General Equilibrium Modelling and Trade

Policy AnalysisMarco FugazzaDITC, UNCTAD

15 September 2006

2

Outline

• Why are economic models needed?• What kinds of models are commonly used

for trade policy analysis?• Basics of CGE Modelling• What is involved in a policy simulation?• What should we know of trade

liberalization CGE simulations?• An Application• How can modeling better assist policy

making?

3

Why are economic models needed?

4

Why are economic models needed?

• 1. “Without theory, practice is but routine born of habit.”

• 2. “(S)He who loves practice without theory is like the sailor who boards ship without a rudder and compass and never knows where he may cast.“

3. “Being denied a sufficiently secure experimental base, economic theory has to adhere to the rules of logical discourse and must renounce the facility of internal inconsistency. A deductive structure that tolerates a contradiction does os under the penalty of being useless since any statement can be derived flawlessly and immediately from that contradiction. In its mathematical form, economic theory is open to an efficient scrutinity for logical errors.”

• 4. “In attempting to answer the question ‘Could it be true?’, we learn a good deal about why it might not be true.”

5

• 1. “Without theory, practice is but routine born of habit.” ~ Louis Pasteur

• 2. “(S)He who loves practice without theory is like the sailor who boards ship without a rudder and compass and never knows where he may cast.“ ~ Leonardo da Vinci

• 3. “Being denied a sufficiently secure experimental base, economic theory has to adhere to the rules of logical discourse and must renounce the facility of internal inconsistency. A deductive structure that tolerates a contradiction does os under the penalty of being useless since any statement can be derived flawlessly and immediately from that contradiction. In its mathematical form, economic theory is open to an efficient scrutinity for logical errors.” ~ Gérard Debreu (Nobel Prize winner,1983)

• 4. “In attempting to answer the question ‘Could it be true?’, we learn a good deal about why it might not be true.” ~ Kenneth Arrow (Nobel Prize winner, 1972)

6

Why are economic models needed?

• The use of economic models should help improve policymaking. How?

1. They provide a theoretically consistent framework for analyzing trade policy questions

2. Models can provide a handle on complicated questions

3. Models can help give greater intellectual support for a chosen trade policy

4. The use of models can provide a common “language” for policy discourse or debate

5. But models should complement rather than substitute for policy making

7

Models commonly used for trade policy

analysis

8

Models used for trade policy analysis

• Simulation models: they help answer “What if” types of questions (+ projections): Partial Equilibrium Models, General Equilibrium models

• Econometric Models : – gravity models: reduced form: can be used to

establish whether certain economic variables have an effect on a variable of interest (Does GSP increase trade?)

– Macro-econometric models: tools for projections of aggregates but no info on the industrial structure of the economy + may lack micro-foundations

• Simulation (econometric) models are deterministic (stochastic)

9

A Partial Equilibrium Analysis

Price

Wheat

Pw

Pw(1+t)

Impact of wheat market on rest of the economy can be neglected

DD

DS

10

A General Equilibrium Analysis

Households Firms

Factor services of production

Factor incomes

InvestmentsSavings

spending on goods and services

goods and services

REST OF THE WORLD

impo

rts

expo

rts

FDI

11

• Nature of policy change

– Does it cut across many markets or sectors?• Potential impact of change

– Are there economy-wide impacts?• Constraints imposed by availability of data and

resources (financial and skills)

– PE data and models: free– CGE data: single country (SAM) could be

free, multiple country (GTAP: from $ 360 to $ 4600)

– CGE models: free (GTAP) but may need software for mathematical programming to run (LINKAGE, MIRAGE)

GE or PE analysis?

12

Basics of CGE Modelling

13

A Typology of CGE Modeling

Static: regions, sectors, factors, economic agents

+ set of economic behaviors & relationships

Dynamic=Static features+ explicit inter-temporal features

Micro-Simulation Models: representative

agents hypothesis “removed”

14

• Input Output Economics & SAMs• Behavioral Relationships/ Agents

– Supply– Demand– Trade

• Government• Pricing and Imperfect substitutes• Policy – tax equivalents• Closure

– Accounting identities– Endogenous/exogenous variables– Macroeconomic assumptions– Exchange rate determination

