You are on page 1of 32

No.

30 Social indicators:
problems of defi nitio n
and of selection

cn
a
0
C
a
.I

0
-cnm
.I

0
0
cn
a
IC
CI

cn
L
a
Q
CO
Q
'D
C
CO

Unesco
REPORTS AND PAPERS
IN THE SOCIAL SCIENCES
The Reportsand Papers are intended to present to a restricted
public of specialistsdescriptive or documentary material as and
when it becomes available during the execution of Unescospro-
gramme in the field of the social sciences.They will consist of
either reportsrelating to the Regular Programme of Unesco and
its operational programmes of aid to Member States or documen-
tation in the form of bibliographies,repertoriesand directories.
The authorsalone are responsible for the contentsof the
Reports and Papers and their views should not necessarily be
taken to represent those of Unesco.
These documents are published without strict periodicity.
Currently available.
SS/CH11 - International Repertory of Institutions Conducting Popu-
lation Studies (bilingual:English/French), 1959.
SS/CH15 - International Co-operation and Programmes of Economic
and Social Development (bilingual: English/French), 1961.
-
SS/CH17 International Directory of Sample Survey Centres (outside
the United States of America) (bilingual: English/French),
1962.
-
SS/CH18 The Social Science Activities of S o m e Eastern European
Academies of Sciences, 1963.
-
SS/CH19 Attitude Change :a review and bibliography of selected
research, 1964 (out of print in English, available in French).
-
SS/CH20 International Repertory of Sociological Research Centres
(outside the U.S.A.)(bilingual: English/French), 1964.
-
SS/CH22 Institutions Engaged in Economic and Social Planning in
Africa (bilingual: English/French), 1966.
SS/CH23 - International Repertory of Institutions Specializing in
Research on Peace and Disarmament, 1966.
-
SS/CH24 Guide for the Establishment of National Social Science
Documentation Centres in Developing Countries,1969.
-
SS/CH25 Ecological data in comparative research :Report o n a
fiist International Data Confrontation Seminar (bilingual:
English/French), 1970.
SS/CH 26 - Data archives for the Social Sciences :Purposes,operations
and problems, 197 3.
-
SS/CH 21 DARE Unesco computerized data retrieval system for
documentation in the social and human sciences.
SS/CH28 International Repertory of Institutions for Peace and
Conflict Research.
SSlCH 2 9 - The Unesco Educational Simulation Model (ESM).
SS/CH3 0 Social indicators: problems of definition and of selection
Social indicators:
problems of definition
and of selection
Methods and Analysis Division
Department of Social Sciences

Unesco
Published in 1974 by the Unesco Press,
7,place de Fontenoy,75700 Paris,France

