Accounting for Model Uncertainty and Natural Trading Partner

Effects

Trade Creation and Diversion Revisited:

Christian Henn

Theo S. Eicher Chris Papageorgiou

IMF University of Washington

IMF

2

Road Map Reexamine PTAs’ impacts on trade flows

Introduce New Methodology: Address Model Uncertainty Allow for Natural Trading Partners

Generate New Results: Identify PTAs’ tangible benefits Identify individual PTA’s trade

creation/diversion Highlight relevance of comprehensive

approach

3

A Brief History of Trade Flow Determinants

Ricardo (Technology) Heckscher / Ohlin (Factor Endowments) Brander / Spencer / Krugman (IRS, Market

Structure) Rivera Batiz / Romer / Young (Endogenous

Growth) Melitz/Antras (Intra-Industry Cost

Heterogeneity)

4

Trade Flows and Trade Restrictions

Trail Blazers: Static Models Kruger (1972), Findlay Wellisz (1982), Hillman

(1982)

Second Generation: Endogenous Protection Models Grossman Helpman (1995) etc.

Preferential Trade Agreements (PTAs)

5

Trade Flows and Trade Restrictions

Preferential Trade Agreements (PTAs) Viner (1950) (Static Trade Creation/Diversion)Stepping Stones vs. Stumbling Blocks Krugman (1991, 1993)

PTAs cannot be trade creating in the absence of intercontinental transport costs.

Prohibitive transport costs => PTA trade creation dominates Finite transport costs => PTA trade diversion dominates

Frankel, Stein and Wei (1995, 6, 8) model continuum of transport costs.

The more remote trading partners are from the rest of the world, the more likely they are to form RTAs due to less potential trade diversion.

The more “natural” trading partners are, the more likely an FTA will be formed by the countries’ governments due to more potential trade creation.

6

Trade Flows and Trade Restrictions

Stumbling BlocksLobbies as stumbling blocks, b/c common external tariff (Panagranya & Findlay 1996)

Diverting PTAs are politically more likely (GH 1995)Larger PTAs have monopoly power (Deardorff and Stern 1994)

Given differences in factor endowments, trade with a few countries is sufficient to maximize gains from trade. Deardorff and Stern and Haveman (1994)

PTA’s inhibit further multilateral tariff reduction, (Krishna 1998)

7

Trade Flows and Trade Restrictions

Stepping Stones (“complementary effect of pref. tariffs Bagwell Staiger 1998)preferential tariff induces trade diversion, which is costly, so external tariffs declines to compensate and shift imports back to their original source)

PTAs allow lower external tariff b/c (tariff revenue competition, PTA free trade benefits allow for lower tariff revenues, so external tariff falls. RTA may promote external liberalization, Richardson 1993)

Baldwin (1993, 1995, 1997) & Levy (1996) Non-member exporters lobby forces joining if diversion is

large Larger PTAs shifts more power to the export lobby Larger PTA even more attractive to join (more trade

creation) PTA induced harmonization, allows new revenues that overcome

trade fixed costs Freund (2000)

8

PTA Effects Can Have Many Sources

That Are Difficult to Disentangle

Possible Scenarios: Among members: Trade Creation / Trade Diversion Between members: Trade Diversion / Open Block Between Non Members: Trade Diversion /Open

Block

Which Theory is Empirically Relevant? Which Effect(s) Dominate? How do we sift through the many models for

evidence??

9

PTA Effects Can Have Many Sources

That Are Difficult to Disentangle

Possible Scenarios: Among members: Trade Creation / Trade Diversion Between members: Trade Diversion / Open Block Between non members: Trade Diversion /Open Block

Which Theory is Empirically Relevant? Which effect(s) dominate? How do we sift through the many models for evidence??

This is the DEFINITION of Models Uncertainty

10

Model Uncertainty: Why do we care?

