Johansen's legacy to CGE modelling: Originator and guiding light for 50 years

Accepted Manuscript Title: Johansen’s legacy to CGE modelling: originator and guiding light for 50 years Author: Peter B. Dixon Maureen T. Rimmer PII: DOI: Reference:

S0161-8938(16)00019-3 http://dx.doi.org/doi:10.1016/j.jpolmod.2016.02.009 JPO 6262

To appear in:

Journal

of

Policy

Modeling

Please cite this article as: Dixon, P. B., and Rimmer, M. T.,Johansen’s legacy to CGE modelling: originator and guiding light for 50 years, Journal of Policy Modeling (2016), http://dx.doi.org/10.1016/j.jpolmod.2016.02.009 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ELSEVIER MANUSCRIPT TRANSMITTAL FORM –Please link to item group IG002855 SI: CGE

Modelling (PAPER 2) Name of Journal: Journal of Policy Modeling Editor ref. no.:

ip t

Special Issue: Yes

cr

Manuscript title: Johansen’s legacy to CGE modelling: originator and guiding light

Disc enclosed: No, Submitted electronically

an

Name of first author as on manuscript: Peter Dixon

us

for 50 years

ed

PHONE: (+61 3) 9919 1464 FAX: EMAIL: [email protected]

M

CORRESPONDING AUTHOR/ADDRESS: Peter Dixon

Number ms. Pages: 32 (include every page in count before listing items below)

ce pt

Number of Figures: 0

Number of Figure Captions: 0 Number of Tables: 0

Ac

Number of Colour Illustrations: 0

Keywords: Listed on Manuscript: CGE modelling; Leif Johansen Editor’s notes: None

Editor’s Name

o Yes X No Media format: o PC o Mac o WordPerfect X Microsoft Word o TeX o Other___________ Publication item type: o ABS Abstract o ADD Addendum o ANN Announcement o BRV Book Review o CAL Calendar o CNF Conference paper o COR Correspondence o DIS Discussion o EDI Editorial o ERR Erratum 0 FLA Full length article o IND Index o LIT Literature alert o MIS Miscellaneous o NWS News o PNT Patent report o PRV Product review o REV Review article o SCO Short communication o SSU Short survey

date

Page 1 of 33

Johansen’s legacy to CGE modelling: originator and guiding light for 50 years

ip t

by Peter B. Dixon and Maureen T. Rimmer Centre of Policy Studies, MonashUniversity

us

cr

March 27, 2012

an

Abstract

ed

M

Fifty years ago the Norwegian economist, Leif Johansen, gave us what is now recognised as the first CGE model. While Johansen was first, he is not the father of the whole field. CGE modelling in different styles sprang largely independently from several sources. This paper describes Johansen’s model and how his style of CGE modelling took root in Australia in the 1970s and from there spread to the rest of the world. Today, thousands of economists from nearly every country are undertaking Johansen-style CGE modelling to elucidate policy questions in trade, taxation, environment, labour markets, immigration, income distribution, technology, resources, micro-economic reform and macro stabilization.

ce pt

Key words: CGE modelling; Leif Johansen JEL codes: C68; B23; B31

Ac

Contact author: Peter B. Dixon Email: [email protected]



This is an abridged version of Dixon and Rimmer (2010).

2

Page 2 of 33

Johansen’s contribution to CGE modelling: originatorandguiding light for 50 years

Peter B. Dixon and Maureen T. Rimmer

cr

Centre of Policy Studies, MonashUniversity

ip t

by

us

1. Introduction

an

Johansen (1960a) is one of those rare books that is a complete break with previous literature and the first contribution to a major branch of economics,

M

computable general equilibrium (CGE) modelling.

CGE belongs to the economy-wide class of models, that is, those that provide

ed

industry disaggregation in a quantitative description of the whole economy. The original economy-wide model was Leontief’s input-output system (Leontief, 1936, Following Leontief, the next stage of economy-wide modelling was the

ce pt

1941).

programming models of Sandee (1960), Manne (1963) and others. Input-output and programming models lacked clear descriptions of the behaviour of individual agents.

Ac

In input-output modelling, the economy organized production of each commodity (the vector X) to satisfy a vector of final demands (the vector Y) with given technology specified by the input-output coefficient matrix (A). In programming models, the economy organized production to maximize a welfare function subject to Leontief’s technology specification and subject to constraints on the availability of primary factors.

3

Page 3 of 33

What Leif Johansen did was identify behaviour by individual agents. This is the defining feature of CGE models that distinguishes them from the earlier economywide models.1Households in Johansen’s model maximize utility subject to their budget constraint. Industries choose inputs to minimize costs subject to production-

ip t

function constraints and the need to satisfy demands for their outputs. Capitalists allocate capital between industries to maximize their returns. The overall outcome for

cr

the economy is determined by the actions of individual agents co-ordinated through

us

price adjustments that equalize demand and supply in product and factor markets. Johansen conception proved remarkably popular and versatile. In the 50 years

an

since the publication of his book, CGE modelling has attracted a huge group of researchers. The GTAP2 network alone connects 7,500 CGE practitioners in 150

M

countries. The issues to which CGE modelling has been applied include

macro,

welfare,

industry,

regional,

labour-market,

distributional

and

ce pt

environmental variables of

ed

the effects on

taxes, public consumption and social security payments; tariffs and other

Ac

interferences in international trade; environmental policies; technological change; international commodity prices; interest rates; wage setting arrangements and union behaviour; mineral discoveries (the Dutch disease); immigration; microeconomic reform; and major projects.

1

Definitions of CGE modelling can be found in Bergman (1985: 137) and Dixon and Parmenter (1996: 5) 2 This is the Global Trade Analysis Project centred at Purdue. We discuss GTAP in section 5.

4

Page 4 of 33

While most of these issues have been analysed in single-country, single-period models, there are now numerous CGE models which are either multi-regional or multi-period (dynamic) or both. By going multi-regional, CGE modelling has thrown light on both intra-country and inter-country regional questions. By going dynamic,

ip t

CGE modelling can broaden and deepen its answers to all the questions with which it has been confronted. It has also re-entered the forecasting arena, rather belatedly

cr

following the lead of Johansen and his colleagues (Johansen 1974, ch. 10; Bjerkholt

us

and Tveitereid 1985; and Schreiner and Larsen 1985). Early CGE modellers including Johansen (1960a, 1974), Hudson and Jorgenson (1974), Adelman and Robinson

an

(1978) and Taylor et al. (1980) gave their results a real time dimension, either historical or future. The subsequent generation of modellers worked mainly in a

M

comparative static framework. As described by Johansen in connection with projections made with his model in 1969 by the Norwegian Ministry of Finance for

ed

the period 1963 to 1990, forecasts generate interest and constructive feedback: “Several institutions, enterprises and persons approached the Ministry of Finance

ce pt

to get more details or suggest alternative assumptions. There was a clear need to continue the explorations and the Ministry invited interested persons to take part in an informal working group. Some 30 persons representing organisations,

Ac

research institutes, private firms and other ministries participated. This group met regularly with members of staff of the Ministry in 1970.” [Johansen 1974: 225-6].

