es w
orks
wel
l the
oret
ical
ly, b
ut r
athe
r ba
dly
viol
ates
the
fact
s” (
p.
11).
Tab
le 1
sum
mar
izes
maj
or s
tudi
es t
hat
exam
ined
whe
ther
pre
fere
nces
are
ind
eed
iden
tical
and
hom
othe
tic.
In
one
of t
he m
ost
com
preh
ensi
ve s
tudi
es,
Hun
ter
and
Mar
kuse
n (1
988)
test
ed th
e nu
ll hy
poth
esis
of
hom
othe
tic p
refe
renc
es f
or 1
1 go
ods
and
serv
ices
in
34 c
ount
ries,
and
reje
cted
the
nul
l hy
poth
esis
for
7 e
xpen
ditu
re c
ateg
orie
s
レジュメ集３－6 ページ
2
(incl
udin
g fo
od, i
n ac
cord
ance
with
the
Eng
el l
aw).
On
the
othe
r ha
nd, D
owric
k an
d
Qui
ggin
(19
94)
test
ed t
he n
ull
hypo
thes
is o
f id
entic
al p
refe
renc
es b
y ap
plyi
ng t
he
reve
aled
pre
fere
nce
appr
oach
to
10 a
nd 3
8 ex
pend
iture
cat
egor
ies
in 6
0 co
untri
es.
Surp
risin
gly,
th
e nu
ll hy
poth
esis
w
as
viol
ated
on
ly
twic
e (o
ut
of
1700
pa
irwis
e
inte
rnat
iona
l com
paris
ons)
with
10
expe
nditu
re c
ateg
orie
s, an
d w
as n
ever
vio
late
d w
ith
38 e
xpen
ditu
re c
ateg
orie
s.
Gen
eral
ly, t
here
is li
ttle
doub
t tha
t the
idea
lized
wor
ld o
f ide
ntic
al a
nd h
omot
hetic
pref
eren
ces i
s a p
oor m
atch
to th
e ac
tual
con
sum
ptio
n pa
ttern
s in
the
wor
ld.
How
ever
, it
is m
uch
less
cle
ar w
heth
er d
evia
tions
from
this
rest
rictiv
e de
man
d st
ruct
ure
(whi
ch a
llow
for n
on-h
omot
hetic
and
non
-iden
tical
pre
fere
nces
) can
pro
vide
a b
ette
r exp
lana
tion
of
trade
flow
s, in
com
paris
on w
ith th
e tra
ditio
nal f
ocus
on
supp
ly-s
ide
diffe
renc
es o
nly.
Mar
kuse
n (1
986)
was
an
early
atte
mpt
of a
pply
ing
such
a le
ss re
stric
tive
dem
and
stru
ctur
e to
ana
lyzi
ng tr
ade
betw
een
rich
Nor
th a
nd p
oor S
outh
cou
ntrie
s. H
e su
gges
ted
an ‘e
clec
tic m
odel
’ tha
t ble
nded
scal
e ec
onom
ies,
diff
eren
ces i
n en
dow
men
ts a
nd n
on-
hom
othe
tic p
refe
renc
es, a
nd th
is m
odel
exp
lain
ed th
e lo
w le
vel o
f int
er-in
dust
ry tr
ade
betw
een
capi
tal-a
bund
ant N
orth
and
labo
r-ab
unda
nt N
orth
. C
hung
(200
0) c
onsi
dere
d
how
the
incl
usio
n of
non
-hom
othe
tic p
refe
renc
es in
the
Hec
ksch
er-O
hlin
-Van
ek th
eory
coul
d ex
plai
n th
e va
st a
mou
nt o
f ‘m
issi
ng tr
ade’
, poi
nted
by
Tref
ler (
1995
).
Subs
eque
ntly
, Rei
mer
and
Her
tel (
2003
) em
piric
ally
dem
onst
rate
d th
at a
llow
ing
for n
on-
hom
othe
tic p
refe
renc
es g
reat
ly im
prov
es th
e fit
bet
wee
n pr
edic
ted
and
obse
rved
fact
or
cont
ent o
f tra
de1 .
1 Spe
cific
ally
, the
pap
er fo
und
that
non
-hom
othe
tic p
refe
renc
es c
an e
xpla
in a
way
abo
ut o
ne-f
ifth
of th
e “m
issi
ng tr
ade”
and
one
-third
of t
he “
endo
wm
ents
par
adox
” m
yste
ries.
レジュメ集３－7 ページ
4
This
pap
er a
lso
aim
s to
dem
onst
rate
the
impo
rtanc
e of
dem
and
side
diff
eren
ces
for e
xpla
inin
g in
tern
atio
nal t
rade
flow
s.Its
maj
or n
ovel
con
tribu
tion
is th
at d
eman
d
diff
eren
ces a
re c
onsi
dere
d as
an
inde
pend
ent s
ourc
e of
com
para
tive
adva
ntag
es, r
athe
r
than
a ‘f
ix’ t
o im
prov
e th
e pr
edic
tive
perf
orm
ance
of s
uppl
y-si
de th
eorie
s of i
nter
natio
nal
trade
2 . Th
e re
leva
nce
of in
tern
atio
nal d
eman
d di
ffer
ence
s for
inte
rnat
iona
l tra
de fl
ows i
s
exam
ined
with
a ri
ch d
atas
et o
f 24
cons
umpt
ion
cate
gorie
s in
115
coun
tries
.
Maj
or fi
ndin
gs c
an b
e br
iefly
sum
mar
ized
as f
ollo
ws.
Whi
le th
e as
sum
ptio
n of
iden
tical
and
hom
othe
tic p
refe
renc
es im
plie
s tha
t goo
ds a
re c
onsu
med
in th
e sa
me
expe
nditu
re sh
ares
, I fo
und
a su
bsta
ntia
l dis
pers
ion
in th
ese
shar
es a
cros
s the
wor
ld.
Whi
le p
revi
ousl
y th
is c
oncl
usio
n w
as re
ache
d by
est
imat
ing
a pa
rticu
lar d
eman
d sy
stem
,
in th
is p
aper
I us
e a
mor
e ge
nera
l, m
odel
-fre
e ap
proa
ch to
reac
h th
e sa
me
conc
lusi
on.
Spec
ifica
lly, I
app
lied
a te
st o
f mul
tivar
iate
out
liers
dire
ctly
to c
onsu
mpt
ion
shar
es, w
ith
an im
porta
nt a
dvan
tage
that
the
outc
ome
is n
ot c
ondi
tiona
l on
the
choi
ce o
f a sp
ecifi
c
dem
and
syst
em.
The
larg
est n
umbe
r of u
nusu
al (‘
outly
ing’
) con
sum
ptio
n sh
ares
was
iden
tifie
d fo
r exp
endi
ture
s on
food
(in
agre
emen
t with
the
Enge
l law
). A
lso,
mos
t of
outly
ing
cons
umpt
ion
shar
es re
pres
ente
d co
untri
es w
ith lo
w le
vels
of p
er c
apita
tota
l
cons
umpt
ion
(with
cut
off v
alue
of a
bout
5,0
00 in
tern
atio
nal d
olla
rs),
indi
catin
g th
at th
e
iden
tical
and
hom
othe
tic p
refe
renc
es m
ay in
deed
be
an a
ccep
tabl
e as
sum
ptio
n fo
r
expl
aini
ng tr
ade,
but
onl
y am
ong
rich
coun
tries
.
In it
s sec
ond
part,
the
pape
r exa
min
es th
ese
devi
atio
ns fr
om id
entic
al a
nd
hom
othe
tic p
refe
renc
es, a
nd a
naly
ses w
heth
er u
nusu
ally
larg
e (o
r sm
all)
norm
aliz
ed
cons
umpt
ion
shar
es c
an b
e as
soci
ated
with
unu
sual
ly la
rge
(or s
mal
l) no
rmal
ized
exp
orts
2 In th
is re
spec
t, th
e pa
per i
s mos
t clo
sely
rela
ted
to H
unte
r (19
91),
who
est
imat
ed th
at th
e in
trodu
ctio
n of
no
n-ho
mot
hetic
, but
iden
tical
, pre
fere
nces
may
acc
ount
for a
s muc
h as
one
-qua
rter o
f tra
de fl
ows.
5
and
impo
rts.
This
is d
one
with
the
sign
test
and
med
ian
test
from
the
sem
inal
pap
er b
y
Bow
en, L
eam
er a
nd S
veik
ausk
as (1
987)
. I f
ound
a h
ighl
y si
gnifi
cant
pos
itive
rela
tion
betw
een
cons
umpt
ion
shar
es a
nd im
port
shar
es, i
n ag
reem
ent t
o th
e pr
ior e
xpec
tatio
n th
at
unus
ually
stro
ng d
eman
d fo
r a g
ood
is li
kely
to re
sult
in e
xtra
impo
rts.