Individual researchers typically emphasize a single model as they seek support for a particular regressor (the alternative is “Null Hypothesis”, or “no effect”)

Inferences procedures based on a single model overstate the precision of the inferences procedures do not account for the

additional uncertainty surrounding the validity of model.

standard errors understate uncertainty

Est. relationship, past studies Positive None Negative

AFTAij 3 2 ANZCERTAij 1 APECij 3 APij 3 2 CACMij 4 2 CARICOMij EEAij EFTAij 3 5 EUij 9 9 LAIAij 4 2 MERCOSURij 2 3

Trade Creation

NAFTAij 1 3 AFTAi 2 1 1 ANZCERTAi APECj APj 1 2 CACMj 2 2 CARICOMi EEAi EFTAi 1 1 EUi 2 1 LAIAi 2 2 MERCOSURi 2 2

Trade Diversion / Open Bloc

NAFTAi 1 2 1

13

Previous Approach To Model Uncertainty In Trade Flow Empirics

Ghosh and Yamarik (2004, JIE) First (and to date only) attempt to

account for model uncertainty in trade flow / PTA estimation

Use Extreme Bounds Analysis (EBA) (Leamer 1978, 1983, 1985)

14

Previous Approach To Model Uncertainty In Trade Flow Theory

Ghosh and Yamarik (2004, JIE) First (and to date only) attempt to account

for model uncertainty in trade flow / PTA estimation

Use Extreme Bounds Analysis (EBA) (Leamer 1978, 1983, 1985)

Conclusion: No evidence of trade creation or diversion for any PTA! Relaxed extreme bounds pick up only trade diverting PTAs!!

15

Extreme Bound Analysis-- Assessment --

Lack of Statistical Theory Backbone Reduce model space artificially to avoid running all

regressions. (Why?) Why has each model (no matter how terrible) equal

weight…? What is the robustness criterion? (see growth theory) How relevant is the analysis if the model space holds

billions of models and the researcher searches only over less than 0.0000001% of the model space?

Severe omitted variable bias

Could be (and has been) considered data mining…

16

Bayesian Model Averaging (BMA)

BMA Intuition Model Selection: estimating the performance of

different models in order to choose the best one. Model Averaging: average predictions from

different models to achieve improved performance Posterior Estimate: weighted average over all

models, where weights are given by model quality. BMA is a) theory based, b) based on objective

criteria, and c) proof exists that BMA delivers best predictive performance (Raftery 1995)

17

Bayesian Model Averaging:Quick details for the math hungry

1

( | ) ( )( | )

( | ) ( )k k

k K

l ll

p D M p Mp M D

p D M p M

( | ) ( | , ) ( | )k k k k k kp D M p D M p M d

D is the data, Mk is a model in some model space M, K is number of models, is a quantity of interest

Nice: priors largely wash out with 15000 observations

• posterior probability for model Mk (the “weight”) is

• is the integrated likelihood (over all regressors)

• Posterior Mean

DMpMDEDE k

K

k K |,||0

18

Econometric Framework Basic Building Block: Gravity Equation

Can be derived from a variety of trade models Deardorff (1998)

Successful in explaining implied trade flows (Frankel Romer 1999)

1 2

3 4 5

log( ) log( ) log( )

+ + + +

ijt it jt ij

ijt ijt it ijt

Trade GDP GDP Distance

Z RTA RTA

a b b

b b b e

= + × +

Zijt: proxies for trade costs / trade theory covariates

19

Definition of PTA Dummies

Trade Creation dummies (PTAij) 1 only if both trade partners are members of a

respective PTA 0 otherwise

Trade Diversion/Open Block Dummies (PTAi) 1 if one and only one trade partner is a member of a

respective PTA 0 otherwise

1 2

3 4 5

log( ) log( ) log( )

+ + + +

ijt it jt ij

ijt ijt it ijt

Trade GDP GDP Distance

Z RTA RTA

a b b

b b b e

= + × +

20

Trade Theory Covariates (Zijt)

Geography Border, Remoteness, Landlocked, Island, Area

Historical Ties Language, Common Colonizer, Colony

Exchange Rate / Trade Policy Sachs dummy, Currency Union, Floating FX rate, FX

volatility Factor Endowments / Development

GDP p.c. Log Differences in

GDP p.c. Education Population Density

21

Data Identical to “no effect” data of Ghosh and Yamarik

(2004) Avg Bilateral Trade Flows, 186 countries, 14,522

observations, 3,420 bilateral trade pairs, five-year intervals (1970-1995)

12 major PTAs: Europe: EU, EFTA, EEA Pacific Rim: APEC, ASEAN, NAFTA, ANZCERTA Latin America: CACM, CARICOM, LAIA, AP

(Andean Pact), MERCOSUR

22

European PTA MembersAbbreviation Name of RTA Start Member countries EEA European Economic

Area 1994 Austria, Belgium, Denmark, Finland, France, Germany,

Greece, Luxembourg, Iceland, Italy, Ireland, Liechtenstein, Netherlands, Norway, Portugal, Spain, Sweden, United Kingdom.