CGE models are now used to generate forecasts of the prospects for different industries, labour force groups and regions. These forecasts feed into investment decisions by private and public sector organizations.

5

Page 5 of 33

This paper is concerned with the Johansen legacy. In section 2, we describe the distinctive features of his 1960 model. We review how he used the model and what he found out. Our main aim in that section is to define what we mean by Johansen-style CGE modelling. Section 3 looks at some other early styles of CGE Sections 4 and 5 show how Johansen-style modelling took root in

ip t

modelling.

Australia and then spread to the rest of the world. Section 6 contains concluding

cr

remarks.

us

2. The Johansen approach

Johansen built his model of Norway around a 22-industry input-output table

an

for 1950. He derived input-demand equations by assuming that industries are price-

M

takers that choose their inputs to minimize the cost of producing any given level of output subject to a production-function constraint reflecting Cobb-Douglas

ed

substitution between capital and labour, Hicks-neutral primary-factor-saving technical change and Leontief treatment of intermediate inputs. Imported inputs in Johansen’s

ce pt

model are either identical to domestic inputs or treated as non-competitive. Johansen assumed that households choose their consumption of each commodity to maximize an additive utility function subject to their budget constraint. Following the lead of

Ac

his senior colleague, Ragnar Frisch (1959), Johansen exploited additivity to dramatically simplify the estimation of own- and cross-price elasticities. Among the empirically important details that Johansen capsulated in specifying production and consumption behaviour is the distinction between buyers’ prices (that motivate consumption decisions) and sellers’ prices (that motivate production decisions): he modelled the gap between the two by introducing a trade-services commodity (a combination of transport, wholesale and retail margins). In handling investment, Johansen distinguished between maintenance and expansion.

6

He modelled

Page 6 of 33

maintenance as proportional to existing stocks of structures and equipment, and he specified expansion of these stocks exogenously. Public expenditure, net exports3 and inventory accumulation by commodity are exogenous in Johansen’s model, and, consistent with a long-run focus, total employment is also exogenous. There is one

ip t

type of labour which is perfectly mobile across industries, although wages differ by industry according to exogenously given relativities. The overall level of the nominal

cr

wage is exogenous and acts as the numeraire. For capital, Johansen assumed that an

exogenously given relative rates of return.

us

exogenous aggregate quantity is allocated across industries in accordance with Rather uncomfortably, the

an

structure/equipment composition of this aggregate capital stock morphs costly to suit the requirements of the using industries, becoming more structure-intensive if the

M

industrial composition of activity shifts towards industries whose capital is structureintensive and vice versa.

ed

Throughout his book Johansen provided thoughtful commentary on the He had a strong,

ce pt

theoretical limitations of his model, but was not apologetic. immediate and practical purpose.

“The model presented in this study is in many respects unsatisfactory when judged

Ac

from a purely theoretical point of view. It is, however, constructed with an eye to the possibility of it being implemented by existing statistics. Without this possibility the model would hardly be very interesting.” [Johansen 1960a: 1.] “I do believe that the numerical results also give a rough description of some important economic relationships in the Norwegian reality.” [Johansen 1960a: 3.]

3

Net exports (exports less imports) are exogenous for all commodities except non-competitive imports which are endogenously determined as Leontief inputs to production and as a commodity in the household utility function.

7

Page 7 of 33

He argued that people who merely use the data to illustrate the working of a model are giving themselves an excuse for not using the data seriously.

He thought that

underestimating the value of the data is as serious a sin as drawing conclusions that are too strong.

ip t

Fifty years on, the exact specification of Johansen’s model is of only historical importance. It was revolutionary at its time and even today, it would be respectable.

cr

However, there are two aspects of the model that remain of immediate relevance to

us

today’s CGE community and which are the defining features of what we will refer to as Johansen-style CGE modelling.

The first is the way in which Johansen presented his model as a rectangular

an



M

system of linear equations in change and percentage-change variables and solved it by matrix inversion.

The second, and related aspect, is how he used the linear representation and the

ed



linear solution method: to clarify properties of the model; to elucidate real world

ce pt

issues; and to check the validity of the model. 2.1. Johansen’s solution method

The computational form of Johansen’s model is a linear system of 86

Ac

equations and 132 variables. He expressed it as: B*   L* 

(1)

where B and L are matricesof dimensions 86 by 86 and 86 by 46; and  andare column vectors with dimensions 86 and 46 ofchanges or percentage changes in the model’s 86 endogenous variables and 46 exogenous variables.

8

Page 8 of 33

Johansen derived most of the equations in (1) from underlying levels forms. For example, in (1) he represented the Cobb-Douglas relationship: 



X j  A j * N j j *K j j *exp

 j*t

(2)

ip t

between the output in industry j (Xj), labour and capital inputs (Nj and Kj) and the rate of technical progress (j) as

cr

x j   j *n j   j *k j   j

(3)

us

where xj, nj and kj are percentage rates of change in Xj, Nj and Kj, and j and jare

an

positive parameters summing to one for most industries but less than one for a few industries, most notably agriculture. For consumer demands, Johansen specified a

simplified version of what he wrote is:

M

percentage change form directly without giving an underlying levels form. A slightly

where

(4)

ce pt

j

ed

ci  v   i, j *p j  E i *  y  v 

ci, pj, y and v are percentage rates of change in consumption of good i, the price of good j, total household expenditure and the number of households, and

Ac

i,jand Ei are parameters representing the elasticity of demand for i with respect to the price of j and the elasticity of demand for i with respect to total household expenditure.

From (1), Johansen solved his model as   T* 

(5)

where T is the 86 by 46 matrix given by B-1*L.

9

Page 9 of 33

2.2 The T matrix: clarifying properties of the model; elucidating real world issues; and checking the validity of the model Johansen was fascinated by the T matrix in (5). He devoted most of chapters 7 and 8 of his book to discussing it and also displayed the entire matrix as a foldout

ip t

table. His focus on the T matrix explains much about the way in which he was able to understand his model and assess it against reality. In this section we describe how

cr

Johansen used the T matrix and some of the things he found out.

us

The T matrix shows the sensitivity (usually an elasticity) of every endogenous variable with respect to every exogenous variable. Johansen regarded the 3,956

an

entries in the T matrix as his basic set of results and he looked at every one of them.