Surp
risin
gly,
ther
e w
as o
nce
agai
n a
posi
tive
rela
tion
betw
een
cons
umpt
ion
and
expo
rt sh
ares
. I
inte
rpre
t thi
s fin
ding
as a
noth
er p
iece
of e
vide
nce
abou
t ‘ho
me
mar
ket e
ffec
t’, in
trodu
ced
by K
rugm
an (1
980)
and
rece
ntly
ver
ified
em
piric
ally
by
Dav
is a
nd W
eins
tein
(200
3),
Han
son
and
Xia
ng (2
004)
. A
noth
er n
otew
orth
y re
sult
occu
rred
whe
n th
e sa
mpl
e w
as
rest
ricte
d to
the
subs
et o
f 38
rich
coun
tries
(usi
ng a
s a c
utof
f the
val
ue o
f 500
0
inte
rnat
iona
l dol
lars
for t
otal
per
cap
ita c
onsu
mpt
ion)
. Th
e id
entif
ied
rela
tions
hips
betw
een
cons
umpt
ion
patte
rns a
nd tr
ade
wer
e no
long
er si
gnifi
cant
, onc
e ag
ain
indi
catin
g
that
dem
and
diff
eren
ces m
ay n
ot p
lay
an im
porta
nt ro
le fo
r tra
de o
f ric
h ec
onom
ies.
This
pap
er is
org
aniz
ed a
s fol
low
s. S
ectio
n 1
disc
usse
s the
impa
ct o
f con
sum
er
pref
eren
ces o
n in
tern
atio
nal t
rade
. Se
ctio
n 2
desc
ribes
test
s to
exam
ine
the
rela
tions
hip
betw
een
dom
estic
dem
and
and
inte
rnat
iona
l tra
de, a
nd d
escr
ibes
dat
a so
urce
s. M
ajor
resu
lts a
re re
porte
d in
sect
ion
3. S
ectio
n 4
cont
ains
maj
or c
oncl
usio
ns a
nd d
irect
ions
for
futu
re re
sear
ch.
Sect
ion
1. C
onsu
mer
pre
fere
nces
and
inte
rnat
iona
l tra
de
To d
efin
e IH
P, c
onsi
der a
repr
esen
tativ
e co
nsum
er w
ho c
hoos
es b
etw
een
good
s x a
nd y
.
Hom
othe
tic p
refe
renc
es a
re d
efin
ed a
s fol
low
s: fo
r all
0, t
he c
onsu
mer
pre
fers
a
bund
lex
to a
bun
dle
yif
the
cons
umer
pre
fers
bun
dle
x to
y.
In
wor
ds, t
he o
rder
ing
of
レジュメ集３－8 ページ
6
cons
umer
’s p
refe
renc
es d
oes n
ot c
hang
e if
cons
umpt
ion
bund
les a
re sc
aled
by
a sc
alar
.
Hom
othe
tic p
refe
renc
es w
ith
1 ar
e ill
ustra
ted
by F
ig.1
.
Fig.
1.
Hom
othe
tic p
refe
renc
es
To d
efin
e id
entic
al p
refe
renc
es, c
onsi
der r
epre
sent
ativ
e co
nsum
ers i
n H
ome
and
Abro
ad a
nd th
eir r
espe
ctiv
e so
cial
indi
ffere
nce
curv
es.
Pref
eren
ces a
re id
entic
al a
nd
hom
othe
tic if
, up
to a
scal
ing
fact
or
> 0
, the
shap
e of
soci
al in
diffe
renc
e cu
rves
rem
ains
the
sam
e.
The
assu
mpt
ion
of IH
P gr
eatly
sim
plifi
es th
e de
man
d si
de o
f con
vent
iona
l tra
de
theo
ries.
Con
sum
er p
refe
renc
es c
an b
e de
duce
d fr
om a
sing
le in
diffe
renc
e cu
rve,
rega
rdle
ss o
f inc
ome
leve
ls.
Mor
eove
r, w
hen
cons
umer
s hav
e di
ffere
nt in
com
es, b
ut fa
ce
the
sam
e pr
ices
(due
to p
ostu
late
d fr
ee tr
ade)
, the
con
sum
ers w
ould
dem
and
good
s in
a
prop
ortio
n th
at is
con
stan
t alo
ng th
e ra
y fr
om th
e or
igin
(in
Fig.
1, t
his i
dent
ical
prop
ortio
n in
Hom
e an
d Ab
road
is e
vide
nt in
the
iden
tical
slop
e of
Oc 1
and
Oc 2
). In
oth
er
wor
ds, i
t is o
nly
rela
tive
pric
es o
f goo
ds x
and
y, a
nd n
ot in
com
e or
tast
es, t
hat d
eter
min
e
the
ratio
in w
hich
x a
nd y
are
con
sum
ed in
Hom
e an
d Ab
road
. Si
nce
the
conv
entio
nal
y
x
c 1
c 2
3c
O
HO
ME
ABRO
AD
7
trade
theo
ry a
ttrib
utes
diff
eren
ces i
n re
lativ
e pr
ices
to o
nly
supp
ly-s
ide
diffe
renc
es a
mon
g
natio
ns (i
n pr
oduc
tivity
, fac
tor e
ndow
men
ts, e
tc),
the
use
of IH
P as
sum
es a
way
any
impa
ct o
f dem
and-
side
diff
eren
ces o
n in
tern
atio
nal t
rade
.
Con
vers
ely,
if p
refe
renc
es a
re n
on-id
entic
al o
r non
-hom
othe
tic, d
eman
d-si
de
diffe
renc
es c
ould
be
a so
urce
of c
ompa
rativ
e ad
vant
ages
. Fi
g. 2
illu
stra
tes t
he c
ase
of
non-
iden
tical
pre
fere
nces
in H
ome
and
Abro
ad.
Fig.
2.
Non
-iden
tical
pre
fere
nces
as a
sour
ce o
f com
para
tive
adva
ntag
es
Bot
h co
untri
es sh
are
the
sam
e pr
oduc
tion
poss
ibili
ty fr
ontie
r. H
owev
er,
cons
umer
s in
Hom
e pr
efer
x to
y, w
hile
pre
fere
nces
in A
broa
d ar
e th
e ot
her w
ay a
roun
d.
Not
e th
at n
ow th
e ra
tio o
f x a
nd y
, as w
ell a
s the
ir co
nsum
ptio
n sh
ares
, are
no
long
er
iden
tical
in H
ome
and
Abro
ad.
Rel
ativ
ely
high
dem
and
for x
inH
ome
mak
es x
rela
tivel
y
mor
e ex
pens
ive.
The
sam
e ho
lds f
or y
in A
broa
d. T
here
fore
, Hom
e ha
s the
com
para
tive
adva
ntag
e in
y, w
hile
Abr
oad
has t
he c
ompa
rativ
e ad
vant
age
in x
. In
shor
t, co
untri
es
**
yx
PP y
xP
P
ABRO
AD
HO
ME
xO
c 1
c 2
y
レジュメ集３－9 ページ
8
expo
rt th
e go
od th
at th
ey d
o no
t dem
and
so m
uch,
and
impo
rt th
e go
od th
ey d
eman
d
rela
tivel
y m
ore.
To sh
ow h
ow n
on-h
omot
hetic
pre
fere
nces
can
giv
e ris
e to
inte
rnat
iona
l tra
de,
supp
ose
that
pre
fere
nces
in H
ome
and
Abro
ad a
re id
entic
al, b
ut p
refe
renc
es fo
r x a
re n
on-
hom
othe
tic.
Let x
be
an in
ferio
r goo
d. I
f inc
ome
leve
l in
Hom
e is
hig
her r
elat
ivel
y to
Abro
ad, t
hen
Hom
e w
ould
hav
e th
e co
mpa
rativ
e ad
vant
age
in x
(due
to lo
wer
dom
estic
dem
and
for t
his g
ood)
, whi
le A
broa
d w
ill h
ave
com
para
tive
adva
ntag
e in
y.
In g
ener
al,
the
rela
tivel
y ric
h co
untry
will
exp
ort t
he g
ood
whi
ch h
as a
rela
tivel
y lo
wer
inco
me
elas
ticity
of d
eman
d, re
flect
ing
diff
eren
t con
sum
ptio
n sh
ares
for x
and
y in
Hom
e an
d
Abro
ad d
ue to
the
non-
hom
othe
tic d
eman
d.
Sect
ion
2. D
ata
and
empi
rica
l tes
ts.
To e
valu
ate
the
effe
ct o
f diff
eren
t con
sum
ptio
n sh
ares
on
natio
nal e
xpor
ts a
nd im
ports
,
this
pap
er w
ill te
st w
heth
er d
evia
tions
from
‘typ
ical
’ con
sum
ptio
n sh
ares
is re
late
d to
sim
ilarly
def
ined
dev
iatio
ns fo
r exp
orts
and
impo
rts, w
ith t
ypic
al le
vels
repr
esen
ted
by
med
ian
shar
es o
f a g
iven
com
mod
ity a
cros
s cou
ntrie
s.
Let c
ount
ries b
e de
note
d by
n(
1,...
,)
nN
, and
com
mod
ities
by
k(
1,...
,)
kK
.
The
valu
e of
con
sum
ptio
n, e
xpor
t and
impo
rts b
y de
fined
by
uppe
rcas
e le
tters
C,X
, and
M, a
nd le
t the
cor
resp
ondi
ng lo
wer
case
lette
rs d
enot
e th
e co
rres
pond
ing
shar
es in
the
tota
l as f
ollo
ws:
,
,,
1kn
kn
Kk
nkC
cC
,,
,,
1kn
kn
Kk
nkX
xX
,,
,,
1kn
kn
Kk
nkM
mM
.
