Home Bias for Intermediate and Finished Goods: Evidence ...trade/apts/2006/pdf/Hayakawa.pdf · Home...
-
Upload
nguyentuyen -
Category
Documents
-
view
223 -
download
2
Transcript of Home Bias for Intermediate and Finished Goods: Evidence ...trade/apts/2006/pdf/Hayakawa.pdf · Home...
Home Bias for Intermediate and FinishedGoods: Evidence from East Asia
Kazunobu Hayakawa ∗†
Graduate School of Economics, Keio University, Japan
Abstract
In East Asia, international production/distribution networks havedeveloped and have explosively increased intra-regional trade particu-larly in intermediate goods. This paper examines whether home biasfor intermediate goods gets smaller than that for finished goods in the1990s. In regression results, we find that, in East Asia, while the homebias for intermediate goods was at the same level as that for finishedgoods in the former half of the 1990s, the home bias for intermediategoods has dramatically decreased since the latter half of the 1990s,being lower than that for finished goods in 2000.
Keywords: home bias; fragmentation; agglomeration; East AsiaJEL Classification: D23; F15; R12; R15
∗Correspondence address. Graduate School of Economics, Keio University, Mita 2-15-45, Minato-ku, Tokyo 108-8345, Japan. E-mail: [email protected]
†This paper is to be presented at the Asia Pacific Trade Seminars held in Kobe Uni-versity, Japan on July 15-16, 2006. I would like to thank Fukunari Kimura, Tomoya Mori,and Ryuhei Wakasugi for their helpful comments and suggestions. Any errors are mine.
1
1 Introduction
Do trade costs in intermediate goods get smaller than those in finished goods?
In East Asia, the development of international production/distribution net-
works in the 1990s has explosively increased intra-regional trade particularly
in intermediate machinery goods. The intra-regional trade in machinery
parts and components has increased much more rapidly than the trade in
finished machinery goods in East Asia (Kimura et al., 2006). On the con-
trary to the case of trade volume between intermediate and finished goods,
differences in trade costs between them have remained unknown.
In empirical analysis, trade costs have been measured as home bias or
border effect. Home bias, of which Wei (1996) is a pioneer work, is defined
as a country’s import from itself in excess of its import from other countries
after taking into account some elements affecting international transactions.
Following Wei (1996), many papers measure the home bias among countries
(e.g., Helliwell, 1997; Head and Mayer, 2000; Wolf, 2000; Poncet, 2003; Head
and Zignago, 2005). The estimation of home bias comes to be taken as a
useful tool in measuring trade costs.
However, to our knowledge there is no study that carefully distinguish
home bias between intermediate and finished goods transactions despite the
fact that there are quantitatively and qualitatively different costs between
them1. Tariff rates, for example, are substantially much lower in interme-
1Hayakawa (2006) measures the home bias for intermediate goods among East Asiancountries from 1985 to 1995 with using Asian International Input-Output Table. He
2
diate goods than in finished goods. Due to import-substitution policy in
developing countries, tariff rates are truly much lower in intermediate goods
than in finished goods. Furthermore, trade liberalization for semi-conductor-
related parts and components in the 1990s as well as duty drawback system
remarkably decreases the payment of tariffs by intermediate producers.
Compared with trade costs in finished goods there are potentially the huge
costs in intermediate goods. The geographical separation between finished
and intermediate goods producers or between intermediate goods producers
imposes larger costs in searching potential business partners, consulting de-
tailed specs of parts, controlling product quality and delivery timing, solving
disputes over contracts, and monitoring business partners. Although the sep-
aration between finished goods producers and consumers also incurs a part
of those costs, this must simultaneously lead to a further increase of the costs
between finished and intermediate goods producers. As a result, in the na-
ture of the relationship among finished goods producers, intermediate goods
producers and consumers, non-policy barriers in intermediate goods may be
always higher than those in finished goods.2
The purpose of this paper is to examine whether home bias for intermedi-
ate goods gets smaller than that for finished goods in East Asia. To this end,
finds that the home bias in each East Asian country has remarkably declined since 1985.However, the paper does not focus on the difference in home bias between them andprimarily uses different specifications with the ones used in this paper.
2Hummels (2001) pointed out that “when countries specialize in stages of productionand trade intermediate goods, the inventory-holding and depreciation costs for early-stagevalue added accrue throughout the duration of the production chain.”
3
this paper estimates the equation in which the dependent variable is a ratio
of inter-regional to intra-regional import values by goods in order to avoid
some cumbersome issues. First, appropriate data on price index in each good
are definitely unavailable. Second, total expenditure on intermediate goods
become complicated if we allow firms in intermediate goods sector to input
intermediate goods themselves. Then, data availability on the total expen-
diture turns out to be problematic. To avoid these problems, we employ the
method of “log odds ratios” used in Head and Mayer (2000), whose formula-
tion enables us to cancel out the variables relating to the total expenditure
and the price index.