M

How could he cope with this volume of results? As many current CGE modellers find, it is easy to become overwhelmed and confused when confronted with a huge

ed

number of results. Johansen’s management strategy was to use a one-sector back-ofthe-envelope (BOTE) model as a guide. He set this up in chapter 2.4 The BOTE told

ce pt

him what to look for and what to expect in his full-scale model for the effects of exogenous changes in aggregate capital, aggregate employment, the number of households, technologies, exogenous demands and the price of non-competitive

Ac

imports. The effective use of BOTE models to organize, understand and explain results is a skilful art: the BOTE model must strike a balance between being readily accessible andbeing descriptive of the main model. Johansen was the first master BOTE builder (a well known Norwegian specialty!) in CGE modelling. Perhaps because of the degree of difficulty, he has not had many peers in this area.

A referee drew our attention to Johansen’s (1960b) paper on “Rules of thumb …”. That paper describes a stand-alone theoretical model with a stylized empirical implementation. This is not what we mean by a BOTE model. Rather, a BOTE model is set up alongside a full-scale empirical model and serves as a guide to the interpretation of results from the full-scale model. 4

10

Page 10 of 33

Johansen’s examination of the T matrix starts with a discussion of the elasticities of industry prices and outputs with respect to aggregate capital(K) and employment (N): entries in theprice and output rows of the k and n columns. On the basis of the BOTE model, he explained why the full model shows negative entries in

ip t

the price rows of the k-columnand positive entries in the price rows of the n-column. An increase in K raises the marginal product of labour requiring an increase in the

cr

real wage rate. With the overall nominal wage rate set exogenously, this must be

us

brought about by a decrease in the price level. An increase in Nlowers the marginal product of labour requiring an decrease in the real wage rate, brought about by

an

anincrease in the price level.

For the output rows, Johansen’s BOTE model suggests that the entries in the

M

k- and n-columns should lie in the (0,1) interval: this follows from a macro version of equation (3)5. With one exception, this expectation is fulfilled: the T matrix shows a

ed

negative entry for the elasticity of equipment output with respect to an increase in

ce pt

aggregate employment. Following up and explaining exceptions is an important part of the BOTE methodology. In this way we can locateresult-explaining mechanisms in the full model that are not present in the BOTE model. In other words, we can figure out what the full model knows that the BOTE model doesn’t know. In the case of the

Ac

elasticity of equipmentoutput with respect to aggregate employment, the explanation of the negative result is that an increase in employment changes the composition of the economy’s capital stock in favour of structures and against equipment. This morphing of the capital stock reduces maintenance demand for equipment, thereby reducing the output of the equipment industry. How does an increase in employment lead to a decrease in the equipment-intensity of the economy’s capital stock? The 5

That is x   *n   *k   where  and  are parameters with values between 0 and 1.

11

Page 11 of 33

answer is that it leads to a reallocation of the economy’s capital towards housing, a highly structure-intensive form of capital. With aggregate capital fixed, there is a consequent reduction in the economy’s stock of equipment capital. What explains the reallocation of the economy’s capital towards housing? An increase in employment Whereas the

ip t

stimulates consumption of all goods including housing services.

increases in the outputs of labour-intensive consumption goods are accommodated

cr

with increased labour input and reduced capital input, increased output of housing

us

services can be achieved only by increases in the housing stock.

One of the most interesting parts of T is the 22 by 22 sub-matrix, T(x,z),

an

relating movements in industry outputs (x) to movements in exogenous demands (z). At the time when Johansen was writing, Leontief’s input-output model, with its

M

emphasis on input-output multipliers, was the dominant tool for quantitative multisectoral analysis. In Leontief’s model, if an extra unit of output from industry j is

ed

required by final users, then production in j must increase by at least one unit and

ce pt

production in other industries will increase to provide intermediate inputs to production in j. Further rounds of this process can be visualized with suppliers to j requiring extra intermediate inputs. Thus, in Leontief’s picture of the economy, industries are in a complementary relationship, with good news for any one industry

Ac

spilling over to every other industry. Johansen challenged this orthodoxy. His T(x,z) sub-matrix implies diagonal derivatives ( X j Z j ) that in most cases are less than one and off-diagonal derivatives that are predominantly negative.

Rather than

emphasising complementary relationships between industries, Johansen emphasised competitive relationships. In Johansen’s model, expansion of output in one industry drags primary factors away from other industries. Only where there are particularly

12

Page 12 of 33

strong input-output links did Johansen find that stimulation of one industry (e.g. food) benefits another industry (e.g. agriculture). In chapter 8, Johansen used the T matrix to decompose movements in industry outputs, prices and primary factor inputs into parts attributable to observed changes in

ip t

six sets of exogenous variables: aggregate employment; aggregate capital; population; Hicks-neutral primary factor technical change in each industry; exogenous demand

In making his

cr

for each commodity; and the price of non-competing imports.

us

calculations, he applied shocks to all 46 exogenous variables. These shocks represent average annual growth rates for the period around 1950. Johansen computed the

an

effects [ ( j) ] on the 86 endogenous variables of movements in the jth set of

( j)  T( j)* ( j)

M

exogenous variables as , j = 1, …, 6

(6)

ed

where ( j) is the vector of shocks applied to the jth set of exogenous variables; and

ce pt

T(j) is the sub-matrix of T formed by the columns corresponding to the jth set of exogenous variables.

In discussing the ( j)s , Johansen paid particular attention to the behaviour of

Ac

agricultural employment. This was a contentious issue among economists in 1960. On the one hand, diminishing returns to scale suggested that relative agricultural employment would grow with population and perhaps even with income despite low expenditure elasticities for agricultural products. On the other hand, agriculture was experiencing rapid technical progress, suggesting that employment in agriculture might not only fall as a share of total employment but might even fall in absolute terms. Johansen was able to separate and quantify these conflicting forces. The

13

Page 13 of 33

( j)s for capital, employment and population growth showed relatively strong

increases in agricultural employment, consistent with diminishing returns to scale interacting with increased consumption of food. However, the dominant effect on agricultural employment was technical change. This was strongly negative, leaving

ip t

agricultural with net declining employment. Another aspect of the decomposition results that interested Johansen was the

cr

role of rapid capital accumulation. The results indicated that this was the major

us

source of real wage growth. He worried that increases in the capital/labour ratio were reducing rates of return, raising questions about the sustainability of capital growth

an

and therefore of real wage growth. In the 1973 addendum to his book, Johansen continued to worry about these issues and wondered whether long-run movements in

M

rates of return can be described adequately in a model with Cobb-Douglas production

ed

functions and Hicks-neutral technical change.

Having completed his examination of the individual ( j) vectors, Johansen

ce pt

compared the aggregate effects [  j ( j) ] of the six sets of exogenous shocks with reality for 1948 to 1953. He assessed this validity test as follows: “A first glance at the figures confirms that some general tendencies are roughly

Ac

the same in the computed and in the observed figures.

Some interesting

differences are, however, also revealed.” [Johansen 1960a: 152]

Johansen used these differences to pinpoint weaknesses in his model and to organize a discussion of real-world developments. For agriculture he found that the computed growth rate in output closely matched reality but that the computed growth rate in employment was too high while that in capital was too low. This led to a discussion of reasons, not accounted for in the model, for exodus of rural workers to the cities.