9
Let
kc,
kx a
nd
km
den
ote
med
ian
valu
es o
f con
sum
ptio
n, e
xpor
t and
impo
rt
shar
es o
f goo
d k
acro
ss a
ll N
cou
ntrie
s. T
erm
s ,kn
c,
,kn
x a
nd
,kn
m a
re m
ade
scal
e-
inva
riant
by
the
follo
win
g no
rmal
izin
g tra
nsfo
rmat
ions
:
,,
ˆln
()
kn
kn
kc
cc
,,
,ˆ
ln(
)k
nk
nk
xx
x,
,,
ˆln
()
kn
kn
km
mm
(1
)
Not
e, in
par
ticul
ar, t
hat i
f cou
ntrie
s dem
and
the
sam
e sh
are
of e
ach
good
(as t
hey
shou
ld u
nder
IHP)
, tha
n th
e te
rm
,kn
c w
ould
be
zero
. A
lso,
a n
egat
ive
valu
e of
,ˆ kn
c
indi
cate
s tha
t con
sum
ptio
n of
com
mod
ity k
in c
ount
ry n
is lo
wer
than
the
‘ben
chm
ark’
shar
ekc
(def
ined
by
med
ian
valu
es fo
r the
full
set o
f N c
ount
ries)
. C
onve
rsel
y, p
ositi
ve
valu
e of
,ˆ kn
cm
eans
that
com
mod
ity k
in c
ount
ry n
is c
onsu
med
in re
lativ
ely
high
prop
ortio
n th
an th
e be
nchm
ark
inte
rnat
iona
l lev
el
kc fo
r com
mod
ity k
3 . Tr
ade
shar
es
,ˆ k
nx
and
,
ˆ kn
m h
ave
the
sam
e in
terp
reta
tion.
(a) D
ata
sour
ces
Con
sum
ptio
n sh
ares
,ˆ kn
c w
ere
com
pute
d fr
om th
e la
test
ben
chm
ark
stud
y of
the
Inte
rnat
iona
l Com
paris
on P
rogr
am (I
CP)
. Th
e IC
P or
igin
ated
in th
e 19
60s i
n th
e
Uni
vers
ity o
f Pen
nsyl
vani
a, a
nd a
t pre
sent
is c
oord
inat
ed b
y th
e W
orld
Ban
k. I
ts p
rimar
y
goal
is th
e ca
lcul
atio
n of
pur
chas
ing
pow
er p
ariti
es fo
r maj
or e
xpen
ditu
re c
ateg
orie
s of
natio
nal a
ccou
nts.
As a
resu
lt, th
e da
taba
se p
rovi
des n
ot o
nly
nom
inal
con
sum
ptio
n
expe
nditu
res (
in d
omes
tic c
urre
ncy)
, but
als
o th
e co
rres
pond
ing
set o
f pur
chas
ing
pow
er
3Th
e us
e of
ben
chm
ark
leve
ls in
(1)
mak
es th
ese
varia
bles
con
cept
ually
sim
ilar t
o ID
IOD
EM v
aria
ble
in
Dav
is a
nd W
eins
tein
(200
3), w
here
it w
as d
efin
ed a
s a “
mea
sure
of t
he e
xten
t of i
dios
yncr
atic
der
ived
de
man
d [th
at] …
mea
sure
s the
ext
ent t
o w
hich
the
rela
tive
dem
and
for a
goo
d w
ithin
an
indu
stry
diff
ers
from
that
in th
e re
st o
f the
wor
ld”
(p. 7
).
レジュメ集３－10 ページ
10
parit
y (P
PP) e
stim
ates
. Im
porta
ntly
, gre
at c
are
is g
iven
in IC
P to
ens
ure
that
goo
ds o
f
sim
ilar q
ualit
y ar
e co
mpa
red
acro
ss c
ount
ries.
For t
his p
aper
, I u
sed
the
late
st a
vaila
ble
ICP
data
base
for 1
9964 .
It in
clud
es
cons
umpt
ion
data
for 1
15 c
ount
ries (
the
larg
est n
umbe
r eve
r in
the
ICP’
s his
tory
). In
tota
l, 24
con
sum
ptio
n ca
tego
ries w
ere
exam
ined
5 . Th
e fu
ll lis
t of c
ount
ries a
nd
expe
nditu
re c
ateg
orie
s is p
rovi
ded
in A
ppen
dix
tabl
es A
1 an
d A
2.
The
orig
inal
sam
ple
of 1
15 c
ount
ries i
nclu
des a
larg
e nu
mbe
r of d
evel
opin
g
coun
tries
that
may
hav
e re
lativ
ely
unre
liabl
e es
timat
es o
f con
sum
ptio
n ex
pend
iture
s. T
o
iden
tify
a su
bset
of c
ount
ries w
ith re
lativ
ely
high
qua
lity
of d
ata,
I us
ed v
aria
nce
scor
es
from
Hes
ton
et a
l. (2
004)
. Th
e sc
ores
rang
e fr
om 0
to 5
, with
hig
h sc
ores
iden
tifyi
ng
coun
tries
with
hig
h co
nsis
tenc
y in
PPP
est
imat
es a
cros
s diff
eren
t ben
chm
ark
stud
ies.
The
‘hig
h qu
ality
’ sub
set o
f cou
ntrie
s had
var
ianc
e sc
ores
4 o
r 5, a
nd c
onta
ined
76
coun
tries
.
To c
ompu
te e
xpor
t and
impo
rt sh
ares
,
ˆ kn
x a
nd
,ˆ k
nm
, I u
sed
Wor
ld T
rade
Ana
lyze
r
data
base
, rel
ease
d by
Sta
tistic
s Can
ada
(200
1).
The
data
base
con
tain
s nom
inal
bila
tera
l
trade
flow
s at 4
-dig
it le
vel o
f Sta
ndar
d Tr
ade
Inte
rnat
iona
l Cla
ssifi
catio
n (S
ITC
, rev
. 2)
for a
lmos
t all
coun
tries
in th
e IC
P sa
mpl
e of
cou
ntrie
s. T
rade
dat
a tu
rned
out
mis
sing
for
only
7 c
ount
ries f
rom
the
orig
inal
sam
ple
of 1
15 c
ount
ries,
and
for o
nly
4 co
untri
es fr
om
the
‘hig
h qu
ality
’ sub
set o
f 76
coun
tries
6 . Th
eref
ore,
the
impa
ct o
f con
sum
ptio
n on
trad
e
patte
rns c
ould
be
anal
yzed
with
108
and
72
coun
tries
. To
cal
cula
te tr
ade
shar
es
,ˆ k
nx
and
4 The
dat
aset
was
dow
nloa
ded
from
http
://pw
t.eco
n.up
enn.
edu/
Dow
nloa
ds/b
ench
mar
k/be
nchm
ark.
htm
l 5
I exc
lude
d tw
o ex
pend
iture
cat
egor
ies f
rom
the
orig
inal
dat
aset
(‘re
stau
rant
s, ca
fes a
nd h
otel
s' an
d ‘o
ther
go
ods a
nd se
rvic
es’)
bec
ause
it is
com
mon
to in
clud
e in
such
resi
dual
s cat
egor
ies e
xpen
ditu
res t
hat a
re
diff
icul
t to
clas
sify
(Kra
vis e
t al.,
198
2, p
. 49)
, mak
ing
inte
rnat
iona
l com
paris
ons f
or th
ese
expe
nditu
re
cate
gorie
s les
s rel
iabl
e.6 T
he li
st o
f cou
ntrie
s with
mis
sing
trad
e da
ta is
pro
vide
d in
App
endi
x ta
ble
A3.
11
,ˆ k
nm
, I d
id n
ot u
se tr
ade
figur
es fo
r cou
ntry
n to
the
wor
ld, b
ut o
nly
to th
e ag
greg
ate
of
coun
tries
that
wer
e in
clud
ed in
this
stud
y (s
uch
as 1
08 c
ount
ries o
f the
full
sam
ple)
.
It w
as m
ore
diffi
cult
to m
atch
con
sum
ptio
n ca
tego
ries i
n th
e IC
P da
tase
t with
SITC
cat
egor
ies i
n th
e W
orld
Tra
de A
naly
zer.
Bec
ause
ther
e is
no
stan
dard
con
cord
ance
for m
atch
ing
thes
e tw
o pr
oduc
t cla
ssifi
catio
ns, I
com
pile
d a
new
con
cord
ance
for 1
5
cons
umpt
ion
cate
gorie
s in
the
ICP
data
set7 .
The
rem
aini
ng 9
con
sum
ptio
n ca
tego
ries
wer
e ei
ther
non
trada
bles
, or n
o re
liabl
e m
atch
cou
ld b
e fo
und
for t
he in
the
SITC
clas
sific
atio
n.
Take
n to
geth
er, t
he fu
ll sa
mpl
e of
115
cou
ntrie
s acc
ount
ed in
199
6 fo
r abo
ut 7
8.9
perc
ent o
f the
glo
bal G
DP,
and
89.