Our findings are summarized as follows. First, in East Asia, while the
home bias for intermediate goods was at the same level as that for finished
goods in the former half of the 1990s, the home bias for intermediate goods
has dramatically decreased since the latter half of the 1990s, being lower
than that for finished goods in 2000. Second, in 2000, the home bias for
intermediate goods in almost all East Asian countries becomes as low as or
becomes lower than that in Singapore. In particular, it is surprising that
the home bias in China is as low as in Singapore and that the home bias in
Malaysia and the Philippines is lower than in Singapore.
The rest of this paper is organized as follows. In section 2, we specify our
estimation equations. Section 3 reports data and econometric issues. The
regression results are reported in section 4. Section 5 concludes.
4
2 Empirical specification
This section derives regression equations for transactions of both finished and
intermediate goods. It always comes to a discussion what kinds of models we
should choose in specifying empirical equations. While it becomes common
to employ agglomeration or Dixit-Stiglitz monopolistic competition model
in analyzing industrial distribution and international trade in Europe, it
remains unsettled in East Asia.
As pointed out in Ando (2006), traditional comparative advantage theory
can no longer explain the observed trade pattern in East Asia: not only de-
veloped but developing countries intensively engage in back-and-forth trans-
actions of machinery parts and components. Both horizontal intra-industry
trade model, e.g., Helpman and Krugman (1985), and vertical intra-industry
trade model, e.g., Flam and Helpman (1987), is not a match for it. Rather,
it comes to be considered fragmentation theory, e.g,. Jones and Kierzkowski
(1990), is a better fit (see, for example, Ando and Kimura, 2005; Kimura et
al., 2006).
In this paper, we use two familiar models rather than the fragmentation
model in order to make our estimation/measurement easier and to shed light
on some distinctive features in each good in East Asia. In intermediate
goods sector, Krugman-Venables model (Krugman and Venables, 1995) is
employed. Also in East Asia, a substantial amount of fixed costs seems to
be required to produce intermediate goods, especially high-tech parts and
5
components like liquid crystal display. In addition, electric machinery parts
are well differentiated and specialized, like application specific integrated
circuit. To incorporate these features into empirical specifications, we adopt
monopolistic competition model into intermediate goods sector.
As the result of these features, the strong effect of not consumers’ demand
linkages but cost linkages among intermediate goods can be observed in East
Asia, resulting in the development of agglomeration. Samut Prakan and the
Eastern Seaboard in Thailand, Penang and Shah Alam in Malaysia, and
others are the examples (Kimura, 2006). Therefore, we employ Krugman-
Venables vertical linkages model in intermediate goods sector, which has
indeed horizontal linkages rather than vertical linkages for the sake of sim-
plicity.
On the other hand, in final assembly process, inexpensive labors would
be a more important factor than the capability to incur fixed costs. Indeed,
in East Asia, finished goods sector distributes across countries according to
comparative advantages. For example, China has attracted a large number
of factories to produce finished goods due to its abundant labors. To place
emphasis on the role of comparative advantages and to enable our model to
well describe the observed two-way transactions on finished goods, this paper
employs Armington model in finished goods sector.
The representative consumer in each region is assumed to have a two-
tier utility function, which becomes standard in international trade and new
economic geography literature. The upper tier is a Cobb-Douglas function
6
of the utility derived from consumption of finished goods. Specifically, we
apply the following utility function:
Ur =l∏
i=1
(Ci
r
)αi
,l∑
k=1
αk = 1,
where Cir is the aggregate consumption of finished good i.
We formalize expenditure allocation across finished machinery goods con-
sisting of multiple products and omit the subscript representing the name of
finished goods for now. The consumer has the following preference specified
as constant elasticity of substitution (CES) function over the products:
Cr =
[R∑
i=1
x(σF−1)/σF
r,i
]σF /(σF−1)
, σF > 1
where R, xr,i, and σF are the number of countries, the demand of country
r for the machinery product produced in country i, and the elasticity of
substitution between finished products, respectively.
In this paper, we do not introduce a preference parameter of consumers in
country i for products/varieties produced in country j, which is denoted by
aij in the previous studies. Although cultural ties such as the use of common
language are often taken as the elements affecting the preference, those ties
are also the components of trade costs, e.g., communication costs. In this
paper, therefore, those elements are put together into trade costs functions,
but the formulation of regression equations do not depend on this setting
and becomes the same one as in the previous studies.
Here, we assume identical technology across countries but assume that
the endowment of productive factors is different across countries. Then, the
7
import demand in country r for each finished product produced in country
j, xr,j, is given by
xr,j = τ 1−σFr,j p−σF
j P σF−1r EF
r , (1)
where pj and Pr denote the price of each variety produced in country j and
the price index in country r, respectively. EFr is total expenditure on finished
goods in country r. Transactions in finished goods between country r and s
is modeled as facing Samuelsonian iceberg costs, τr,s.