14

Page 14 of 33

For communication & transport and for forestry, the computed growth rates for output and primary-factor inputs were too high. In the case of communication & transport, Johansen wondered whether this reflected unsatisfied demand for these government supplied services. For forestry, Johansen thought that he may have set the income

ip t

elasticity of demand for forestry products too high and also that there may have been a taste change, not included in his model, against the use of forestry products as fuel.

cr

By going through his results in this way, Johansen developed an agenda for model

us

improvement (ch. 9).

In the 1973 addendum, Johansen (1974, ch. 10) reported implementation of

an

some of the items on this agenda. He also reported further applications of the model including a validation test in which projections for 1950 to 1963 covering output,

M

capital, labour and prices by industry are compared with reality. However, at the theoretical level, rather little change took place in the model during the 1960s.6 In

ed

particular the treatment of trade remained unchanged and underdeveloped. This was

ce pt

surprising considering that exports and imports each represented about 40 percent of Norwegian GDP. It appeared that Johansen did not have a solution to the flip-flop problem.This refers to unrealistic and unstable levels of industrial specialization. It occurs in long-run simulations with models in which import/domestic price ratios are

Ac

allowed to play a role in import/domestic choice and imported and domestically produced units of commodity i are treated as perfect substitutes. It is also a problem when export prices are taken as given and long-run supply curves are flat. Johansen discussed flip-flop in the addendum in connection with the model of Taylor and Black (1974). He recognized that Taylor and Black avoided extreme flip-flop by adopting a short-run closure (fixed capital in each industry) thereby giving supply curves a 6

Development of the model accelerated in the late 1970s and 80s with an emphasis on capital-labourenergy-material substitution, see Longva et al. (1985).

15

Page 15 of 33

positive slope. With his focus on long-run tendencies, Johansen could not adopt the Taylor and Black approach. Instead hecontinued to treat exports and competitive imports exogenously.7 3. Other starting points for CGE modelling

ip t

While Johansen was the first to plant a seed in what has now become the CGE

cr

forest, it was not the only seed. Largely independently of Johansen’s work, CGE

Jorgenson and two groups at the World Bank.

us

modelling had several other starting points, associated with Herbert Scarf, Dale

an

Scarf’s work was published in the late 1960s and early 1970s without reference to Johansen.8It drew inspiration from Arrow and Debreu (1954) and others

M

who applied fixed-point theorems to demonstrate the existence of equilibria in a broad class of algebraically specified general equilibrium models. Scarf did not make a

ed

contribution to the economic content of CGE modelling: he provided a solution algorithm. His technique was never the most effective method of solving CGE

ce pt

models and,after considerable attention in leading journals in the 1970s9, was largely abandoned in the 1980s in favour of much simpler methods. Starting in the early 1970s Jorgenson and his associates focused on energy.10

Ac

They solved what can be clearly recognised as CGE models by iterative methods mimicking a Walrasian search for an equilibrium price vector. As with Scarf, there is nothing to suggest a debt to Johansen. With associates, Jorgenson has made a series of path-breaking contributions to the economic content of CGE modelling. Hudson 7

This was also the approach of Hudson and Jorgenson (1974) for the U.S. Adelman and Robinson (1978) in their study of Korea set exports of some commodities exogenously and fixed the share of exports in domestic output for other commodities. For most imports, Adelman and Robinson fixed the import share in domestic demand. Taylor et al. (1980, ch. 7), in their study of Brazil, exogenized exports and related imports to industry outputs and final demands via exogenous coefficients. 8 See Scarf (1967 and 1973). 9 See, for example, Shoven and Whalley (1972, 1973 and 1974). 10 See for example, Hudson and Jorgenson (1974).

16

Page 16 of 33

and Jorgenson (1974) were the first to include in a CGE model flexible functional forms for production (econometrically estimated translog functions). Jorgenson and Yun (1986a & b) embedded Jorgenson’s well known theory of investment (Jorgenson 1963) into a CGE study of U.S. tax policy. The models of Jorgenson and Wilcoxen

ip t

(1993) and Ho and Jorgenson (1994) included econometrically estimated consumer demand functions based on the theory of exact aggregation [Jorgenson et al. (1980)].

cr

A model incorporating many of Jorgenson’s innovations is described in Jorgenson et

us

al. (2012).

Working at the World Bank, Adelman and Robinson (1978) made pioneering

an

contributions to CGE modelling which again were largely independent of Johansen. While they were aware of his work, it was clear that the style of their modelling owed

M

little to Johansen. Equations were presented and discussed in non-linear levels form; no use was made of the T matrix; little emphasis was given to explaining results via

ed

BOTE models; and no attempt was made to match structural results against observed

ce pt

outcomes over a time horizon of more than two years from the base period11.By contrast, Taylor and his associates, also at the World Bank, drew more heavily on Johansen. Echoing Johansen, Taylor et al. (1980) emphasised the performance of their model in reproducing industry growth rates for an historical period (1959-71)

Ac

and set out a linear-percentage-change BOTE model to assist in the explanation of results.

Since the 1970s, two of the CGE seedlings have grown and spread throughout the world: those starting with Johansen in Norway and Adelman-Robinson at the World Bank. In the latter case, the spread occurred directly from the World Bank, 11

Adelman and Robinson (1978: 62-76) showed that when their model was fed base-period values for the exogenous variables then it closely reproduced base-period values for the endogenous variables. This indicates that they successfully calibrated their model but it does not test the predictive or explanatory power of the model.

17

Page 17 of 33

particularly through the efforts of Sherman Robinson. In the case of Johansen, the seed took a curiously long time before spreading. Johansen’s approach was not pursued outside Norway in a sustained way for well over a decade after the publication of his seminal work in 1960. There were some important one-off flurries

ip t

using his technique in the mid-1970s (see for example Taylor and Black (1974), Staelin (1976), Bergman (1978) and Keller (1980), but the place where Johansen-style

cr

modelling really flourished outside Norway was 16,000 kilometres away in

us

Australia.12

For both the Johansen and Adelman-Robinson styles of modelling, special

an

purpose software played a key role in their wide-spread adoption: GEMPACK for the Johansen-style models from Australia and GAMS for the models from the World

M

Bank.13 GEMPACK and GAMS allow economists to specify their models in the computer in a form that is close to ordinary algebra and then call upon standard

ed

equation solvers to produce solutions. From the point of view of economists wishing

ce pt

to enter the CGE field, GEMPACK and GAMSdramatically reduced the required level of knowledge of numerical methods and computing, and also reduced the time required in writing and checking computer programs.

Both software systems

facilitated communication by allowing models to be transferred conveniently between

12

Ac

users and between sites.