• Solution– Equilibrium– Linearization– Percent change variables

CGE Standard Model Elements I

15

CGE Standard Model Elements II

• Calibration/Benchmarking• Aggregation

– Agents– Goods/Sectors

• Experiments– Welfare Measures– Projections– What if

• Extensions– Imperfect Competition, IRS– Product Differentiation– Dynamics

• Results Comparisons

16

Input-Output economics & SAMs

• Production= Intermediates + Value Added

• Production= Intermediate demand + Final Demand

• +• Macroeconomic accounting identities to

capture income flows, tax incidence, trade and payments, and savings-investment balances

• = > SAMs capture `circular flow’ of income and expenditure

17

Households Firms

Factor services of production

Factor incomes

InvestmentsSavings

spending on goods and services

goods and services

REST OF THE WORLD

impo

rts

expo

rts

FDI

Input-Output economics & SAMs

18

• Whether neoclassical, strucuralist, neo-Keynesian, or Monetarist, a CGE modeler must respect accounting identities and equilibrium conditions. Hence, most applied work is based on a social accounting matrix to benchmark (calibrate) a model and to represent relevant accounting identities.

• SAMs capture equilibrium conditions

• Walras’ law applies

Input-Output economics & SAMs

19

Decision Making and Institutions

• Linkages in SAMs are accounted for by modelling the decision-making process of the firm, the consumer, as well as other economic agents and institutions: production and demand structure

• Trade results from that decision-making processes and their interaction with institutions:

• Production- Exports + Imports=Consumption

20

Closing the Model

• Need to define a numéraire (walras law allows to “drop” one market)

• Assumption about the adjustment mechanism in factor and commodity markets

• Macro closure– Macro accounting balance (gvt expenditure

and deficit; aggregate saving and investment; balance of trade and -real- exchange rate)

– Macro adjustment mechanism (exogenously determined)

21

• Johansen closure: investment is exogenous and consumption is the adjustment variable

• Keynesian closure: nominal wage is fixed and employment is the adjustment variable (unemployment)

• Kaldorian closure: wages could be less or equal to the marginal product of labor (exploitation of labor model)

• Classical closure: prices and wages are the adjustment variables (constant employment) and investment becomes endogenous and adjusts to total savings available

• Foreign borrowing (Robinson): trade balance is endogenous, current account and hence net capital inflows are the adjustment variable

Closing the Model

22

Beyond the Standard Model• Economies of scale, monopolistic competition and

differentiated products• Institutional features of a particular economy (e.g.

tax collection costs)• Specific features of a policy instrument• Increase effort on estimation of substitution

elasticities• Dynamics to account for dynamic aspects (policy

credibility; capital accumulation; FDI; knowledge accumulation and spillovers) and adjustment

• Account for the extensive margin of trade (the “small-shares” issue)

23

CGE Dynamic Models• Recursive:

– solves annually– Current economic conditions (e.g. the

availability of capital) are dependent on past outcomes but are unaffected by forward looking expectations

– Linked with a macro econometric to include exogenously projected changes in demographic trends or in technology: baseline scenario

– Impact of policy change is given with respect to the baseline scenario (sector specific TFP and real GDP growth are solved endogenously)

24

• Forward looking:– Ramsey model, OLG, Infinite lived consumer

with financial market– No extensive baseline scenario: trade

performance-productivity linkage + gvt investment on infrastructure and TFP linkage + investment in education and labor productivity linkage

– Could account for transitionary disequilibrium states (true adjustment process?)

CGE Dynamic Models

25

Micro-Macro Models• Combination of a Micro Simulation model (base

on Household surveys: fiscal and labor) and a CGE model

• Ideal to assess the impact of macroeconomic (trade) policies and shocks on poverty/ inequality: MAMS (maquette for MDG simulation)

• Two types of combination: – Fully-integrated: the household model built

directly into the CGE : CGE model with heterogeneous agents (high complexity)

– Sequential (top-down): CGE simulation results are passed on to an household model (macro and micro need not to be reconciled but possible lack of coherence)

26

Micro Simulation Models• Micro-accounting models: “the day after”

approach– Capture 1st order effects– No behavioral response

• Behavioral Models– Capture 2nd order effects– Behavioral response– Reduced form or Structural

• Dynamic versus Static models– Dynamic is usually associated with aging of

information– Dynamic could be behavioral

27

What is involved in a policy simulation?