Finished goods are produced with a constant-returns to scale technol-
ogy. The producer of each country combines a composite index aggregated
across varieties of intermediate inputs and primary productive factors, e.g.,
labor and physical capital, with Cobb-Douglas manner. The composite en-
ters the production function for each producer through a CES aggregator.
Specifically, we have the following production function3:
xr =(LF
r
)1−µ (DI
r
)µ,
DIr =
[R∑
i=1
∫ mi
0zr,i(j)
(σI−1)/σIdj
]σI/(σI−1)
, σI > 1,
where µ, LFr , and DI
r are a linkage parameter between finished and inter-
mediate goods, a Cobb-Douglas aggregator of primary factors employed by
each country to produce finished output xr, and a quantity index aggregated
across varieties of intermediate goods. mi, zr,i(j), and σI are the number of
3Formally, we should include a constant term, e.g., µ−µ(1 − µ)µ−1, but this plays norole in the analysis as long as we assume that the term is identical across countries.
8
varieties produced in country i, the demand of country r for j-th variety pro-
duced in country i, and the elasticity of substitution between intermediate
goods, respectively.
We again assume identical technology across firms of intermediate va-
rieties and across countries. Then, import demand in country r for each
intermediate variety produced in country j, zr,j, is given by
zr,j = t1−σIr,j q−σI
j ΠσI−1r EI
r , (2)
where qj and Πr denote the price of each variety produced in country j, and
the price index in country r, respectively. EIr is total expenditure on inter-
mediate inputs in country r. Transactions on intermediate goods between
country r and s is modeled as facing Samuelsonian iceberg costs, tr,s.
The market structure in intermediate goods sector is assumed to be
Chamberlinian monopolistic competition. The producer of each intermediate
variety inputs not only primary productive factors but a bundle of intermedi-
ate varieties. Specifically, we suppose the following production function (see
footnote 3):
fI + cIzr =(LI
r
)1−ν (DI
r
)ν,
where fI , cI , ν, and LIr are a fixed cost of setting up a plant, a variable cost,
the horizontal linkages among intermediate varieties, and a Cobb-Douglas
aggregator of primary factors employed by each firm of intermediate varieties
to produce output, zr, respectively. Here it is assumed that the elasticity of
substitution among the varieties is the same for finished goods producers
9
as it for intermediate goods producers. Then, denoting the total value of
production of finished and intermediate goods in country r as n̄r and m̄r,
respectively,
EIr = µn̄r + νm̄j.
We here need to sidestep two difficult data issues. First, appropriate data
on price index in each sector are definitely unavailable. In the recent two
decades, data availability of price indices always annoyed empirical analysts.
In this paper, to make things worse, we need to obtain price index data for
intermediate and finished goods sector separately. Second, when firms in
the intermediate sector use each other’s output as intermediate input, the
regression equation for intermediate goods sector becomes a non-(log-)linear
equation as in Hillberry and Hummels (2002), which is also quite different
from the equation for finished goods sector. To avoid these difficulties, we
employ the method of “log odds ratios” used in Head and Mayer (2000).
They estimate an equation in which the dependent variable is a log ratio of
inter-national to intra-national input values, and this formulation enables us
to cancel out the variables relating to the total expenditure and the price
index.
From equations (1) and (2), we obtain a ratio of total input values in
country r for the goods produced in country j to the values for the goods
produced domestically as
Xr,j ≡ pjxr,j
prxr,r
=
(τr,j
τr,r
)1−σF(
pj
pr
)1−σF
, (3)
10
Zr,j ≡ qjmjzr,j
qrmrzr,r
=(
mj
mr
) (tr,jtr,r
)1−σI(
qj
qr
)1−σI
. (4)
This formulation relates the decisions of consumers and finished variety pro-
ducers in country r on how to allocate expenditure between home-made and
foreign-made finished goods and between home-made and foreign-made in-
termediate goods, respectively.
We again encounter the issue of data availability. As well as price index
above, the data on the number of firms in finished goods sector and in in-
termediate goods sector separately are unavailable. As in Head and Mayer
(2000), therefore, rather than using other variables as a proxy for the num-
bers, we eliminate the firm number variables from the regression equations
by using the following relationship.
Remember that we assume identical technology across firms and coun-
tries. Denoting the quantity produced by each firm in intermediate goods
sector as z̄, we obtain m̄r = z̄qrmr. Substituting this relationship into equa-
tion (4), the equation is re-written as
Zr,j =(
m̄j
m̄r
) (tr,jtr,r
)1−σI(
qj
qr
)−σI
. (5)
Furthermore, in order to avoid simultaneity problem between zr,j and m̄j,
as in Head and Mayer (2000), we move m̄j to the LHS of the equation.
(Zr,j
Mr,j
)=
(tr,jtr,r
)1−σI(
qj
qr
)−σI
, (6)
where Mr,j ≡ m̄j/m̄r.