EFTA European Free Trade Arrangement

1960 Iceland, Liechtenstein (1991), Norway (1986), Switzerland Former: Denmark (1960-72), United Kingdom (1960-72), Portugal (1960-85), Austria (1960-94), Sweden (1960-94), Finland (1986-94).

EU European Union 1958 Austria (1995), Belgium, Denmark (1973), Finland (1995), France, Germany, Greece (1981), Luxembourg, Ireland (1973), Italy, Netherlands, Portugal (1986), Spain (1986), Sweden (1995), United Kingdom (1973).

23

Results: Europe Specification 1

Time Fixed Effects

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Trade Creation EEAijt 0.01 0.26 0.19 29% EFTAijt 0.00 0.02 0.26 2% EUijt 0.00 0.03 0.14 3% Trade Diversion Open Block EEAit 0.00 0.01 0.08 1% EFTAit 1.00 0.35*** 0.05 43%

EUit 1.00 0.56*** 0.04 75%

Clear, strong open Block Effects.

24

Pacific Rim PTA membersAbbreviation Name of RTA Start Member countries ANZCERTA Australia – New

Zealand Closer Economic Relations Trade Agreement

1983 Australia, New Zealand

APEC Asia Pacific Economic Community

1989 Australia, Brunei, Canada, China (1991), Chile (1994), Taiwan (1991), Hong Kong (1991), Indonesia, Japan, South Korea, Malaysia, Mexico (1993), New Zealand, Papua New Guinea (1993), Peru (1998), Philippines, Singapore, Thailand, United States, Vietnam (1998).

ASEAN Association of South East Asian Nations

1967 Brunei (1984), Cambodia (1998), Indonesia, Laos (1997), Malaysia, Myanmar (1997), the Philippines, Singapore, Thailand, Vietnam (1995).

NAFTA Canada-US Free Trade Arrangement / North America Free Trade Agreement

1988 Canada, United States, Mexico (1994).

25

Results: Pacific Rim

Specification 1

Time Fixed Effects

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Trade Creation AFTAijt 0.00 -0.22 0.54 -20% ANZCERTAijt 0.01 0.89 0.96 144% APECijt 1.00 1.48*** 0.15 338% NAFTAijt 0.01 -0.89 0.84 -59% Trade Diversion, Open Bloc AFTAit 0.03 0.17 0.11 19% ANZCERTAit 1.00 -0.47*** 0.10 -37% APECit 1.00 0.55*** 0.06 73%

NAFTAit 1.00 -0.63*** 0.10 -47%

Now significant trade creation and diversion and open block

PTA’s matter a lot!

Puzzling high APEC…

Large Trade Diversion for Nafta!

26

Latin American PTA Membership

Abbreviation Name of RTA Start Member countries AP Andean Community /

Andean Pact 1969 Bolivia, Colombia, Ecuador, Peru, Venezuela (1973),

Former: Chile (1969-76) CACM Central American

Common Market 1960 Costa Rica (1963), El Salvador, Guatemala, Honduras,

Nicaragua. CARICOM Caribbean

Community/ Carifta 1968 Antigua and Barbuda, Bahamas (1983), Barbados, Belize

(1995), Dominica (1974), Guyana (1995), Grenada (1974), Jamaica, Montserrat (1974), St. Kitts and Nevis, St. Lucia (1974), St. Vincent and the Grenadines, Suriname (1995), Trinidad and Tobago.

LAIA Latin America Integration Agreement

1960 Argentina, Bolivia (1967), Brazil, Chile, Colombia (1961) Ecuador (1961), Mexico, Paraguay, Peru, Uruguay, Venezuela (1966).

MERCOSUR Mercado Comun Del Sur

1991 Argentina, Brazil, Paraguay, Uruguay

27

Results: Latin America

Time Fixed Effects

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Trade Creation APijt 0.01 -0.05 0.27 -5% CACMijt 1.00 2.25*** 0.23 848% CARICOMijt 1.00 2.08*** 0.41 702% LAIAijt 0.91 0.46*** 0.13 58%

MERCOSURijt 0.12 1.66 0.70 424% Trade Diversion, Open Bloc APit 0.52 -0.19* 0.06 -17% CACMit 0.85 -0.18** 0.05 -17% CARICOMit 1.00 -0.74*** 0.07 -52% LAIAit 1.00 -0.40*** 0.07 -33%

MERCOSURit 0.79 0.42** 0.12 52%

Orthodox results ofStrong trade creationAnd trade diversionIn Latin America

28

Summary part I Using:

identical data statistically sound approach to model uncertainty considers all possible models (including best

model) Derive quality-weighted averaged estimates

We overturn the bleak “PTAs don’t matter“ results and show that several have trade creating, trade diverting and open block effects.