In Norway, Johansen is revered and the name MSG (Multi-Sectoral Growth), originally used by him, has been adopted for a series of models developed by Statistics Norway since the 1960s. However, the Norwegians appear to have moved away from the Johansen style. In discussing the latest MSG model, Holmoy and Strom (2012) explain that “… MSG6 has little in common with the original MSG model”. 13 GEMPACK was developed starting about 1980 by Ken Pearson, specifically to solve Johansen-style models. The first version of GEMPACK was used for teaching in 1984, and shortly after that was adopted by Australian CGE modellers. The first GEMPACK manuals were published in 1986 (see Codsi and Pearson, 1986). Early journal descriptions of GEMPACK are Pearson (1988) and Codsi and Pearson (1988).GAMS was developed starting in the mid-1970s by Alex Meeraus and Jan Bisschop at the World Bank. Its original focus was non-linear optimization problems. The adaption of GAMS to CGE modelling took place at the World Bank in the mid-1980s. The first GAMS-based CGE model was by Condon et al. (1987). For a comparison of GEMPACK and GAMS, written by key people in their development, see Horridgeet al. (2012)

18

Page 18 of 33

4. Johansen in Australia In 1975, the Australian Government set up the IMPACT Project. Its initial task was to build an economy-wide model of Australia for assessing the effects of reductions in tariffs on manufactured imports.

The modellers at IMPACT were

ip t

attracted to Johansen’s simple and effective methods. They presented their ORANI model (Dixon et al. 1977, 1982) as a linear system in changes and percentage

cr

changes, Johansen’s equation (1). This simplified the interpretation of the model and

us

facilitated teaching. They adopted Johansen’s solution, equation (5). This enabled them to solve in the 1970s a model with much larger dimensions than was possible

an

with other styles of CGE modelling. Johansen’s use of BOTE models was also taken up and extended at the IMPACT Project. Over the last 37 years, the IMPACT team

M

has taken on various institutional guises, and since 1991 has formed the Centre of Policy Studies (CoPS) at Monash University.

The MONASH family of models

ed

(Dixon and Rimmer 2002) evolved from ORANI and continues to reflect the

ce pt

Johansen style. MONASH modellers have followed Johansen in using the T matrix: to understand properties of their models; to explain periods of history via decomposition analysis; and to perform validation exercises.

Ac

While Johansen’s techniques were simple and effective, adopting them came at a cost. His solution equation (5) gives only an approximation to effects implied by the underlying non-linear model: (5) produces solutions that are subject to linearization errors.14 At the time that Johansen was developing his model, incurring this cost was a computational necessity. Later CGE pioneers were keen to avoid

14

Johansen recognised this problem and reported (Johansen, 1974) experience with a method implemented in the late 1960s by Spurkland (1970) for calculating accurate solutions. Spurkland’s method used (5) to obtain an approximate solution and then moved to an accurate solution via a general non-linear-equation method such as Newton’s algorithm. Spurkland’s method sacrificed Johansen’s simplicity and was rather awkward to implement. It was not widely adopted.

19

Page 19 of 33

linearization errors and perhaps this caused them to overlook the strengths of Johansen’s approach.

Linearization errors were avoided in ORANI and later

MONASH models without sacrificing Johansen’s simplicity and interpretability. This was done by adopting a multi-step extension of Johansen’s approach which became

ip t

known as the Johansen/Euler method. In a two-step Johansen/Euler computation, the effect of a shock such as a 50 per cent increase in a tariff was computed by first

cr

applying the Johansen method to work out where the economy would go with a 25 per

us

cent increase, then computing a new T matrix for that situation, and finally applying the Johansen method for a second time to impose the second half of the 50 per cent

an

tariff increase. As explained in Dixon et al. (1982, sections 8 and 35)15, accurate solutions can be obtained even for large shocks by the Johansen/Euler method with a

M

small number of steps supplemented by an extrapolation.

The multi-step solution method was not the only innovation in the creation of

ed

ORANI and later MONASH models. As described in Dixon and Rimmer (2010) and

ce pt

Dixon et al. (2012), other innovations were: the treatment of imports and competing domestic products as imperfect substitutes16(which effectively dealt with the flip-flop problem); the incorporation of policy-relevant detail requiring large dimensionality (e.g. transport and other margins, purchasers and producer prices, and a regional

Ac

dimension); closure flexibility allowing, for example, generation of short-run results (with real wages and capital fixed exogenously), long run results (with employment and rates of return fixed exogenously) and transition between short- and long-run via

15

This material is also presented in Dixon et al. (2012). Dixon and his colleagues named this the Armington assumption in recognition of Armington (1969) who introduced the idea in the context of a 15 country model in which each country produced one good but consumed 15. Subsequently, the Armington assumption was adopted almost universally in CGE modelling. 16

20

Page 20 of 33

partial adjustment mechanisms; and inclusion of complex functional forms (e.g. CRESH production functions and CRETH transformation frontiers). Within 10 years of the first account of ORANI, there were 202 published applications (see Powell and Lawson 1990). These covered a wide range of topics

ip t

and were undertaken by or for many different organizations including Australian federal government agencies, State and regional authorities in Australia, universities

cr

and private businesses.Hundreds of further applications have been made in Australia

us

with the MONASH models, in the areas of forecasting, adjustment costs, validation, technological change, trade, environment, taxation, microeconomic reform and labour

an

markets.

M

Why has the Johansen-style CGE modelling achieved such broad-based acceptance and application in Australia? It is impossible to give precise reasons but

ed

Dixon and Rimmer (2010) identifies two factors that were certainly important. The first was a favourable policy and institutional environment in the 1970s when the

ce pt

IMPACT Project was set up. The second was that the Johansen formulation and the Australian extensions gave the ORANI and MONASH models: (a) the capacity to carry credibility-enhancing detail; (b) flexibility in application; (c) transferability

Ac

among institutions; and (d) interpretability of both theory and results by users with limited training in mathematics and computing. 5. Taking Johansen from Australia to the rest of the World From the early 1980s, insights gained from Australia’s Johansen-style modelling began to be exported to other countries. The first exports were associated with the appointments of Australian CGE modellers to positions in foreign

21

Page 21 of 33

institutions.17Further exports occurred via foreign students trained in Australia, many of whom built and applied Johansen-style models for developing countries, see for example Naqvi (1997). Since 1991 CoPS has been undertaking major model development and

ip t

application tasks for organizations outside Australia. As shown on its website, CoPS has now completed modelling projects for about 20 countries and is currently engaged

cr

in ongoing projects in Vietnam, Malaysia, Finland, the Netherlands, China and the

us

U.S. In all of these projects, Johansen-style models implemented in GEMPACK software have been constructed and applied, often in collaboration with officials in

an

policy-making government departments.

