Transcript Trade policy analysis: choosing the appropriate methodology

Trade policy analysis: choosing
the appropriate methodology
Marc Bacchetta
ERSD/WTO
Outline
Research based policy making
 Criteria for choosing a methodology
 Key methodologies

– Modelling assumptions
– Required resources
Research based policy making

Research based information on policies is
needed at different stages of the policy
making process
– Proposal
– Policy dialogue with stakeholders
– Implementation
Dialogue
Dialogue between researchers and policy
stakeholders is crucial
 Researchers may help policy makers
specify questions
 Researchers should guide the choice of
appropriate methodologies

Methodology: the question
 Which
approach is best suited to
answer the question at stake given
existing constraints ?
Choice of methodology is not
necessarily straightforward
The question should dictate the choice of a
methodology ...
 But there are various constraints:

– Time and resources
– Sunk costs (incl. familiarity with certain
methodologies, institutional constraints)
– Fashion
Key approaches
Descriptive statistics
 Modelling approaches:

– Ex ante vs ex post approaches
– Econometric vs simulation models
– Partial vs general equilibrium
Ex ante vs ex post approaches

Ex ante simulation involves projecting the effects
of a policy change on a set of economic variables
of interest
– Can answer “what if” questions

Ex post approaches use historical data to conduct
an analysis of the effects of past trade policy
– Can answer “what if” questions if estimated
parameters are used for simulation (which assumes
that past relations continue to be relevant)
Econometric vs simulation models

In econometric models, parameter values are
estimated using statistical techniques
– Parameter values come with confidence intervals
– Parameter estimation is resource intensive
– Results are specific to one country or group of
countries

In simulation models, parameters are typically
drawn from a variety of sources and some are
calibrated
Descriptive statistics (1)
Trade flows analysis
 Example: assessment of trade patterns
and/or trade performance
 Conceptual tools

– Revealed comparative advantage, intraindustry trade, export diversification, etc.

Empirical tools
– Entropy indices, revealed comparative adv.
Trade performance and
competitiveness assessment


Trade in services
Merchandise trade
– Trade performance



Growth and pattern in the direction of trade
Composition of trade
Export concentration and principal products
– Competitiveness, specialization and complementarity




International competitiveness of exports
Export specialization
Trade complementarity and intensity
...
Indices


Export specialization
index
Trade
complementarity
index
ESI ijk 
X ijk X ijt
M jk M jt
TCIij   RCAik  RCD jk  M wk M w 
k
– with
X ik X i
RCAik 
M wk M w
RCD jk 
M jk M j
M wk M w
Descriptive statistics (2)
Trade policy analysis
 Example: trade policy assessment
 Conceptual tools

– Effective protection, non tariff barriers, bound
and applied tariffs, tariff escalation, etc.

Empirical tools
– Averages, dispersion indices, coverage ratios,
etc.
Trade policy assessments

Trade policy reviews
– Trade and investment regime
– Trade policies by measure and by sector

Tariff profiles

NAMA tariff simulations
Econometric estimation of gravity
equations
Can be used to study the impact of trade
policy variables on trade flows
 Example: effect of regional trade
agreement
 Ex post: not well suited for making
predictions
 No welfare effects

Econometric estimation of gravity
equations
Reasonable data requirements
 Reasonable entry costs
 Cost of econometric software package ?
 Importance of integrating theory with
estimation

Econometric estimation of economic
consequences of trade
Example: assessment of the distributional
effects of trade policy
 Ex post
 Variable level of econometric
sophistication
 Variable data requirements

Partial equilibrium simulations
Focuses on a specific market or product
and ignores interactions with other markets
 Best suited for the analysis of sectoral
policies, or when interactions with other
markets are expected to be limited
 Allows to include more market relevant
details than GE models
 Ex: assessing the welfare effect of a
reduction of the tariff on wheat

Partial equilibrium simulations

A number of partial equilibrium models
have been developed to simulate
international trade policy changes
– Those include SMART, ATPSM, SWOPSIM

Data requirements are manageable
– Elasticities are crucial
General equilibrium simulations




GE explicitly accounts for all the links between
sectors of an economy – households, firms,
governments and the rest of the world
Imposes constraints s.t. income equals
expenditure
Trade off between detail and breadth of coverage
Assesses effects of policy changes on aggregate
and sectoral variables, including:
– Income, production, employment, relative factor and
product prices, etc.
General equilibrium simulations
Single or multiple country models
 Highly intensive in data and parameters

– SAM, behavioural parameters, elasticities
– GTAP provides data and a simple model
Entry cost is significant
 High risk of misinterpretation

Paper on Turkey

Glenn Harrison, Thomas Rutherford and David Tarr
(2003) “Trade Liberalization, poverty and Efficient
Equity,” Journal of Development Economics, Vol. 71 (1),
June 2003, 97-128.

Policy application—joining a Customs Union with the
European Union (not membership).

Small Open Economy model with 40 households, 20
rural and 20 urban.

Economic Theory suggests that for developing countries
trade liberalization should shift production toward labor
intensive products. That should be pro-poor.

In Turkey the authors did not find this. They estimated
that the poor lost. Why?

They show that estimation of factor intensities is crucial
to a sensible link between trade policy and poverty
analysis.

Input-Output tables notoriously fail to accurately report
factor intensities. Capital’s share is a residual. In
Agriculture, labor’s earnings are underreported, so it is
the most capital intensive sector in IO tables. Services
sectors are also problematical.

In Turkey locomotives are reported as 100% labor;
textiles are capital intensive.

Perverse results may be obtained if estimates are based
on IO table factor intensities. The lesson learned for
future applications is that factor intensities should be
estimated outside of the input-output model.

The authors caution consumers of these models to ask
producers how they have obtained the values of their
factor intensities.

The study identifies sectors where an increase in the tariff
rates will increase the welfare of the poor. The authors
caution, however, that political economy reasons suggest
that these models should not be used for industrial policy
to help the poor. Likely to be abused like the many other
arguments to depart from tariff uniformity.
Partial vs general equilibrium models
Capturing economy wide linkages
Consistency wrt budget constraints
Capturing disaggregated effects
Capturing complicated policy mechanisms
Use of timely data
Capturing short and med. term effects
Capturing long term effects
Past performance in projecting impacts
CGE PE
X
X
X
X
X
X
X
Conflicting results
Policy makers don’t like conflicting results
 In most cases different results reflect
different assumptions
 Such differences are difficult to avoid
 What matters is the presentation of the
results

References
Piermartini, R. and R. Teh (2005)
Demystifying modelling methods for trade
policy, WTO Discussion Paper No. 10
 Bowen, H.P., A. Hollander and J.-M.
Viaene (1998) Applied international trade
analysis, University of Michigan Press
 Abler, D. (2006) Approaches to measuring
the effects of trade agreements, CATPRN
Paper 2006-1
