Uploaded by yoriz06

Empirics of International Trade Models

Empirics of International
Trade Models
• Empirical works international trade
• Empirics can mean various things in economic studies
• Ex-ante
• Simulation, e.g. partial and general equilibrium modeling
• To answer what-if questions, e.g. policy analysis: what will happen to imports if tariff
is eliminated
• Ex-post
• Academic exercise: testing hypothesis
• Policy analysis: to complement simulation-based analysis in determining parameters
• Here, we looks at ex-post empirical works
• Some empirical works on GFT
• Other works on Ricardian and H-O Model
Gains from Trade: Empirics (I)
• Frankel and Rose (1999) takes a huge question: ‘Does trade cause
growth?’ with more attention to the endogenous nature of trade
• Key idea: IV: how far that country is from large (ie rich) potential trade partners
• First-stage regression has two parts
• First is based on well-known gravity equation
• FR (1999) aggregate the previously estimated gravity regression over all of
country i ’s imports from all of its bilateral partners, j
• This constructed variable Tî is then used as an instrument for how much a
country is actually trading
• That is, the real first-stage regression is to regress Ti (exports plus imports over
GDP) on T̂i and population and area
Gains from Trade: Empirics (I)
• FR (1999) run the regression ‘Does trade cause
• These are surprisingly big effects. Possible
• The IV results are still biased upwards. Countries
that are close to big countries are rich not just
because of trade, but because of spatially
correlated true determinants of prosperity
• ‘Openness’ is proxying for lots of true treatment
effects of proximity to neighbors, e.g. MNCs,
technology transfer, not only trade
• The dynamic effects of ‘openness’ accumulated
over a long period of time, are larger than the
static one-off effects of opening up to trade
Gains from Trade: Empirics (II)
• Comparative Advantage is a
fundamental theoretical idea in
Economics, yet testing it is hard. Why?
• Problem 1: The theory is too weak to
test in real world (where more then 2
countries or goods).
• Problem 2: Latent variable problem:
The statement is about trading
behavior but is based on autarky
• Problem 3: Periods of autarky rarely
• Bernhofen and Brown (2004) exploits the
closed economy of Japan in 1858, and its
subsequent opening up to trade in 1859,
as a natural experiment to test for Law of
• It solves problem No. 3
• Further attractive features of this setting
• Relatively simple economy
• Subsequent opening up was plausibly
exogenous to economic change in Japan
• Suppose 1858 is autarky and 1859 is not
• They can compute p1858.t1859 and check
whether it’s bigger than value of trade
in 1859
Measuring Gains from Trade
• Under the assumption t1858 = t1859, it
puts an upper-bound on GT
• cf1858 is the counterfactual consumption
of Japan in 1858 if it were open to trade.
• Of course, by WARP, cf1858 was not
affordable in 1858 or else it would have
been chosen.
• ΔW measures the amount of income that
would have made counterfactual cf1858
These translate into 5.4-9.1 % of GDP
Empirics of Ricardian Models
• MacDougall (1951) made use of newly available
comparative productivity measures (for the UK and
the USA in 1937) to ‘test’ the intuitive prediction of
Ricardian models
• His hypothesis was that given the U.S. wage rate in
1937 was approximately twice that in Britain, US
firms should have an export advantage in
manufacturing sectors in which US labor
productivity exceeded twice the level in UK
• Using simple measures of labor productivity, he
found that, for the exports of U.S. and U.K. to the
same third world countries, for 20 of the 25
products considered the ratio of US exports to UK
exports exceed one, while in the rest of the cases
this ratio was less than one
• Thus, this simple prediction was confirmed for the
bulk of the products in the study
• Golub and Hsieh (2000) are interested in testing
the Ricardian model by comparing the US to the
UK, as MacDougall did
• The coefficient of interest is β.
• The intuition of the Ricardian model suggests that
β1 > 0 and β2 > 0.
• But there is no explicit multi-country Ricardian
model that would generate this estimating
Empirics of HO Model
• The H-O model is probably the most influential model in all of Trade.
So how do we assess how useful a description of the real world it is?