29

Results: Control Variables

Specification 1 Specification 2

Time Fixed Effects

Time & Country-Pair Fixed Effects

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

log(DISTANCEij) 1.00 -1.19*** 0.02 -1% 0% log(GDPit GDPjt) 1.00 0.88*** 0.01 1% 1.00 1.13*** 0.05 1%

Core Gravity log(gdpit gdpjt) 1.00 0.55*** 0.02 1% 0.02 0.16 0.10 0%

SACHSit+SACHSjt 1.00 0.35*** 0.03 42% 1.00 0.13*** 0.03 14% CUijt 1.00 1.40*** 0.29 307% 0.01 -0.64 0.67 -47% FLOATijt 0.00 -0.01 0.02 -1% 0.20 -0.05 0.02 -5%

Economic Policy

VOLATILITYijt 0.25 0.006 0.002 0% 0.01 -0.002 0.002 0% abs(gdp_DIFF) 1.00 0.18*** 0.02 0% 1.00 -0.31*** 0.05 0% abs(DENS_DIFF) 1.00 0.23*** 0.01 0% 0.75 0.25** 0.09 0%

Development, Factor Endowment abs(SCHOOL_DIFF) 0.01 0.02 0.02 0% 0.03 -0.06 0.04 0%

BORDERij 1.00 0.53*** 0.10 69% REMOTEij 1.00 342*** 39.79 342% LANDLOCKij 1.00 -0.42*** 0.04 -34% log(AREAi AREAj) 0.92 -0.03** 0.01 0%

Geography

ISLANDij 0.02 -0.05 0.03 -5% COMLANGij 1.00 0.47*** 0.05 60% COMCOLij 1.00 0.77*** 0.07 116%

History COLONYij 1.00 1.44*** 0.12 320%

Strong effects from gravityand other country specific

Controls

The only variables that do not receive support are

- FLOAT- ED DIFF- ISLAND

30

The Importance of Fixed Effects

(Natural Trading Partners) Hummels and Levinsohn (1995)

Trade is largely specific to country-pairs We may not know why, or cannot account for it

with our covariates, no matter how many controls we include.

We introduce country pair fixed effects Isolate whether countries trade a lot with each

other because of PTA’s or because they are natural trading partners

Cheng and Wall, 1999 (EU, Nafta, Mercsur); Egger and Pfaffermayr, 2003 (gravity only)

31

The Importance of Fixed Effects

(Natural Trading Partners) Introduce Country Pair Fixed Effects

Example: similarities in economic and social institutions, such as corruption or rule of law, or simple economic infrastructure such as telecommunications.

A concrete example would be France-Germany, with excellent transport links (unobserved) and a PTA (observed), vs. Afghanistan-Kazakhstan, with bad transport links (unobserved) and no PTA (observed).

ij

ijtitijtijtjtitijtijt PTAPTAZYYT 5431

~loglog

ijtitijtijtjtitijtijt PTAPTAZYYT 5431

~loglog

32

EuropeTrade Effects Purged of Natural Trading Partner Effects

Now EU is Trade Creating Strong Open Block Effect

Specification 1 Specification 2

Time Fixed Effects Time & Country-Pair Fixed

Effects

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Trade Creation EEAijt 0.01 0.26 0.19 29% 0.00 0.06 0.17 6% EFTAijt 0.00 0.02 0.26 2% 0.03 0.52 0.32 69% EUijt 0.00 0.03 0.14 3% 1.00 0.66*** 0.17 93% Trade Diversion Open Block EEAit 0.00 0.01 0.08 1% 0.20 -0.14 0.06 -13% EFTAit 1.00 0.35*** 0.05 43% 0.98 0.26** 0.08 29%

EUit 1.00 0.56*** 0.04 75% 0.21 0.16 0.07 17%

33

EuropePTA Flows Purged of Natural Trading Partner Effects

Only after controlling for natural trading pairs, the EU is Trade Creating EU countries naturally under-trade relative

to the prediction of the standard gravity model. This “missing trade” is a standard feature of the basic gravity model, see e.g. Pollak (1996) or Rose (2004).