4 of
glo
bal e
xpor
ts o
f goo
ds a
nd se
rvic
es (W
orld
Ban
k,
2002
). T
he c
orre
spon
ding
shar
e in
the
wor
ld’s
pop
ulat
ion
was
smal
ler (
50.7
per
cent
),
mai
nly
due
to th
e ex
clus
ion
of C
hina
and
Indi
a fr
om th
e IC
P da
tase
t.
(b) E
mpi
rica
l tes
ts
As d
iscu
ssed
in se
ctio
n 1,
if th
ere
are
dem
and
diffe
renc
es (e
ither
due
to n
on-id
entic
al
pref
eren
ces,
or n
on-h
omot
hetic
pre
fere
nces
), co
untri
es n
o lo
nger
con
sum
e go
ods i
n
iden
tical
pro
porti
ons.
In a
dditi
on, u
nusu
ally
stro
ng d
eman
d fo
r a g
ood
wou
ld c
hang
e th
e
com
posi
tion
of c
ompa
rativ
e co
sts a
cros
s cou
ntrie
s, an
d le
ad to
incr
ease
d im
ports
and
redu
ced
expo
rts o
f thi
s goo
d, a
s dis
cuss
ed in
pre
viou
s sec
tion.
In
this
sect
ion,
I de
scrib
e
test
s tha
t exa
min
e th
is a
ssum
ed re
latio
nshi
p.
Thes
e te
sts a
re b
ased
on
the
sign
and
med
ian
test
, int
rodu
ced
in th
e se
min
al st
udy
by B
owen
et a
l. (1
987)
. Th
e si
gn te
st c
ompa
res t
he si
gn o
f nor
mal
ized
rela
tive
7 The
con
cord
ance
is re
porte
d in
App
endi
x ta
ble
A4.
レジュメ集３－11 ページ
12
cons
umpt
ion
shar
e ,ˆ kn
c a
nd th
e si
gn o
f cor
resp
ondi
ng e
xpor
t sha
re,
ˆ kn
x, o
r the
sign
of
impo
rt sh
are
,ˆ k
nm
:
,,
ˆˆ
()
()
kn
kn
Sign
cSi
gnx
for
exp
orts
(2a)
,,
ˆˆ
()
()
kn
kn
Sign
cSi
gnm
for
impo
rts
(2
b)
whe
re1,
...,
;1,
...,
.n
Nk
K
The
test
com
pare
s the
sign
pat
tern
of d
evia
tions
from
an
inte
rnat
iona
l ben
chm
ark
leve
l for
con
sum
ptio
n sh
ares
, with
sim
ilarly
def
ined
dev
iatio
ns in
exp
ort o
r im
port
shar
es.
As i
n B
owen
et a
l. (1
987)
, the
test
focu
ses o
n th
e pr
opor
tion
of c
orre
ct si
gn m
atch
es th
at
are
pred
icte
d by
(2a)
and
(2b)
, tha
t is,
oppo
site
sign
s for
,ˆ kn
c a
nd
,ˆ k
nx
, and
the
sam
e si
gns
for
,ˆ kn
c a
nd
,ˆ k
nm
.
In th
e ra
nk te
st, t
he fo
llow
ing
rela
tions
hips
are
exa
min
ed:
,,
,,
,,
,,
ˆˆ
ˆˆ
ˆˆ
ˆˆ
kn
lnk
nln
kn
lnk
nln
cc
xx
cc
xx
for e
xpor
ts
(3a)
and
,,
,,
,,
,,
ˆˆ
ˆˆ
ˆˆ
ˆˆ
kn
lnk
nln
kn
lnk
nln
cc
mm
cc
mm
o
r im
ports
(3
b)
whe
re
1,...
,;
,1,
...,
.n
Nk
lK
In a
dditi
on to
com
parin
g ,ˆ kn
c w
ith
,ˆ k
nx
and
,
ˆ kn
m tw
o at
a ti
me,
I co
mpa
red
thei
r ran
king
acro
ss a
ll K
goo
ds fo
r eac
h co
untry
1,...
,n
N, a
nd th
e si
gnifi
canc
e of
the
rank
corr
elat
ion
was
eva
luat
ed b
y th
e Sp
earm
an ra
nk c
orre
latio
n co
effic
ient
.
In e
ach
case
,
exac
t p-v
alue
s for
the
coef
ficie
nt
wer
e ca
lcul
ated
by
algo
rithm
of B
est a
nd R
ober
ts
(197
5).
To sa
ve sp
ace,
I w
ill re
port
only
cou
ntrie
s whe
re th
e ra
nk c
oeff
icie
nt
was
13
sign
ifica
nt a
t 5 p
erce
nt si
gnifi
canc
e le
vels
, diff
eren
tiatin
g be
twee
n co
rrec
t ran
king
s
(neg
ativ
e be
twee
n ,ˆ kn
c a
nd,
ˆ kn
x, a
nd p
ositi
ve b
etw
een
,ˆ kn
c a
nd,
ˆ kn
m),
and
inco
rrec
t
rank
ings
.
U
p to
now
, the
thre
e te
sts u
sed
in th
is p
aper
(sig
n te
sts (
2a) a
nd (2
b), r
ank
test
s
(3a)
and
(3b)
and
Spe
arm
an ra
nk c
orre
latio
n) a
re v
ery
sim
ilar t
o th
e ap
proa
ch o
f Bow
en
et a
l. (1
987)
to e
valu
ate
the
empi
rical
per
form
ance
of H
ecks
cher
-Ohl
in-V
anek
theo
ry o
f
trade
. Th
ere
is o
ne im
porta
nt d
iffer
ence
. B
owen
et a
l. (ib
id.)
just
repo
rted
the
shar
e of
‘suc
cess
ful m
atch
es’ i
n si
gn a
nd ra
nk te
sts.
In c
ontra
st, t
his p
aper
eva
luat
es th
e st
atis
tical
sign
ifica
nce
of th
ese
shar
es.
Bow
en e
t al.
(ibid
.) re
ject
ed th
e po
ssib
ility
of t
estin
g th
e
stat
istic
al si
gnifi
canc
e of
thei
r tes
ts b
y cl
aim
ing,
“th
e ra
nk a
nd si
gn p
ropo
sitio
ns b
ased
on
the
prop
ortio
n of
"suc
cess
es" d
o no
t ref
er to
any
spec
ific
alte
rnat
ive
hypo
thes
is a
nd th
us
leav
es u
ncle
ar th
e ch
oice
of s
igni
fican
ce le
vel.
With
out k
now
ing
the
prop
ortio
n of
succ
esse
s exp
ecte
d un
der a
spec
ific
alte
rnat
ive
hypo
thes
is, j
udgi
ng th
e re
lativ
e
perf
orm
ance
of t
he H
-O-V
mod
el is
larg
ely
impr
essi
onis
tic”
(p. 7
95).
Still
, it i
s pos
sibl
e to
def
ine
anot
her e
xpec
ted
prop
ortio
n, b
ut n
ot u
nder
the
alte
rnat
ive
hypo
thes
is, b
ut u
nder
the
null
that
pos
tula
tes n
o re
latio
n be
twee
n co
nsum
ptio
n
and
corr
espo
ndin
g tra
de sh
ares
. W
ith n
o re
latio
n be
twee
n do
mes
tic d
eman
d an
d tra
de,
the
prop
ortio
n of
‘suc
cess
ful’
and
‘uns
ucce
ssfu
l’ m
atch
es in
the
sign
and
rank
test
shou
ld
be e
qual
to 0
.5.
Bas
ed o
n th
is o
bser
vatio
n, it
is p
ossi
ble
to te
st w
heth
er th
e m
edia
n sh
are
of, s
ay, ‘
succ
essf
ul m
atch
es’ i
n a
coun
try n
(1,
...,
nN
) for
all
K g
oods
is si
gnifi
cant
ly
diff
eren
t fro
m 0
.5.
This
can
be
done
by
the
Wilc
oxon
test
for t
wo
paire
d sa
mpl
es, w
here
the
seco
nd sa
mpl
e is
just
the
colu
mn
with
K c
ells
with
0.5
. Es
sent
ially
, thi
s is a
univ
aria
te v
ersi
on o
f the
Wilc
oxon
test
, abo
ut a
spec
ific
valu
e of
pop
ulat
ion
med
ian.
So
レジュメ集３－12 ページ
14
in a
dditi
on to
repo
rting
p-v
alue
for t
he n
ull h
ypot
hesi
s tha
t the
pro
porti
on o
f suc
cess
ful
mat
ches
was
0.5
, I w
ill a
lso
repo
rt th
e m
edia
n va
lue
of th
ese
‘suc
cess
ful m
atch
es’ F
or
this
reas
on, r
epor
ts o
f app
lyin
g th
is te
st w
ill in
clud
e th
e m
edia
n le
vel o
f ‘su
cces
sful
mat
ches
’ acr
oss a
ll N
cou
ntrie
s, an
d a
p-va
lue
for t
he te
st.
