Agricultural Trade Protectionism in Japan: A Survey. Delbert A. Fitchett. No. 29. ...... Notes: Libya and Saudi Arabia are included only in the total. Growth rates.
Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized
WDP41 March 1989
41 U
World Bank Discussion Papers
Patternsof Development, 1950to 1983
Moshe Syrquin and Hollis B. Chenery
RECENTWORLDBANKDISCUSSION PAPERS No. 1.
PublicEnterprises in Sub-Saharan Africa. JohnR. Nellis
No. 2.
RaisingSchoolQualityin Developing Countries:What Investments BoostLearning?BruceFuller
No. 3.
A Svstemfor Evaluating the Performance of Government-Invested Enterprises in the Republicof Korea. YoungC. Park
No. 4.
CountryCommitment to Development Projects.RichardHeaverand ArturoIsrael
No. 5.
PublicExpenditure in LatinAmerica: Effectson Poverty.Guy P. Pfeffermann
No. 6.
CommunityParticipation in Development Projects:The WorldBankExperience.SamuelPaul
No. 7.
International FinancialFlowsto Brazilsincethe Late1960s: An Analysisof DebtExpansion and PaymentsProblems.PauloNogueiraBatista,Jr.
No. 8.
Macroeconomic Policies, DebtAccumulation, andAdjustment in Brazil,1965-84.CelsoL. Martone
No. 9.
The SafeMotherhood Initiative:Proposals for Action. BarbaraHerz and AnthonyR. Measham LAlsoavailablein French(9F)and Spanish(9S)]
No. 10.
Improving UrbanEmployment and LaborProductivity. Friedrich Kahnert
No. 11.
Divestiture in Developing Countries.ElliotBergand MaryM. Shirley
No. 12.
EconomicGrowthand the Returnsto Investment.DennisAnderson
No. 13.
Institutional Development and Technical Assistance in Macroeconomic PolicyFormulation: A Case Studyot Togo. SvenB. Kjellstrom andAyite-Fily d'Almeida
No. 14.
ManagingEconomicPolicyChange: Institutional Dimensions.GeoffreyLamb
No. 15.
DairyDevelopment and Milk Cooperatives: The Effectsof a DairyProjectin India. George mergosand Rogerb aade
No. 16.
Macroeconomic Policiesand Adjustment in Yugoslavia:SomeCounterfactual Simulations. Fahrettin Yagciand StevenKamin
No. 17.
PrivateEnterprise in Africa: Creatinga BetterEnvironment.KeithMarsdenand ThereseBelot
No. 18.
RuralWaterSuPplYand Sanitation:Time for a Change. AnthonyA. Churchill, with the assistance of Davidde Ferranti, RobertRoche,CarolynTager, AlanA. Walters,and AnthonyYazer
No. 19.
The PublicRevenueand EconomicPolicyin AfricanCountries:An Overviewof Issuesand Policy Options. DennisAnderson
No. 22.
Demographic Trendsin Chinafrom 1950to 1982. KennethHill
No. 23.
FoodImportDependence in Somalia:Magnitude, Causes,and PolicyOptions. Y. HosseinFarzin
No. 24.
The Relationship of ExternalDebtand Growth: Sudan'sExperience, 1975-1984.Y. Hossein Farzin
No. 25.
The Poorand the Poorest:Some InterimFindings.MichaelLipton
No. 26.
RoadTransport Taxationin Developing Countries:The Designof User Chargesand Taxesfor Tunisia.DavidNewbery,GordonHughes,WilliamD.O. Paterson, and EsraBennathan
No. 27.
Tradeand Industrial Policiesin the Developing Countries of EastAsia. AmarendraBhattacharya and JohannesF. Linn
No. 28.
Agricultural TradeProtectionism in Japan: A Survey. DelbertA. Fitchett
No. 29.
Multisector Framework for Analysisof Stabilization and Structural Adjustment Policies:The Caseof Morocco. AbelM. Mateusand others
No. 30.
Improvino the Oualityof Textbooksin China. BarbaraW. Searleand MichaelMertaughwith AnthonyReadand PhilipCohen
(Continued on theinside backcover.)
4 1K zI
WorldBankDiscussionPapers
Patternsof Development, 1950to 1983
Moshe Syrquin and Hollis B. Chenery
The World Bank Washington, D.C.
Copyright (C) 1989 The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing March 1989 Discussion Papers are not formal publications of the World Bank. They present preliminary and unpolished results of country analysis or research that is circulated to encourage discussion and comment; citation and the use of such a paper should take account of its provisional character. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. Any maps that accompany the text have been prepared solely for the convenience of readers; the designations and presentation of material in them do not imply the expression of any opinion whatsoever on the part of the World Bank, its affiliates, or its Board or member countries concerning the legal status of any country, territory, city, or area or of the authorities thereof or conceming the delimitation of its boundaries or its national affiliation. Because of the informality and to present the results of research with the least possible delay, the typescript has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to Director, Publications Department at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to photocopy portions for classroom use is not required, though notification of such use having been made will be appreciated. The complete backlist of publications from the World Bank is shown in the annual Index of Publications, which contains an alphabetical title list and indexes of subjects, authors, and countries and regions; it is of value principally to libraries and institutional purchasers. The latest edition of each of these is available free of charge from Publications Sales Unit, Department F, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'1ena, 75116 Paris, France. Moshe Syrquin is professor of economics at Bar Ilan University (Israel); Hollis B. Chenery is Thomas B. Cabot professor of economics at Harvard University and institute fellow at the Harvard Institute for International Development. Both are consultants to the World Bank. Library of Congress Cataloging-in-Publication
Data
Syrquin, Moises. Patterns of develoDment,1950 to 1983 / Moshe Syrquin, Hollis B. Chenery. p.
cm. --
(World Bank discussion papers ; 41)
ISBN 0-8213-1153-0 1. Economic history--1945- 2. Economic development. Holis Burnley. II. Title. III. Series. HC59.S8985 1988 330'.9'04--dcl9
I. Chenery, 88-31727 CIP
ABSTRACT
The main purpose of this paper is to provide more accurate measures of the dimensionsof structural transformationduring the process of development,by estimatinglong-run patterns of development for the period 1950-83. By including the turbulent decade after 1973 it tries to assess the stability of estimates of long run transformationand the robustnessof inferences derived from data about the pre-1973 period. The relativelylong time-seriesfor a large number of countries allow a more detailed examinationof the relation between cross-section and time-seriesestimates. The typology of development patterns used in previous studies is elaboratedand expanded. The study focuses on processes of resource allocation, specificallyon the structuresof final demand, trade, productionand employment. The samples consist of up to 108 economies over the period 1950-83. ACKNOWLEDGEMENTS This paper summarizesthe results of research project 673-51 sponsoredby the EconomicAnalysis and ProjectionsDepartmentwith collaborationfrom the Economic Research Department. Arabinda Kundu and Shujiro Urata were part of the research team. We would like to thank Yosi Deutsch for help and advice on the econometricpart, and Delfin Go, Bertha Namfua, and Narayana Poduval for help with computer work.
Contents
I.
3 Scope of the Study. .....*........................................ 3 ................................................. A. Objectives 5 as ................ B. Issues in Comparative Analysis.......... 7 ....................... Data and Procedures C. Econometric . 13 ................... *.. . ........ ... D. Working Hypotheses . 14 - Sources of uniformity ........................................ .14 ....... ........................ - Sources of diversity
II.
Dimensions of the Structural Transformation..............15 ... . 15 . ..... ....... A. The Period 1950-83.................. 18 ........................ B. The Transformation as a Whole 21 ............. .... .... - Resource allocation processes 26 ............. view integrated - Industrialization - an - Manufacturing - disaggregated results..............29
III.
34 Typology of Development Patterns............... 38 .... . . A. - Performance by type ..................... 43 ...... allocation B. - Variation in patterns of resource 47 C. - Classification of trade patterns. .................
III.
61 Changes Over Time ................o................................... 61 ............... A. Stability of Development Patterns.., - Accuracy of the new estimates......................63 65 - Principal time trends... .......................... 68 B. Average Time-Series: 1950-1983...................... 70 ....... - Individual time-series.. 75 .... - Average time-series..... ..........
Conclusions ..................................................... 81
ANNEXES A.
B. C.
Notes on the classification of countries.................90 .. *..*....96 Tests of homogeneity...... ... o ............... 99 ........... inflow The impact of variations in the capital
Statistical Appendix* .. *..................................
*
102
The statistical appendix and the three annexes prepared by A. Kundu are available from the Socio-Economic Data Division of the International Economics Department. a. Patterns in oil importing and oil exporting countries (A. Kundu).* b. Energy consumption and patterns of development (A. Kundu).* c. Patterns with ICP figures (A. Kundu).* - v
-
Intercountrycomparisonsare an important source for studying the associationof changes in economic structureand the level of development. In a series of studies beginning the the 1950's, Simon Kuznets establisheda number of empirical generalizationsabout long-term changes in economic structure that are a concomitantand actually define modern economic growth (1966, 1971). Kuznets also showed that the association between the interrelatedprocesses of change and the level of income found in the long-term experienceof the industrializedcountries,could also be observed in cross-countycomparisonsfor a given period. Previous comparative studies of structuralchange had focused on single processes such as the pattern of consumption (Houthakker,1957) and the sectoralcompositionof employment (Clark, 1940). In the series of detailed and comprehensivestudies published between 1956 and 1967, Kuznets presented the structuraltransformationas a whole rather than as a set of separate phenomena.
In Patterns of Development (Chenery and Syrquin 1975, henceforth cited as C-S) we chose a large set of the processes that characterizemodern economicgrowth and extended the approach in an econometricstudy for over 100 countries for the period 1950-1970. The processes studied centered around those most likely included in a minimal definitionof the structural transformation: accumulationof physical and human capital and shifts in the compositionof demand, trade, output and factor use. Also included were some socio-economicprocesses,such as urbanization,demographictransitionand
-2-
changes in income distribution,which appeared to be correlated with the level of development. The results in C-S presented a view of the transformationas a transitionfrom an economic structure representativeat low income levels to one typical for high income countries. The transformationof the economic structurewas further described in Chenery (1979), and elaborated in a recent comparativestudy of Industrializationand Growth (Chenery,Robinson and Syrquin, 1986). The patterns of industrialchange in C-S summarize the relationshipsthat exist along growth paths, where income is the measure of development. The patterns can be interpretedas reduced forms from a more general model. In Chenery and Syrquin (1986a,b)we presented a disaggregated simulationmodel of industrializationthat goes back to some of the underlying relations determining industrialchange; and also examined the behavior during the postwar period of about 40 countries that can be classifiedas semi-industrial. This group of countries showed an accelerationof growth and structuralchange since 1950. On the average, during the 3 decades 1950-1980, the transformationin semi-industrialcountries resembled the one that took place in industrialcountries over a period twice as long, when the latter were at a similar development stage. The use of patterns for country analysis has evolved from simple static comparisonsof actual structure to the predicted one, to analyses of long-run transformationin a comparativeframework. Particularlyuseful for assessing the specificfeatures of individualcountrieshas been the typology of allocationpatterns presented in Chenery and Taylor (1968) and in C-S and expanded in Chenery and Syrquin (1986b). In the latter study, the typology
- 3-
was used as a frame for a detailed examinationof the transformationin the postwar period in a sample of semi-industrialcountries. Chenery (1982) and Wood (1986) used the patterns approach to illustratethe distinguishingcharacteristicsof industrializationin large countries. Examples of comparative studies of long-term growth and structural change in individualcountries can be found in Syrquin (1986a and 1986c) on Israel, and Colombia;lOfer (1987) on the USSR; and World Bank (1985) on China.
I.
SCOPE OF THE STUDY
The present study relies on informationfor 108 economies during the period 1950-83 on aspects of economic structurerelated to the sectoral allocationof resources:demand, trade, production and factor use.
A.
Objectives
The main purpose of this study is to provide more accurate measures of the dimensions of the transformation,by reestimatinglong run patterns of developmentfor the period 1950-83. By including the turbulent decade after 1973 we try to assess the stabilityof estimates of long run transformation and the robustnessof inferencesderived from data about the pre-1973 period. The available informationgives a sample larger by about 50 percent from the sample in C-S. The relativelylong time series (over 30 years) for a large number of countries allows us to further explore the relation between cross-sectionand time-seriesestimates.