• One immediate obstacle is that the theory’s predictions aren’t that
• The 2 × 2 model makes precise predictions, but not much of this generalizes
to higher dimensional settings
• If G ≤ F, production (and hence trade) is determinate. Hence the ‘Goods
Content of Trade’ (GCT) (or pattern of trade) is determinate
• If G > F, production (and hence trade) is indeterminate. But the (Net) Factor
Content of Trade (NFCT) is determinate, HOV prediction
Empirics of HO Model
• Theory: Tc (p,Vc ) = R(p).(Vc − scVw)
• Baldwin (1971) was the first to explore the implications of this equation
• He could have either:
• Taken data on Tc (p,Vc ), R(p) = [A(w)]−1, and (Vc − scVw), to check this prediction
exactly. As we’ll discuss next lecture, one can obtain data on A(w) from input-output
• Or, regressed Tc (p,Vc ) on R(p) = [A(w)]−1 to check whether the estimated coefficients
take the same signs/magnitudes as (Vc −scVw)
• Or, regressed Tc (p,Vc ) on (Vc − scVw) to check whether the estimated coefficients take
the same signs/magnitudes as R(p) = [A(w)]−1
• Baldwin (1971) did 2. Leamer (1984) did a version of 3
Leontief Paradox
• Leontief, using data on United States from 1947, computed the amount of labor
and capital required to produce $1 million of US exports and $1 million of US
imports. Since this was not feasible, Leontief calculated the labor and capital
required to produce US goods that are similar to US imports also referred to as
import-competing goods
• Since during those times US was the most capital-abundant (or labor scarce) as
compared to the rest of the world, the expectation was that exports would be
capital-intensive as compared to the imports
• Leontief found that the capital-labor ratio in the US imports exceeded that in the
US exports by about 23 percent. This result is called the Leontief paradox
Krugman’s Monopolistic Model
• Each firm produce differentiated products
• Increasing returns to scale and
• Monopolistic competition
• An important result is that relative country size determines volume of
• Different than the result of HO model
Testing the hypothesis
• Specification below is used to test the hypothesis
• It is then translated to econometric specification below
Results of Estimations
Simulation Modelling to Find Some Testable
• Baier and Bergstrand (JIE 2004) empirically
test factors affecting the decision of two
countries in forming FTAs
• The premise is that if preferential trade
liberalization between two countries increases
welfare, both will be likely to engage in FTA
• They build model to simulate the impact of
trade liberalization between two countries
Model with 6 countries in three continents
Differ in endowments, sizes
Two sectors with different factor intensities
Trade costs, transportation cost
• Simulation is to have trade cost eliminated
between different pairs of countries
• Several hypothesis can be inferred
• First, FTAs between natural partners (same
continents) are more welfare improving
• Distance negatively affect the likelihood of
forming FTAs
• Second, net welfare gain is increasing as the
size of both countries increases
• Countries with bigger sizes tend to form FTAs
• Third, net welfare gain increases as the more
similar the sizes
• There are several other hypothesis can be
• They use those hypothesis to form empirical
model and
Introduction: Gravity Modeling
• “Gravity” is the workhorse of empirical international trade.
• Origins with Tinbergen (1962).
• Thousands of published articles and working papers since then.
• “Some of the clearest and most robust findings in empirical economics.”
(Leamer & Levinsohn, 1995)
• Numerous applications looking at different types of trade
• costs, and their impacts on trade flows:
Transport costs.
Tariffs and non-tariff barriers.
Regional integration agreements, currency unions
Time delays at export/import and trade facilitation.
Governance, corruption, and contract enforcement
Gravity Model
• Newton’s theory of gravity postulates that the gravitational force between
two objects is directly proportional to each of their masses, and inversely
proportional to the square of the distance between them
• Economic “mass” is proxied by a country’s GDP
• The distance between countries is taken as an indication of the level of
trade costs they face.
• Think of a larger economy pulling more goods from its trading partners;
similarly, a larger economy pushes more goods towards its trading
• But this process is impeded by trade costs (e.g., transport, tariffs, etc.)
Gravity Model for Trade
X ij  c
YiY j
d ij
• Take logs, add some reduced form coefficients, and include an error
term to get the familiar empirical gravity model
Gravity in Action
• Using some basic data, we can quickly get a sense of just what the
gravity model is picking up
• Bilateral trade: 2005, nominal USD.
Source: Comtrade via WITS.
• Export and importer GDP: 2005, nominal USD
Source: World Development Indicators.
• Bilateral distance.
Source: CEPII distance dataset.
Imports and GDP
Imports and Distance
Theoretical Foundation of Gravity Model
• Theory-based gravity model
X ij  
 Yi E j
  P1
 i j
• Looks like gravity, but also has some important
differences with respect to the basic model.
• Selection of variables.
• Interaction between trade costs and the substitution
• Inclusion of the multilateral resistance terms.
Notes on Gravity Model
X ij  
 Yi E j
  P1
 i j
• Each observation should represent a unidirectional flow
between a pair of countries, e.g. exports from i to j
• GDP should be in nominal, not real, terms. In fact, the
destination variable should be representation of demand
• We need to do something to take account of price.
• Standard price indices are not aggregated in the way implied by
theory; a poor proxy for the true (ideal) price indices
• The model depends on trade costs, but they are not
directly observable as an aggregate. We need to build
them up by specifying the term as observables factors
Notes : Multilateral Resistance
 i  
• Bilateral trade is not merely about the two countries
• Other countries characteristics influence bilateral relations
• Two countries far away from RoW tends to trade more
• Australia and New Zealand
• Those are represented in “multilateral resistance” factor
• Can also be perceived as “market potential”
• It is often measured by the sum of partners’ real GDPs divided
by bilateral distance