ISBN 92-3-101161-8
French edition ISBN 92-3-201161-1

Printed in the workshops of Unesco


Printed in France
0 Unesco 1974 [B]
P reface

The present paper needs to be seen within the context of potential. The concept used has been broad enough to
the Unesco project on human resources indicators. cover,under the label human resourcesindicators,
In 1967 a special meeting of experts was held by such indicatorsas are described in other research pro-
Unesco in Warsaw with the purpose of discussing alterna- grammesas social indicators,socio-economicindicators,
tive approachesto the elaboration of a system of indica- indicators of modernization,identified as those indicators
tors of human resources development,and of giving ad- which are supposed to affect both labour productivity
vice on the choice of methodology. The meeting had and the quality of working life.
discussed papers on the concept and theory of develop- The ranking,grouping and clustering of countries on
ment,on procedures of evaluating high-levelmanpower the basis of a number ofhuman resourcesindicatorswere
data and methods for establishing an order of countries effected by mathematical methods devised for the
ranked according to their development levels. purpose.
In the years that followed,the project developed with Studies of gaps and disparities between countries
the help of an Advisory Panel which met a few times. through the use of these various methods have been com-
Working groups discussed the work carried out,which pleted by similar studieswithin individual countries.
was mainly effected through a number of studies, and The project on human resources indicators ended in
made an evaluation of the various methods which have December 1973. The last in a series of meetings to be
been devised or adapted to deal with specific problems, held under this project took place in December 1973 at
the most important of which has been the identification the Institute of Development Studies at the University of
and selection of indicators. It is this last topic which Sussex,United Kingdom. Two new projects have been
constitutes the main body of the present paper. The launched,the objectives of which have been clearly
paper also explores the problem of definition,which has defined.
been centralto the whole project,whether of develop- The first project relates to the identificationof key
ment on which it hinges,or of the indicators themselves indicatorsof social and economic change,for the pur-
with which it deals. pose ofappraising socioeconomicprogress towards the
In 1969 when the Advisory Panel met for the first objectives of the United Nations Second Development
time,its first task was to review a tentative list of human Decade.
resources indicators. It was also expected to resolve some The second project relates to the use of socio-economic
controversialpoints and formulate operationalrecom- indicators in development planning, Such indicatorswill
mendations for future research. The other topics dis- be inore suited to the requirementsof planners in an in-
cussed were the concept of human resources,the weight- dividual country situation,than those which are highly
ing of componentindicators ofhuman resources,and aggregated and tailored for international comparison.
the r61e of human resources in development. The indicators of relevance to this particular project will
At its second meeting,held in 1970,the Panel dis- be drawn from lower levels of aggregation in both the
cussed papers and studieswhich were undertaken on the geographical and the social space. They will relate to
basis of recommendationsmade at the first meeting. some of the major sets of determinantsof distributionof
The emphasis shifted however from definitionsof con- development and welfare such as: geographical location
cepts,from the descriptive aspects of human resources and ecology,social characteristics,and economic function.
and the mathematical problems of selecting and weight-
ing indicators,to the interrelationshipsbetween human
resources and development. Human resources were not 1. See a list of these in Annex on page 27.
understood in the project merely as labour or labour (These mimeographed papers are avai )le on request).

3
A third project which will be launched in 1975 will methodology of indicator identificationand selection,
relate to indicators of the quality of life and of the which with concept definition,has occupied a privileged
environment. place in the old project. The present paper illustrates
A common concern of these new projects is the this effort.

4
Table of contents

Part I Social Indicators: Problems of Methodology and Selection


.

by Serge Fanchette ................................................. 7

Part I1 A Method for the Selection of a Compact Set of Variables


.
by Zdzislaw Hellwig ................................................. 11

Part 111 A method of Establishing a List of Development Indicators ...................


.

by Branislav Ivanovic ................................................ 21

Annex .................................................................. 27

5
Part I

Social Indicators:
Problems of Methodology and Selection
by Serge Fanchette
Methods and Analysis Division
Department of Social Sciences
U nesco