-4-
In section II.C we present a typology of developmentpatterns that elaborates our earlier work in this area (Cheneryand Taylor 1968, C-S, Chenery and Syrquin 1986b). Depending on the problem being analyzed and the objectives of the study, alternativetypologiescan be designed by combining structural featureswith policy aspects. In our presentationwe emphasize policies related to the external sector and their links with size and resource availability. This choice reflects the importanceof trade policies as an instrumentto influence resource allocation in the period studied. The processes analyzed are a subset of those in C-S, with some significantextensions. In C-S only two sectors producing commoditieswere distinguished:primary and industry. In the present study primary is now divided into agricultureand mining, and within industry,manufacturingand constructionare shown separately. Our recent comparative study (Chenery,Robinson and Syrquin 1986), shows the importanceof disaggregatingthe manufacturingsector for analyzing industrializationand the principal alternative sequences of the process. Chenery (1960), Chenery and Taylor (1968) and Maizels (1963), estimated the relation of industrialstructure with the level of development based on informationthat did not go beyond the early 1960's. In an unpublished report Prakash and Robinson (1979) extended the period to 1973. In the present study we make use of the Prakash - Robinson data base and extend the analysis to 1981. For the structuresof demand and productionwe estimate uniform patterns for shares in GDP in current and in constant prices. Differences in the time series between the two formulationare analyzed wherever we can establish systematicchanges in relative prices during the period.
-5-
B. Issues in ComparativeAnalysis Comparativeanalysis of structuraltransformation,while useful, has its own limitations. Some are addressed here and in other related work (Syrquin 1988), while others are inherent in the analysis and have to be recognizedas qualifications. Patterns of developmentrelating changes in strudture to the level of developmentprovide a concise picture of average long run transformationbut are not well suited to incorporatemarket behavior. Such a task requires a price endogenousmodel within a country. The two approachesare complementary. Intercountrycomparativeanalysis in a general equilibrium framework can provide orders of magnitude for key relations and parameters used in country models. The price endogenousmodels can be used to check the sensitivityof the average stylized facts to variation in relative prices and other variables not explicitlyconsidered in studies of long run transformation. (See Ranis 1984, and Chapter 11 in Chenery, Robinson and Syrquin, 1986). Changes in structure and performanceare interrelated. The patterns approach focuses primarily on the associationof the level of developmentwith those changes in economic structure necessary to sustain further growth. The patterns by themselvesdo not reveal the impact of structureon performance. Structuralchange is not sufficientfor explaininggrowth, but neither is the pure supply side approach that ignores changes in structure. Intercountrycomparisonsare of help in establishingstylized facts about the transformation,that give the expected changes in structureas a country develops. They are of little help in analyzing stagnationin countrieswith very low income. But even for such countries, the average
-6-
patterns of transformationare useful as indicatorsof feasible paths derived from the experienceof other countries (World Bank, China Report 1985). Patterns of development,based on intercountrycomparisons,are average relations showing the expected transformationduring the transition from a low income, agrarian economy to an industrialurban economy with substantiallyhigher per capita income. The same overall pattern of transformationcan accomodate significantdifferences in the timing and sequencingof particularaspects of change. The various paths may reflect differencesin initial conditions (size, resources), in the historical environment (world markets, wars) and in economic policies in the relevant period. "The search for uniform features of developmentalmost inevitably leads to a division of countries into more homogeneousgroups" (Chenery, 1986 p. 18). Statisticalconstraintsand the issues explored, limit the number of in a typology. But such limits of groups that is useful to distinguis'h conveniencedo not establish the "...notion that there are a limited number of different patterns of growth" (Papanek, 1977 p. 276). A closely related issue has to do with the level of generality or aggregagationat which the analysis is conducted. Are the associations representedby patterns of development (for example, the shift from primary productionto manufacturing)necessary conditionsfor sustained growth, or are they merely statisticalcorrelationsfor a given sample and period? At a broad level "...
the standard pattern of economic developmentis something
like a historicallikelihoodor near-necessity. The modern world could have evolved somewhat differently,but since it did not, it would be extraordinarilydifficult to change the standard pattern now" (Solow, 1977 p. 493).
- 7 -
Uniformitiesat a broad level of aggregation can hide wide variation in the behavior of individualcomponents. This notion resembles the concept of substitutabilityin Gerschenkron (1962) but modified to recognize that its nature depends not only on relative backwardnessbut also on the factors mentioned above as responsible for variations in development patterns. In a study on comparative long run economic growth in France and Britain,O'Brien and Keyder (1978, p. 196) conclude that " ...
there is more
than one way of transition from an agriculturalto an industrial economy and from rural to urban society". Earlier in their book they complain that "Economic theory lends no support to assumptions ...
that there is one
definable and optimal path to higher per capita incomes and still less to the implicitnotion that this path can be identifiedwith British as it proceeded from 1780 to 1914". (p. 18, quoted in industrialization Crafts, 1984). O'Brien and Keyder do not seem to question the notion of a transitionfrom "an agriculturalto an industrialeconomy" but only the view that the path of transition is unique. This assumptionof a unique and optimal path which they attack is nowhere implied in the analysis of the transformationbased on intercountrycomparisons. Recognizing the possibility of substitutabilityand of a typologyof development patternswould prevent economichistoriansof identifyingcountries " ...
as backward because their
coal and iron and cotton productionare relativelylow" (p. 16).
C. EconometricProceduresand Data To facilitatecomparisonwith previous studies we followed generally the methodology in C-S. The variables studied are listed in table 1, and the economies in table 2. As in C-S we excluded most of the communist economies
- 8 Table 1: CHARACTERISTICS ANALYZED ANDSAMPLESIZES
Variable
Final Demand (Share of GDP) Current price shares Constant price shares Private consumption Government consumption Investment Export Imports Food consumption current constant Merchandise Trade (Share of GDP) Exports of merchandise Primary products Fuels, Minerals and Metals Other primary Manufactures Imports of merchandise Primary products Manufactures
Symbol
No. of Period
No. of Countries
Observations
1950-83 1950-83
107 103
3019 2531
1953-82 1960-82
54 36
1126 662
1962-83
98
1829
1950-83 1950,55
104 92
2360 1921
1953-54 58-81
70
1043
1950-80
108
2710
1950-83
107
2513
1950,55 60-82
92
1764
C G I E M FCN
ECR EP EFMM EOP EM MMR MP MM
Production (Share of GDP) Current price shares Constant price shares
60-83 Agriculture Mining Manufacturing Construction Utilities Services
VA VN VM VC VU VS
Manufacturing (Share of GDP) ISIC
code Food, beverages and tobacco Textiles, apparel and leather Wood and furniture Paper and printing Chemicals, petroleum and rubber Non-metalic minerals Basic metals Metal products and machinery Other
31 32 33 34 35 36 37 38 39
Employment (Share of total) Agriculture Industry Services
LA LI LS
Relative Prices (1970-100) Demand Consumption Government Investment Exports Imports
PC PG Pi PE PM
Production Agriculture Mining Manufacturing Construction Utilities Services
PVA PVN PVM PVC PVU PVS
- 9 Table 2:
ECONOMIES INCLUDED IN THE STUDY
Type
Population 1965 (mill)
Afghanistan Algeria Angola Argentina
SP SP SM LP
11.115 11.923 5.347 22.283
223.8 2,111.6 831.9 1,982.7
Austria Bangladesh Belgium Benin Bolivia
SM LM SM SM SP
7.255 60.482 9.448 2.332 3.841
10,106.0 130.2 12,005.5 335.1 758.7
8,495.9
Brazil Burkina Faso Burma Burundi Cameroon
LP SP LP SP SP
84.292 4.595 24.250 3.131 5.825
2,001.0 218.2 172.8 239.6 737.4
3,243.4
Canada LP Central Afr. Rep. SP Chad -SP Chile SP China LM
19.678 1.735 3.307 8.310 746.800
10,249.4 349.7 112.7 2,399.2 289.6
11,096.8
Colombia LP Congo, PR SP Costa Rica SP Denmark SP Dominican Republic SF
18.488 1.066 1.490 4.758 3.719
1,282.0 987.3 2,048.8 12,616.1 1,158.5
2,756.9
Ecuador Egypt El Salvador Ethiopia Finland
SP LM SM LP SM
5.134 29.389 3.005 22.550 4.564
1,476.2 581.6 735.1 132.4 10,257.6
2,606.3
France Germany Ghana .Greece Guatemala
LM LM SP SM SP
48.758 58.619 7.767 8.572 4.615
12,213.0 13,331.4 374.0 4,301.6 1,077.3
9;849.7 10,224.7
Guinea Haiti Honduras Hong Kong Hungary
SP SP SP SM SM
4.137 3.950 2.304 3.598 10.153
299.1 274.2 636.8 5,467.7 2,033.0
India Indonesia Iran Iraq Ireland
LM LP LP SP SM
487.324 104.756 24.078 7.976 2.876
235.6 473.4 2,271.0 3,043.2 5,093.9
Israel Italy Ivory Coast Jamaica Japan
SM LM SP SM LM
2.563 51.987 4.159 1.749 98.883
4,749.7 7,059.8 1,212.6 1,097.3 8,906.4
Jordan Kenya Korea, Rep. Lebanon Liberia
SM SM LM SM SP
1.962 9.521 28.7i)9 2.151 1.139
1,137.3 412.8 1,606.6 1,841.7 521.7
Economy
GNP Per Capita 1980 1980 (USS) (ICP$)
Level of Income Lower Upper Low Middle Middle Income Income Income (29) (29) (29)
Industrial (19)
x x x 3,795.6
x
x x
x x x
x
1,502.2 x x x x 798.1
x x x x
3,508.4
x x x x
3,032.5 9,569.8 1,950.3
1,402.9 273.3 8,465.0
x x x x x x x x x x x
5,175.6 2,306.9
x x x x
1,137.2
x x x
4,881.0 551.7 1,060.0
x x x x
5,576.8 6,596.7 7,966.3 1,299.3
x x x x
8,414.4
622.9 2,541.4
x
x x x x x x
-
10
-
GNP Per Capita 1980 1980 (ICP$) (US$)
TYpe
Population 1965 (mill)
Libya Madagascar Malawi Malaysia Mali
SP SP SP SP SP
1.623 6.080 3.919 9.531 4.558
10,899.4 366.5 197.6 1,653.3 199.4
Mauritania Mexico Morocco Mozambique Nepal
SP LP SM SP SP
1.085 43.500 13.323 7.263 10.344
433.1 2,615.4 948.2 322.7 133,8
Netherlands New Zealand Nicaragua Niger Nigeria
SM SP SP SP LP
12.377 2.628 1.613 3.510 58.490
11,910.9 7,288.0 784.5 326.6 991.7
9,359.7
Norway Pakistan Panama Papua New Guinea Paraguay
SM LM SM SP SP
3.723 52.414 1.269 2.141 2.019
13,572.2 311.8 1,861.7 815.1 1,359.2
10,882.4 986.1 3,301,0
Peru Philippines Portugal Puerto Rico Rwanda
SP LP SM SM SP
11.230 31.771 9.199 2.594 3.250
1,117.7 730.6 2,462.9 3,386.9 222.9
2,491.2 1,725.8 3,894.4
Saudi Arabia Senegal Sierra Leone Singapore Somalia
SP SP SP SM SP
4.793 3.919 2.304 1.887 2.816
12,640.3 501.6 349.4 4,509.3 274.8
South Africa Spain Sri Lanka Sudan Sweden
LM LM SP SP SM
19.467 32.056 11.133 12.359 7.734
2,666.4 5,615.9 271.2 367.3 14,738.4
Switzerland Syrian Arab Rep. Taiwan (China) Tanzania Thailand
SM SP SM SM LP
5.856 5.325 12.443 11.595 31.241
16,620.6 1,512.3 2,268.8 264.2 706.9
Togo Tunisia Turkey Uganda United Kingdom
SP SM LP SP LM
1.704 4.630 31.151 8.432 54.436
430.4 1,369.4 1,312.6 235.3 9,361.1
United States Uruguay Venezuela Yemen, AR Yugoslavia
LP SP SP SP LM
194.303 2.693 9.169 4.659 19.434
11,562.4 3,450.2 3,828.0 441.7 3,045.0
Zaire Zambia Zimbabwe
LP SP SP
19.524 3.643 4.268
Economy
Notes:
203.1 619.4 762.3
Level of Income Upper Lower Middle Middle Low Income Income Income (29) (29) (29)
Industrial (19)
x 565.7 402.1
x x
335.1
x
x
x x
4,840.9 1,302.5
x x x x x x x x
896.3
x x
x x x x
2,210.1 x x
x x x
x
675.0 x
x x x 6,326.3 1,189.6
x x x x x
x x 374.8
x x x x
1,994.9
x x 8,271.2
x
x
x
11,562.4 4,141.6 5,233.1
x x x
3,863.4 x 699.0 933.0
x x
The type refers to the classificationin part II.C and table 7: LP-large, primary oriented; IMLarge, manufacturingoriented; SP-small, primary oriented; SM-small, manufacturingoriented. The groupings by level of income and according to the trade position in oil are from the 192 World DevelopmentReport. The ICP real income figures are preliminaryestimates from phase IV of the International ComparisonsProject.