Pr
ior t
o ap
plyi
ng th
ese
thre
e te
sts,
I als
o ex
amin
ed w
heth
er th
ere
was
a si
gnifi
cant
varia
bilit
y am
ong
orig
inal
con
sum
ptio
n sh
ares
,kn
c fo
r a sp
ecifi
c go
od k
am
ong
N
coun
tries
. Th
e la
ck o
f int
erna
tiona
l sim
ilarit
y in
con
sum
ptio
n sh
ares
indi
cate
s the
lack
of
IHP
acro
ss c
ount
ries,
so b
y ch
ecki
ng th
e po
ssib
le v
aria
bilit
y of
con
sum
ptio
n sh
ares
, we
esse
ntia
lly e
xam
ine
the
feas
ibili
ty o
f thi
s app
aren
tly re
stric
tive
treat
men
t of d
eman
d
cond
ition
s in
the
conv
entio
nal t
rade
theo
ry.
The
varia
bilit
y is
eva
luat
ed b
y a
robu
st te
st
of o
utlie
rs in
mul
tivar
iate
dat
a, b
ased
on
min
imum
cov
aria
nce
dete
rmin
ant (
MC
D)
estim
ator
, due
to R
ouss
eeuw
(198
4).
If c
onsu
mpt
ion
shar
es a
re b
road
ly si
mila
r acr
oss
coun
tries
, thi
s tes
t sho
uld
not i
dent
ify a
larg
e nu
mbe
r of m
ultiv
aria
te o
utlie
rs in
the
data
.
The
MC
D e
stim
ator
is e
ssen
tially
a ro
bust
ver
sion
of M
ahal
anob
is d
ista
nce
MD
i,
a m
easu
re h
ow fa
r an
ith ro
w in
mul
tivar
iate
dat
a de
viat
es fr
om th
e sa
mpl
e’s c
ente
r
(mea
sure
d by
sam
ple
mea
nx
):
1(
)(
)((
))(
)i
ii
MD
xx
xCo
vX
xx
How
ever
, the
Mah
alan
obis
dis
tanc
e is
susc
eptib
le to
out
liers
, and
sign
ifica
nt m
ultiv
aria
te
outli
ers m
ay h
ave
smal
l val
ues o
f MD
i. T
he p
robl
em is
less
serio
us w
ith th
e M
CD
estim
ator
, mai
nly
beca
use
it us
es a
robu
st e
stim
ator
of d
ata
cent
er (d
enot
ed b
y RC
) and
a
robu
st e
stim
ator
of v
aria
nce-
cova
rianc
e m
atrix
RC
ov (f
or d
etai
ls o
f a fa
st a
lgor
ithm
to
com
pute
the
estim
ator
, see
Rou
ssee
uw, a
nd V
an D
riess
en (1
999)
):
1(
)(
)((
))(
)i
ii
MCD
xx
RCRC
ovX
xRC
(4
)
15
In th
is p
aper
, the
MC
D e
stim
ator
was
app
lied
to 2
3 co
nsum
ptio
n ca
tego
ries l
iste
d
in T
able
A28 .
In a
dditi
on, t
he p
ossi
ble
diss
imila
rity
was
exa
min
ed w
ith se
vera
l sub
sets
of
cons
umpt
ion
cate
gorie
s tha
t con
tain
ed e
xpen
ditu
res o
n (1
) foo
d, (2
) bev
erag
es a
nd
toba
cco,
(3) c
loth
ing
and
foot
wea
r, (4
) ren
t and
fuel
, (5)
hou
seho
ld m
aint
enan
ce, (
6)
trans
porta
tion,
(7) r
ecre
atio
n an
d ed
ucat
ion
(the
com
posi
tion
of th
ese
subs
ets o
f
expe
nditu
res i
s sho
wn
in T
able
A2)
.
Sect
ion
3. T
est r
esul
ts
Res
ults
of a
pply
ing
MC
D te
st fo
r con
sum
ptio
n sh
are
data
are
show
n in
Fig
. 3 fo
r the
full
sam
ple
of 1
15 c
ount
ries,
and
in F
ig. 4
for t
he ‘h
igh
qual
ity’ s
ampl
e of
76
coun
tries
. Th
e
figur
es sh
ow ro
bust
dis
tanc
es, w
ith h
oriz
onta
l lin
es in
dica
ting
criti
cal v
alue
s for
the
MC
D
dist
ance
stat
istic
. Th
e nu
ll hy
poth
esis
is th
at c
onsu
mpt
ion
shar
es a
re si
mila
r, w
ith n
o
sign
ifica
nt o
utlie
rs.
Tabl
e 2
repo
rts th
at w
ith a
ll 23
con
sum
ptio
n ca
tego
ries,
the
test
iden
tifie
d 25
and
13
outly
ing
coun
tries
in sa
mpl
e of
115
and
76
coun
tries
, res
pect
ivel
y.
Inte
rest
ingl
y, a
lmos
t all
of th
ese
coun
tries
wer
e co
ncen
trate
d in
a su
bset
of p
oor c
ount
ries,
with
an
appr
oxim
ate
cuto
ff of
abo
ut 5
,000
inte
rnat
iona
l dol
lars
of t
otal
con
sum
ptio
n
expe
nditu
res p
er c
apita
(the
cut
off i
s sho
wn
by v
ertic
al li
nes i
n Fi
g. 3
and
4).
The
num
ber o
f out
lyin
g co
untri
es b
ecam
e ev
en la
rger
for a
subs
et o
f foo
d
expe
nditu
res (
31 a
nd 1
9, re
spec
tivel
y), a
nd d
ivis
ion
by th
e cu
toff
valu
e be
cam
e ev
en
mor
e m
arke
d. I
n ad
ditio
n, th
e nu
mbe
r of o
utly
ing
coun
tries
rem
aine
d ra
ther
larg
e fo
r tw
o
mor
e co
nsum
ptio
n ca
tego
ries (
‘bev
erag
es a
nd to
bacc
o’ a
nd ‘h
ouse
hold
’), b
ut fo
r the
rem
aini
ng su
bset
s of c
onsu
mpt
ion
cate
gorie
s, m
uch
few
er o
utly
ing
coun
tries
wer
e fo
und
8 One
con
sum
ptio
n ca
tego
ry (e
duca
tion)
had
to b
e ex
clud
ed to
avo
id th
e si
ngul
arity
pro
blem
in c
onsu
mpt
ion
shar
es.
レジュメ集３－13 ページ
16
(Tab
le 2
). M
oreo
ver,
thes
e ou
tlier
s wer
e di
strib
uted
muc
h m
ore
even
ly a
mon
g ric
h an
d
poor
cou
ntrie
s.
Bas
ed o
n th
ese
resu
lts, t
hree
con
clus
ions
can
be
mad
e. F
irst,
the
larg
e nu
mbe
r of
outly
ing
coun
tries
with
23
cons
umpt
ion
cate
gorie
s app
aren
tly w
as m
ainl
y du
e to
idio
sync
ratic
food
shar
es.
Seco
nd, t
wo
subs
ets o
f ric
h an
d po
or c
ount
ries c
an b
e
iden
tifie
d. T
hird
, the
subs
et o
f rel
ativ
ely
rich
coun
tries
had
muc
h fe
wer
num
ber o
f
regr
essi
on o
utlie
rs, i
ndic
atin
g th
at d
eman
d pa
ttern
in th
ese
coun
tries
may
inde
ed b
e
desc
ribed
by
IHP,
with
muc
h le
ss si
gnifi
cant
impa
ct fr
om d
eman
d di
ffere
nces
on
trade
.
To v
erify
this
con
ject
ure,
in a
dditi
on to
the
orig
inal
sam
ples
of 1
15 a
nd 7
6 co
untri
es, a
subs
et o
f 38
coun
tries
with
tota
l con
sum
ptio
n ex
ceed
ing
the
cuto
ff le
vel,
will
be
used
in
anal
yzin
g th
e re
latio
nshi
p be
twee
n do
mes
tic d
eman
d an
d in
tern
atio
nal t
rade
flow
s9 .
Tabl
e 3
repo
rts th
e si
gnifi
canc
e of
sign
test
, ran
k te
st a
nd S
pear
man
rank
corr
elat
ion
whe
n th
ey w
ere
appl
ied
to e
xam
ine
the
rela
tion
betw
een
,ˆ kn
c a
nd,
ˆ kn
x, a
nd
betw
een
,ˆ kn
can
d,
ˆ kn
m.
Con
side
r res
ults
for t
he c
ompl
ete
sam
ple
of 1
08 c
ount
ries i
n Ta
ble
3. I
n st
udyi
ng th
e re
latio
nshi
p be
twee
n ,ˆ kn
c a
nd
,ˆ k
nx
(den
oted
by
‘exp
orts
’), t
he m
edia
n
shar
e of
‘suc
cess
ful m
atch
es’ (
unde
r the
ass
umpt
ion
of n
egat
ive
rela
tion
betw
een
,ˆ kn
c a
nd
,ˆ k
nx
) was
0.4
67 a
nd 0
.471
for s
ign
test
and
rank
test
s, re
spec
tivel
y.Th
is in
dica
tes t
he
prev
alen
ce o
f ‘un
succ
essf
ul m
atch
es’,
base
d on
pos
tula
ted
nega
tive
rela
tion
betw
een
,ˆ kn
c
and
,ˆ k
nx
. M
oreo
ver,
the
med
ian
shar
e de
viat
ed si
gnifi
cant
ly fr
om th
e nu
ll le
vel o
f 0.5
for
the
rank
test
(p-v
alue
was
0.0
03),
and
bare
ly m
isse
d 0.