1. DEFINITION AND PURPOSES OF (i) appears to be the main function served by social
SOCIAL INDICATORS indicators in general. (ii) requires the building of a
The first question which arises is that of definition of a system of indicators or the construction of models
social indicator,followedby a corollary question related which should integrate at least some of the main com-
to the purposes. Most of the definitions which have been ponents of social reality. The conceptual difficulties
put forward contain reference to the purposes of social arising from the many complexitiesinvolvingamong
indicatorsthereby making artificial the distinction defini- others the interplay of empirical and theoreticalfactors
tionlpurpose. In the rapidly growing literature on social in this field have limited attemptsto build social models.
indicatorsdefinitionsof great diversity are proposed, Such attempts have either over-simplifiedreality in order
ranging from the term being used to qualify good social to capture some relationshipsthat may exist between a
statistics obtainable as time-seriesto series which can very limited number of social variables,or have attempted
fit explicitly into sociological,social policy or social pro- to introduceintothe model somany parametersbelonging
cesses models. Within this wide spectrum we find the in- to a host of social phenomena as to blur out any existing
dicatorswhich under the broad appellation of social relationships.
are described as normative,further distinction being By-passing(ii) it has been customary to link up (i) and
made between input and output variables,i.e. indicators (i), going from the descriptive to the prescriptive through
of means as distinguished from indicators of results. an intuitive or a heuristic approach,more often of the
There are the structuralvariables characterized by their former than of the latter.
interrelationshipswithin part or the whole of a social The definition of Stuart Rice,especially in its prag-
system,as opposed to those which are amenableto some matic approach,finds an echo in the following quotation
kind of summation into a composite index. And,finally, from Sheldon and Moore (eds.), Indicatorsof Social
we find.those indicatorswhich are not viewed within any Change (NewYork,Russell Sage Foundation,1968,p.4):
system and which will rank as performance indicators, It is especially for those who have undertaken responsi-
socialreporting or socialintelligence,etc. bility for bringing about publicly approved changes that
All these aspects are combined in Stuart Rices defini- the notion of social indicatorsis appealing. Such in-
tion of social indicators: Social indicators are needed dicatorswould give a reading both on the current state of
to find pathways through the maze of societysinter- some segment ofthe social universe and on past and future
connections. They delineatesocial states, define social trends,whether progressive or regressive,according to
problems and trace social trends,which by social engineer- some normative criteria. The notion of social indicators
ing may hopefully be guided towards social goals formu- leads directly to the idea of monitoring social change.
lated by social planning. If an indicator can be found that will stand for a set of
In the above definition three main elements emerge correlated changes,and if intervention can be introduced
which define social indicators and their purpose: (whether on the prime,indicative,variable or on one of
(i) the descriptive functionof the indicators - the its systenlic components), then the programme administra-
description of social states and of trends in social tor may have been provided a powerfulanalytical and
change; policy tool.
(ii) their interconnections- which suggestsa systems The above passage introducesthe hypothesis of a set
approach; ofcorrelated changes for which an indicator could stand;
(iii) the analytical tools they can be to the social planner - a similar concept runs through the second and third sec-
for the monitoring of social change. tions of this paper when it comesto selecting a reduced