-
11
-
and countrieswhere the population in 1965 was less than one million.2/ For each variable x, expressed as a share of GDP (or of total labor force in the case of employment),equation (1) was estimated in two variants: with and without the capital inflow ratio (F).
x = a + 0 lny + 82 (lny)2 + y lnN + y2 (lnN)2 + E6.T. + eF
(1)
where x
=
dependent variable (see table 1),
y
=
per capita GNP in 1980 dollars,
N
=
population in millions,
F
=
imports minus exports of goods and nonfactor services as a
share of GDP, and Ti=
dummy variables for time periods taking a non-zero value as follows: T2 = 1 if t
>
1960,
T2= 1 if t
>
1967,
T3 = I if t
>
1973, and
T4 = 1 if t
>
1979.
The time variables measure uniform shifts of the relations across countries,and are defined in a incrementalway. For example, the coefficient to T2 measures any shift after 1967 over and above the post-1960 one given by the coefficientof T1. The semilog formulation is a convenientone for the analysis of structuralchange because of its adding-up property. The fitted equations and derived predicted values from a common semilog formulationfor the components
- 12 -
of an aggregate add up identicallyto ithefitted equation and predicted value for
provided
the aggregate,
all
estimates
refer
to exactly
the same sample.
Equation (1) was run for pooled samples combining the individualtime series for all countries or for groups of countries according to the typologies described above. In these regressionsmost of the variance to be explained is in general still due to variation among countries, but to a lesser extent than in previous studies. This is so because of the length of the time-seriesin the present study and the substantialgrowth and transformationexhibited in the group of newly industrializingcountries since the early 1950's. The individualtime-seriesare also analyzed directly in two ways. Average time-seriesrelations are estimated (with and without F), by allowing each country to have its own intercept as in equation (2):
x = a.
+
where a.
+ B2 (lny) 2 +
llny
= intercept for country i.
By allowing the variation only.
countries parameters
The estimated estimates,
variables.
Since
growth of population
to have its
each country
between
time-series
terms
(2)
+ eF
MT
and pool the within-countries are weighted
with weights
in time-series
related
analysis
is indistinguishable
from the equation.
own intercept
averages
all
variation
of the individual
to the variance any uniform
we eliminate
change
from a time trend,
of the explanatory such as the we omit the N
- 13 -
Individualtime-series relationswithin countries are also estimated in all cases where a minimal number of annual observationswere available. In these regressionsonly lny appears as explanatory variable. The issue of stability of development patterns is discussed in section III.A. A central question there is whether we can identify a structuralbreak in the relations after 1973. At the aggregate level we address that question by comparing the average cross-countrypatterns before and after 1973. For individualtime-serieswe apply, in a separate paper, the cusum test introducedby Brown, Durbin and Evans (1975) and the recent extension of the method to analyze panel data of Han and Park (1986).
D.
Working Hypotheses The structuraltransformationof an economy, comprises a set of
interrelatedprocesses of change. Shifts in the internal allocationof resources among sectors are the result of the interactionof changes in the compositionof demand, and variationson the supply side. On the demand side the changes are derived from the pattern of income elasticitiesof demand, and on the supply side they are the effect of factor accumulationand productivitygrowth. The demand and supply effects are not totally independentfrom one another. Thus changes in demand between internaland external sources reflect changing comparativeadvantage;while aggregate productivitygrowth incorporatesresource shifts from low productivityto higher productivitysectors. Various models based on cross-countryinformation,have studied the interactionof the various elements leading to change. Recent examples include the price endogenousmodel of Kelley and Williamson (1984) and our
- 14 -
model of industrialization(1980, 1986a, and Syrquin 1986b). This model and the original paper on "Patterns of IndustrialGrowth" (Chenery, 1960) provide the rationale for expecting systematicassociationsof economic structurewith the level of development in cross-countrycomparisons.
Sources of Uniformity The main sources of uniformityare the pattern of final and intermediatedemand and the evolution of comparativeadvantage. In final demand the best established trends are the decline in the share of food in consumptionand the rise in the share of resources allocated to investment. Industrializationusually increases the share of intermediatesin total gross output while varying its composition from primary to manufacturingoutput. The rise in the ratio of capital (human and physical) to labor and the observed higher rate of productivitygrowth in the more modern sectors of the economy, tend to shift the comparativeadvantage from primary activities to industrialones.3/ An additional source of uniformity is the internationalenvironment during the period under observation,which includes imitative (demonstration) effects on consumptionand on development strategies.
Sources of Diversity Of the factors that affect the transformation,the most variable is the extent of participationin the internationaleconomy. The level and compositionof trade and hence the type of specializationare largely determinedby the interplayof structure (size and resource availability)and policy. As argued above and shown in section II.C, differences in the type of
- 15 -
specializationaffect more the timing of the transformationthan its overall nature.
II.
DIMENSIONSOF THE STRUCTURALTRANSFORMATION
In this part we look at the transformationas a whole, and present the dimensionsof change as reflected in intercountrycomparisons of economic structureduring the period 1950-83.
A. The Period 1950-83. A brief description of key features of the internationalenvironment and of average growth performanceduring the period, might be useful as backgroundfor the results in the following sections. The decade following the second World War was a period of reconstructionin Europe and of a significantdrive to promote developmentin various parts of the World: South Europe, Latin America and in a large group of newly independentnations that emerged with the dismantlingof Empires. This drive was influencedby the rivalry among political blocks after the War, and by a change in perceptionsabout the role of the state in fostering economic development. The network of internationaltrade suffered severe blows during the depressionof the 1930's and the War. Memories of the breakdown of this network coupled with the new responsibilitiesadopted by many states for acceleratinggrowth, resulted in development strategiesthat were largely inward orientated. This is the more significantsince by the early 1960's, world trade was expanding at rates that had not been observed for decades.
- 16 -
The main participantsin this expansion were the advanced countries and, since the mid 1960's an increasinggroup of semi-industrialcountries that abandoned the inward strategy in favor of one that emphasizes a greater participationin the internationaleconomy. The four fold increase in the price of oil in 1973 and the collapse of the Bretton Woods System, changed drastically the economic environment. The rise in the price of oil aggravatedinflationarypressures and led to a recession in OECD countries,which was then magnified in less developed economies. The expansion of world trade, and the growth of output and productivityslowed down considerably. Recovery was halted abruptly by the second oil-shock in 1979. This time it was accompanied by interest rates in internationalmarkets that reached unprecedentedlevels. The foreign debt in various countries reached crisis proportions,and the recession became a serious depression in various regions, primarily in Latin America. Average growth in total and per capita income during the three decades 1950-80,was significantlyhigher than in any comparableperiod in recent history. Growth rates were high in almost all regions of the world, except for the group of very low income countries in sub-SaharaAfrica. Average growth rates for the whole period studied,are shown in table 3 for the economies in our sample grouped according to the classificationin the 1985 World DevelopmentReport. The figures are simple averages of least square estimates within countries for as many years as the data permitted. (See note to the table). For the complete sample, income per capita grew on the average at 2.4 percent per year. At this rate after 30 years income per capita would
-
Table 3:
17 -
AVERAGE ANNUAL RATES OF GROWTH DURING 1950-83: COUNTRIES GROUPED BY INCOME LEVEL
Annual Growth Rates (percent)
Group All Low income Lower middle income Upper middle income Industrial
Notes:
Number of Countries
Per Capita Income
GNP
Multiple of Initial Income Per Capita after 30 Years
108 29 29 29 19
2.4 0.8 1.9 3.6 3.2
4.6 3.0 4.7 6.0 3.7
2.04 1.27 1.76 2.89 2.57
Libya and Saudi Arabia are included only in the total. Growth rates computed by least squares regressions for all observations available within a country. 60 countries had 30 or more annual observations, 41 countries had between 24 and 29 observations and 7 had less than 24 observations (see Table 7).
-
18
-
have doubled. The rate was not uniform among countries or groups. Seven of the 29 economies in the low income group had negative rates of growth. In the 22 with positive growth, the average rate equals 1.4 percent. Among LDC's there is, in table 3, a clear accelerationof growth as income goes up, reaching a rate of 3.6 in the upper middle income group. At this rate the initial level would increase by a factor of about 3 in 30 years. A sizable number of countries performed even better than this average rate, multiplying their starting income level by a factor of 4, or even 5. This is very significantfor a study of patterns of development since it implies that, within the period of observation,a number of countries traverseda large segment of the transition range. We can therefore reexamine,with a more solid data base, the concept of a transitionfrom one state to another that we advanced in C-S (p. 135), to replace the notion of a dichotomy between less developed and developed countries. B.
The Transformationas a Whole In C-S the transitionwas representedby the income interval $100 -
$1000 in 1964 US dollars, based on the observation that about 75 to 80 percent of the transformationin structure takes place within this range (p. 19). In this study we define the transitionrange in 1980 US dollars as the interval from $300 to $4000 per capita GNP. These revised figures account for inflation since 1964, and reflect the strong (but apparently little noticed) trends in the pattern of real exchange rates between the two dates. Real exchange rates in developingcountries have tended to depreciate ralative to the average for industrialeconomies, and the lower the income level, the greater the depreciation. (Syrquin 1985, Wood 1987).
-
19
-
Table 4 shows the overall pattern of transformation,in the principal variables studies,derived by estimating equation (1) (omitting F), for a country of average size (N=20). The effects of variations in the capital inflow are discussed in annex C. The estimated regressionsappear in table Sl in the StatisticalAppendix. The predicted values for the selected income benchmarksrefer to the period after 1973. That is, in calculating those levels, T1 = T2 = T3 = 1 and T4 = 0.
Average values in the sample are also
given for countries with per capita income in 1970 below $ 300 ("low income") and countries with per capita income in 1970 above $ 5000 ("high income") except for Libya and Saudi Arabia. The differencebetween these two average values is a measure of the magnitude of change during the transition. It appears in the column before last in table 4. Finally, to bring out the differences in timing, the last column indicates the level of income at which half of the total change has taken place, for variables where change is significantand monotonic. Table 4 is similar to table 3 in C-S. The main difference is that in this work we only study variables related to what we label there "Resource AllocationProcesses",but at a more disaggragatedlevel. The sets of resource allocation processes in table 4 represent the principal featuresof economic transformationidentifiedas industrialzation. The overall picture in table 4 and in figures 1-4, is not much different from the one in C-S. Shifts in the relations primarily after 1973 are discussed in section III.A. In that section we also examine the accuracy of the estimates underlying table 4 as measured by the standard errors of estimate, (SEE).
- 20
Table 4:
-
AVERAGE VARIAT'IONIN ECONOMIC STRUCTURE WITH LEVEL OF DEVELOPMENT FOR POST 1973 PERIOD (Population = 20 million)
Meana under $300
Income per capita (1980 US$) 300 500 1000 2000 4000
Meanb Over $5000
Total change
y at mid point
Final Demand Private Consumption Gov. Consumption Investment Exports Imports
.79 .12 .14 .16 .21
.733 .136 .184 .193 .246
.702 .135 .208 .207 .252
.664 .137 .233 .226 .260
.631 .144 .250 .245 .270
.603 .154 .259 .264 .280
.60 .14 .26 .23 .23
-.19 .02 .12 .07 .02
600 400 400 -
Food Consumption
.39
.387
.345
.291
.239
.189
.15
-.24
1200
.14 .03
.152 .048
.169 .063
.188 .073
.203 .072
.212 .061
.18 .02
.04 -.01
400 -
.10 .01
.091 .013
.086 .020
.079 .037
.069 .061
.057 .094
.05 .11
-.05 .10
1250 2000
.16 .05 .11
.182 .064 .118
.193 .067 .126
.206 .071 .135
.217 .075 .142
.227 .080 .147
.19 .07 .12
.03 .02 .01
Production Agriculture Mining Manufacturing Construction Utilities Services
.48 .01 .10 .04 .06 .31
.394 .050 .121 .044 .067 .324
.317 .066 .148 .049 .074 .346
.228 .077 .181 .055 .081 .378
.154 .075 .210 .061 .088 .412
.097 .061 .236 .067 .093 .447
.07 .01 .28 .07 .10 .47
-.41 .18 .03 .04 .16
700 1200 1000 900 1300
Labor Force Agriculture Industry Services
.81 .07 .12
.749 .092 .159
.651 .132 .217
.517 .192 .291
.381 .256 .363
.242 .326 .432
.13 .40 .47
-.68 .33 .35
1300 1600 1000
Trade Exports: Total merchandise Fuels, minerals, metals Other primary Manufacturing Imports: Total merchandise Primary Manufacturing
a
Approximately $180. $300 in 1970.
Means values for 1960-72 of countries with y under
b
Approximately $7300. $5000 in 1970.
