|
| |
N° 2007-05 |
  |
| March 2007 |
| Trade Costs and the Home Market Effect |
Matthieu Crozet
Federico
Trionfetti |
|
| |
|
Models characterized by the presence of increasing returns, monopolistic competition, and
trade costs typically give rise to a more-than-proportional relationship between a country’s
share of world production of a good and its share of world demand for the same good. This
relation between country’s market size and industrial specialization has become known as
the Home Market Effect after Krugman (1980) and Helpmand and Krugman (1985).
This relationship has a central place in theoretical discussions about international trade. Indeed,
it suggests that larger countries should specialize in increasing return to scale industries;
on the contrary, countries in remote location should specialize in constant return to
scale industries and consequently have lower income and long term economic growth. Moreover,
the Home Market Effect is the main engine of the agglomeration processes emphasized
by the new economic geography models (Krugman, 1991). This literature shows that trade
openness may influence capital flows and accumulation, and thus contribute to widen international
inequalities. Testing for the existence of the Home Market Effect is a first step
toward and empirical validation of the new economic geography framework. Finally, the
HME is so closely associated to the presence of increasing returns to scale (IRS) and monopolistic
competition (MC) that it has been used as a discriminating criterion in a novel
approach to testing trade theory pioneered by Davis and Weinstein (1999, 2003).
However, Davis (1998) shed doubt on the theoretical robustness of this relationship. Indeed,
one pervasive assumption in the literature to date is that of the presence of an “outside good”,
freely traded and produced under constant returns to scale (CRS) and perfect competition
(PC) ; Davis (1998) shows that in the absence of an outside good (i.e. assuming trade costs
in all sectors), the HME may disappear. The assumption of the existence of a freely traded
CRS-PC good is as much convenient as it is at odds with reality. As noted by Head and
Mayer (2004, p. 2634) when discussing this issue in their comprehensive account of the
literature:“... the CRS sector probably does not have zero trade costs or the ability to absorb
all trade imbalances.” The pervasive use of the outside good assumption and its inconsistency
with reality raise the question of what are the consequences of its removal on the HME. The
present paper investigates this question.
We eliminate the outside good from the main model used in the empirical literature on the
HME. This model, in two different variants, has been used in Davis and Weinstein (1999,
2003) and in Head and Ries (2001). We find that, in general, the HME survives when the
outside good is absent but its average magnitude is attenuated. More interestingly, both
variants of the model predict a non-linear relationship between the production share and the
demand share. The non-linearity is characterized by a weak HME (or absence thereof) when
countries’ demand shares are not too different from the world average. The HME becomes
stronger when countries’ demand share become more dissimilar.
In order to put this result to empirical verification, we use the trade and production database
developed by CEPII (see Mayer and Zignago, 2005). Because of numerous missing values,
we finally use a restricted balanced data set for 25 countries and 25 industries, over the period
1990-1996. Nevertheless, the set of countries remains very large; in 1996, it accounted for
more than 78% of world GDP and about 70% of world trade.
In a fist step we estimate a polynomial equation relating countries’ relative demand for each good (i.e. the demand deviation from the sample average), to the corresponding production
deviation. Demand deviations are computed as the sum of sectoral expenditures in all locations
weighted by accessibility to consumers. We use Head and Ries (2001)’s measure
of trade freeness to weight countries demand. The results strongly support the theoretical
predictions: we observe a significant Home Market Effect on average, and a smoothly nonlinear
relationship between demand deviations and production deviations. We perform also
several robustness checks. For instance, we follow Davis and Weinstein (2003) using a two
step procedure to estimate the freeness of trade. First, we perform, for each industry, a gravity
estimation using bilateral trade data; then the coefficients of this regression are used to
compute the bilateral trade barrier. The results remain barely unchanged: the HME is smaller
when absolute value of demand deviations from the average are small.
In a second step, we go further by estimating the critical value of demand deviations beyond
which the HME has a stronger influence on specializations. We thus perform with maximum-
Wald tests for each industry. We show that the relation between demand and production
deviations is strictly linear for two industries only: Wearing apparel and rubber products.
One industry (Footwear) shows a non-linear shape that is clearly consistent with the constantreturn
to scale and perfect competition paradigm. Finally, for eleven industries (out of 25),
we observe a “piecewise” HME: the relationship between output and demand deviations is
proportional for medium-sized demand deviations, and more than proportional for very large
and very low demand deviations. For these eleven industries (that account all together for
more than 62% manufacturing production in our sample) the HME matters only for one fifth
of the observations, and HME is of negligible importance or totally absent for the remaining
observations. |
Non-technical summary |
Résumé
non-technique
en français  |
| Full text |
| |
|
| International trade; test of trade theories; economic geography |
Keywords |
| F1, R12 |
JEL classification |
| |
|
| To visualise the full text document, use Acrobat
Reader |
|
|
|
| |
|
|
|
|
|