7
number ofindicators from a larger set. In the section by A social statistic may rank as a social indicator only in a
Z.Hellwig a clusterhgofdevelopment indicatorswill be specific social field for a specific purpose. Instead of
effected by the use of taxonomic graphs on the scale of drawing up a long comprehensive list of social indicators
distances calculated according to correlationcoefficients. for universal usage it might be appropriate to subject each
In the section by B. Ivanovic a method of establishing a candidate indicator to close scrutiny within the context
list of development indicatorsis proposed,based on the of specific areas of social concern and through rigorous
I-distance. interdisciplinary treatment. The UN system of social and
The definition of Stuart Rice and the quotation from demographic statistics could be seen as the ordering and
Sheldon and Moore appear to circumscribe directly and classification of the raw material in a comprehensive and
precisely the attributes of a socialindicator. However, systematic way as suggested in Group A (iii) on this page.
with the wide interestexpressed in socialindicatorsgrow- Some of the data upon which such a system will focus
ing into what has been called the Social Indicators will initiate the construction of certain indicatorsor may,
Movement,there has been a great laxity in the use of the without further manipulation,serve directly as indicators.
term,to the extent that most of the social statistics which The inductive approach underlying much of the current
are usually collected for administrativepurposes are being data collection and selection should be supersededby a
rebaptized Social Indicators. There is a good deal of deductive approach;the identification ofdemands,of
uncertainty as to where to draw the line. At a consulta- aspirations,of social concernsin general,should influence
tion on social indicators with the secretariatsofinterna- and consequentlyprecede the collecting and marshalling
tional organizations,convened in Geneva by the Con- of relevant data.
ference of European Statisticians,a body of the United The Unesco Social Science work on indicatorshas
Nations Economic Commission for Europe,an eminent until recently focused on human resources indicators.
participant with the view of tidying up terminology These may not all be termed social indicatorsand cer-
proposed a classificationof statisticalseries into six types. tainly most of the indicators in the system would not if
Rices definition were accepted in toto. Many of the in-
GROUP A: dicators regarded as human resourcesindicatorsin
(i) Raw statistical series. Unesco work would,however,fit the definition of social
Key series, e.g. a common sense attempt to pick indicatorsas implied in some of the tables which appear
(ii) in the UN document. This definitionwould be considered
out series which one regards as most interesting or
by many as being too loose,showing a great laxity in the
important for presentation. The Social Trends,
published in the U.K., could be considered as a care- use of the term. The list of human resources indicators
ful,thoughnot rigorous,selection ofseries. which has been drawn up as the project developed does
not refer to the indicators as social indicators. The dif-
(iii) The system of social and demographic statistics;
which is still basically the raw material but put to- ferent fieldswhich have been delineated contain indicators
gether in a comprehensive and systematic way. which are described as educational indicators,health
indicators,employmentindicatorsand the like. They
GROUP B: would be regarded as definitional indicatorsin the
first place,i.e.they refer to such phenomena as are wholly
(iv) Composite indices derived from somehow combin- or partly identical with the indicatum. Other indicators
ing individualseries. This is an application of the may be termed inferential,i.e.they reflect the directly
classicalindex number approach. unobservable or unobserved indicatum indirectly.The
(v) Synthetic representative series derived by multi- educational indicators are termed educational because
variate techniquessuch as discriminant analysis etc. they relate to education and are therefore definitional.
(vi) Series which fit explicitly into social models. But they would also be inferential indicatorsof human
resources,showing the flow aspect of human re-
The United Nations have been attempting to develop sources development. As such they would better qualify
an Integrated System of Demographic and Social Statis- as indicators,i.e.in the second sense,since many believe
tics. In the many documents which have appeared as
working papers for meetings held for that project no
criteria have been proposed by which social indicators 1. T h e participant later added that he would not want his
and socio-demographicstatistics are differentiated. proposals to be regarded as too weighty from the point of
Thus the different tables in one of the documents2 show view of substance and that there was quite a lot wrong with
this classification. H e went on to say that the essence of
in the first and fifth columns statistics which appear as the social indicator,by definition, must be that it is indicative
both itemsof data and social indicatorswithout of something else, and that this attracted him to confining
apparent justification. It is true that most of the series the term to category (vi). H e thought that in practice that
listed in these tables could rank within Social intelligence was asking too much, since people are almost bound to want
or Social accounting. Short of a social model into which to use the term for some of the other categories. But what is
important is that w e think in terms of a hierarchy of types of
these indicators will fit explicitly,one may still use series.
certain criteria,based on specialist knowledge,to select 2. An integrated system of demographic and social statistics, etc.
the social indicators in view of some specific purpose. - -
(E/CN3/432)United Nations Geneva.