Mean values for 1960-72 of countries with y over
-
-
- 21 -
Resource Allocation Processes The principal results in table 4, for the average economy, are first briefly discussed by process and then combined in table 5. Demand: The transformationin final demand is one of the most uniform features of the process of development. On the average the share of private consumption in GDP declineswith the level of income allowing a rising investmentshare and a lower import surplus. Food consumptiondeclines by about 20 percentagepoints while nonfood consumptiongoes up. The shift from consumption to investmenttakes place early during the transition;the decline in food consumption is spread over a wider income range (see last column in table 4). Trade: Only a small part of the variation in aggregate trade can be related to income. In the compositionof exports we do find a systematicshift from primary products to manufactures,mostly in the upper levels of the transition. No such change takes place in imports, for which there is an increase in both components. Industrializationincreases the demand not only for primary imports but also for imports of manufactures. Only in the case of large countries (see below section II.C) do we find a decline in manufactured imports clearly related to early import substitutionin those countries. The changes in final demand and trade reinforce each other and combine with complementarychanges in intermediateuses and productivity growth to produce a more pronounced shift in the structuresof productionand labor use. Production: The share of value added in agriculturedeclines sharply over the transition,while manufacturingand social overhead (construction plus utilities) double their share and the services sector rises its share by
gY" 1
STRUCTURE OF DOMESTIC DEMAND 0.8 0.7 s 0.6 H R 0.5E S 0.4 -
o
FC~~~~~N
F 0.3
P 0.2-
0.0 -
250
500 GNP PER CAPITA
1000
2000
(1980 U.S. Dollars)
4000
- 23 -
a 2.
STRUCTURE OF MERCHANDISE TRADE:EXPORTS 0.22
_
0.20EM
0.18 S 0.16
A
R 0.14
E S 0.12 0 0.10 F G 0.08
_OP ,
__________
D p 0.06
M
-
__
-
-
-
0.04
EM--
0.02 0.00
500
250
4000
2000
1000
GNP PER CAPITA (1980 U.S. Dollars)
"-
2&
STRUCTURE OF MERCHANDISE TRADE: IMPORTS 0.24M
0.220.20S 0.18
H
R 0.16
----
MM -------
00.14
00.12 F
-
G 0.10p 0.08
MP
___
0.06 0.04 0.02 250
500
1000
2000
GNP PER CAPITA (1980 U.S. Dollars)
4000
24 -
-
,Sp-
S
STRUCTURE OF PRODUCTION 0.5s _
0.45
VA
0.40
Vs
_
-
S H 0.35. A
-
R030 S
VM
0.25 20--,---F 0.20 F0
G D 0.15
\
---
---
VC+VU
=
P.o.50
soo50
1000o
2000
4000
GNPPER CAPITA(t980 U.S. Dol lars) 0.05 ---
4
LABORALLOCATION 0.8-
S 0.6 H\0.8. 250
500
IC000
2000
GNP PER CAPITA ( 1980 U.S. Dol l ars ) P0.7.
00X 50.60.42500 F~
~ ~
GPPRCPT
00 18
0040 .Dlas
4000
- 25 -
about 50 percent. The timing of the shift is an average of the early shift in demand and the later one in exports. The nature of the transformationin the productionstructureduring the transitionagrees in general with that predicted in C-S. The principal differenceis the smaller rise in manufacturingand industryin this study, which probably reflects the fall in the manufacturingshares in output and employmentthat has taken place in virtuallyevery single developed country since the late 1960's. The (see for example OECD 1979 phenomenonhas been labeled de-industrialization and Blackaby 1978). A useful benchmarkwas identifiedin C-S as the income level at which the rising industrialoutput share surpassesthe decliningprimary share. In C-S at approximatelyy=300 (1964 US$) both industryand primary account for about 26 percent of total value added. This finding is almost identical in the present study, in which the two broad sectors equal about 25.5 percent of output at y = 1500 (1980 US$). For the more narrow classificationof agricultureand manufacturingthe crossover income level is 1400 (1980 US$) at which each represents19 percent of total product. The slower rise in manufacturingin table 4 (than in C-S) is accompaniedby a faster increase in services. It has been argued (in Kravis, Heston and Summers 1983) that this rise in the output share of sevices is wholly accounted for by the systematicrise in the relativeprice of services with income across countries. They base this argument on the results of phase III of the internationalcomparisonsproject (ICP). However, given that the ICP includes expenditurecategoriesonly and the limited sample studied so far, their results should be regarded as illustrativerequiringfurther
- 26
-
study. We return to this issue in section III.B where we discuss the time-seriesresults for the whole period. Employment: The qualitativechange in the sectoral compositionof the labor force is similar to the one in value added, but there are also importantdifferencesin the magnitude of the shift and in its timing. The decline in the share of agriculturein employmentis more pronouncedthan in production,but since it starts from a much higher level and takes place at a relativelyhigher income level, it leads to a decline in the relative productivityof labor in agriculture(share in value added divided by share in 4 employment). Only by the end of the transition(around $ 3000) does the
trend reverse itself and the gap in average productivitybegins to narrow.
Industrialization- An IntegratedView The sectoralcorrespondenceof the structuresof demand, trade and productionin table 4 is only approximate. Strict comparabilitywould require matching the classificationschemes (ISIC, SITC), and an interindustry framework to allocate expenditurecategoriesto industriesand to account for intermediategoods. The framework of a multisectoralmodel, allows also the study of the interrelationamong the sets of resourceallocationprocesses. A model of industrializationas a system-widephenomenonwas presented in a study of the transformationof the Japanese economy (Chenery,Shishido,and Watanabe, 1962) and was subsequentlyrevised and adapted to simulate the transformationover the complete transitionrange, on the basis of cross-countrydata (Cheneryand Syrquin, 1980, 1986a). As suggestedby those models, the estimates of the processes of change in table 4 can be related through the material balance equations:
- 27 -
Xi = Wi + Di + Ti
()
Vi = Vi Xi
(4)
where the index i refers to a sector, X is gross output, W is intermediate demand, D is final demand, T is net trade (exportsminus imports), V is value added, and v is the value-addedratio. In this study, all the componentsare expressed as shares of GDP which is equal to ZV. = V.
Combining (3) and (4) and dividing throughoutby
VJ:
Vi/V = vi[Wi/V + Di/V + Ti/V]
(5)
Changes in the sectoral shares in value-addedcan be accounted for by changes in the compositionof demand (intermediateand final), changes in the compositionof trade, and changes in the value-addedcoefficientas in equation (6):
,A(Vi/V) = vi[A(Wi/V)+ A(Ti/V)] + (Vi/V) Avi/vi
(6)
(a bar over a variable means that its value is set at the mean of the initial and terminal levels). The elements in equation (6) are not all immediatelyavailable in table 4. For intermediateproduction(v and W) we rely on a recent
- 28 -
comparativestudy of interindustryrelations that derived some systematic patterns of change from data on 83 input-outputtables (Deutschand Syrquin, 1986). We focus on commoditiesproductiononly, and combine agriculture and mining into one sector called primary. We also assume that food consumption generates demands from the primary sector only, and that the manufacturing sector supplies one half of non-food consumptionand investment (the other half represents constructionand other nontradables). With these assumptions we now compute equation 6) for the whole transitionrange from $300 to $4000. The results are summarizedin table 5.5/ Over the course of the transitionthere is a significantshift in value added from primary productionto manufacturingand nontradables. The average patterns in table 5 show a very close correspondencebetween the directly estimated shift (the last row) and the one calculated by the right-hand-sideof equation (6). Changes in domestic demand (Engel effects) account directly for less than one half of the change in structure,and changes in net trade for about ten percent on the average. The contribution of intermediateshas two components. First there is a very significant increase in the demand for manufacturingproducts to be used as intermediates and a decline in the relative use of intermediateinputs from the primary sectors. These trends reflect the evolution from a comparativelysimple to a more diversified,roundaboutsystem with a higher degree of fabricationand specialization. The substitutionof fabricatedmaterials for natural ones is due to changes in technologyand also to changes in relative prices. The second component refers to variationsin the ratio of value-addedto gross output in a sector. In agriculturethis ratio tends to decline with the rise in income, or equivalently,the use of purchased intermediateinputs per unit
- 29 -
of output tends to increase. As shown in table 5, this factor accounts for about one fourth of the decline in the share of primary in total GDP. In an input-ouputmodel, the variation in intermediateuse can be further attributed to changes in final demand, trade, and input-outputcoefficients. Such a decompositioncan be found in the models cited above (for example in Chenery and Syrquin 1980).
Manufacturing- DisaggregatedResults During the process of industrializationthe composition of the manufacturingsector changes considerably. At a more disaggregatedlevel, country specific features and policy become more prominent in determining the pattern of specialization. Large countries can better exploit economies of scale within their domesticmarkets, and can more easily afford a strategy of import substitutionon a wide front. Variation in resource endowments is expected to generate differencesin productionpatternswithin manufacturing, particularlyin small economies. Nevertheless,various studies have shown that a high degree of uniformitystill remains in the pattern of change within the industrialsector (at the two digit level of the old ISIC classification), both among countries and over long periods of time in the advanced countries. Time series for developingcountries were examined in Chenery and Taylor (1968) for the period 1950-63. To account for the expected differences in specializationdue to size and resources,Chenery and Taylor divided the sample into three more homogeneousgroups, and estimated average patterns of change within groups. The typology in the next section can be seen as an extension and refinementof this original effort. In an unpublished study Prakash and Robinson (1979) extended the analysis for the period 1953-73. In
Table
5:
STRUCTURALCHANGEOVER THE TRANSFORMATION
Impact on the share of: Primary Manufacturing 1. Changes in final demand A(D. /V) Food consumption One half of non-foodC and investment
-.20 .08
2. Changes in intermediatedemand A(Wi/V)
-.06
.18
-. 05
.05
-.31
.31
3. Changes in net trade A(TI/V)
= A(Ei/v
- mi/V)
4. Changes in output A(X /V) = (1) + (2)
+ (3)
5. Mean value-addedratio vi
.71
.35
AVi
-.20
.03
Av1./V.
-.27
.09
-. 22 -.08
.11 .015
.44 .16 -.28
.12 .24 .12
6. Changes in the value-addedratio
7. Impliedchanges in value-added share due to: v.
A(X./V)= (5) x (4)
1
1
(v?7V &vi/v 8.