11
We assume that trade costs mainly consist of transport costs incurred
by geographical distance and of home bias. In this paper, as in Wei (1996),
the home bias is defined as a country’s import from itself in excess of its
import from other countries after taking into account some elements affecting
international transactions. We specify relative trade costs as
ln
(τr,j
τr,r
)= home biasF
r + ϕF ln Distancer,j, (7)
ln
(tr,jtr,r
)= home biasI
r + ϕI ln Distancer,j, (8)
where ln Distancer,j ≡ ln dr,j − ln dr,r. dr,j is geographical distance between
country r and j and is measured by greater circle between their respective
capital cities. dr,r is intra-regional distance, of which the most appropriate
definition remains unsettled in the literature, and is calculated as a radius
of surface area4 in country r. Specifically, dr,r ≡√
surface arear/π. As in
the previous papers, we also augment the basic specification in equations (7)
and (8) by including other variables such as common language dummy.
We capture home bias in each country by examining coefficients for im-
porter dummy variables. Final regression equations are given by
ln Xr,j = ς0 +
(R−1∑
i=1
ς1iDi
)+ ς2 ln Distancer,j + ς3 ln
(pj
pr
)+ εr,j, (9)
ln
(Zr,j
Mr,j
)= ι0 +
(R−1∑
i=1
ι1iDi
)+ ι2 ln Distancer,j + ι3 ln
(qj
qr
)+ εr,j (10)
where ε and ε denote a normally distributed random error in each equation.
4The data on surface area are drawn from World Fackbook (Central IntelligenceAgency).
12
Di, for which the coefficient represents home bias in country i, is a dummy
variable taking the value 1 if i = r and 0 otherwise.
3 Data and econometric issues
This section argues on data and econometric issues. We focus on transactions
in finished and intermediate goods among East Asian countries in machinery
sector since machinery goods have played the most important role in the
development of international fragmentation (see Kimura and Ando, 2005).
Those data are obtained from Asian International Input-Output Table pub-
lished by the Institute of Developing Economies5. The data of total produc-
tion values in intermediate goods are also available in the table. Our sample
consists of nine East Asian counties (China, Indonesia, Japan, Malaysia, Re-
public of Korea, the Philippines, Singapore, Taiwan, and Thailand) and the
U.S. in 1990, 1995, and 2000.
In equations (11) and (12), the relative prices of each product, pj/pr and
qj/qr, partly embody differences in productive factor prices between country
r and j based on differences in factor endowments between them. This term
represents one of the most important factors in fragmentation, “differences
in location advantages”. In this paper, we simply use the relative per capita
GDP between trading partners as a proxy for the relative price, that is, for
the difference in the factor prices. Later, we will also use other data into the
5Since the import values for some pairs, e.g., import values of Taiwan and Korea fromChina, are not reported, we exclude those pairs from our sample.
13
relative price.
Furthermore, two points are to be noted here. First, it is necessary to
exclude one country in importer dummy variables in order to avoid dummy
trap. The home bias consists mainly of policy and non-policy barriers to
foreign-made goods. We exclude Singapore dummy since few policy barriers
exist there. Most of the remaining barriers seem to be universal non-policy
barriers, e.g., modular-technique, which are captured by a constant term in
the regression equations. Assuming that the remaining Singapore-specific
barriers and the elasticity of substitution in each good are almost constant
during the period, we investigate the changes in home bias/country specific
barriers by examining changes in coefficients for importer dummy variables.
Second, in general, separate estimation of two regressions may be ac-
companied with correlated estimation errors. That is, the error term in the
intermediate goods equation could possibly be non-orthogonal to that in the
finished goods equation. This correlation is plausible because the unobserv-
able elements, such as nontariff barriers between trading partners, would
simultaneously affect both transactions. However, we here use the method
of not generalized but ordinary least squares (OLS) by equation, because the
same regressors show up in each equation and thus the OLS estimates become
equivalent to the generalized least squares estimates. Estimating these equa-
tions by OLS, we investigate differences and changes in home bias in each
country for finished and intermediate goods by examining the coefficients for
importer dummy variables. To test the differences statistically, we conduct
14
the Wald test with the null hypothesis that respective those coefficients are
identical in both equations.6
4 Empirical results
This section provides some regression results. We report baseline results and
then the results of some robustness checks.
4.1 Baseline results
Here reports OLS regression results of equations (11) and (12). The baseline
estimation results are shown in Table 1. Three important facts are found in
the table.
First, the coefficient for the relative GDP per capita between trading part-
ners has experienced a steady decrease in both goods. The decrease can be
explained as in Kimura et al. (2006); transactions between developing coun-
tries with relatively small income disparity, such as between the Philippines
and China, increased rapidly.
Second, home bias for intermediate goods experienced a dramatic change
during the sample period. From 1990 to 1995, the home bias for intermediate
goods was at the same level as that for finished goods. Except in Japan and
Korea, the Wald test with the null hypothesis that the coefficient for home
bias is identical between intermediate and finished goods is not rejected in all
6Indeed, we perform generalized least squares estimation in order to obtain the covari-ances between the estimates from different equations, which are needed to perform theWald test.