8
that the essence of an indicator is that it is indicative of Thus it would appear that the construction ofthe com-
something else (see footnote on page 8) posite indices,however big the difficultiesmentioned above
The series listed as human resources indicatorsin the and despite the lack of success which has met the many
Unesco project would belong to type (ii) in Group A endeavours,both for developing such indices or for find-
(see page 8). ing a use for them,is stiU envisaged by quite a few organi-
zations or institutesin their programme of work on indica-
tors. It is true that on the whole the construction ofsuch
2. COMPOSITE INDICES indiceshas lost much attraction,because ofthe many in-
adequaciesmentioned by Professor Moser. But somehow
However they may be qualified many series in the Unesco we know that work for their constructionwill go on for
list have been sometimesaggregated into composite in- quite some time. The aggregation that work involves
dices to provide development measures constructedas hinges heavily on selectionand weighting.
better substitutesfor GDP,to establish international com- The division into the various fieldsin the Unesco list
parison and to show movement over time at both within- ofhuman resources indicators was not only a matter of
country and internationallevels. Many sophisticated classification. There was the intention of aggregatingthe
methods (including Wroclaw Taxonomy,the I-distance indices for indicators within each field into a composite
method) have been used to derive such indices. In other index for that field. The studies on the problem of
organizations,a few years back,the eagerness to have a weighting had in view,among other aims, the solutionto
social development measure to match the economic such a problem of aggregation.
development measure,the GNP,led to the construction of The aggregation of several componentindicatorsinto
various indices. The most interesting have been the dif- one composite index has as many opposers as supporters.
ferent level of living indicesconstructedunder the The former believe that a composite index tells us much
aegis of the United Nations and for which considerable less than the individualindices shown separately and is
work at both conceptualand methodologicallevels has thereforeless useful. For those who believe that develop-
been done by the United Nations Research Institute for ment forms a distinctive pattern with progress along its
SocialDevelopment,at Geneva. This quest for some various componentsbehaving rather uniformly and con-
sort of Gross HappinessProducthas been inspired by tinuously,a profile made up of all the elements separately
the conviction that GNP is not everything and does not is more illuminating. It provides a picture of dispersion
in itself reflect the welfare of people. This inspiration measures in the human resources componentsof develop-
has unfortunately become an obsession and GNP,despite ment,thereby pointing to imbalance and bottlenecks in
its many shortcomings,is always present,so much so that the system. On the other hand,if the trajectories of
the system of social accountsbeing envisaged tries to be development are not parallel and substitutability exists
too close an imitation of the economic accounts. Adding to some extentbetween indicatorsof the same field,
up the componentsof production,with a convenientcom- aggregating several componentindicators within that
mon measure like their price,is not attended by the many field would provide a unitary measure which could be of
conceptualproblems underlying the aggregation of com- great value in international comparison.
ponents of welfare,ifnot componentsof the goodlife.
A short article in the 25 December 1971 issue of the
Economistentitled Where the grass is greener,com- 3. THE CONSTRUCTION OF SOCIAL MODELS
menting on the construction of a Social thermometer
which is the equivalent of the gross national product, The hierarchy of series listed on page 8,which proceeds
said: Certainly,social statistics are improving and in- in ascending order of relevance to the social indicator
creasing; statisticiansare making ever more ambitious concept,shows the last type of series as comprising
attempts to compare like with unlike and to marshal1 those which fit explicitly into social models. Accord-
prioritiesin the social services into neat statisticalbalances. ing to a strong body of opinion,these series would better
But the task of selecting and weighting for the calculation qualify than others for the term social indicator. Some
of an umbrella index of the quality of life makes the cal- of the protagonists in the quarrelover the definition
culation of GNP seem like schoolboy arithmetic.... of a social indicatorwould reserve the term to a series
Discussing the different definitions of a social indicator, which satisfies expressly the followingconditions,i.e.
Professor Claus Moser,Director,Central Statistical Office, that,to rank as an indicator,it must have a place within
London,wrote recently: ...in some writings,it is held some sociological or social policy model. Economic in-
that socialindicators should necessarily be combinations dicatorsfit into economic models,constructedon the
of series,i.e. index number in the conventionalsense, basis of economic theories. A social indicatorshould
This is an undesirable restriction of the term. Apart from likewise find an interpretation only within the context
anything else, the technical objections to such indices - of a theory. The analogy in this respect to the economic
-
most basically the problem of weighting are substantial.
All the inadequaciesof concepts,measurement,inter-
indicatorhas to be pursued to a logical conclusion,
which appears to many social scientistsas the ultimate
pretation are compounded,and in our approach,work on phase of work in the field of social indicators before
index number combinationsis envisaged for later stages.... they can become useful tools of planning,namely their