Value-added
-.30
.125
shares
Predictedshare at: y=300 y=4000 Changes in shares
- 31 -
the present study we started with the data base from Prakash-Robinson,and added annual observationsthrough 1981. The final data set used in this study refers to the period 1953-81,and although the coverage for the 1950's is very sparse,the series are long enough to study the evolutionwithin developing countriesand to assess the stabilityof the relationsafter 1973. The analysisof the time series and of the stabilityof the estimates is presented in part III. In this section we focus on long run industrialchange in the average economy. In this study we follow the two-digitlevel of aggregation of the revised ISIC. At this level, nine separatebranchesare distinguished (see table 1). To assure compatibilitybetween the aggregageshare for manufacturingfrom the national accounts (VM) and the disaggregatedfigures (based on industrialcensusesand surveys),we imposed the national accounts figure for the aggregateand adjustedproportionatelythe data of the subsectorsfor all observations. Table 6 shows the standardvariation in industrialstructurewith the level of development,in the same format as the aggregate results in table 4. There have been various attempts in the literatureto group industrialsectors into homogeneouscategoriesdiffering in the demand for their products, their technologyor their dynamism. Hoffmann (1958) stressed the systematicdecline in the ratio of consumerto producergoods, while at the EconomicCommissionfor Latin America (1964) the labels became more emotive: dynamic and vegetativebranches. In table 6 and figure 5 we present two groupingswhich we have used in previouswork. In Chenery and Syrquin (1986a),14 manufacturingsectors are distinguishedand the results are presentedat a four sector level. In table 6 we further combine food products and consumer goods into light industry,and producergoods and machinery into
-32
-
Table 6: AVERAGE VARIATION IN INDUSTRIAL STRUCTURE WITH LEVEL OF DEVELOPMENTFOR POST-1973 PERIOD (Population = 20 million)
ISIC Code
Mean under $300
3
.119
.120
.151
.188
.219
.244
.269
.150
1200
31
.028
.042
.045
.047
.046
.042
.040
.012
-
32
.034
.026
.030
.033
.034
.032
.029
-
-
33
.002
.004
.005
.006
.008
.010
.014
.012
2000
34
.005
.003
.004
.007
.011
.016
.025
.020
3000
35
.018
.024
.030
.036
.040
.042
.034
.016
400
36 37
.005 .006
.005 .005
.008 .009
.011 .013
.012 .017
.013 .019
.014 .020
.009 .014
700 1000
38 39
.018 .003
.010 .001
.019 .001
.032 .003
.048 .003
.066 .004
.087 .006
.069 -
2500 -
31-34,39
.072
.076
.085, .096
.102
.104
.114
.042
900
35-38
.047
.044
.066
.092
.117
.140
.155
.108
1500
31,32,39 33,35,36 34,37,38
.066 .024 .029
.069 .033 .018
.076 .043 .032
.083 .053 .052
.083 .060 .076
.078 .065 .101
.074 .062 .132
.038 .103
500 2500
Sector
Manufacturing Food, beverages and tobacco Textiles and clothing Wood and products Paper and Printing Chemicals and rubber Non-metalic minerals Basic metals Metal products and machinery Other Light industry Heavy industry Early Middle Late
Income per capita (1980 US$) 300 500 1000 2000 4000
Mean Over $5000
Total change
y at mid point
- 33 -
heavy industry. The figures show that as income rises the composition of manufacturingshifts from light to heavy industry. The early increase in light industry is generally the result of domesticdemand and the opportunitiesfor import substitutionwhich are exhausted at an early stage. Static comparisonsof relative labor productivityand capital intensity within manufacturing(Syrquin 1986b), suggest a higher use of capital in heavy industry, particularlyin producer goods, accompaniedby higher levels of labor productivityand wages. Part of this difference is due to a higher level of skills, especially in some branches of the machinery sector. Over time, the growth of labor productivityand of total factor productivitytends to be higher in heavy than in light industry. Economies of scale are more prevalent in heavy than in light industry and correspondingly, the weight of small scale firms in smaller. Chenery and Taylor (1968) examined 11 branches of manufacturingand grouped them into three categories "accordingto the stage at which they make their main contributionto the rise of industry" (p. 409). Early industries are establishedat low income levels to satisfy the essential demands of the population. They are characterizedby simple technologiesand low income elasticitiesof demand. Their share in GDP remains static during the transition (within manufacturingtheir share goes down significantly), although there are some recent exceptionswhere the output of some branches in this group expanded rapidly for exports. Middle industriestypically double their share in GDP early in the transitionbut show little further increase. A large proportionof their output is used as intermediateinputs by other sectors. This source of demand expands fast in the lower income levels when the matrix of interindustry
- 34 -
relations becomes more dense. Income elasticitiesfor the finished products from the group of middle industriesare generally above unity. The group of late industriesaccounts for virtually all of the increase in the manufacturingshare in the latter stages of the transformation. This group includes investmentgoods, some intermediates,and durable consumer goods with high income elasticitiesof demand. Some indicationabout the changes in intermediatedemands for the productsof the aggregate groups in table 6, can be obtained from unpublished calculationsdone for the study of Deutsch and Syrquin (1986). The change in the ratio of intermediateuses of manufacturesto GDP over the range $300 to $4000, equals 18 percentage points. Of these, only 3 points originate in light industrywhile the other 15 points come from heavy industry. The classificationin Deutsch and Syrquin allows only an approximatematching with the early-middle-latedivision. The approximateallocation shows no change in the ratio of sectoral intermediatedemands to total GDP in the early group, an increase of 7 percentagepoints in the middle industriesand an increaseof 11 points in late industries. The sources of structuralchange in industry are analyzed in some detail in Chenery,Robinson, and Syrquin (1986, part II). Alternative patterns of specializationappear below in section II.C.
C.
Typology of Development Patterns Average patterns of developmentare a useful starting point. They
provide an initial reference point stressingthe uniformitiesof the transformation. Various other factors influence the processes of change. Our hypothesisis that these factors affect primarily the timing and sequence of
- 35 -
.za
S.
(A) ADDEDIN MANUFAC'TURING STRUCTURE OF VALUJE 0.26
VM
0.24 (J.22
s 0 20 A 0 18 R E 0 16 S014
HEAVY
0 0.124 LIGHT G0.I0
-------------------
.0
po 0
- ----
-
0.06 0.04
0.02
_
500
250
2000
1000
GNP PER CAPITA (1980 U.S.
4000
Dollars)
y... 6&
(B) STRUCTURE OF VALUEADDEDIN MANUFACTURING 0.12. 0.11 LATE
0.10 S. H
009 0 .EARLY
A 0.08
-
R E 0.07
MIDDLE ,
0. 06 F
0.05
.-- ,-
G00.04 P 0.03 0.020.01 0.00
250
500
1000
GNP PER CAPITA (1980 U.S.
2000 Dollars)
4000
- 36 -
structuralchange and less its overall nature. In this section we focus on selectedcharacteristicsof economic structureand policy that have a systematiceffect on resourceallocation,and present a simple typologyof developmentstrategies. We then exploit the large samples to estimate separateregressionsfor each group. The approach in this study extends our earlier typologyof allocation patterns (Cheneryand Syrquin, 1975 and 1986b. See also Chenery and Taylor, 1968 and Chenery, 1979). It is based on a statisticalanalysis of the level and orientationof exports, and incorporatesstructuralcharacteristicsand policy. The typologyrecognizesthree dimensions: a)
Size: Economies are separated into small and large on the basis of
their populationsize in 1965. Other measures of size and their relation to populationare discussed in Perkins and Syrquin (1986),where the very large countriesare singledout for further analysis (Chenery,1979 chapter 3, and 1982, and Wood 1986, also focus on the distinguishingfeaturesof the very large countries). b)
Openness: We distinguishbetween an inward and an outward
orientationaccording to the level of merchandiseexports relativeto the value predictedby the regressions. To account for the negative association between size and openness, separateregressionswere run for large and small countries,and N was left as an expLanatoryvariable in equation (1). To minimize the impact of potential breaks in the regressionsafter 1973 (see section III.A), within each group we further divided the sample by time - pre 1973 and post 1973 - and estimated separateregressionsfor each subperiod. The relative export level (EL=E/Ewhere E here refers to merchandiseexports) was calculatedfor 1965 and 1980 (table 7).
- 37 -
c)
Trade Orientation: Direct measures of the availabilityof natural
resourcesare hard to come by, and besides, the level and compositionof exports reflect the combinedeffect of the abundance of resources and trade policy. To capture the combined effect of the two, we define a trade orientationindex (TO) that measuresthe deviation of the actual trade bias (TB) from the one predictedfor a country of similar income and size. The trade bias is defined as TB = (Ep - Em)/E where E here stands for merchandise exports (=Ep+Em).
It is a measure of the compositionof commodity exports
normalizedby total merchandiseexports. The TO index equals:
TO = TB - TB = (E - E )/E - (E - E )/E p m p m
The index considersthe pattern of specializationas well as the level of total exports (for further discussionsee the appendix to Chenery and Syrquin, 1986b). The TO index was calculatedfor 1965 and 1980 on the basis of separateregressionsby size (small and large) and period (pre and post 1973). On the basis of these three dimensions- size, the relativeexport level, and the trade orientationindex - countrieswere assigned into one of the various types in table 7. In classifyingcountries the statistical measureswere supplementedby dynamic considerationsand an evaluationof their trade strategies. Thus, if a country experienceda significantshift in policy or in the two indicatorsbetween 1965 and 1980, we tended to classify it by its position in the terminalyear or by the directionof the shift. More details about the classificationappear in annex A. In table 7 the countries are first divided by size into large (L) and small (S), then by trade orientationinto primary (P) and manufacturing(M),
-
38 -
and finally by openness into inward (In), neutral (N), and outward (0). The inward and neutral classes appear together under the inward label. They are distinguishedin the table but not in the regressions.
Performanceby Type Before contrastingthe patterns of resource allocaltion in the various groups in the typology of table 7, we present a rough picture of comparativeperformanceduring the period 1950-83,based on unweighted averages of growth rates of total GDP (table 8). There is great variance within groups and factors other than those dividing the groups in the table, influence the rate of growth. Still, the differencesare of interest though the above caveats should be remembered in any causal interpretation. During the period 1950-83 Large countries performed better than small ones, a manufacturingspecializationoutperformeda primary specialization, and an outward orientation exhibited higher growth than an inward orientation. The superiorityof the outward orientationtook place within all four types in the table (LP, LM, SP, and SM).
-39-
Table 7:
A TYPOLOGY OF TRADE PATTERNS A.
PRIMARY
n
Large
F
TO
EL
gy
65
80
65
80
65
80
Inward Argentina Brazil x Burma Colombia x Ethiopia Mexico Philippines x Thailand x Turkey United States
26 24 34 34 26 34 34 34 34 34
1.2 4.4 1.9 2.3 1.4 3.2 2.6 3.6 3.1 1.9
46 120 78 53 43 50 103 90 40 55
32 58 37* 72 86 22* 111 132 37 70
33 30 19 11 22 25 8 7 25 106
29 -6 18* 23 42 14* -27 -11 8 46
-2.8 -2.2 5.9 -1.0 1.0 1.0 -0.03 1.3 1.4 -0.7
2.2 2.1 1.2* -0.5 5.2 2.7* 5.5 5.4 8.0 1.0
Outward Canada Indonesia Iran Nigeria Zaire
34 28 26 31 34
2.7 3.3 5.7 2.2 -0.1
85 80 129 207 375
143 270 210* 171* 157*
17 7 35 24 -1
47 22 89* 49* 26
0.2 0.5 -6.4 2.1 -9.8
-1.7 -8.3 -9.1* -1.4* 19.1*
MANUFACT. Inward Bangladesh
62
24
0.4
China
25
4.6
France x India Pakistan Spain
34 34 24 33
3.7 1.5 2.8 4.2
34 34 34 34 34 34 34 26
3.1 4.0 4.0 6.4 4.9 2.3 2.0 4.7
Outward Egypt Germany Italy Japan Korea S.Africa United Kingdom Yugoslavia
-102*
-108
---
---
---
---
---
88 62 70 40
97 120 82 57
-3 -115 -54 4
13 -118 -69 -29
-0.9 2.2 8.3 3.4
1.7 3.9 11.5 2.6
105 127 119 131 39 135 121 94
138 140 121 84 157 186 130 75
-25 -18 -38 -25 -96 -18 -20 -74
22 -5 -41 -42 -112 -71 -3 -57
3.8 -0.3 -2.6 -1.4 7.4 0.3 0.7 -0.3
15.4 0.4 2.9 0.9 7.6 -8.2 -2.5 4.0
26* ---
7.3*
14.7
-40-
Table 7:
A TYPOLOGY OF TRADE PATTERNS (Cont'd) b.
PRIMARY
Inward Afghanistan Australia Bolivia Burundi Central Africa Rep. x Chad x Chile Costa Rica x Denmark x Dominican R. Ghana x Guatemala x Haiti Madagascar x Mali x
n
Small
gy
EL
TO
651 Y
80
65
F 80
65
80
23 34 25 24
0.2 2.3 1.1 2.5
49 85, 76
--44 88* 53*
--49 26 - 4
--27 43* 8*
--2.4 5.0 2.2
1.9 5.6* 10.9*
0.2 -2.8 0.8 2.5 2.9 2.6 -0.8 2.0 0.5 -0.7 0.9 -0.3 0.3 1.6 1.2 2.3 1.6 0.9 -0.3 -0.6 0.6 3.3 1.0 4.6
78 103 57 59 86 56 97 77 740 72 98 --107* 66 80 43 62 97 100 70 81 54 53 12*
70 115* 69 56 72 61 121* 88 --105 114* --79 61 67 35* 84 76* 95 36 48 67* 28 3
-89 4 46 6 16 31 14 10 -66* 8 - 5 ---75* 69 24 16 28 2 27 - 8 - 2 20 47 ---
-25 -12* 40 -11 - 3 12 1* 5 --- 4 -23*
Nepal New Zealand Nicaragua Paraguay Peru Rwanda Senegal x Somalia Sudan Syria Uruguay Yemen
24 23 26 24 34 34 34 34 25 26 34 31 26 34 34 34 34 26 24 24 30 26 34 14
10.6 7.5 -1.1 10.0 1.5 2.3 9.6 2.8 8.2* 6.4 11.2 ---4.5* 2.2 3.1 0.8 1.2 4.8 4.0 3.5 0.7 0.04 -7.0 28.8*
16.5 19.8* 4.2 10.3 1.1 9.7 -0.9* 2.9 --16.9 22.7* ---7.2 1.2 18.9 5.5* -4.0 11.4* 17.2 21.1 10.9 12.5* 6.2 64.3
Outward Algeria Camaroon Congo Ecuador Guinea Honduras Iraq Ivory Coast Liberia Libya Malawi Mauritania Malaysia
34 26 24 34 24 34 29 24 24 24 30 24 29
2.3 2.0 2.8 3.4 1.3 0.9 4.3 1.8 0.7 3.4 2.2 1.8 4.0
90 124 64* 77 --109 155 147 151 146 147 114 211
118 141 125 95 --143 208* 141* 155 160 256 98* 217
24 17 12* 34 --21 50 32 12 72 - 2 16 17
31 29 25 54 --20 62* 31* 24 95 -21 21* 2
Mozambique
-60 45 17 27* -15 6* 8 21 - 5 43* - 4 ---
3.0 0.5 17.1 3.0 ---0.5 -15.0 -1.4 -10.2 -20.7 12.1 -13.5 -4.4
-1.9 -4.6 0.1 0.2 --7.4 -13.4* -0.2* 0.05 -35.2 14.3 22.7* -2.6
- 41 -
Table 7:
A TYPOLOGY OF TRADE PATTERNS (Cont'd) b.