05 si
gnifi
canc
e le
vel f
or th
e si
gn
test
(with
p-v
alue
0.0
53).
On
the
othe
r han
d, fo
r im
ports
the
prev
alen
t out
com
e w
ere
9 Thi
s new
subs
et o
f rel
ativ
ely
rich
cou
ntrie
s is l
iste
d in
Tab
le A
1.
17
‘suc
cess
ful m
atch
es’ b
etw
een
,ˆ kn
c a
nd,
ˆ kn
m (m
edia
n sh
ares
for 1
08 c
ount
ries w
ere
0.60
0 an
d 0.
562
for s
ign
and
rank
test
s).
The
devi
atio
n of
thes
e m
edia
n sh
ares
from
the
null
valu
e of
0.5
was
hig
hly
sign
ifica
nt, w
ith p
-val
ues l
ess t
han
0.00
1 in
bot
h te
sts.
The
appl
icat
ion
of S
pear
man
rank
cor
rela
tions
reve
aled
the
sam
e pa
ttern
. Fo
r
corr
elat
ions
bet
wee
n,ˆ kn
c a
nd
,ˆ k
nx
, the
re w
ere
10 c
ount
ries w
ith si
gnifi
cant
, b
ut th
e
sign
of c
orre
latio
n w
as c
orre
ct in
onl
y 2
coun
tries
. Fo
r im
ports
, the
out
com
e w
as th
e
oppo
site
: out
of 1
1 co
untri
es, 1
0 ha
d ex
pect
ed p
ositi
ve si
gn.
With
‘hig
h qu
ality
’ sam
ple
of 7
2 co
untri
es, t
he g
ener
al p
atte
rn re
mai
ned
esse
ntia
lly th
e sa
me,
but
with
one
not
ewor
thy
chan
ge.
The
prev
alen
ce o
f ‘in
corr
ect
mat
ches
’ for
exp
orts
rem
aine
d, b
ut it
was
no
long
er si
gnifi
cant
at 0
.05
sign
ifica
nce
leve
l
(p-v
alue
s for
the
sign
and
rank
test
s wer
e 0.
096
and
0.09
3, re
spec
tivel
y).
The
mos
t dra
mat
ic c
hang
e oc
curr
ed w
hen
the
sam
ple
of c
ount
ries w
as re
stric
ted
to 3
8 ric
hest
cou
ntrie
s. B
oth
for e
xpor
ts a
nd fo
r im
ports
, the
sign
test
and
rank
test
wer
e
insi
gnifi
cant
, with
ver
y la
rge
p-va
lues
. Th
e nu
mbe
r of c
ount
ries w
ith si
gnifi
cant
rank
corr
elat
ions
dro
pped
to o
ne fo
r exp
orts
, and
non
e fo
r im
ports
. A
ppar
ently
, the
pre
viou
sly
iden
tifie
d si
mila
rity
of c
onsu
mpt
ion
shar
es fo
r thi
s sub
set o
f cou
ntrie
s gre
atly
redu
ced
the
scop
e fo
r a si
gnifi
cant
impa
ct b
y de
man
d on
inte
rnat
iona
l tra
de.
In su
m, w
hile
dev
iatio
ns
from
the
idea
lized
wor
ld o
f ide
ntic
al a
nd h
omot
hetic
pre
fere
nces
app
ear t
o be
rela
ted
with
trad
e flo
ws o
f cou
ntrie
s with
bro
ad sp
ectru
m o
f inc
ome
leve
ls, t
he im
pact
of d
eman
d
on tr
ade
is m
uch
less
vis
ible
for r
ich
coun
tries
, to
whi
ch th
e as
sum
ptio
n of
iden
tical
and
hom
othe
tic p
refe
renc
es m
ay b
e a
feas
ible
one
.
(to b
e co
ntin
ued)
レジュメ集３－14 ページ
18
Fig
3. T
est o
f mul
tivar
iate
out
liers
with
MC
D e
stim
ator
(sam
ple
of 1
15 c
ount
ries
)
19
Fig
4. T
est o
f mul
tivar
iate
out
liers
with
MC
D e
stim
ator
(sam
ple
of 7
6 co
untr
ies)
レジュメ集３－15 ページ
20
Tab
le 2
. T
he n
umbe
r of
out
lyin
g co
untr
ies i
n co
nsum
ptio
n sh
ares
Sa
mpl
e of
11
5 co
untri
es
Sa
mpl
e of
76
cou
ntrie
s
Tota
l Po
or
Ric
h
Tota
l Po
or
Ric
h
1. A
ll 23
con
sum
ptio
n sh
ares
25
23
2
13
11
2
Sub-
cate
gori
es:
2.
Foo
d
31
31
0
19
19
0 3.
Bev
erag
es &
toba
cco
13
12
1 12
9
3 4.
Clo
thin
g &
foot
wea
r 9
8 1
5 4
1 5.
Ren
t & fu
el
10
9 1
6
5 1
6. H
ouse
hold
19
14
5
8
6 2
7. T
rans
porta
tion
7 5
2 7
5 2
8. E
duca
tion
& re
crea
tion
5 3
2
3 1
2 N
ote:
(a
) mul
tivar
iate
out
liers
wer
e te
sted
by
the
MC
D e
stim
ator
(4);
(b) T
he d
ivis
ion
into
rich
and
poo
r cou
ntrie
s is d
one
by th
e cu
toff
val
ue o
f 5,0
00 in
tern
atio
nal
dolla
rs o
f tot
al c
onsu
mpt
ion
expe
nditu
res.
Tab
le 3
. T
he si
gnifi
canc
e of
sign
test
, ran
k te
sts,
and
Spea
rman
ran
k co
rrel
atio
n.
The
shar
e of
‘cor
rect
mat
ches
’ in:
The
num
ber o
f si
gnifi
cant
rank
(a
) sig
n te
st
(b) r
ank
test
co
rrel
atio
ns:
Con
sum
ptio
n sh
ares
in:
Sam
ple
med
ian
p-va
lue
Sa
mpl
e m
edia
n p-
valu
e
Cor
rect
In
corr
ect
(a) s
ampl
e of
108
cou
ntri
es
Ex
ports
0.
467
0.05
3
0.47
1 0.
003
2
8
Impo
rts
0.60
0 <0
.001
0.56
2 <0
.001
10
1 (b
) sam
ple
of 7
2 co
untr
ies
Ex
ports
0.
467
0.09
6
0.47
6 0.
093
2
5
Impo
rts
0.60
0 <0
.001
0.59
0 <0
.001
7 1
(c) s
ampl
e of
38
coun
trie
s
Expo
rts
0.53
3 0.
709
0.
452
0.27
3
1 1
Im
ports
0.
533
0.58
8
0.50
0 0.
942
0
0
21
App
endi
x ta
ble
A1.
Lis
t of f
ull s
ampl
e of
115
cou
ntri
es
Alb
ania
Gre
nada
a, b
Phili
ppin
es
Ant
igua
& B
arbu
daG
uine
aPo
land
a
Arg
entin
aaH
ong
Kon
ga,b
Portu
gala,
b
Arm
enia
Hun
gary
cQ
atar
c
Aus
tralia
a,b
Icel
anda,
bR
oman
iaa
Aus
triaa,
bIn
done
siaa
Rus
siaa
Aze
rbai
jana
Iran
aSe
nega
la
Bah
amas
cIr
elan
da,b
Sier
ra L
eone
B
ahra
inIs
rael
a,b
Sing
apor
ea,b
Ban
glad
esh
Italy
a,b
Slov
akia
a,b
Bar
bado
scJa
mai
caSl
oven
iac
Bel
arus
Japa
na,b
Spai
na,b
Bel
gium
a,b
Jord
ana
Sri L
anka
a
Bel
izea
Kaz
akhs
tana
St K
itts N
evis
a
Ben
ina
Ken
yaa
St. L
ucia
a
Ber
mud
acK
orea
Rep
ublic
a,b
St.V
ince
nt &
Gre
nadi
nesa
Bol
ivia
aK
yrgy
zsta
naSw
azila
ndB
otsw
ana
Latv
iaa
Swed
ena,
b
Bra
zila
Leba
nona
Switz
erla
nda,
b
Bul
garia
Lith
uani
aaSy
rian
Ara
b R
epub
lic
Cam
eroo
nLu
xem
burg
a,b
Tajik
ista
nC
anad
aa,b
Mac
edon
iaTa
nzan
iaC
hile
aM
adag
asca
raTh
aila
ndC
ongo
Mal
awia
Trin
idad
Tob
agoa
Cot
e D
'Ivoi
reM
alia
Tuni
sia
Cro
atia
Mau
ritiu
sa,b
Turk
eya
Cze
ch R
epub
lica,
bM
exic
oaTu
rkm
enis
tan
Den
mar
ka,b
Mol
dova
aU
krai
nea
Dom
inic
aaM
ongo
liaU
nite
d K
ingd
oma,
b
Ecua
dor
Mor
occo
Uru
guay
a
Egyp
taN
epal
aU
SAa,
b
Esto
niaa
Net
herla
ndsa,
bU
zbek
ista
nFi
jiaN
ew Z
eala
nda,
bV
enez
uela
Finl
anda,
bN
iger
iaa
Vie
tnam
a
Fran
cea,
bN
orw
aya,
bY
emen
G
abon
Om
anc
Zam
biaa
Geo
rgia
aPa
kist
ana
Zim
babw
e G
erm
anya,
bPa
nam
aa
Gre
ecea,
bPe
rua C
ount
ry w
as in
the
‘hig
h qu
ality
’ sub
set o
f 76
coun
tries
b C
ount
ry w
as in
the
‘hig
h in
com
e’ su
bset
of 3
8 co
untri
es
レジュメ集３－16 ページ
22
App
endi
x ta
ble
A2.