9
integration as parametersin social models or models of nation,to the quality oflife are not always correlated
social change,built upon social theory and policy. But positively with economic growth. As a result,socio-
prior to the launching of that phase and implicit in the economicmodels,despite the appellation,will rarely
preliminary work remaining to be done is the analysis seek to include these,and will most times lean heavily
by social scientists ofall the currently collected com- towards the purely economic side. Such models,despite
ponents of social accmting,the indicators,so as to their complexity are still within reach of some of the
bring out their real & cmtextual meaning. There is more enterprising model builders; the relatively easy
the constant shift in interest,and some indicators conceptualization of the ingredientindicators,their
either age quickly or else are discovered irrelevantor ready measurability and availability,coupled with the
useless. Other indicatorsstep to the front of the stage, mammoth capacity ofsome third generation computers
propelled by concerns of real urgency,or simply by to extricate relationships,have led to a proliferation of
fashion. For over a decade now quality of life has. models,on which often by hindsight theoriesare built.
become a major preoccupation without there emerging One can conjecture that with the softer, societalin-
any indicators which purport to measure it however dicatorswhich would go into social models dealing with
-
crudely indicatorswhich,it is expected,would dethrone
GNP or rate of economic growth in the public eye. These
real social progress,improvementof the quality of life
and the like,the difficultiesfacing the social model
economic indicators are still faring well,though ancillary builder would make the effortsof the econometrician
economic indicatorslike the rate of inflation,loss of pur- pale into insignificance. All the interacting,conflicting,
chasing power,unemploymentor inequality of income reinforcing,inhibitingvariables changing their directions
distribution,which should be more relevant,seldom enjoy as time unfolds would all have to be taken in and would
with development economists the status they should have indeed make macro-economic models look like schoolboy
in planning. There has been the failure to develop indi- arithmetic,to use the expression of the article of the
cators which might be conceived as describing the quality Economist. The difficultiesare such that the number
oflife. Indicatorswill eventually be constructed for the of those is increasingwho consider the difficultiesas
description of the environment in its physical,social and insuperableand societalmodels as a wild dream!
cultural aspects and in the interreactionsthereof. There An important task which remainsbefore the elabora-
is still a long way to go for the construction or identifica- tion of such models,if one followed the hierarchy sug-
tion of the indicatorsrelating to todays concerns. That gested in the listing of types of series on page 8is the
work lags far behind the emergence of these concerns selection of series,within the inductive approach,among
which have been asserting themselves for more than two those that are currently collected and are sufficiently
decades. One no longer can invoke fashion to explain reliable. Such selectionshave been based on common
the appearance of these new preoccupations,or con- sense,intuition or on interdisciplinary treatmentand will
jecture that with the passage of time some of them will vary with the specificpurpose the series are intended to
fade away. The environment and the quality oflife, serve. When it comes to measuring development levels,
for which interestis relatively new,will assume more both at within-country and internationallevels,there is
importance within the economy of penury which appears the necessity ofreducing the set of development indicators
more likely from now on to govern the existence of man usually available. Parts I1 and I11 propose methods for
on space-shipEarth. For those who are responsible for the selectionof key or core indicators of socio-
the badly needed dashboards,the struggle will have to be economic development. The exerciseswhich have been
carried on at two levels: at the higher one,constructing worked out in the presentation of the methods have more
the indicatorsto serve as dashboard or danger signals illustrativethan substantivevalue. They had to rely on
relative to these concerns,and at a lower one scrutinize, those series which are usually available for a sufficient
test,analyse the mass of data which are regularly col- number of countries. One should not read too much into
lected at great cost in a variety of fields,and weed out the selection of the original series,which was governed
those for which no utilization has ever been recorded more by the availability of data than by the conceptual
and the collectionever been questioned. validity ofthe series themselves.The methods are evidently
An area which has received a great deal of attention applicable to series which may better qualify themselves
has been the study ofinterrelationshipsbetween socio- as socialindicators. As such they would prove a useful
economic factors,reflected by many of the favourite tool for the study and selection of indicatorswhich as
indicators,and economic growth. The selection of in- suggested in Sheldon and Moore stand for a set of cor-
dicatorsis relatively much easier to achieve in this case. related changes.
Many indicatorswhich relate to the welfare ofa