PRIMARY
n
Small
g
EL 65
TO 80
65
F 80
65
80
6.5 19.9 -33.9 2.5 -0.6 5.9 -1.2 -10.5 -14.9 ----
8.7 6.2* -32.7 10.0 22.6 14.9 2.1* -8.2 19.5* ----
Outward Niger Papua Saudi Arabia Sierra Leone Sri Lanka Togo Uganda Venezuela Zambia Zimbabwe
24 24 21 20 34 24 34 34 34 24
-1.5 1.9 6.6 0.8 2.0 2.1 -0.5 2.3 2.0 1.4
53 57 224 131 309 76 160 134 243 ---
148 109* 199 99 216 103 60* 122 153* ---
18 14 76 -101 - 2 8 8 39 39 ---
20 33* 105 -78 -37 12 0* 33 47* ---
MANUFACT. Inward Austria Angolax Benin x El Salvador x Finland Greece x Israel Jamaica x Jordan Morocco x Norway x Panama x Portugal Sweden Tunisia x Tanzania x Upper Volta x
34 31 25 34 34 34 34 34 14 34 34 34 34 34 23 26 25
4.3 -1.3 0.7 1.1 3.7 5.3 4.3 2.1 6.5 1.4 3.2 3.6 4.6 2.7 4.5 1.9 1.1
72 --52 112 70 22 41 72 50* 70 79 --85 73 62 186 67
72 --106* 137 82 35 82 99 55 57 87 --76 73 105 80 115
-68 --8 2 -26 27 -66 -22 -21* 1 - 8 ---84 -33 -1 -26 -4
-61 ---2* -21 -35 -16 -80 -80 -18 -42 44 ---78 -45 -10 -31 -21
0.6 --8.6 2.4 1.7 11.3 13.5 3.6 56.7* -1.3 0.9 --4.6 0.9 13.1 -0.9 7.6
Outward Belgium Hong Kong Hungary Ireland Kenya Lebanon Netherlands Puerto Rico Singapore Switzerland Taiwan
34 26 24 34 34 25 34 34 24 34 34
3.3 7.5 5.9 2.8 2.1 1.1 3.0 3.6 7.1 2.3 5.6
115 218 154* 96 148 --132 --314 86 ---
145 220 143 125 108 --126 --378 87 ---
-67 -113 -86* 1 -20 ---34 ---26 -68 ---
-45 -94 -75 -33 -30 ---13 ---36 -63 ---
-0.2 3.7 7.0 4.6 7.4* 2.2 9.0 13.1 -0.7 11.4 -----0.7 0.4 -----12.1 9.3 0.8 3.5 ------
n: g : xY *:
number of annual observations on income per capita. average annual growth rate of per capita GNP. Neutral (see text). 1975
1.9 --22.7* -0.9 0.8 6.8 12.4 2.1 49.4 11.1 -6.2 --15.1 1.9 5.4 13.1 27.3
-42-
Table 8 ANNUAL GROWTH RATE OF GDP 1950-83: SIMPLE AVERAGES (Percent)
Size Strategy
Large No. of Countries
g y
Small No. of Countries
g y
Primary inward outward
10 5 (LP)
4.94 5.12 5.00
27 23 (SP)
3.58 5.01 All primary 4.24 (P) 4.42
Manufact. inward outward
6 8 (LM)
4.73 5.26 5.04
17 10 (SM)
4.74 5.73 All manufact. 5.11 (M) 5.09
(L)
5.02
(S)
4.54
All inward All outward
4.28 5.22
ALL
4.67
Notes: Growth rates within countriesare OLS estimates. The number of annual observationsvaries from 14 to 34 (see table 7).
- 43 -
Variation in Patterns of Resource Allocation A simple way to assess the effects of size, specializationand openness on the patterns of structuralchange, is to compare the results of separateestimates of equation (1). For each of the three dimensionswe compare the predicted values at the income levels representingthe end points of the transitionrange ($ 300 and $ 4000) for various measures of structure. The effects of size and resources for example are not independent of each other. But as a first approximationwe examine separatelyeach of the three sources of diversity-size,specializationand openness. (Tests of homogeneity appear in annex B.)
Size: The importanceof internationaltrade is much lower in large than in small countries,and the differencebetween the two groups increaseswith the level of income (figure 6). An interestingdifferencerelates to the share of manufacturedimports. In large countries this share is not only lower but it declines during the initial stages of the transition,reflecting the early import substitutionin large countriesmade possible by their larger domestic markets. The differences in trade patterns result in similar differences in the structuresof productionand employment. At low income levels, large countries are more industrializedthan small ones. The contrast is particularlymarked in heavy industrywhich is not economical in small economies at this income level. At higher income levels there is a high degree of convergence,implying a more pronounced transformationin smaller countries.
Trade Orientation: Ignoring the differences in size, the type of specializationhas only a small average effect on the level of trade but (by
-
44 -
definition)a very significantone on its composition. A relative specializationin manufacturingis characterizedby higher exports of manufacturesand lower primary exports, absolutelyand not only in relative terms. The impact on productionand employment is similar to the one of size, but in this case the divergence between the two types is magnified during the transition. The pattern of specializationrepresents, in part, a strategy of development,unlike size which is a structuralfeature whose variation over time is insignificantrelative to its variation across countries.
Openness: The criterion for classifyingcountries as inward or outward, was not the share of trade but the relative export level that measures departures in actual trade from the predicted one. In analyzing this dimension we also incorporatedthe pattern of specializationin a simple way. To the separate regressionswithin the inward and outward groups, we added an additive dummy variable taking the value of one for manufacturingorientationand zero for primary orientation. Within groups the predicted values for manufacturingand primary orientationdiffer by a constant (the coefficientof the dummy variable) at all income levels. The figures in table 9 show the ratio of predicted values of inward to outward for the two types of trade orientation. The last two columns give the coefficientsof the dummy variables.
The principal differencesdue to openness and its interactionwith the orientationof exports, as reported in table 9 are: -
In the more inward-orientedeconomies all categoriesof trade are significantlylower than in the outward group.
Aq~6
EXPORTS: S AND L 0.36 0. 34 0.32 0.30S.o
S
A 0.28 R E 0.26S 0.-240 F 0 22 G 0.20 D P 0. 18 LARGE E--
0.160. 140.12 250
500 GNP PER CAPITA
2000
1000 (1980
U.S.
Dollars)
4000
-46-
Table 9: EFFECTS OF SIZE, TRADE ORIENTATION AND OPENNESS ON PREDICTED VALUES OF STRUCTURE AT TWO INCOME LEVELS ($300 and $4000)
Variable
Orientation: Manuf/Primary (4) (3) 4000 300
Size: Large/Small (2) (1) 4000 300
Coefficient of Openness: dummy variable Inward/Outward for manuf.(X100) (Manuf) (Primary) (10) (9) (7) (8) (5) (6) Inw. Outw. 4000 300 4000 300
EFMM EOP EM MP MM
159 49 104 77 68
58 34 66 65 44
30 93 425 134 119
29 64 445 218 139
16 63 73 69
15 93 92 89 65
46 38 66 73
58 64 47 116 69
0.4 -2.6 4.1 3.5 3.9
-13.6 -1.0 13.3 5.9 4.7
E M-E I
66 55 111
51 41 98
109 207 106
134 112
51 575 77
52 86
65 125 88
64 357 95
4.9 2.7 2.9
1.8 6.5 0.6
VA VM VS
87 129 102
111 115 92
98 115 96
108 123 92
105 118 107
94 136 114
113 60 97
129 87 106
-0.9 -0.6 -1.7
-3.6 9.3 1.7
LA LI LS
88 138 146
97 103 100
91 143 122
98 124 86
96 114 111
75 107 117
110 72 88
106 90 106
-2.7 3.0 -0.3
-12.2 8.6 3.6
Light ind. 127 Heavy ind. 656
89 145
104 124
116 145
173 142
149 125
92 34
95 78
-0.9 1.1
3.4 5.0
Notes:
The figures are ratios of predicted values (x) times 100. Size:
x
small = 100.
Orientation: Openness:
x primary = 100.
x open = 100.
Predicted values calculated for a level of population of 6 million in small countries, 60 million in large countries, and 20 million in other divisions. A blank indicates a negative or negligible value in one of the types.
- 47 -
The share of manufacturingin output and employment in the more closed economies relative to the open ones is higher when trade shows a primary orientation(cols. 5-6) and lower when the specializationfavors manufactures (cols. 7-8). The former case is representativeof the import substitution strategy, primarily in the case of large countries (see below). When the absence of natural resourcesor policy considerationhave led to a trade orientationtoward manufactures,it is in the more open economieswhere we find a higher rate of industrialization. Within the inward and the outward groups an estimate of the average difference due to trade orientationis given by the coefficientof the dummy variable for manufacturingorientation,in the last two columns of table 9. The magnitude of the coefficientsis distinctlyhigher in the outward group. In this group the level of exports is similar for the two types of specializationbut the compositionis very different. When a country's trade is manufacturing oriented, exports of manufactures(as a share of GDP) are 13 percentagepoints higher while primary exports are lower by a similar amount. In the structure of economicactivity, this differenceshows up in shares of manufacturing output and employmentthat are higher by about 9 percentagepoints.
Classificationof Trade Patterns The classificationof countries by scale and trade orientationin table 7 was used as a basis for computing separateregressionsof allocationpatterns. Four groups were identifiedaccording to their size and trade orientation:large, primary oriented (LP); large, manufacturingoriented (LM); small, primary oriented (SP); and small, manufacturingoriented (SM). The terminologyand approach first appeared in Chenery and Taylor (1968), and was revised in our 1975 study (C-S),
- 48 -
except for the division of the large countries into LP and LM which were previously consolidatedinto one group. In this study we show separate results for LP and LM, however, a note of caution is in order about the LM group. The composition of this group can lead to some peculiar results. Among its 14 countries it includes four very large and very poor (Bangladesh,China, India, and Pakistan) and four high income countries with very high shares of industry (France, Germany, Japan and the UK). In between we find a group with very rapid rates of industrializationduring the period (Italy, Korea, Yugoslavia). A comparisonof the principal results appears in tables 10 and 11 and in figures 7-10. Table 10 compares the predicted levels for various indicatorsat one point in the middle of the transition range ($1000). Table 11 shows the magnitude of the transformationin structure for the four types. Some of the results in table 10 mirror the ones in table 9. The main difference is that now we look at the combined effect of size and trade orientation. Large countries export a much smaller share of output than small ones. Within large countries there are some interestingdifferencesbetween LP and LM types, subject only to the above caveat about the compositionof the LM group. The relative abundance of natural resources in the typical LP country, is reflected in its trade composition. Foreign exchange requirementsare derived from primary exports, and there is little need for primary imports. Many countries in this group followed an import substitutionstrategyduring most of the period since 1950. One result of this policy was a failure to develop manufacturedexports which also shows up in the shortfallof light industry relative to the average pattern, and in the relatively low share of industrial employment.
-49-
Table 10: ALTERNATIVEPATTERNS OF SPECIALIZATION: COMPARISONAT y = $1000
Index: Share of Type - Share
Variable
Average,Pattern (Share)"a Index
of Average Pattern x 100 SP SM LP LM
EFMM EOP EM MP MM
7.3 7.9 3.7 7.1 13.5
100 100 100 100 100
125 147 49 97 113
18 106 389 177 180
78 63 35 52 67
19 29 262 152 74
E M-E I
22.6 3.4 23.3
100 100 100
115 76 91
141 341 105
63 50 98
67 221 120
VA VN VM VS
22.8 7.7 18.1 37.8
100 100 100 100
102 109 85 108
105 66 95 104
113 109 97 96
93 43 116 102
LA LI LS
51.7 19.2 29.1
100 100 100
112 83 90
85 123 111
102 88 104
89 132 98
9.6 9.2
100 100
104 60
95 76
89 95
123 137
Light ind. Heavy ind. a)
Shares of GDP except for the employmentvariableswhich are shares of total labor force.