Con
sum
ptio
n ca
tego
ries
. I.
FOO
D
III.
CLO
THIN
G&
FO
OTW
ARE
VI.
18.M
EDIC
AL A
ND
HEA
LTH
SER
VIC
ES1.
Bre
ad a
nd c
erea
ls
11. C
loth
ing
VII
.TRA
NSP
ORT
ATIO
N2.
Mea
t 12
. Foo
twea
r 19
. Per
sona
l tra
nspo
rtatio
n eq
uipm
ent
3. F
ish
IV.R
ENT
& F
UEL
20
. Ope
ratio
n of
tran
spor
tatio
n eq
uip.
4.
Milk
, che
ese
and
eggs
13
. Gro
ss re
nt
21. P
urch
ased
tran
spor
t ser
vice
s 5.
Oils
and
fats
14
. Fue
l and
pow
er
22. C
omm
unic
atio
n 6.
Fru
its, v
eget
able
s, po
tato
es
V.H
OU
SEH
OLD
V
III.
REC
REAT
ION
& E
DU
CAT
ION
7. O
ther
food
15
. Fur
nitu
re, f
loor
cov
erin
gs23
. Rec
reat
ion
and
cultu
ral s
ervi
ces
II.B
EVER
AGES
& T
OBA
CC
O
16. H
ouse
hold
text
iles
24. E
duca
tion
8. N
on-a
lcoh
olic
bev
erag
es
17. H
ouse
hold
app
lianc
es
9.
Alc
ohol
ic b
ever
ages
10
. Tob
acco
App
endi
x ta
ble
A3.
Mis
sing
dat
a in
Sta
tsC
anad
a 1.
Sam
ple
of 1
08 c
ount
ries
Ant
igua
& B
arbu
daB
otsw
ana
Dom
inic
a G
rena
daSt
. Luc
ia
St.V
ince
nt &
Gre
nadi
nes
Swaz
iland
2. S
ampl
e of
72
coun
trie
s D
omin
ica
Gre
nada
St. L
ucia
St
.Vin
cent
& G
rena
dine
s
App
endi
x ta
ble
A4.
Con
cord
ance
bet
wee
n IC
P co
nsum
ptio
n ca
tego
ries
and
SIT
C
(rev
. 2) t
rade
cla
ssifi
catio
n.
IC
P ca
tego
ry
SITC
trad
e cl
assi
ficat
ion
1.B
read
and
cer
eals
042.
Mea
t01
3.Fi
sh03
4.M
ilk, c
hees
e an
d eg
gs02
, exc
l. 02
35.
Oils
and
fats
023+
091+
41+4
26.
Frui
t, ve
geta
bles
and
pot
atoe
s05
, exc
l. 05
837.
Oth
er fo
od05
83+0
6+07
+09
(exc
l. 09
1)8.
Non
-alc
ohol
ic b
ever
ages
111
9.A
lcoh
olic
bev
erag
es11
210
.To
bacc
o12
11.
Clo
thin
g26
+84
12.
Foot
wea
r85
13.
Furn
iture
, flo
or c
over
ings
659+
81+8
214
.H
ouse
hold
text
iles
658
15.
Hou
seho
ld a
pplia
nces
775
23
Ref
eren
ces
Bal
lanc
e, R
ober
t, Fo
rstn
er, H
elm
ut; M
urra
y, T
racy
, 198
5. “
On
Mea
surin
g C
ompa
rativ
e
Adv
anta
ge: A
Not
e on
Bow
en´s
Indi
ces”
Wel
twir
tsch
aftli
ches
Arc
hiv,
121,
346
-
350.
Bes
t, D
.J. a
nd D
.E. R
ober
ts (1
975)
“A
lgor
ithm
AS
89: T
he U
pper
Tai
l Pro
babi
litie
s of
Spea
rman
's rh
o”.
Appl
ied
Stat
istic
s,24
, 377
–379
.
Bow
en, H
arry
P.,
Leam
er, E
dwar
d E,
and
Sve
ikau
skas
, Leo
(198
7).
“Mul
ticou
ntry
,
Mul
tifac
tor T
ests
of t
he F
acto
r Abu
ndan
ce T
heor
y,”
Amer
ican
Eco
nom
ic R
evie
w,
77 (5
), 79
1-80
9.
Chu
ng, C
hul.
2000
. “Fa
ctor
Con
tent
of T
rade
: Non
hom
othe
tic P
refe
renc
es a
nd ‘M
issi
ng
Trad
e’”,
mim
eo, U
nive
rsity
of M
ichi
gan.
Dav
is, D
onal
d R
. and
Wei
nste
in, D
avid
E. (
2003
) "M
arke
t Acc
ess,
Econ
omic
Geo
grap
hy
and
Com
para
tive
Adv
anta
ge: A
n Em
piric
al T
est"
. Jou
rnal
of I
nter
natio
nal
Econ
omic
s, vo
l. 59
, pp.
1-2
3.
Dav
is, D
onal
d R
., an
d D
avid
E. W
eins
tein
(200
1a).
“An
Acc
ount
of G
loba
l Fac
tor
Trad
e”, A
mer
ican
Eco
nom
ic R
evie
w, D
ecem
ber,
91(5
), pp
. 142
3-14
53.
Dav
is, D
onal
d R
., an
d D
avid
E. W
eins
tein
(200
1b).
“The
Fac
tor C
onte
nt o
f Tra
de”,
NB
ER w
orki
ng p
aper
863
7. F
orth
com
ing
as a
cha
pter
in th
e H
andb
ook
of
Inte
rnat
iona
l Tra
de, e
dite
d by
E. K
wan
Cho
i and
Jam
es H
arrig
an.
Dav
is, D
onal
d R
., D
avid
E. W
eins
tein
, Sco
tt B
radf
ord,
and
Kaz
ushi
ge S
him
po. 1
997.
“Usi
ng In
tern
atio
nal a
nd Ja
pane
se R
egio
nal D
ata
to D
eter
min
e W
hen
the
Fact
or
Abu
ndan
ce T
heor
y of
Tra
de W
orks
”, A
mer
ican
Eco
nom
ic R
evie
w, 9
7: 4
21-4
6.
レジュメ集３－17 ページ
24
Dow
rick,
Ste
ve, a
nd Q
uigg
in, J
ohn,
199
4. “
Inte
rnat
iona
l Com
paris
ons o
f Liv
ing
Stan
dard
s and
Tas
tes:
A R
evea
led
Pref
eren
ce A
naly
sis”
, Am
eric
an E
cono
mic
Revi
ew, 8
4(1)
, 332
-341
.
Han
son,
Gor
don
and
Xia
ng, C
hong
(20
04).
“The
Hom
e-M
arke
t Effe
ct a
nd B
ilate
ral
Trad
e Pa
ttern
s,” A
mer
ican
Eco
nom
ic R
evie
w, 9
4 (4
), 11
08-1
129.
Hes
ton
Ala
n, S
umm
ers,
Rob
ert a
nd A
ten,
Bet
tina
(200
4), “
Dat
a App
endi
x fo
r a S
pace
-
Tim
e Sy
stem
of N
atio
nal A
ccou
nts:
Pen
n W
orld
Tab
le 6
.1 (P
WT
6.1)
”, C
ente
r for
Inte
rnat
iona
l Com
paris
ons a
t the
Uni
vers
ity o
f Pen
nsyl
vani
a (C
ICU
P),
http
://pw
t.eco
n.up
enn.
edu/
Doc
umen
tatio
n/ap
pend
61.p
df.
Hor
iba,
Yut
aka,
197
9. “
Test
ing
the
Dem
and
Side
of C
ompa
rativ
e Adv
anta
ge M
odel
s,"
Amer
ican
Eco
nom
ic R
evie
w, 6
9, 6
50-6
61.
Hun
ter,
Lind
a an
d Ja
mes
Mar
kuse
n, 1
988.
“Pe
r Cap
ita In
com
e as
a B
asis
for T
rade
", in
Rob
ert F
eens
tra (e
d), E
mpi
rica
l Met
hods
for I
nter
natio
nal T
rade
, MIT
Pre
ss,
Cam
brid
ge.
Hun
ter,
Lind
a. 1
991.
“Th
e C
ontri
butio
n of
Non
hom
othe
tic P
refe
renc
es to
Tra
de”,
Jour
nal o
f Int
erna
tiona
l Eco
nom
ics,
30, p
p.34
5-35
8.
Kra
vis,
I. B
., Z.
