Kengoay. Pdf
Methodology: A Modified Gravity Model Approach
Yüklə 353 Kb.
|
|
3
Methodology: A Modified Gravity Model Approach
The gravity equation has experienced something of a revival in recent years and provides a natural framework and a useful multivariate approach for assessing the impact of regional trading blocs on the level and direction of bilateral trade flows. In addition to its continued success at predicting bilateral trade patterns it has seen renewed interest from the new economic geography literature as well as increased confidence in its traditionally weak theoretical foundations arising primarily from the so-called “new trade” theory. 8 Indeed, Elbadawi (1997) argues that the successful outcome of the gravity equation may stem from its ability to capture factors that conventional factor endowment theory cannot easily accommodate such as
8 See Evenett and Keller (2002) for a discussion of theories that explain the gravity equation’s success. The underlying theory can be found in Deardorff (1984) and Helpman and Krugman (1985). See Anderson (1979), Bergstrand (1985) and Sanso et al. (1993) for further discussion.
10 intra- industry trade (by considering, for example, economic size and scale economies) and the dramatic reallocation of resources following trade liberalisation (by introducing dummy variables). 9
An example of a gravity equation where trade is positively related to two countries incomes and negatively related to the distance between them (reflected by transport costs) takes the following form (in natural logarithms); 10
ij ij j i j i ij u D N N Y Y M log
log log
log log
log log
log 5 4 3 2 2 0 + + + + + + = α α α α α α
(1)
where M ij is the imports from country i to j, Y is income, N is population, D ij is the distance between economies i and j and u ij is the log normally distributed error term where E(log u
)=0. Assuming no PTAs, equation (1) explains trade between countries i and j and behaves as a counterfactual.
In this paper we modify the basic gravity equation where our estimating equation in natural logs (with expected signs in brackets) is shown in equation (2). 11
9 Previous developments included the addition of a population variable to reflect the role of scale economies (e.g. Linnemann 1966), a dummy variable to represent the existence of preferential trade arrangements (e.g. Aitken 1973), the difference of per capita income and the absolute income level as variables to capture the so-called Linder Hypothesis and differences in factor endowments respectively (see e.g. Elbadawi 1997 and others). Linder (1961) argued countries with similar income level have incentives to trade with each other. In addition to distance and adjacency variables, recent studies also include a geographical measure of “remoteness” although our concentration on the ASEAN region makes its inclusion unnecessary. 10 This is a variation on the original Tinbergen (1962) and Pöylönen (1963) specifications. 11 All regressions are estimated using OLS in TSP. Estimations using a Tobit model (to account for the fact that trade values are bounded from below at zero) resulted in very similar results. See Soloaga and Winters (2001) for a discussion. 11 ( ) ( ) ( ) ( ) ( ) ( ) ( ) ) - or (
)
- or
(
log
)
- or
( ) (
) (
)
- or
(
(-)
log log
) (
) (
) (
) (
log
log log
log log
log 11 10 9 8 7 6 5 4 3 2 1 0 + + + + + + + + + + + + − + + + + + + + + + + = ∑ ∑ ∑ Yüklə 353 Kb. Dostları ilə paylaş:
|