- 50 -
By contrast in the LM group, overall trade is still low but manufactured exports are substantiallyhigher than in the average pattern, as is the share of light industry in GDP. The exploitationof economies of scale is reflected in the high shares of investmentand heavy industry. In small countries trade is more important but again its composition differs according to the pattern of specialization. In the SP economy high trade derives from primary exports that more than offset the shortfall in manufactured exports,while exactly the opposite is true of the typical SM economy. The high level of manufacturedexports in the SM country is accompaniedby an equally high share of manufacturedimports - the exact opposite of the LP pattern. The high level of manufacturedimports reflects input requirements,as well as final imports which are a concomitantof the higher degree of specializationand integrationin the internationaleconomy of the resource-poorSM country. Table 11 compares the magnitude of change during the transitionin the four types. Manufacturedexports increase everywherebut mostly in the manufacturing-oriented groups. The high level of specializationin small countries leads to a greater transformationin the structuresof productionand employment. Some of the early rise in industry in large countries,afforded by size and often prompted by policy, is not reflected in the table. To facilitatea comparisonwith the results in Chenery and Taylor (1968), the dimension of the transformationwithin manufacturingis shown for the early-middle-latepartition of industrialbranches. Beyond the $300 mark there is little change in the share of early industries,except in SM countrieswhere the subsequentrise is closely linked to the developmentof exports. Industrializationduring the transitionis mostly concentratedin the group of late industries,a group characterizedby relativelyhigh capital intensity and economies of scale.
-51-
Table 11:
DIMENSIONS OF THE TRANSFORMATION BY TYPE (Changes in Shares)
SP
Merchandise exports Primary Manufacturing Merchandise imports Primary Manufacturing
Type SM
LP
LM
2.6 2.2
-3.4 15.6
-7.3 3.7
-1.6 6.4
-1.1 1.2
5.3 6.7
-1.2 -1.6
0.8 -1.4
Value added in Agriculture Manufacturing
-33.2 10.2
-30.6 13.7
-23.7 8.5
-27.6 14.0
Labor force Agriculture Industry
-54.1 21.3
-50.7 26.1
-45.7 18.4
-39.6 22.1
1.5 2.5 5.5
5.2 5.0 9.6
-0.3 1.4 7.6
0.1 3.8 9.8
Manufacturing Early Middle Late
Note:
Changes in predicted values from y = $ 300 to y countries, and N = 60 in large ones.
=
$ 4000.
N
6 in small
- 52
-
In the four-way typology analyzed in this section the degree of openness has not been explicitlyconsidered. Further splittingthe types by this dimension would reduce the samples too much for statisticalanalysis. Instead,we make recourseonce again to a dummy variable for estimatingthe average impact of an outward orientationwithin types. The coefficientsof the dummies for the more open groups and their t ratios are reported in table 12. In most of the cases the coefficientsare highly significant statistically. Trade ratios in the more open groups are strikinglyhigher than in the more inward oriented ones. The increase in exports associatedwith greater openness,takes place in primary exports in SP and LP types, and in manufactured exports in SM and LM. Greater openness is associatedwith lower trade deficits and higher investmentshares, implyinghigher saving proportions. The only exception is the SM group where the effects are not significant. In SP and LP countries that are relativelyopen, we find that the share of mining increasesat the expense of manufacturingand services. In SM and LM, manufacturingis higher when the economy is outward orientedand agriculture lower. The higher shares of manufacturingreflect a higher share of light industryin SM and of heavy industryin LM. A positive associationbetween outward orientationand industrial employmentis found only in the SM type economy.
- 53 -
'7% 70
EXPORTS: SP AND SM 0.40 -
0.38 0.3b
S
,
H 0.34
A 0. 32 -
E S
0.30 0 F 0.28 SP,
G
-
-
---- --- - - - ---
0.24 0.22
0.20
___
250
500
2000
1000
GNP PER CAPITA (1980
.71
U.S.
4000
Dollars)
7.
EXPORTS: LPAND LM 0.20 LM
0.19o .18-
H0. 17
A
0.15 R0oq1 F0.
141
G 0. 12 -
0.10 250
500
2000
1000
GNP PER CAPITA (1980
U.S.
Dollars)
4000
54 -
-
~8. STRUCTURE OF TRADE: SP EMR
0.24
0.220.20-
S 0.18/ H A 0.16
R
E 0.14
S
EOP
0.12 F o.1
EFMM
.G 0.08
,
P0.06 0.04 ---------
0.02 0.02
4000
2000
1(100
500
250
GNP PER CAPITA (1980 U.S. Dollars)
STRUCTURE
OF TRADE: SM
0.30
0.28
E
0.26 0.24-
S 0.22 A 0.20
-.
0.18
S
0.16
o 0. 14 F 0.12 G o.io D P 0.08
-EOP
0.06 0.04 0.02
EFMM
0.00
-
250
-
-
-
-
500
-
-
-
-
-
--
-
1000
-
-
__
__
___
___
2000
GNP PER CAPITA (1980 U.S. Dollars)
__
___
4000
__
- 55 5 & STRUCTURE OF TRADE: LP 0.13 EMR
0.120.111-
S
0.10
-
H 0.09 A R 02. 8
E S 0.07 0 0.06 F 0.05 -
D
_,-EFMM
0.040.03 E
0.02 P~~~~~N
o
PE*API-A(180U.
a
:.
EM
STRUCTURE OF TRADE. LM 0. 160. 15
…
0.01
GD0. 12
E 0.09. o0.06
O
0.02
______-_EFMM
250
500
1000
2000
GNP PER CAPITA (1980 U.S. Dollars)
4000
- 56 -
STRUCTURE OF PRODUCTION:SP 0.50
Vs
0.45
--
X
0.40-
E A s 0.35\
.-
_
0F 0.25\
G 0.20
\-
0
VC+VU
-----
0.10 0.05
250
500
2000
1000
4000
GNPPERCAPITA(1980 U.S. Dollars)
7~. 9&
STRUCTUREOF PRODUCTION:SM 0.50
Vs 0.45-
0.40-
S A 0.35 R
0 0.25 F
V v
P
Vc+VU 0.15
----
0.10 0.05
__
_
250
_
_
_
_
_
_
_
500
_
_
_
_
_
_
_
1000
_
_
_
_
_
_
_
_
_
2000
_
GNP PER CAPITA (19801U.S. Dollars)
_
_
_
_
4000
_
- 57 -
7J 7
9;
STRUCTURE OF PRODUCTION:LP o.50 0. 45 0.40 -
S H
A 0.35 R E
oF 0.25
VA
GVM
-0.
0.1 D~~~~~~
-----
C-+VU_-_-
-
0.15. 0.10
,5o
1000
500
2000
4000
GNP PER CAPITA(1980U.S Dollars)
STRUCTURE OF PRODUCTION:LM 0.50 0.45
VS
0. 40A 0 35
_
_
R
E 0.20 F
0.025
-
D
VCzoV
0 0.15 -......
.
0.10 0.05 250
500
1000
2000
GNPPERCAPITA(1980 U.S. Dollars)
4000
-
58 -
STRUCTURE OF VAf lIE ADDED IN MANUF: SP 0.10 0.09
.
0.08 S H 0.07
LATE
A
R 0.06 E MIDDLE
0.05
0.04 D
0.03-
-
---
.-
P 0.02
-
-
0.01. 0.00
_
i
_
_
_
_
_
_
_
o.oo
p
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_ l
_
_
4000
2000
1000
500
250
_
GNPPER CAPITA(1980 U.S. Dollars)
STRUCTURE OF VALUE ADDED IN MANUF: SM 0.10_ EARLY
0.09)0.06
s007 A R 0.06 S
,
/
0.05
-
L1
1
/I -
O 0.04 F
G 0.03,
p 0.02
-;
0.01 ,
0.00 -0 .0 1
_
_
_
250
_
_
_
_
_
_
_
Soo
_
_
_
_
_
_
1000
_
_
_
_
_
_
_
2000
GNPPERCAPITA(1980 U.S. Dollars)
_
_
_
_
_
4000
_
- 59 -
x?.1tog
STRUCTURE OF VALUEADDEDIN MANUF: LM 0.140.13. 0.12-
s0 A
R 0.10 E
EARLY
S 0. 09 /
0 0.08
F
MIDDLE
,-
0.07 G PD0.06 p~~~~~~~
0.05 0.04 0.03
4000
2000
1000
500
250
GNPPERCAPITA(1980 U.S. Dollars)
A,
tfo.
STRUCTURE OF VALUEADDEDIN MANUF:LP 0.10 0.09. /
0.08
H 0.07 A
R 0
EAR
0.06, MIDDLE ---
0.05
-'
.
F 0.04 LATE
G D 0.03
/
/
0.02
-
0.01 0.00
, 250
_
,_
,_I
500
1000
2000
GNPPERCAPITA(1980 U.S. Dollars)
4000
-60-
Table 12: EFFECT OF OUTWARD ORIENTATIONBY TYPE
Dummy variablefor "open" group Coefficient(X100) t ratio SP SM LP LM SP SM LP
C C I E M M-E FCN
-11.4 2.0 5.8 13.0 9.4 -3.6 -7.7
2.8 -2.3 0.2 15.4 16.1 0.7 0.3
-4.9 1.6 1.3 14.0 12.0 -2.0 0.9
-6.2 4.3 1.2 6.2 5.6 -0.6 -5.9
24.0 5.3 7.1 6.3 14.8 0.5 23.6 17.6 16.1 17.2
EP EM MP Mm
12.7 0.4 1.6 4.8
3.1 15.8 6.0 12.2
15.2 1.2 1.7 7.1
1.4 4.7 1.9 2.2
VA VN VH vC VU Vs
-4.0 7.8 -2.4 0.9 1.5 -3.8
-5.3 0.1 4.2 -0.2 -1.4 2.5
3.7 11.2 -9.0 0.1 -1.3 -4.7
-2.2 1.6 3.2 -0.2 1.0 -3.4
LA LI LS
3.8 -1.9 -1.9
-7.5 5.8 1.7
4.4 0.1 -4.5
-5.1 0.8 4.2
Light Ind. Heavy Ind.
-4.1 0.1
1.3 0.7
-3.4 -5.3
0.4 2.8
* Regressionswere run at the 9 sector level.
LM
6.9 4.8 2.7 17.5 15.9
8.6 9.3 2.1 9.3 8.6
0.8
0.3
11.8
19.9 5.8 2.1 15.7 6.0 13.6 13.7 15.9
16.5 5.4 5.8 14.9
3.3 8.9 5.1 4.6
6.2 13.5 7.9 6.5 9.4 8.2
7.2 0.4 7.5 1.1 4.1 3.9
3.7 18.8 20.5 0.4 7.0 6.4
3.3 3.5 4.5 0.9 4.1 4.9
5.8 8.6 7.0 10.5 4.2 2.9
4.2 0.3 5.9
6.0 1.4 7.3
7.5
*
- 61 -
III. CHANGES OVER TIME
The analysis of the structuraltransformationin part II had a long run view, made possible by the wide variation in the level of development across countries. Our sample has an extensive coverage of economies over a relatively long period of time. In this part we take a closer look at changes over time in the uniform relations,and changes within countries. We first summarize the principal time trends in allocation patterns,and then compare time-series estimateswithin countries, to the estimates in part II which are predominantly cross-sectional.
A.
