Evidence-Based Trade Policymaking Capacity Building Programme
ARTNeT Interactive Gravity Modeling Tool
Witada Anukoonwattaka (PhD) UNESCAP 26 July 2011
Outline
Background on gravity model of trade and their applications Theoretical gravity and fixed effect issues ARTNeT Interactive gravity model database Demonstration and exercises Conclusions
Part I
Background on gravity model of trade and their applications
What is Gravity Model
Gravity model is a very popular econometric model in international trade The name came from its utilizing the gravitational force concept as an analogy to explain the volume of bilateral trade flows
Proposed by Tinbergen (1962)
Initially, it was not based on theoretical model, but just intuition only Later on, a range of rigorous theoretical foundation has been given.
The most well-known benchmark so far is Anderson and van Wincoop (2003).
Gravity Analogy
Gravity force equation
Fij = G MiM j
2 Dij
Trade Gravity: Basic Yi Y j X ij = C tij
X ij = exports (or trade) from i to j , C = contstant, Y = economic mass ( GDP), t = trade costs between two countries distance, adjacency, .., " policy factors".
Gravity force between two objects depends on their masses and inversely proportional to the square of distance between them.
Export (or trade) between two countries depends on their economic masses and negatively related to trade costs between them.
Basic gravity model of trade:
X ij = C Yi Y j t ij
Larger countries trade more than smaller ones Trade costs between two trade partners reduce trade between them. Empirical equation for basic gravity model:
ln X ij = b0 + b1 ln( Yi ) + b2 ln( Y j ) + b3 ln( t ij ) + eij b1 , b2 > 0; b3 < 0 A 1% change in Yi is associated with a b1 % chage in X ij .
Proxies for trade costs
Distance Adjacency Common language Colonial links Common currency Island, landlocked Institutions, infrastructures, migration flows,.. Bilateral tariff barriers
Why is it so popular?
Intuitively appealing Fits with some important stylized facts Easily to use real data to explain trade flows with respect to policy factors we are interested in. Estimation using OLS
Examples of Applications
Effects of regional integration on trade
Do RTAs boost trade between members? ln X ij = b0 + b1 ln(Yi ) + b2 ln(Y j ) + b3 ln(tij ) + b4 (dummyRTAij ) + eij
Do RTAs reduce exports from non - members? ln X ij = b0 + b1 ln(Yi ) + b2 ln(Y j ) + b3 ln(tij ) + ... + b4 ( dummy BothInRTAij ) + b5 (dummy OneInRTAij ) + eij
See, World Bank (2005) for example.
Examples of Applications
Effects of trade facilitation on trade
How much can tradefacilitati on boost bilateraltrade? ln X ij = b0 + b1 ln(Yi ) + b2 ln(Yj ) + b3 ln(tij ) + b4 ln(timeiX ) + eij
See, Djankov etal.(2010) for example.
Examples of Applications
Effects of institutional weakness on trade
How does corruption affect trade? ln X ij = b0 + b1 ln(Yi ) + b2 ln(Y j ) + b3 ln(tij ) + b4 ln(corruptioni ) + eij
See, Anderson & Marcouiller(2002) for example.
Examples of Applications
Estimating trade potential
ln X ij = b0 + b1 ln(Yi ) + b2 ln(Y j ) + b3 ln(tij ) + .... + eij
Fist step: estimate the model to get estimated coefficients Second step: Use estimated coefficients give predicted Xij
+b ln( Y ) + b ln( Y ) + b ln( t ) + ... =b X ij 0 1 i 2 j 3 ij
Third: Trade potential is the gap between predicted and actual Xij
Part II
Theoretical gravity and Fixed effect issues
Why do we need theoretical gravity?
Major weaknesses of basic gravity: No theoretical background Failure to take into account of GE effects
Nothing from the third party can affect trade between the two partners Cannot say about welfare effects Cannot say about resource allocations within an economy
A brief introduction to theoretical gravity
Enormous literature has lied theoretical foundations to gravity model
Ex. Anderson (1979), Bergstrand (1990), Anderson&Van wincoop (2003), Helpman et al (2008).
Some of them have strong implications on the data and estimation method
Complex empirical issues that OLS cannot handle For ex. see the gravity course on Ben Shepherd website at [Link]
AvW (2003)
The most formal benchmark for theoretical gravity model so far
Bringing the gravity model a step closer to GE effects Accounting for relative price effects on trade flow Things affecting relative price can influence bilateral trade flow. No matter the things happen between the two trading partners or happen with third parties.
AvW (2003): Empirical implications
Relative trade costs matter, not only absolute trade costs,
Not only distance between countries, but also their distance from RoW matter for their bilateral trade
Two types of trade costs have to be taken into account
Trade costs between i and j Trade costs of i and j with third parties
This idea introduces a multilateral trade resistance (MTR) term to the gravity regression
MTR terms are empirically unobservable One approach to take into account MTR in the regression is using Fixed Effects Dummies.
ln X ij = b0 + b1 ln(Yi ) + b2 ln(Y j ) + b3 ln(tij ) + b4 ln(MRTi ) + b5 ln(MRT j ) + eij
Using fixed Effects Dummies
An approach to (partly) take into account MTR in the regression, while using OLS Fixed effects dummies
Exporter & Importer dummies Year dummies Sector dummies
With fixed effects dummies :
ln X ij = b0 + b1 ln(GDPi ) + b2 ln(GDPj ) + b3 ln(tij ) + ... b4 dummy _ countryi + b5 dummy _ country j + eij
DONT make these MISTAKES: Xij = trade between ij (wrong!)