15
countries. Therefore, in this period, although trade costs are often supposed
to be much larger in finished goods due to active dual track approach, that
is, trying to foster both import-substituting and export-oriented industries
at the same time, the international transactions of intermediate goods indeed
required high trade costs in a mass. As argued in Section 1, these high costs
would be given rise to by the inadequateness of timely delivery and uncer-
tainty regarding the coordination of a series of activities from production to
the shipment of end products.
However, the home bias for intermediate goods decreased remarkably
from 1995 to 2000, being lower than that for finished goods in 2000. This
dramatic reduction must be due not only to tariff-related trade policies, e.g.,
trade liberalization for semi-conductor-related parts and components in the
latter half of the 1990s, but also to the effect of a policy change in develop-
ing countries to the “accept everybody” policy for incoming foreign direct
investments in the latter half of the 1980s or the early 1990s (Kimura, 2006).
That is, investment liberalization attracts multinational firms which form
production/distribution networks, resulting in an explosive increase in the
international transactions of intermediate goods particularly with their home
countries and with intra-network firms locating in other countries.
Third, in 2000, the home bias for intermediate goods in almost all coun-
tries becomes as low as or becomes lower than that in Singapore. In partic-
ular, it is unbelievable that the home bias in Singapore is as low as in China
and is higher than in Malaysia and the Philippines. Indeed, these develop-
16
ing countries input foreign-made intermediate goods more than home-made
intermediate goods. The import values particularly from Japan are outstand-
ing in the countries. Therefore, a large number of international transactions
by Japanese firms within the networks may augment foreign-made interme-
diate inputs and underestimate the estimates of home bias in the developing
countries.
Next, to investigate more closely the changes in home bias coefficients, we
take the log difference in equations (3) and (6) between time-k and (k − 1)
as follows.
ln
(Xr,j,k
Xr,j,k−1
)= (1− σF ) ln
(τr,j,k/τr,r,k
τr,j,k−1/τr,r,k−1
)+ (1− σF ) ln
(pj,k/pr,k
pj,k−1/pr,k−1
),
ln
(Zr,j,k/Mr,j,k
Zr,j,k−1/Mr,j,k−1
)= (1− σI) ln
(tr,j,k/tr,r,k
tr,j,k−1/tr,r,k−1
)− σI ln
(qj,k/qr,k
qj,k−1/qr,k−1
).
Since “Distance” variable is time-unchangeable, final regression equations are
simply given by
ln
(Xr,j,k
Xr,j,k−1
)= ς0 +
(R−1∑
i=1
ς1iDi
)+ ς2
(ln
(pj,k
pj,k−1
)− ln
(pr,k
pr,k−1
))+ εr,j,(11)
ln
(Zr,j,k/Mr,j,k
Zr,j,k−1/Mr,j,k−1
)= ι0 +
(R−1∑
i=1
ι1iDi
)+ ι2
(ln
(qj,k
qj,k−1
)− ln
(qr,k
qr,k−1
))+ εr,j.(12)
The regression results are shown in Table 2. In the result of the period
1990-1995, we find that home bias for finished goods declines more rapidly
than for intermediate goods. Particularly in Indonesia, Japan, Korea, and
Thailand, the Wald test is rejected at 1% significant level. On the other hand,
in the period 1995-2000, Indonesia, Malaysia, and the Philippines have a more
rapid decrease in home bias for intermediate goods, but China has the reverse
17
result. Consequently, we can say that, in East Asian developing countries,
the decrease in home bias for intermediate goods was not outstanding in the
former half of the 1990s but turns out to be remarkable in the later half of
the 1990s.
4.2 Robustness checks
We here report two regression results. In the first, we introduce some inde-
pendent variables as done in the previous studies. The second relates to the
term of relative product prices. In baseline results, we use the relative per
capita GDP as a proxy for a difference in productive factor prices since the
product price embodies comparative advantages and since, as pointed out
in Mayer and Zingnago (2005), detailed variables, e.g., wage rates, are more
incomplete and noisy. Here a more appropriate proxy is employed.
First, we add three variables to extract some components of trade costs.7
First, language variable is a dummy variable taking unity if the same language
is spoken by at least 9% of the population in both countries and zero other-
wise. Second, we introduce colony variable taking unity if the two countries
have ever had a colonial link. Third, religion variable is a dummy variable
which takes unity if the two countries have the same representative religion
and zero otherwise. These cultural ties would reduce the business costs of
multinational firms. The two former variables are obtained from CEPII home
7This paper does not use common colonizer dummy and contiguity dummy variables,which are often introduced in the previous studies, since those variables are highly corre-lated with “Distance” variable in East Asia.