Stability of DevelopmentPatterns The stabilityof cross-countryrelations over time was already addressed
in C-S. The unstable conditions in the internationaleconomy since the early 1970s have given greater weight to this issue. A variety of reasons could affect the temporal stabilityof intercountrypatterns of development. For the present discussionthey can be grouped under two headings: omitted variables and structuralbreaks. Omitted Variables: Besides income and size, other factors influence the patterns of resource allocation. The impact of variables that are correlatedwith income over time is reflected in part, in the estimated income effects. Some long run processes of change proceed over time independentlyof variations in income. For example, changes in the level of technology,the internationalenvironment,or the strategiesof development,may lead to shifts in the dependent variables. To the extent that those long run processes of change can be assumed to be universal and
62 -
to affect all countries alike, their effect would be captured by the time dummy variables in equation (1). In a more general model the time-shiftvariables would be replaced by the processes for which they stand as proxies. Some of the omitted variables vary primarily among countries and are relatively invariant within a country over time. If these variables are correlatedwith income across countries,as in the case of the "exchange-ratedeviation-index"6/or because of some historical reason, the cross-country patterns will differ from time-seriesestimates and this differencemay end up as part of the time trend in the intercountryestimates. The interpretationof the shift in this case is quite different from the previous one. Finally,we have the case of random or unanticipatedshocks, such as the quadrupling of the price of oil in 1973. If the impact of the shocks is uniform for all countries, it will appear in the time-shiftvariable. If it is random, it may impair the accuracy of the estimates. The case of a differentialimpact across countries is considered in the sequel as a structuralbreak. Structuralbreaks: Estimates of cross-countryrelationsmay shift over time because of changes in the structure of a model (changes in structure in the econometricsense). This source of variation is not always different from the case of changes in omitted variables just mentioned. When the impact of the change in an omitted variable is different for different groups of countries, an additive time-shiftvariable will fail to represent the differentialeffect. If there is reason to believe that the effect of the change (in oil prices for example) varies systematicallywith income or some other characteristics,we could introduce interactionterms or split the sample and estimate uniform time-shifts within groups. As part of this project separateregressionswere estimated for countries grouped by income level according to the classificationin the 1985
- 63 -
World DevelopmentReport, and for a two way partition of developingcountries into oil importing and oil exporting countries. A summary of results appears in annex D. An alternativeapproach to examine the interactionof time shifts with income, adopted in this section, is to estimate separateregressionsfor the time periods before and after 1973, and to compare predictedvalues at various income levels. Cross-countrypatterns are not well suited to incorporatedynamic relations. Processes of adjustmentthat are distributedover time, may cause the static relations to shift, transformingpart of what really is a long run income effect into a time trend. Summing up the discussion:it is importantto allow for temporal shifts in the estimationof cross-countryrelations,but the results should be evaluated with care since they are open to more than one interpretation. Accuracyof the New Estimates What effect did the events of the early 1970s have on the estimates of developmentpatterns? We first address this questionby comparing the goodness of fit of the estimates in C-S, which covered the period 1950-70,to the new ones which extend the period to 1983. Table 13 presents the standarderror of estimate (SEE) of the regressionsfor all those variables that appear in both studies. The first two columns show the SEE's in the present study and in C-S. The figures are strikinglysimilar and suggest that if there is any difference,it favors the new study. To focus directly on the events around 1973 we estimated separate regressionsfor the pre 1973 and post 1973 periods,but to eliminate effects due to the compositionof the sample we only includedcountries with informationfor all the years between 1960 and 1982 (1962-1979in the case of merchandise trade). SEE's from regressionsbased on this compatiblesample, for the whole
-64 -
Table 13:
GOODNESS OF FIT
ALL
Standard Error of Estimate ALL C-S Compatible Pre 1973
C G I E N
.087 .047 .065 .118 .121
.086 .043 .051 .119 .120
.077 .044 .067 .115 .116
.074 .038 .061 .111 .109
.080 .052 .073 .121 .125
n*
2954
1508/1432'
1518
858
660
FCN n
.051 1100
.048 642
.041 573
.046 360
.029 213
EP EM n
.095 .073 1782
.091 .064 413
.079 .078 1098
.077 .074 671
.082 .083 427
) ) ) )
.091
.028 .075 2311
.026 .084 1325
.082 .066 .051 .020 .029 .062 1518
.091 .064 .049 .021 .032 .062 858
.070 .069 .052 .017 .025 .059 660
.118 .061 .084 2746
.116 .064 .089 165
.114 .064 .081 1098
.118 .064 .080 671
.107 .063 .082 427
VA VN viM VC Vii VS n LA LI LS n
.092 .077 .058 .020
.060
n = number of observations.
Post 1973
- 65 -
period and for the subperiodsappear in the last 3 columns in table 13. As expected, the accuracy after 1973 is poorer for variables related to trade, but only by a small margin. SEE's for the pattern of productionare actually lower after 1973. The results in table 13, and the resemblenceof the overall transformationin part II to that based on earlier data, dispel the notion that the instabilityof the 1970s has invalidatedprevious estimates of development patterns. On the contrary, the overall picture of uniformity of the structural transformationappears to be quite robust. Principal Time Trends The estimated shifts will now be presented,first assuming them to be uniform for all countries,and then allowing for nonuniformitiesby contrasting pre-1973 and post-1973 regressions. Uniform shifts: The time dummies in equation (1) are designed to capture uniform changes in the level of the regressions. As explained in part I the dummies measure incrementaladditive shifts. 1950-60 is the base period, T1 gives the expected shift after 1960, T2 the additionalchange after 1966 (over and above the one measured by T1), T3 and T4 stand for the incrementalshifts after 1973 and 1979 respectively. T1 differentiatesthe 1950s from the rest of the period. Problems related to the quality of the informationfor this decade, might be reflected in the coefficientof T1.
The first oil shock and accompanyingchanges
in the internationaleconomy suggested the year 1973 as a natural dividing point for T3.
Similarly, 1979 was selectedfor T4 on account of the second oil shock
and the onset of deep recessionsin a large number of developing countries, primarily in Latin America.
- 66
-
Uniform time shifts were also estimated in C-S.
The T1 variable there
compared 1950-54 to 1965-69. The coefficientsof the time dummy variables are shown in table 14. (The t ratios appear in table S-1 in the Statistical appendix). T1 of C-S is also shown for comparison. The uniform time shifts up to the late 1970s, reinforce the income related shift from food consumptionto investmentand government consumption,and the increase in trade ratios. The increasein imports substantiallyexceeds the addition to exports resulting in higher proportionsof capital inflow. The time-relatedrise in trade shares after 1973 encompassesboth primary and manufacturedexports and imports. If in the case of demand the results were similar to those in C-S, for the pattern of production there are some important differences. In the pre-1973 period we observe a very large shift from agricultureto all other sectors. This fall in the share of agriculturaloutput comes on top of the effect of rising income, and can be explained by the nature of technologicalprogress and the substitutionof fabricatedproducts for natural materials. The size of the effect - much larger than that in C-S - is probably due in part to data problems in the 1950s. After 1973, the share of manufacturinggoes down reflecting the spread of the de-industrialization phenomenon in advanced economies. The exogenous shift after 1979 from tradables to nontradablescombines the effects of the depression and worsening terms of trade in oil importing countries,with the changes in structure in oil exporting economies commonly identifiedas "Dutch-disease" effects. Non-uniform shifts: To evaluate how uniform the time shifts are, and the stabilityof a unique relation for the whole period, the predicted values from the separate regressionsrun for the pre-1973 and post-1973 periods are compared in
-67-
Table
Demand
14:
UNIFORMTIME TRENDS
Coefficient T1+T2
of Time Variables (percent) T3 T4 Tl(C-S)
C G I E M M-E
-3.0 2.8 2.2 0.9 2.9 2.0
-1.0 1.2 2.6 3.4 6.1 2.7
1.2 1.0 -0.4 1.6 3.4 1.8
-2.1 2.2 1.3 0.2 1.6 1.4
FCN
-2.7
0.2
-1.3
-2.5
0.5 -0.6 0.7 -0.3 1.1
1.8 0.5 1.1 2.2 2.8
-0.3 0.0 1.8 2.3 1.7
Production VA VN VM vC vU VS
-7.1 2.2 2.6 1.3 1.7 0.7
0.1 1.4 -0.7 0.1 -0.4 -0.6
-1.6 0.3 -0.6 0.2 0.5 1.3
Employment LA LI LS
-0.8 -1.3 2.2
-0.9 -0.2 1.1
-1.2 0.2 1.0
Manufacturing Light Ind. Heavy Ind.
-1.0 0.3
-0.6 -0.4
0.1 0.2
Trade* EFMM EOP EM MP Mm
* Only T2 since the data start in 1962.
) ) ) )
-1.5 -1.0 0.6 1.9
- 68 -
table 15, at three income levels: $300, $1000, and $4000. In general the regressionsare not very different for both periods. In most cases the hypothesis of homogeneity cannot be rejected (see annex B). It is still of interest to locate the range of major discrepancies,as in the comparisonsin table 15. Since the emphasis is on breaks around 1973, the coefficientsof T3 and T4 are also shown in the table. These figures differ slightly from the ones in table 14 because they were estimated from compatible samples as explained above. There are some cases where significantnon-uniform shifts appear to be present. At low income levels ($300) food and total private consumptionshift upward after 1973, in contrast to the drop at middle and higher income levels. The most significantdifferencesare related to trade. The increase in manufacturedand total exports after 1973 (holding income and size constant) is positively correlatedwith income, while the import surplus increases most at low income levels. The shares of manufacturedoutput and industrialemployment are almost the same before and after 1973 at low and middle incomes,but fall significantly in the richer countries. The counterpart is a positive time shift in services employment in advanced countries.
B. Average Time-Series:1950 - 1983
So far we have consideredprimarily the cross-countrydimension of our data set, and made use of the variation over time to determine time trends in the cross-countryrelations. In this section we switch our attention to the time-serieswithin countries but in a comparativeframework.
-69-
Table 15: COMPARISONOF ESTIMATEDPATTERNSFOR PERIODSBEFORE AND AFTER 1973,AND UNIFORMTIME SHIFTS (Percent)
x(73+) -
x (73-)
Coefficient
of:
300
1000
4000
T3
M-E
0.1 1.8 3.5 2.1 7.7 5.6
-3.4 1.5 4.4 5.5 8.0 2.5
-2.7 1.5 2.4 6.4 7.6 1.2
-1.2 1.3 2.6 3.6 6.4 2.8
0.8 1.1 -0.4 1.8 3.4 1.6
FCN
2.3
-0.4
-1.1
-0.2
-1.0
EMER EP EM MMER NP NM
2.7 2.3 0.3 4.5 2.2 2.4
3.8 2.5 1.2 5.0 2.2 2.9
4.0 2.0 2.1 4.8 1.8 3.1
3.2 1.8 1.3 4.6 2.1 2.5
1.6 1.0 0.6 2.2 0.8 1.4
VA
-0.5
-0.4
0.3
-0.1
-1.4
VN
0.8
1.8
1.8
1.1
1.0
vm VC VU VS
0.0 0.9 0.2 -1.2
0.1 1.1 0.3 -2.8
-1.3 0.3 -0.4 -0.7
-0.6 0.2 -0.2 -0.4
-1.2 0.1 0.3 1.2
LA LI LS
-0.1 0.7 -0.6
-1.7 0.1 1.6
-0.8 -1.9 2.7
-0.5 -0.7 1.2
-0.6 -0.2 0.8
Light Ind. Heavy Ind.
0.1 -0.2
-0.2 0.0
-0.8 -0.3
-0.5 -0.4
C C
I E M
T4
---
--
70-
Cross-countrypatterns can be interpretedas long run adjustment paths, reflectingthe accumulateddevelopment experienceof various decades and even centuries characterizedas modern economic growth. This was the way we presented the estimates of the structuraltransformationin part II. One reason behind the cross-sectionalemphasis in previous studies,was the limited availability of comparabletime-seriesinformationfor a large number of countries. In C-S we compared time-seriesand cross-countryestimates based on data for some 40 countriesover about a 20-year period. Our present data set expands the coverage to about 100 countries over three decades. Average time-seriesrelations are derived in two ways. First we estimate simple time-seriesrelationswithin each country and second, we estimate average time-seriespatterns by pooling the time series and introducingcountry dummy variables. IndividualTime-Series Within each country and for each real variable x, we estimate equation (7)
x =
a
+
B ln y
(7)
In the present study we have informationon shares of demand and output in GDP, in both current and constant prices. Their ratio is a measure of the price of an aggregate relative to the CDP deflator. We set all relative prices equal to 100 in 1970 and estimate for each relative price p, equation (8).
ln p = a + bt (t = time)
(8)
-
71 -
The coefficientb is an estimate of the annual rate of change in the relative price. Table 16 summarizesthe results. The first two columns give the mean and standarddeviation of the individualestimtes of n and b.
The number of estimates
range from a minimum of 42 (countries)for the employmentvariables, to a maximum of 106 for the componentsof demand. To get an indicationof the distributionof the estimatedparameters, the next four columns show the number of cases in each of four size intervals (two positive and two negative). Finally a second-stage regressionwas run for the estimated coefficientsas dependent variables with the log of per capita income in 1980 as explanatory variable. The results appear in the last two columns. The principal results are now briefly summarized. 1.
There is considerablevariation in time-series income slopes across
countries. Regarding the sign of the estimates, in most cases one sign dominates although we can always find exceptions. 2.
Looking at the simple unweightedmeans of the income slopes, we find that
the nature of the implied transformationin economic structure is quite similar to the cross-countryone in part II, but somewhat larger in magnitude. (The income effects in equation (7) may in some cases reflect exogenous time shifts). 3.
On the average increases in income during the period were accompanied by
a drop in the share of food consumptionmatched by an increase in investment. The shares of aggregate trade show a much stronger increasewith income than in the cross-countrypatterns. Both primary and manufacturedexports have average positive income slopes. In the case of primary the average is influencedby the very large increasein the share of exports of fuels.
-72-
Table 16: TIME-SERIESRELATIONS:MEANS AND DISTRIBUTION
Time-series coefficient of ln y or t
Distributionof slope coefficientsby size intervals
Mean
S.D.