10
Part I I

A Method for the Selection


of a CompactSet of Variables
by Zdzislaw Hellwig
Graduate School of Economics
Wroclaw, Poland

1. EMPIRICAL MODELS OF SOCIO-ECONOMIC all variables (the number of which is limited only
DEVELOPMENT by data availability) and the many different subsets
of variables selected the best way so as to throw
It is a known fact that the problem of economic growth light on the many facetsof the object under study.
belongs to the most complex,difficult,and thereby con- (iv) Having selected optimal sets of variables one can
troversial part of theoreticaland applied economics. This then define countries in vector terms. This allows
problem has attracted a lot of scientific interest and atten- the division of the whole universe of countries into
tion and has been tackled from many view-pointsby many a number of sub-groupswhich are more homo-
outstandingeconomistsand mathematicians. The per- geneous than the universe itself. This clustering
spective has ranged from a historical or descriptive pur- process is of real importance because it reveals the
pose at one end of the spectrum to abstract and sophisti- main differences and proximities between particu-
cated relationshipsof a causal nature,at the other. lar countries. Statisticalanalysis can now allow a
An analysis of the practical function of purely descrip- deeper investigationof the internal nature of the
tive or even more advanced mathematical models usually phenomena concerned (for instance by applying
based upon trend extrapolation will show that while they variance analysis and related techniques).
may be useful for pedagogical reasons they do not help (v) When countries have been represented by vectors
to solve some basic problems of economic development they can be more easily compared. This may be
which planners have to face in their every-daywork. This effected by defining the distance of one country
has led some investigatorsto look for alternative approaches from another. This distance can simultaneously
to the important problems of economic development and play the r61e of a yardstick of socio-economic
growth. One of these approaches has been presented in development (by finding how far a country lags
the series of studiespublished as working documents under behind the ideal country) and may be looked
the Unesco project on human resources indicators. upon as a composite index of the overall achieve-
Implicitin the approach is a concept which will be ment of that country in a specific sphere of its
roughly sketched as follows: activity. Recent studieshave shown that the use
of certain metrics allows the possibility to go up
6) The object ofany scientific study has to be pre- and construct more general composite indices,
cisely defined. In cases where political,administra-
tive,geographical or social structuresare being in- solving thereby a very important problem of
vestigated the most concise and convenientdescrip- measure aggregation,or to go down,i.e. split a com-
tion is by the way of vectors. The object of study, posite structure into some more simple and elemen-
a country for example,or any of its administrative tary parts (see for example the study by S.Fanchette,
parts,is usually a very complex geo-politicaland proposing a method for the aggregation and dis-
socio-economicobject. Precision and concision which aggregation of socio-economicmeasures based on
are both very important here can find expression in the Euclidean metric).l In this respect,the pos-
the vector concept. sibility of both analysis and synthesishas been
(ii) If any socio-economicunit is to be presented and demonstrated.
described in vector terms then the most essential
1. Study XX: Distance-based analysis, numerical taxonomy and
part of the exercise will obviously be the appropriate classification of countries according to selected areas of socio-
selection of the vector components. -
economic development, by Serge Fanchette, Unesco mimeo.
(iii) One has to distinguish between the optional list of (Ref. SHC/WS/237, M a y 1972).