18
page8, and the last variable is constructed using World Factbook.9
The regression results of the equation added these dummy variables are
shown in Table 3. We can immediately see that most of the dummy variables
are not significantly estimated. The decrease in the coefficient for common
language dummy would be explained by the same manner as the decrease in
the coefficient for the relative product prices; transactions between countries
with different languages remarkably increased due to the entry of network-
forming firms. On the other hand, the results in home bias are quantitatively
and qualitatively unchanged with the results in Table 1. This implies that
these cultural elements occupy only a small portion of the home bias in each
country.
Second, we employ the unit value indices of machinery exports from each
East Asian country to the World. In this paper, product prices in both
finished and intermediate goods contain not only the composite of productive
factor prices but price index of intermediate goods. Since GDP per capita
is a proxy only for the composite, an omitted variable bias may yield in our
estimates.
The data of the unit value indices are drawn from “Compilation and Ap-
plication of Trade Indices: in East Asian Countries and Regions” published
by Institute of Developing Economies in 2003. The data in the United States
8http://www.cepii.fr/anglaisgraph/bdd/distances.htm#9We specified a religion which has the majority of the country as a representative reli-
gion in each country and then classified the representative religion into Buddhist, Taoist,Hindu, Jewish, Muslim, Orthodox, and Christian.
19
are unavailable in this book. Although industrial classification is the same
as that in this paper, we cannot get the unit value indices of finished ma-
chinery goods and machinery parts separately. Therefore, we use the unit
value indices of machinery exports for both intermediate and finished goods
equations. Furthermore, since these indices are incomparable at level among
countries, we estimate log difference equations, i.e., equations (11) and (12).
The results are shown in Table 4 and are qualitatively unchanged with
the ones in Table 2. That is, home bias for intermediate goods did not
decrease more rapidly in the former half of the 1990s but experiences a more
rapid decrease in some developing countries in the later half of the 1990s. In
particular, the decrease in the Philippines is outstanding.
Other regressions are also performed. We use the price level of GDP ex-
pressed relative to the United States and obtain its data from the Penn World
Tables v.6.1. as a proxy for relative product prices and obtain a qualitatively
the same result as the one in Table 1. The equation for finished goods sector
derived from a monopolistic competition model is also regressed since re-
cently the location decision of finished goods sector according to consumers’
demand comes to be observed, e.g., in Thailand. Although the estimated
coefficient for relative product prices was somewhat different from that in
Table 1, the results in home bias coefficients, particularly the relationship in
home bias between finished and intermediate goods, were quite the same.
Notice that we assume that home bias in Singapore has been almost un-
changed. If home bias for finished goods in Singapore declines more rapidly
20
than for intermediate goods, our conclusion may change somehow. However,
tariff rates, which must be the ones of the significant portions of policy bar-
riers, in both intermediate and finished machinery goods have been almost
zero in Singapore. Of course, Singapore does not already adopt import-
substitution policy in the 1990s. Rather it is natural also in Singapore to
suppose that the home bias for intermediate goods has experienced a more
rapid decline than for finished goods due to the decrease of non-policy barri-
ers in intermediate goods as argued in Section 1. Therefore, we can conclude
that home bias for intermediate goods transactions decreased more notably
than for finished goods transactions in the latter half of the 1990s.
5 Concluding remarks
This paper examines whether home bias for intermediate goods gets smaller
than for finished goods in East Asia in the 1990s. We find that, in East Asia,
while the home bias for intermediate goods was at the same level as that for
finished goods in the former half of the 1990s, the home bias for intermediate
goods has dramatically decreased since the latter half of the 1990s, being
lower than that for finished goods.
Recently, spatial economy comes to focus on the work of vertical linkages
between intermediate and finished goods. In the literature, absolute and
relative levels of transportation costs in intermediate and finished goods play
an important role in determining industrial location in an economy. The
estimates obtained in this paper will present illustrative parameter values of
21
the transportation costs for intermediate and finished goods.
References
[1] Ando, M., 2006, Fragmentation and Vertical Intra-industry Trade in East
Asia, forthcoming in North American Journal of Economics and Finance.
[2] Ando, M. and Kimura, F., 2005, The Formation of International Produc-
tion and Distribution Networks in East Asia, In Takatoshi Ito and Andrew
Rose, eds., International Trade in East Asia, University of Chicago Press.
[3] Flam, H. and Helpman, E., 1987, Vertical Product Differentiation and
North-South Trade, American Economic Review 77: 810-22.
[4] Hayakawa, K., 2006, Measuring Barriers to International Division of La-
bor in East Asia, KUMQRP Discussion Paper Series, No. 2005-031, Keio
University.
[5] Head, K. and Mayer, T., 2000, Non-Europe: The Magnitude and Causes
of Market Fragmentation in Europe, Weltwirschaftliches Archiv, 136:
285-314.
[6] Helliwell, J., 1997, National Borders, Trade and Migration, Pacific Eco-
nomic Review 2(3): 165-85.
[7] Helpman, E. and Krugman, P. R., 1985, Market Structure and Foreign
Trade, Cambridge: The MIT Press.