28

What is involved in a policy simulation?

Economy before

trade policy change

Economy after

trade policy change

Difference between the two is attributed to policy change

Policy change

29

What is needed for a policy simulation?

Inputs OutputsMODEL / Closure

30

What are the inputs?

• Baseline data:– trade flows – levels of protection – input-output structure: national income

aggregates

• Measure of responsiveness of economic agents to price changes (i.e. elasticities)

• Policy - negotiating scenario– Sectors (Agriculture, NAMA, etc.)– Depth of liberalization

31

What are the Outputs?• Configuration of the economy after policy

change• Overall income gains/losses from policy

change• Sources of income gain

– Sectoral (agriculture vs. NAMA)– Policy instrument (market access or

domestic support)• Winners or losers (at the country level)• Changes in pattern and volume of trade and

income• “Story” to explain how inputs and model

combine to determine the output/outcome

32

Tracing Differences in Results

• Deterministic – outcome is completely determined by choice of inputs and model (no “residuals”)

Inputs + MODEL Outputs

Differences in simulation results = differences in choice of inputs and model/closure

“Story” must explain why the choice of inputs and model is appropriate/optimal for the policy question of interest

33

Towards an “objective” look at trade liberalization CGE simulations

34

Doha Round CGE Simulations

• Common results:• Multilateral liberalization is beneficial

at the global level• There are potential gains for

developing countries • Developing countries own

liberalization is an important source of their gains

• Removing subsidies may damage net food importer countries

35

Doha Round CGE Simulations

• Results differ among studies on how gains are redistributed

• 1. What share of the benefits goes to developing countries?

• 2. What share comes from agriculture liberalization? From NAMA?

36

What are the gains for developing countries?

0

500

1000

1500

2000

2500

Carnegie, 2006 World Bank,2006

WorldBank,2003

MichiganModel, 2003

Developing countries Developed Countries

Full liberalization scenario. Million $ 1997

30% 55%46% 21%

37

What share of the gains comes from agriculture

liberalization ?

-20%

0%

20%

40%

60%

80%

100%

Carnagie, 2006 World Bank,2006

World Bank,2003

MichiganModel, 2003

Agriculture Manufacturing Services

Results based on the full liberalization scenario

38

How important are assumptions?

• Examine one scenario with differing assumptions

• Scenario 50% cut in all tariffs and subsidies– Standard closure– Fixed trade balance– Fixed wages (unskilled unemployment)

in developing countries– Double trade elasticities (substituability

between domestic and foreign products)

39

Global welfare by closure

0

50

100

150

200

250

300

$b

Standard TB 2xArm Labour

Scenario

Source: UNCTAD

40

Fixed Trade Balance Welfare compared with standard closure

-30

-20

-10

0

10

20

30

40

50

$b

Scenario

Source: UNCTAD

Japan

USA

41

Trade elasticities Welfare compared with standard closure

0

5

10

15

20

25

30

35

$b

Scenario

Source: UNCTAD

Japan

42

Fixed wages Welfare compared with standard closure

0

10

20

30

40

50

60

70

80

90

$b

Scenario

Source: UNCTAD

China

43

Other assumptions we ignore

• CRTS• Perfect competition• Static/dynamic• Technology• Productivity

• Key elasticities e.g. K/L substitutability

• Aggregation

44

An Application: The potential for South-

South Trade

45

Setting a Policy Simulation

• Qualifying the general issue of interest: “what is at stake ?” analysis

• Choice of aggregation• Model specification and Closure• Choice of scenario to be simulated• Presentation of the results• Interpretation of the results

46

What is at stake?