M

CoPS’ most elaborate venture in exporting Johansen-style modelling services is its collaboration with departments of the U.S. government. Starting in 2000 with

ed

the MONASH model of Australia as a template, CoPS worked on the construction of USAGE (U.S. Applied General Equilibrium), a policy-oriented model of the U.S.

ce pt

economy. After 10 years work at CoPS and contributions by economists at the U.S. International Trade Commission, USAGE has become the most detailed CGE model to have been constructed anywhere in the world.

Exploiting the computational

Ac

simplicity of the extended Johansen method and the data handling convenience of the GEMPACK software, USAGE can accommodate500 U.S. industries, 51 regions (50 states plus the District of Columbia), 700 occupations and 23 sources for U.S. imports and destinations for U.S. exports.

The model is dynamic and integrates

microeconomic detail with a fully articulated macro specification covering taxes,

17

Early examples are the appointments of David Vincent and Philippa Dee to fellowships at the Kiel Institute which led to ORANI-like models of Chile, West Germany, Korea, Colombia, Ivory Coast and Kenya: see Dicke et al. (1983, 1984, 1989); Vincent (1982, 1985) and Dee (1986). Later examples are the appointments of Robert McDougall and Terrie Walmsley, both formerly at CoPS, to senior positions in the Global Trade Analysis Project (GTAP) at Purdue University.

22

Page 22 of 33

tariffs, public expenditures, public debt, the balance of payments, foreign assets and foreign liabilities. It has been used to track economic performance from 1992 and to produce detailed baseline projections and policy simulations to 2020. USAGE is now widely recognized in Washington DC. It has been applied not

ip t

only on trade issues for the U.S. International Trade Commission (see USITC 2004,

Commerce and Homeland Security. These topics include

partial replacement of imported crude petroleum by domestically produced

us



cr

2007, 2009), but also on other topics for the U.S. Departments of Agriculture,

biofuels (see Dixon et al. 2007; and Gehlhar et al. 2010);

a security-related closure of U.S. ports (see Dixon et al. 2011a);



a serious H1N1 epidemic (see Dixon et al. 2011c);



the implications for industries and regions of the 2009-10 U.S. recession and the

M

an





ed

likely effects of the Obama stimulus package (see Dixon and Rimmer 2011); and different approaches to the problem in the U.S. of illegal immigration (see Dixon

ce pt

et al. 2011b).

A final conduit of Johansen-style modelling from Australia to the rest of the world is the Global Trade Analysis Project (GTAP) founded in 1992 by Tom Hertel at

Ac

Purdue University (see Hertel 1997, 2012). Hertel visited the IMPACT Project for the academic year 1990-91. During his visit, he was exposed to the ORANI model and the GEMPACK software. He was particularly impressed by the SALTER model, a Johansen-style multi-country model being constructed at the Industries Assistance Commission in Canberra (IMPACT’s main sponsoring agency) using ORANI as the theoretical starting point and GEMPACK software for computations. Hertel took SALTER back to Purdue and this formed the basis for the first version of the GTAP model of the global economy.

23

Page 23 of 33

GTAP is an extraordinarily successful project. As well as the GTAP model, the Project provides: a database of input-output tables and trade flows for about sixty commodities and one hundred countries; training courses attended by people from around the world; an annual conference at which about 200 papers are delivered

ip t

mainly on trade issues using the GTAP data and model; and a website on which several thousand participants from nearly every country exchange data and ideas on

cr

CGE modelling.

us

With regard to the model, GTAP has retained a large scale but relatively simple Johansen-style model as its core analytical and teaching device. The model is

software as they occur.

an

implemented in GEMPACK18and takes full advantage of developments in this Via GTAP and the GEMPACK software, thousands of

M

economists from every part of the world are undertaking Johansen-style modelling: they areusing the linear representation of a CGE model to clarify its properties; to

ce pt

6. Concluding remarks

ed

elucidate real world issues; and to check model validity.

Since its inception in 1960 with the publication of Johansen’s book, CGE modelling has proved a remarkably fruitful technique for combining data with

Ac

economic theory to project implications for macro, industry, regional, occupational, environmental and distributional variables of a wide range of policy changes and other shocks to the economy. Johansen was first in the CGE field and provided a particular style of CGE modelling based on a linear representation of the theory and a linear solution method.

18

There is now a version of the GTAP model in GAMS. As noted by Rutherford (2006), simplifications were made in creating this version. For example, the CDE expenditure functions used in standard GTAP were replaced in the GAMS version by a Cobb-Douglas specification. It appears that outside the Johansen framework, advanced functional forms such as CDE cause difficulties.

24

Page 24 of 33

The main objection to his methods, especially in North America, was that Johansen solutions were only approximations. This objection was overcome in Australia by 1980 through the simple and effective multi-step Johansen/Euler procedure that eliminated linearization errors. By adopting the Johansen style, Australian CGE

ip t

modellers were able to make rapid progress. In the 1970s they created CGE models with: price-sensitive treatments of international trade; policy-relevant levels of detail;

cr

flexible closures; and the ability to handle complex functional forms.

us

Facilitated by Johansen’s linear representation implemented with GEMPACK software and by the use of BOTE models, Australia’s ORANI and MONASH models

an

have been taught, transferred and interpreted. By the 1980s, Johansen-style modelling had been widely adopted by government departments and consulting firms in

M

Australia and was being used in public debates on almost every important economic issue. While ORANI was comparative static, MONASH is dynamic with an explicit

ed

baseline or business-as-usual forecast. In common with Johansen’s experience in

ce pt

Norway, we have found in Australia that baselines facilitate discussion with policy advisers and business people. These discussions are a critical part of confronting CGE results with reality and play an important role in the development of CGE

Ac

modelling.

From Australia, Johansen-style modelling spread to the rest of the world

through the activities of members and former members of IMPACT and the Centre of Policy Studies and through the GTAP Project. Throughout this history of quantitative contribution to economic analysis and policy, Leif Johansen’s book published 50 years ago has been a guiding light. References

25

Page 25 of 33

http://www.sciencedirect.com.ezproxy.lib.monash.edu.au/science?_ob=ArticleURL& _udi=B6V82-4GPVXDB-1&_user=542840&_handle=V-WA-A-W-EMsSAYWA-UUA-U-AAZWEDUZCA-AAZUCCAVCA-AYVCEYAYC-EU&_fmt=summary&_coverDate=11%2F30%2F2005&_rdoc=6&_orig=browse

ip t

&_srch=%23toc%235858%232005%23999729991%23612107%21&_cdi=5858 &view=c&_acct=C000027659&_version=1&_urlVersion=0&_userid=542840&

cr

md5=dfceda435340e037ccd6b36310686179

us

Adelman, I. and Robinson, S. (1978) Income Distribution Policy in Developing Countries: A Case Study of Korea. New York: Oxford University Press.

an

Armington, P.S. (1969) The geographic pattern of trade and the effects of price changes. IMF Staff Papers XVI: 176-199.