22
[8] Hillberry, R. and Hummels, D., 2002, Explaining Home Bias in Con-
sumption: The Role of Intermediate Input Trade, National Bureau of
Economic Research, Working Paper #9020.
[9] Hummles, D., 2001, Time as a Trade Barrier, GTAP Working Papers
1152, Center for Global Trade Analysis, Department of Agricultural Eco-
nomics, Purdue University.
[10] Kimura, F., 2006, International Production and Distribution Networks
in East Asia: 18 Facts, Mechanics, and Policy Implication, mimeo, Keio
University.
[11] Kimura, F., Takahashi, Y. and Hayakawa, K., 2006, Fragmentation and
Parts and Components Trade: Comparison between East Asia and Eu-
rope, KUMQRP Discussion Paper Series, No. 2005-030, Keio University.
[12] Mayer, T. and Zignago, S., 2005, Market Access in Global and Regional
Trade, CEPII, Working Paper, No. 2005-02.
[13] Poncet, S., 2003, Measuring Chinese Domestic and International Inte-
gration, China Economic Review, 14: 1-21.
[14] Wei, S-J., 1996, Intra-National Versus International Trade: How Stub-
born Are Nations in Global Integration?, National Bureau of Economic
Research, Working Paper #5531.
[15] Wolf, H.C., 2000, Intranational Home Bias in Trade, The Review of
Economics and Statistics, 82(4):555-563.
23
Table 1: Baseline results
1990 1995 2000Fin. Int. Fin. Int. Fin. Int.
distance -0.26 ** -0.93** -0.17 * -0.64** -0.12 * -0.55**(0.16) (0.18) (0.15) (0.12) (0.13) (0.11)
relative pergdp 1.07** ** 0.65** 0.78** ** 0.38** 0.49** ** 0.08(0.09) (0.09) (0.08) (0.06) (0.07) (0.06)
China -13.33** -10.67** -9.69** * -6.65** -4.70** ** -0.41(1.04) (1.13) (1.00) (0.78) (0.50) (0.41)
Indonesia -7.68** -9.16** -3.87** * -6.77** -2.99** -2.40**(0.89) (0.97) (0.86) (0.67) (0.50) (0.41)
Japan -5.19** * -2.89** -3.13** -2.22** -1.42** * 0.27(0.68) (0.74) (0.66) (0.51) (0.44) (0.36)
Korea -6.77** * -4.40** -4.38** -3.29** -3.11** ** -0.33(0.66) (0.72) (0.63) (0.49) (0.44) (0.36)
Malaysia -3.62** -5.33** -1.83* -2.60** -1.00* ** 1.16**(0.76) (0.83) (0.74) (0.57) (0.46) (0.37)
Philippines -6.25** -8.80** -3.52** -6.09** -1.24* ** 0.97*(0.79) (0.87) (0.77) (0.60) (0.50) (0.40)
Thailand -5.53** -5.77** -2.60** -3.90** -0.55 0.05(0.79) (0.86) (0.76) (0.59) (0.47) (0.39)
Taiwan -3.87** -3.46** -2.21** -2.32** -1.18** * 0.34(0.63) (0.68) (0.60) (0.47) (0.44) (0.36)
USA -2.08** -2.53** -1.70* -1.83** -0.14 ** 2.05**(0.72) (0.78) (0.70) (0.54) (0.49) (0.40)
cons 1.43 3.62** -0.02 2.02** -1.12 1.97*(0.95) (1.03) (0.91) (0.71) (1.13) (0.92)
R-sq 0.8469 0.6774 0.7835 0.7055 0.6933 0.6641Obs. 87 87 89 89 90 90
Notes: “Fin.” and “Int.” mean intermediate and finished goods equation, respectively.A dependent variable is a log ratio of “total production values weighted-” inter-nationalto intra-national import values. Regional names represent importer dummy variables. **shows 1 % and * shows 5 % significant. The inside of a parenthesis is a White consistentstandard error. The column between “Fin.” and “Int.” reports the result of the Waldtest with the null hypothesis that each coefficient is identical in intermediate and finishedgoods equations.
24
Table 2: Log-difference results
1990-1995 1995-2000Fin. Int. Fin. Int.
relative pergdp 1.79* * 0.21 0.32 ** -1.15**(0.76) (0.64) (0.43) (0.34)
China 1.95** 1.55** 3.17** ** 1.18**(0.44) (0.37) (0.51) (0.40)
Indonesia 2.46** ** 0.40 -0.74 ** 1.20**(0.43) (0.36) (0.47) (0.37)
Japan 1.97** ** 0.04 1.38* 0.75**(0.43) (0.36) (0.44) (0.35)
Korea 1.70** ** 0.29 0.64 1.28**(0.45) (0.38) (0.45) (0.35)
Malaysia 0.96* 1.34** -0.24 ** 1.26**(0.43) (0.36) (0.44) (0.35)
Philippines 1.33** 1.02** 0.78 ** 4.13**(0.46) (0.39) (0.44) (0.35)
Thailand 2.00** ** 0.29 0.87 1.56**(0.43) (0.36) (0.47) (0.37)
Taiwan 1.28** 0.57 0.45 0.93*(0.46) (0.39) (0.45) (0.36)
USA -0.15 0.01 0.83 0.62(0.53) (0.45) (0.47) (0.37)
cons -0.48 0.30 -0.67* -0.59*(0.31) (0.26) (0.31) (0.25)
R-sq 0.4192 0.3244 0.4944 0.668Obs. 69 69 71 71
Notes: See notes in Table 1. Dependent and independent variables except dummy vari-ables are expressed by the log difference.