EU Open Block Trade Creation does not survive but is simply due to similarities among trading partners

34

Time Fixed Effects Time & Country-Pair Fixed

Effects

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Trade Creation APijt 0.01 -0.05 0.27 -5% 0.02 0.56 0.37 74% CACMijt 1.00 2.25*** 0.23 848% 0% CARICOMijt 1.00 2.08*** 0.41 702% 0.00 0.12 0.71 13% LAIAijt 0.91 0.46*** 0.13 58% 0%

MERCOSURijt 0.12 1.66 0.70 424% 0.00 0.37 0.51 45% Trade Diversion, Open Bloc APit 0.52 -0.19* 0.06 -17% 0.03 -0.17 0.10 -15% CACMit 0.85 -0.18** 0.05 -17% 0% CARICOMit 1.00 -0.74*** 0.07 -52% 0.39 -0.29 0.11 -25% LAIAit 1.00 -0.40*** 0.07 -33% 0%

MERCOSURit 0.79 0.42** 0.12 52% 0.01 0.11 0.09 11%

Latin America PTA Flows Purged of Natural Trading Partner Effects

Counterintuitive Effects Disappear All Effects are Natural Trading Partner Effects

35

Pacific Rim PTA Flows Purged of Natural Trading Partner Effects

Time Fixed Effects Time & Country-Pair Fixed

Effects

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Post. Incl.

Prob.

Post. Mean

Post. Std. Dev.

Post Mean in %

Trade Creation AFTAijt 0.00 -0.22 0.54 -20% 0.00 -0.22 0.39 -19% ANZCERTAijt 0.01 0.89 0.96 144% 0.00 0.30 0.92 35% APECijt 1.00 1.48*** 0.15 338% 0.01 0.14 0.12 16% NAFTAijt 0.01 -0.89 0.84 -59% 0.00 0.60 0.65 81% Trade Diversion, Open Bloc AFTAit 0.03 0.17 0.11 19% 1.00 0.40*** 0.08 49% ANZCERTAit 1.00 -0.47*** 0.10 -37% 0.10 -0.18 0.09 -17% APECit 1.00 0.55*** 0.06 73% 1.00 0.23*** 0.05 26%

NAFTAit 1.00 -0.63*** 0.10 -47% 0.99 -0.31*** 0.08 -26%

Counter Intuitive APEC Trade Creation vanishesAFTA and APEC Trade Creation confirmed

NAFTA Trade Creation is maintained but smaller

!

36

Pacific Rim PTA Flows Purged of Natural Trading Partner Effects

NAFTA 26% trade diversion?

37

Trade Creation/Diversion Importance of Comprehensive Approach

Most PTA Studies estimate Trade Creation for one/a few PTAs Marginal Effects (too diverse to average) vs Global Effects

Most studies don’t distinguish Trade Creation/Open Block

Example1: Two countries are in two different PTAs: what matters for trade flows is the Open Block effect / Trade Diversion of BOTH PTAs (70 countries, most high income countries are in PTAs)

Example2: All countries (PTA or not) get open block effects from all other PTA’s. So: pairwise trade diversion may turn into multilateral trade creation!

38

Trade Creation/Diversion

Most PTA interactions are Trade Creating EVEN NAFTA trade can be trade creating with Nafta non members

NAFTA/APEC ANZCERTA/APEC AFTA/APEC APEC EU EFTA

NAFTA 0%

ANZCERTA -26% 0%

AFTA 10%

=((1+49%)*(1-26%))-1 49% 0%

Rest of APEC -26% 0% 49% 0%

EU -7%

=((1+26%)*(1-26%))-1 0%

87%

=((1+26%)*(1+49%))-1 26% 93%

EFTA 20%

=((1+29%)*(1+26%)* (1-26%))-1

63%

=((1+26%)*(1+29%))-1

142%

=((1+26%)* (1+49%)*(1+29%))-1

63%

=((1+26%)*(1+29%))-1 29% 0%

Rest of the World -7%

=((1+26%)*(1-26%))-1 26%

87%

=((1+26%)*(1+49%))-1 26% 0% 29%

39

Conclusions 1) Address Model Uncertainty:

obtain correct PTA’s effects: PTA’s do matter (!) 2) Purge Natural Trading Partner contamination.

Eliminate Counterintuitive Results (!) Isolate actual PTA effects: (most PTA effects

smaller) 3) Comprehensive Approach

Identifies crucial interactions that can overcome “false trade diversion” (US / ASIA)

Overwhelmingly, PTAs are stepping stones to freer trade