Ken
esse
y, A
. Hes
ton
and
R. S
umm
ers (
1975
), A
Syst
em o
f Int
erna
tiona
l
Com
pari
sons
of G
ross
Pro
duct
and
Pur
chas
ing
Pow
er. B
altim
ore:
Joh
ns
Hop
kins
Uni
vers
ity P
ress
.
Leam
er, E
dwar
d E.
200
2. “
Link
ing
the
Theo
ry w
ith th
e D
ata:
Tha
t's th
e C
ore
Prob
lem
of
Inte
rnat
iona
l Eco
nom
ics.”
For
thco
min
g in
Jam
es H
eckm
an a
nd E
dwar
d Le
amer
,
eds.
Han
dboo
k of
Eco
nom
etri
cs, V
ol. 6
.
25
Mar
kuse
n, J
ames
R. (
1986
). "E
xpla
inin
g th
e V
olum
e of
Tra
de: a
n Ec
lect
ic A
ppro
ach,
"
Amer
ican
Eco
nom
ics R
evie
w, 7
6 (D
ecem
ber)
, 100
2-10
11.
Mat
suya
ma,
Kim
inor
i (2
000)
“A
Ric
ardi
an M
odel
with
a C
ontin
uum
of
Goo
ds u
nder
Non
hom
othe
tic P
refe
renc
es:
Dem
and
Com
plem
enta
ritie
s, In
com
e D
istri
butio
n
and
Nor
th-S
outh
Tra
de,"
Jou
rnal
of
Polit
ical
Eco
nom
y, 1
08 (
Dec
embe
r 20
00):
1093
-112
0.
Pena
D.,
and
Prie
to F
.J. (2
001)
, "M
ultiv
aria
te O
utlie
r Det
ectio
n an
d R
obus
t Cov
aria
nce
Mat
rix E
stim
atio
n" (w
ith d
iscu
ssio
n), T
echn
omet
rics
, 43
(3),
286-
300.
Rei
mer
, Jef
frey
J. 2
002.
“In
term
edia
te G
oods
and
the
Fact
or C
onte
nt o
f Tra
de,”
mim
eo,
Purd
ue U
nive
rsity
.
Rou
ssee
uw, P
.J. (1
984)
, "Le
ast M
edia
n of
Squ
ares
Reg
ress
ion,
" Jou
rnal
of t
he A
mer
ican
Stat
istic
al A
ssoc
iatio
n, V
ol. 7
9, p
p. 8
71-8
81.
Rou
ssee
uw, P
.J. a
nd V
an D
riess
en, K
. (19
99),
"A F
ast A
lgor
ithm
for t
he M
inim
um
Cov
aria
nce
Det
erm
inan
t Est
imat
or",
Tec
hnom
etri
cs, 4
1 (3
), 21
2-22
3.
Stat
istic
s Can
ada,
200
2.W
orld
Tra
de A
naly
zer,
2002
CD
-RO
M.
Otta
wa:
Sta
tistic
s
Can
ada.
Tref
ler,
Dan
iel,
1995
. "Th
e C
ase
of th
e M
issi
ng T
rade
and
Oth
er M
yste
ries"
, Am
eric
an
Econ
omic
Rev
iew,
v. 8
5, 1
029-
1046
.
Wor
ld B
ank,
200
2. W
orld
Dev
elop
men
t Ind
icat
ors 2
001,
CD
-RO
M. W
ashi
ngto
n, D
.C.:
Wor
ld B
ank.
レジュメ集３－18 ページ
International Labor Standards and Their Harmonization
By
Kenzo Abe Laixun ZhaoOsaka University Kobe University
Introduction
What is labor standards (LS)?ILO: freedom of communication, unionization, nondiscrimination, no child labor, no forced laborPopular Claim: Weak LS = artificially low wages, augmenting the competitiveness of poor countries. Firms gain a competitive edge by applying low LS;WTO should set LS to protect the interest of workers: labor unions, humanitarian groups and NGOs urge for LS to be included in WTO negotiations.Some advocate for trade sanctions against countries that do not enforce a set of agreed LS~~”Social Clause”
p. 1
レジュメ集３－19 ページ
Introduction continued
By the claims of different groups, we can infer that workers benefit from LS improvements.
However, this has not been modeled in the literature.
This paper
LS= Productive work safety, work environment: lighting, ventilation, temperature, overtime work,
and so on;
LS not only increases costs, but also productivity
LiteratureSrinivasan (1995); Brown Deardorff andStern (1996): GE models;Martin and Maskus (2001); Bagwell and Staiger (2001)Rodrik (1996); Elliot (2000)
Bhagwati (1995) and Basu (1999)LS enters consumer utility or national welfare directly. LS raises utility instead of productivity or workers’ happiness.
p. 2
レジュメ集３－20 ページ
Present paper
Incorporate productive LS in oligopoly.LS raises productivity so that firms & workers benefit, instead of increasing consumer utility, or raising cost only;Countries adopt different LS due to differences in technology, wages, market structure;Imposing trade restrictions to force the foreign country to raise LS does not work—2 parts of the tariff: “stick” & “carrot”;A policy game of LS between governments: North chooses an even higher or lower LS than under free choices depending on N’s valuation of Southern LS;
Present paper continued
The commonly abided by LS that maximizes world welfare is not binding;A minimum LS, binding for both countries exists, but it lowers Northern profits and welfare;First best policy calls for different LS to be applied in the two countries;In terms of both efficiency and equity, technology transfer is a better alternative, which does not require any intervention from governments or international organizations.
p. 3
レジュメ集３－21 ページ
Basic Model Setup
2 firms in 2 countries: North & South.Identical products, sold in North only, with a tariff , against imports from SouthFirm i producesYi, using labor Li only.
Unit cost of LS:
Implications: costly to obtain LS; LS contributes to production
' "( ) / , 0, 0i i i i i if Y a f fL
ic
Nt
Profits:
Northern Welfare (importer)
Southern Welfare
Tariff:
“Reservation tariff” & “valuation tariff”
( , , , )i S N S N i i i i i j iY Y pY w L c t Y
( ) ( )N N N S N S N Su Y Y Y Y p t Y
S S
0 1N St
p. 4
レジュメ集３－22 ページ
Timing~~3 Stages:1st stage: Northern government imposes tariff;
2nd stage: LS is chosen, 3 regimes~~Free choice of LS, policy game, harmonization
--LS chosen freely, or by governments, or harmonized;
3rd stage: output competition—Cournot
Game solved backwards
Equilibrium of free LS
Proposition 1: Suppose that the initial equilibrium is completely symmetric. An increase in the wage-technology ratio of the North raises (reduces) the Northern LS relative to the Southern one if
( , ; ) 0N N SA
( , ; ) 0S N SAFigure 1
'( ) 16 ( 1)9
N S i i i
N
d B r f fdr D p
i i
i i
p r fr f
( ) 1
p. 5
レジュメ集３－23 ページ
F
H
G
Figure 1
O
A
B
T
K
I
45o
GNR
NR
0SR
SR
SR
GNR
S
N
Proposition 2: An increase in the Northern reservation tariff raises (reduces) the Northern (Southern) LS.
Proposition 3: An increase in the Northern government’s valuation of Southern LS raises (reduces) the Southern (Northern) LS.
p. 6
レジュメ集３－24 ページ
Equilibrium of a Policy Game
( , ; ) 0N N S N
N N N N
d dRd d
( , ; ) 0S N S S
S S
'1 ( ) 0;3
< 0
N S N NN
N
d Y Y r fd
dRd
Proposition 4: In the absence of the import tariff, the equilibrium of a policy game between the two governments involves in a higher (lower) LS in the North (South) than under free choices of LS.
Proposition 5: Under a sufficiently high tariff, , the policy game results in higher LS for the South and a lower one for the North than under free choices of LS.
1Nt P
レジュメ集３－25 ページ
p. 7
International Harmonization
World-welfare maximization: ILO chooses a single, internationally abided by LS,
Max
Equilibrium:
w S N
( , ) ( , ) ( , )w N w S wW
( , ; ) ( , ; ) 0N w w S w wA A
Lemma 1: Given , the world welfare maximizing LS lies on the segment between points H and J in Figure 2(a), which is not binding for the North.
Proposition 6: Under the uniform, minimum LS that is binding for both countries, world welfare is not maximized.
See Figure 2
w S N
p. 8
レジュメ集３－26 ページ
I
RN
RS
F
H
W
L
K
(a)
(b)
J
S
N
FN
HN
IN
HS
FS
FS
IS
HS
Iw
Hw
w
Fig. 2
Different Labor Standards
ILO chooses 2 different LS for North and South respectively,
Proposition 7: (i). In the South, the first best LS is lower than the one in the regime of free choices of LS; (ii). In the North, it lies in the area betweencurves RN and , but to the left of RS. Thus, the first best combination of LS must lie in the area ABFG inFigure 1.
( , ; ) ( , ; ) ( , ; )N S N N S S N SW
p. 9
レジュメ集３－27 ページ
Technology Transfer~~to be revisedCostly tech. transfer: unit cost e
Proposition 8: Under technology transfer from the North to the South, the Southern firm voluntarily raises its LS, and the Northern profits and welfare are higher.See Figure 1
(1 )S S N
Efficiency, Equity and Human Rights ^^ in progress
Thank You!
p. 10
レジュメ集３－28 ページ
Related Documents