Developed Developing Least developed

% % %

Source

Developed 2.1 9.2 11.1

Developing 3.9 7.2 14.4

Least developed 3.1 7.2 8.3

Total 2.9 8.1 13.6

Trade weighted average applied tariffs (inc. preferences) by development status

Source: Computed from TRAINS/WITS (2004)

47

Aggregation• South-South trade is the focus:

– keep as many southern countries as possible (21)

– Identify those sectors with the highest protection among developing countries (20) and/or with little access to norther markets

– Adjust the country-groups selection in accordance

• General hints:– Aggregation is usually 20*20 (max 30*30)– Different aggregations affect differently the

level of protection an distortion that will characterize the simulation exercise (could hide/highlight gains and losses ) and thus the expected gains from the policy simulation

48

Model and Closure• Model characteristics are likely to depend on

your CGE skills• Standard GTAP is widely used but more and

more imperfect competition in manufactures• Still prevalence of static models because of

high computational resources required for the baseline scenario in recursive models

• No a priori concerning the favorite closure• With a focus on developing countries fixed

wage (flexible employment) for unskilled labor could be sensible

• Standard non-standard: fixed trade balance for all countries but the USA

49

Choice of Scenario• Pre-simulation to account for the aging of data

and policy (e.g. China accession to the WTO, end of ATC)

• Political Relevance: Potential of south-south trade relevant in the context of GSTP negotiations (43 countries + opening to the group of 77 + China)

• If not related to “realistic” scenarios (Doha negotiations) identify sector and country relevant scenarios (manufactures, agriculture and overall liberalization or tariff cuts/ across different regions)

• In general tariff cuts are simulated together with export taxes and subsidies

• Compensation of changes in tariff revenues (e.g. income is made endogenous) could be a major concern for developing countries

50

Presentation of Results• Present both absolute variations and

proportional variations with respect to relevant initial values (welfare as a percentage of initial GDP)

• Present the results of the same scenario obtained with at least one different closure from the favorite one (annex)

• Results could be presented in aggregate form for sake of clarity with reference o fully disaggregated results in the text

• Present the results obtained in a benchmark simulation: usually full trade liberalization

51

Interpretation of Results• Are you sure you understand what is going on?

(should be the case if good preliminary analytical work done for aggregation)

• Results are your “story” and must reflect a good coherence between your scenario(s) and the various component of your modeling approach

• Multiple scenarios and comparative analysis are less sensible to modeling specificities: everything is relative

• Simulations of a specific agreement/policy scenario must be based on the most “realistic” computational framework

• Make sure that what your telling is in line with your assumptions (e.g. do not talk about changes in labor demand when employment is assumed to be fixed)

52

How can modelling better assist policy making?

• A. Improving data– trade transaction costs – more disaggregated sectors– better disaggregation of regions/countries– protection in services– TNC activities: FDI vs. Outsourcing– Households surveys to account for poverty

and inequality impact

53

.... Cont’d• B. Improving modelling of:

– Better treatment of services– Adjustment costs– Functioning of factor (labour) markets– Tariff revenue implications of trade

liberalization– Extensive margin of trade (potential for

diversification + productivity gains)• C. Improving confidence in simulation

results– Sensitivity analyses: within and across

models– Ex-post verifications– Use also focused models

54

“Demystifying Modelling Methods for Trade Policy”, Roberta Piermartini and Robert Teh, Discussion Paper No. 10, World Trade Organization, Geneva, Switzerland, September 2005 (http://onlinebookshop.wto.org/shop/article_details.asp?Id_Article=661)

“Structure of GTAP” ,Thomas W. Hertel and Marinos E. Tsigas, Chapter 2 in T.W. Hertel (ed.), Global Trade Analysis: Modeling and Applications, Cambridge University Press, 1997. (https://www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=413)

Global Trade Analysis Project (GTAP):http://www.gtap.agecon.purdue.edu/

“LINKAGE Technical Reference Document”, Dominique van der Mensbrugghe, DECPG, World Bank, December 2005 (http://siteresources.worldbank.org/INTPROSPECTS/Resources/334934-1100792545130/LinkageTechNote.pdf)

References

55

ReferencesMichigan Model of World Production and Trade:http://www.fordschool.umich.edu/rsie/model/description.html

“Mirage, a Computable General Equilibrium model for Trade Policy Analysis”, Bchir E., Y. Decreux, J-L. Guérin, S. Jean, CEPPI http://www.cepii.fr/anglaisgraph/workpap/pdf/2002/wp02-17.pdf

World Scan Dynamic Model of the World of the Netherlands Bureau of Economic Policy analysis (CPB):http://www.cpb.nl/nl/pub/bijzonder/20/bijz20_c.pdf

Harrison/Rutherford/Tarr Multi-Regional Global Trade Model:http://dmsweb.badm.sc.edu/Glenn/ur_pub.htm