M

Arrow, K.J. and Debreu, G. (1954) Existence of an equilibrium for a competitive

ed

economy. Econometrica 22: 265-90.

Bergman, L. (1978) Energy policy in a small open economy: the case of Sweden. RR-

ce pt

78-16. International Institute for Applied Systems Analysis, Laxenberg, Austria. Available at http://www.iiasa.ac.at/Admin/PUB/Documents/RR-78-016.pdf Bergman, L. (1985) Extensions and applications of the MSG-Model: a brief survey.

Ac

In Production, Multi-Sectoral Growth and Planning (F.R. Forsund, M. Hoel and S. Longva, Eds.) Amsterdam: Contributions to Economic Analysis 154, NorthHolland Publishing Company, 127-161.

Bjerkholt, O. and Tveitereid, S. (1985) The use of the MSG model in preparing a ‘perspective analysis 1980-2000’ for the Norwegian economy. In Production, Multi-Sectoral Growth and Planning (F.R. Forsund, M. Hoel and S. Longva, Eds.) Amsterdam: Contributions to Economic Analysis 154, North-Holland Publishing Company, 271-87. 26

Page 26 of 33

Codsi, G. and Pearson, K.R. (1986) GEMPACK documents numbers 2, 3, 5-8 and ll18. IMPACT Project, available from CoPS, Monash University, Victoria, Australia, 3800. Codsi, G. and Pearson, K.R. (1988) GEMPACK: General-purpose Software for

Science in Economics and Management 1: 189-207.

ip t

Applied General Equilibrium and Other Economic Modellers”, Computer

cr

Condon, T., Dahl, H. and Devarajan, S. (1987) Implementing a computable general

us

equilibrium model on GAMS: the Cameroon model. Report No. DRD290, World Bank, Washington DC.

an

Dee, P.S. (1986) Financial markets and economic development: the economics and politics of Korean reform. Kiel Institute Study 198, J.C.B. Mohr (Paul Siebeck)

M

Tubingen.

ed

Dicke, H., Gupta, S., Mayer, T. and Vincent, D. (1983) The short-run impact of fluctuating primary commodity prices on three developing countries Colombia,

ce pt

Ivory Coast, Kenya. World Development 11: 405-16. Dicke, H., Gerken, D., Mayer, T. and Vincent, D. (1984) Stabilization strategies in primary commodity exporting countries: a case study of Chile. Journal of

Ac

Development Economics 15: 47-75. Dicke, H., Donges, J.B., Gerken, E. and Kirkpatrick, G. (1989) Effects of agricultural trade liberalization on West Germany’s economy. Chapter 4 in Macroeconomic Consequences of farm support policies (A.B. Stoeckel, D.P. Vincent and S. Cuthbertson, Eds.). Durham and London: Duke University Press, 125-149. Dixon, P.B. and Parmenter, B.R. (1996) Computable General Equilibrium Modelling for Policy Analysis and Forecasting. Chapter 1 in Handbook in Computational

27

Page 27 of 33

Economics, Vol. 1 (H.M. Amman, D.A. Kendrick and J. Rust, Eds.). Amsterdam: Elsevier, 3-85. Dixon, P.B. and Rimmer M.T. (2002) Dynamic General Equilibrium Modelling for Forecasting and Policy: a Practical Guide and Documentation of MONASH.

ip t

Contributions to Economic Analysis 256. Amsterdam: North-Holland Publishing Company.

cr

Dixon, P.B. and Rimmer M.T. (2010) Johansen’s contribution to CGE modelling:

us

originator and guiding light for 50 years. CoPS/IMPACT Working Paper G203. See http://www.monash.edu.au/policy/ftp/workpapr/g-203.pdf .

an

Dixon, P.B. and Rimmer M.T. (2011) Simulating the U.S. Recession with and without the Obama package: the role of excess capacity. Economic Modelling 28(1-2):

M

602-613.

ed

Dixon, P.B., Parmenter, B.R., Ryland, G.J. and Sutton, J. (1977) ORANI, A General Equilibrium Model of the Australian Economy: CurrentSpecification and

ce pt

Illustrations of Use for Policy Analysis. Vol. 2 of the First Progress Report of the IMPACT Project, Australian Government Publishing Service, Canberra. Dixon, P.B., Parmenter, B.R., Sutton, J. and Vincent, D.P. (1982) ORANI: A

Ac

Multisectoral Model of the Australian Economy. Contributions to Economic Analysis 142. Amsterdam: North-Holland Publishing Company.

Dixon, P.B., Osborne, S. and Rimmer, M.T. (2007) The economy-wide effects in the United States of replacing crude petroleum with biomass. Energy and Environment 18(6): 709-722. Dixon, P.B., Giesecke, J., Rimmer M.T. and Rose, A.Z. (2011a) The Economic Costs to the U.S. of closing its borders: a computable general equilibrium analysis. Defence and Peace Economics 22(1): 85–97. 28

Page 28 of 33

Dixon, P.B., Johnson, M. and Rimmer, M. T. (2011b) Economy-wide effects of reducing illegal immigrants in U.S. employment. Contemporary Economic Policy 29(1): 14-30. Dixon, P.B., Lee, B., Muehlenbeck, T., Rimmer, M.T, Rose, A.Z. and Verikios, G.

ip t

(2011c) Effects on the U.S. of an H1N1 epidemic: analysis with a quarterly

7(1,75), 2010 DOI: 10.2202/1547-7355.1769. See

us

http://www.bepress.com/jhsem/vol7/iss1/75

cr

CGE model. Journal of Homeland Security and Emergency Management

Dixon, P.B., Koopman, R.B. and Rimmer, M.T. (2012) The MONASH style of CGE

an

modelling: a framework for practical policy analysis. In Handbook of Computable General Equilibrium Modeling (P.B. Dixon and D.W.

M

Jorgenson, Eds.). Amsterdam: Elsevier.

ed

Frisch, R. (1959) A Complete Scheme for Computing All Direct and Cross Elasticities in a Model with Many Sectors. Econometrica 27: 177-196.

ce pt

Gehlhar, M., Somwaru, A., Dixon, P.B., Rimmer, M.T., and Winston, R.A. (2010) Economywide Implications from US Bioenergy Expansion. American Economic Review: Papers & Proceedings 100: 172-177.

Ac

Hertel, T.W. (Ed.) (1997) Global Trade Analysis: Modeling and Applications. Cambridge, U.K.: Cambridge University Press.

Hertel, T.W. (2012) Global applied general equilibrium analysis using the GTAPframework. In Handbook of Computable General Equilibrium Modeling (P.B. Dixon and D.W. Jorgenson, Eds.). Amsterdam: Elsevier. Ho, M.S. and Jorgenson, D.W. (1994) Trade Policies and U.S. Economic Growth. Journal of Policy Modeling 16(2): 119-46.