We need a clear distinction between exporter effects and importer effects
USE GDP per capita rather than GDP (wrong!)
Price Index has already been included in MTR terms
Part III
ARTNeT Interactive gravity model database
ARTNeT online gravity
ARTNet has developed an online gravity interface to make it easier for policy researcher to get started with gravity Available online and FREE at [Link] Can be used for
Estimate basic gravity models using a wide variety of policy data Analyze trade potentials Download pre-formatted data that can be plugged into statistical software for more extensive analysis
Data and sources
Cover various countries and time periods Exports and imports GDP, GDP per cap, growth, population Geographical, historical , and cultural links Cover trade policy data:
trade credits, trade facilitation, logistics, infrastructure Behind the border barriers Aid for trade
Current datasets
General trade dataset: 1994-2007
Use for general analysis and for calculate trade potential
Trade facilitation dataset: 2006
Use for analysis focusing trade effects of trade facilitation, customs and border procedures, and behind the border barriers
Aid for trade dataset: 2002-2007
Use for analysis of trade effects of aid flows
Example 1: Basic Regression
Select ARTNeT data on trade dataset
Use for general analysis and for calculate trade potential
Set the dependent (LHS) variable to log(exports) Standard independent (RHS) variables
Partner and reporter GDP Distance Common language, geographical contiguity, colony, common colonizer
Run the regressions and interpret the results
Part IV
Demonstration and Exercises
Example 1: Trade potential
Calculate estimated to actual trade ratios for the following country pairs:
Exports from Thailand to USA for 2007, and averaged over 1994-2007 Exports from USA to Thailand for 2007, and averaged over 1994-2007
Example 1: Trade projection
Using 2007 as the base year, calculate trade projections for:
Thailands exports to Cambodia, China, Indonesia, India, and Japan, based on an expected 2% fall in reporter and partner GDP Thailands exports to USA, Canada, Germany and Japan, based on expected 1% fall in international transport costs
Example 2: Trade Effects of Tariffs
Select trade facilitation dataset
Use for general analysis and for calculate trade potential
Set the dependent (LHS) variable to log(exports) Independent (RHS) variables
Partner and reporter GDP per capita Partner and reporter populations Distance Common language, geographical contiguity, colony, common colonizer Simple avg tariffs in the partner (importing) country
Run the regressions and interpret the results Analyze trade impact of a 5% cut in tariffs
Part V
Conclusions
Value of the interactive database
A wide variety of trade-related data, in particular on trade facilitation Additional data can easily be plugged in Saves researchers a lot of time and resources in terms of data collection and preliminary analysis Extensions form basic models are possible such as fixed effects ARTNet has developed an online gravity interface to make it easier for policy researcher to get started with gravity Available online and FREE Can be used for
Estimate basic gravity models using a wide variety of policy data Analyze trade potentials Download pre-formatted data that can be plugged into statistical software for more extensive analysis
Things to aware
Cant say about welfare / resource allocation
It is not based on general equilibrium It is based on historical data
Cant be a primary tool for forecast / counterfactuals
It is purely based on historical data
Other tools (ex. CGE) have to use in conjunction to handle these issues OLS may not give an efficient estimates if
if there is zero trade flow. Reverse causality exist (endogeneity problems) Multi-dimension data
Suggestions
Robustness check
Whether model specification and estimation techniques used are appropriate
Being modest when using the gravity results in your analysis
the gravity results give ideas about the relationship, but do not expect precise $ figure
If an analysis really need $ number, do it for a range rather than a single number (ex. Between SD and +SD) Asking questions that gravity can answer well
Ex. Trade potential based on 2007 data does not say about current/future potential
If there are serious doubts about reverse causality, treat the conclusions as association, not causal links Apply gravity analysis in conjunction with other approaches such as CGE. Keep updating the up-to-date model and techniques
References
Tinbergen,J. (1962). Shaping the World Economy: Suggestions for an International Economic Policy: 20th Century Fund, NY. Anderson, J. and E. Van Wincoop (2003). Gravity with Gravitas: A solution to the Border Puzzle, AER, 93: 170-192. World Bank (2005). Global Economic Prospect: Trade, Regionalism, and Development. Djankov, S., C. Freund & C. S. Pham (2010). "Trading on Time," The Review of Economics and Statistics,92(1):166-173. Anderson, J. E. and D. Marcouiller (2002). Insecurity and the Pattern of Trade: An Empirical Investigation. Review of Economics and Statistics, 84(2), 342-52. Anderson, J. E. (1979). A theoretical foundation for the gravity equation. AER, 69(1), 106-116. Bergstarnd, J.H. (1990). The H-O-S model, the Linder Hypothesis, and the determinants of bilateral intra-industry trade. Economic Journal, 11(4):1216-1229. Helpman, E., Melitz, M. and Rubinstein, Y. (2008), Estimating trade flows: Trading partners and trading volumes, Quarterly Journal of Economics, 123, 441-487.