25
Table 3: Regression results in equations with some dummy variables
1990 1995 2000Fin. Int. Fin. Int. Fin. Int.
distance -0.20 ** -0.95** -0.15 * -0.67** -0.10 -0.59**(0.16) (0.18) (0.16) (0.12) (0.14) (0.11)
relative pergdp 1.04** ** 0.66** 0.78** ** 0.39** 0.50** 0.09(0.09) (0.09) (0.08) (0.06) (0.07) (0.06)
China -13.15** -10.48** -9.80** * -6.63** -4.84** ** -0.31(1.05) (1.16) (1.02) (0.78) (0.51) (0.41)
Indonesia -7.32** -9.01** -3.84** * -6.85** -3.01** -2.46**(0.94) (1.03) (0.91) (0.70) (0.54) (0.43)
Japan -5.36** ** -2.41** -3.34** -1.97** -1.53** ** 0.49(0.72) (0.79) (0.70) (0.54) (0.48) (0.38)
Korea -6.74** * -4.19** -4.49** -3.18** -3.19** ** -0.23(0.67) (0.73) (0.64) (0.49) (0.45) (0.36)
Malaysia -3.25** -5.31** -1.71* -2.75** -0.94* ** 1.04**(0.79) (0.87) (0.77) (0.59) (0.47) (0.38)
Philippines -5.89** * -8.70** -3.41** * -6.22** -1.16* ** 0.85*(0.84) (0.92) (0.82) (0.63) (0.52) (0.42)
Thailand -5.60** -5.34** -2.88** -3.67** -0.78 0.27(0.83) (0.91) (0.80) (0.61) (0.51) (0.40)
Taiwan -3.91** -3.16** -2.32** -2.20** -1.24** ** 0.45(0.64) (0.70) (0.61) (0.46) (0.44) (0.35)
USA -1.82* -2.41** -1.59* -1.92** -0.06 ** 2.03**(0.75) (0.82) (0.74) (0.57) (0.50) (0.40)
language -0.41 * 0.62* -0.46 * 0.46* -0.28 0.31(0.28) (0.31) (0.27) (0.21) (0.23) (0.19)
colony -0.01 -0.11 -0.39 0.13 -0.54 0.22(0.52) (0.57) (0.51) (0.39) (0.44) (0.35)
religion 0.61* * -0.30 0.44 * -0.42 0.37 -0.45*(0.30) (0.33) (0.29) (0.22) (0.25) * (0.20)
cons 1.07 3.38** -0.02 2.07** -1.22 2.32(1.03) (1.13) (0.99) (0.76) (1.21) (0.97)
R-sq 0.8576 0.695 0.7954 0.7300 0.7067 0.6904Obs. 87 87 89 89 90 90
Note: See notes in Table 1.26
Table 4: Log-difference results: unit value indices of machinery exports
1990-1995 1995-2000Fin. Int. Fin. Int.
unit value indices 0.10 0.47 0.28 * -0.30(0.39) (0.32) (0.26) (0.22)
China 2.08** 1.64** 2.88** 1.82**(0.51) (0.41) (0.52) (0.44)
Indonesia 2.46** ** 0.44 -0.53 * 0.80(0.52) (0.42) (0.50) (0.42)
Japan 2.18** ** 0.19 1.32** 0.77(0.52) (0.42) (0.50) (0.42)
Korea 1.58* 0.63 0.48 1.44**(0.62) (0.51) (0.54) (0.45)
Malaysia 0.99* 1.28** -0.26 ** 1.37**(0.49) (0.40) (0.51) (0.43)
Philippines 1.66** 1.11** 1.16* ** 3.91**(0.49) (0.40) (0.58) (0.49)
Thailand 1.89** ** 0.26 0.75 1.34**(0.49) (0.40) (0.50) (0.43)
Taiwan 1.60** * 0.53 0.58 1.05*(0.51) (0.41) (0.53) (0.45)
cons -0.56 0.25 -0.64 -0.60*(0.35) (0.29) (0.35) (0.30)
R-sq 0.3365 0.3441 0.4556 0.6425Obs. 69 71 69 71
Notes: See notes in Table 1. “unit value indices” denotes relative unit value indices ofmachinery exports. Dependent and independent variables except dummy variables areexpressed by the log difference.
27