29

Page 29 of 33

Holmoy, E. and Strom, B. (2012) CGE assessments of fiscal sustainability in Norway. In Handbook of Computable General Equilibrium Modeling (P.B. Dixon and D.W. Jorgenson, Eds.). Amsterdam: Elsevier. Horridge, M., Meeraus, A., Pearson, K. and Rutherford, T. (2012) Solution

ip t

software for CGE modelling. In Handbook of Computable General Equilibrium Modeling (P.B. Dixon and D.W. Jorgenson, Eds.). Amsterdam:

cr

Elsevier.

us

Hudson, E.A. and Jorgenson, D.W. (1974) U.S. Energy Policy and Economic Growth, 1975-2000. Bell Journal of Economics and Management Science 5(2):

an

461-514.

Johansen, L. (1960a) A Multisectoral Study of Economic Growth, Contributions to

M

Economic Analysis 21. Amsterdam: North-Holland Publishing Company.

ed

Johansen, L. (1960b) Rules of Thumb for the expansion of industries in a process of economic growth. Econometrica 28(2): 258-71.

ce pt

Johansen, L. (1974). A Multisectoral Study of Economic Growth, second enlarged edition, Contributions to Economic Analysis 21. Amsterdam: North-Holland Publishing Company.

Ac

Jorgenson, D.W. (1963) Capital Theory and Investment Behavior. The American Economic Review 53(2), Papers and Proceedings of the Seventy-Fifth Annual Meeting of the American Economic Association: 247-259.

Jorgenson, D.W. and Yun, K.-Y. (1986a) The efficiency of capital allocation. Scandinavian Journal of Economics 88(1): 85-107. Jorgenson, D.W. and Yun, K.-Y. (1986b) Tax policy and capital allocation. Scandinavian Journal of Economics 88(2): 355-377.

30

Page 30 of 33

Jorgenson, D.W. and Wilcoxen, P.J. (1993) Energy, the Environment and Economic Growth. In Handbook of Natural Resource and Energy Economics, Vol.III (A.V. Kneese and J.L. Sweeney, Eds.). Amsterdam: North-Holland, 12671349.

aggregation. American Economic Review 70: 268-72.

ip t

Jorgenson, D.W., Lau, L.J. and Stoker, T.M. (1980) Welfare comparison under exact

cr

Jorgenson, D.W., Goettle, R.J., Ho, M.S. and Wilcoxen, P.J. (2012) Energy, the

us

environment, and U.S. economic growth. In Handbook of Computable General Equilibrium Modeling (P.B. Dixon and D.W. Jorgenson, Eds.). Amsterdam:

an

Elsevier.

Keller, W.J. (1980) Tax incidence: a General Equilibrium Approach, Contributions to

M

Economic Analysis 134. Amsterdam: North-Holland Publishing Company.

ed

Leontief, W.W. (1936) Quantitative Input-Output Relations in the Economic System of the United States. The Review of Economics and Statistics 18(3): 105-125.

ce pt

Leontief, W.W. (1941) The Structure of the American Economy 1919-1929. Massachusetts: Harvard University Press. Longva, S., Lorentsen, L. and Olsen, Ø. (1985) The multi-sectoral growth model

Ac

MSG-4: formal structure and empirical characteristics. In Production, MultiSectoral Growth and Planning, Contributions to Economic Analysis 154 (F.R. Forsund, M. Hoel and S. Longva, Eds.). Amsterdam: North-Holland Publishing Company, 187-240. Manne, A.S. (1963) Key sectors of the Mexican economy 1960-1970. In Studies in process analysis (A.S. Manne and H.M. Markowitz, Eds.) editors. New York: Wiley, 379-400.

31

Page 31 of 33

Naqvi, F. (1997) Energy, economy and equity interactions in a CGE model for Pakistan. Aldershot: Ashgate. Pearson, K.R. (1988) Automating the Computation of Solutions of Large Economic Models. Economic Modelling 5(4): 385-395.

ip t

Powell A.A. and Lawson, T. (1990) A decade of applied general equilibrium modelling for policy work. In General Equilibrium Modeling and Economic

cr

Policy Analysis (L. Bergman, D. Jorgenson and E. Zalai, Eds.). Massachusetts:

us

Basil Blackwell, 241-290.

Rutherford, T. (2006) GTAP6 in GAMS. See http://www.mpsge.org/gtap6/.

an

Sandee, J. (1960) A long-term planning model for India. New York: Asia Publishing

M

House and Calcutta, Statistical Publishing Company.

Scarf, H.E. (1967) The approximation of fixed points of a continuous mapping. SIAM

ed

Journal of Applied Mathematics 15(5): 328-343. Scarf, H.E. (1973) The computation of economic equilibria. New Haven/London:

ce pt

Yale University Press.

Schreiner, P. and Larsen, K.A. (1985) On the introduction and application of the MSG model in the Norwegian planning system. In Production, Multi-Sectoral Growth

Ac

and Planning, Contributions to Economic Analysis 154 (F.R. Forsund, M. Hoel and S. Longva, Eds.). Amsterdam: North-Holland Publishing Company, 241269.

Shoven, J.B. and J. Whalley, J. (1972) A general equilibrium calculation of the effects of differential taxation of income from capital in the U.S. Journal of Public Economics 1: 281-321.

32

Page 32 of 33

Shoven, J.B. and Whalley, J. (1973) General equilibrium with taxes: A computational procedure and an existence proof. Review of Economic Studies 40: 475-489. Shoven, J.B. and Whalley, J. (1974) On the computation of competitive equilibrium in international markets with tariffs. Journal of International Economics 4: 341-

ip t

354. Spurkland, S. (1970) MSG – A tool in long-term planning. Report prepared for the

cr

First Seminar on Mathematical Methods and Computer Techniques, organized

us

by the United Nations Economic Commission for Europe, Varna.

Staelin, C.P. (1976) A general equilibrium model of tariffs in a non-competitive

an

economy. Journal of International Economics 6(1): 39-63.

M

Taylor, L. and Black, S.L. (1974) Practical general equilibrium estimation of resource pulls under trade liberalization. Journal of International Economics 4(1): 35-58.

ed

Taylor, L., Bacha, E.L., Cardoso, E.A. and Lysy, F.J. (1980) Models of growth and distribution for Brazil. New York: Oxford University Press for the World Bank.

ce pt

United States International Trade Commission (2004, 2007, 2009) The Economic Effects of Significant U.S. Import Restraints: Fourth, Fifth and Sixth Updates. Investigation No. 332-325, Publications 3701, 3906 and 4904.

Ac

Vincent, D.P. (1982) Effects of higher oil prices on resource allocation and living standard: The case of Korea. The Developing Economies XX: 279-300.

Vincent, D.P. (1985) Exchange rate devaluation, monetary policy and wages: a general equilibrium analysis for Chile. Economic Modelling 2: 17-32.

33

Page 33 of 33