11
Nowadays one cannot speak of the development of 3. Development of some intuitive approach toward the
any country by isolating it from its geographical problem, formulation of working hypotheses which
environment or from the family of countries to could provide us with some plausible explanation of
which it is bound by geo-politicalties. One may go the internal structure of the whole system of inter-
further still and say that its development cannot be related phenomena.
studied in ignoring the rest of the world. Since we 4. Designing of statistical experiments.
cannot construct a model which will describe and 5. Data collection.
explain on a world scale all the movements on the 6. Data testing.
map of development we may accept the following 7. Introductory statistical data analysis.
fundamentalassumption: it can be judged and 8. Setting up correlation tables.
measured only by making comparisons. There are 9. Clustering statistical units (which are formally under-
two main dimensions of comparison: time and stood as vectors in an N-dimensionalmace).I ,

space,the latter being understood in a very broad 10. Clustering statisticalvariables which are also inter-
sense. By comparing countries we may discover the preted as vectors in an N-dimensionalspace.
pattern of a well-balancedstrategy of devel3p- 11. Selection of the most representative variables.
ment and also point out this country within a 12. Aggregation and construction of some composite
relatively homogeneous group of countries whit 7 indices.
could play for the others the r61e of an empirickLLy 13. Final statistical analysis and building empirical
verified model of development. One could study models.
with the utmost attention the conscious or uncon- 14. Estimation of parameters.
scious development patterns adopted by the leading 15. Final tests of coherence between the theoretical
country. The implication is that instead of trying description of the phenomena involved and their
to construct some more or less primitive (or sophi- true description.
sticated as one may in fact judge) models of de- In the research project on human resources indicators
velopment,it would be easier,cheaper,safer and being carried out by the Unesco Department of Social
better to have a real,living model. The problem is Sciences,steps 1-9and also 12 have already been dealt
only how to detect this model in the number of with to a greater or lesser extent. In this paper we will
families of homogeneous countries. concentrate on stages 10 and 1 1.
Having outlined,and commented upon, this concept,
let us consider the methodological aspects and working 2. S O M E SELECTION CRITERIA
procedure in building such empirical models. In this There are many methods available which could be applied
procedure the selection of variables will play a very for the selection of such a set of variables. One can divide
crucial r6le. the whole range of these methods into two groups accord-
In some modern fields of experimentation,e.g.in ing to the character of the criteria chosen for effecting
biology, medicine, agriculture etc., it is beyond our this selection. If we decide on intuition or on commonly-
reach to develop a very strong and reliable basis for des- held views that one of the variables belonging to the
cribing and explaining interrelationshipsbetween all the complete set of candidate variables should serve as such
phenomena involved. There are two main reasons for a criterion then we refer to an endogenous criterion.
this: This is the case for instance when we choose per capita
(a) The whole set of phenomena is, at least in theory, GNP as the most valuable indicator of economic perform-
unlimited; ance and use this variable as a yardstick for measuring
@) There is the strong influence of some random fac- the relevance of components of a system,of which it is
tors which distorts a pure functional relation- a part,which cuts across economic and social fields,a
ship between identifiable and measurable variables system of which economic performance measured in the
which are known to be involved. value of goods and services produced is one of the
characteristics.
To meet this methodological challenge in some applied Sometimes having decided on the components of a
sciences (sociology and economics offer good illustrations) system reflected by the number of variables chosen,one
a great deal of scientific effort has been recently under- may turn to an external variable to serve as the yard-
taken,aimed at the elaboration of a statistical procedure stick to assess the relevance of each of the candidate
which leads to the construction of more or less sophisti- variables we have defined as likely to be part of the set
cated models which could simulate with some degree of we should select for describing the system. In this case
accuracy the behavioural pattern and performance of we speak of an exogenous criterion. For example a
the object under investigation. composite index of socioeconomic development identi-
Such a procedure may consist of the following fied with the distance from the ideal country of
stages: Wroclaw Taxonomy 1 may serve as such a criterion.
1. An introductory penetration of the field of interest.
2. Elaboration of definitions of some basic phenomena, 1. The ideal country concept is expounded in Study 111, by
facts,variables and categories. Z.Hellwig.
12
This involves a certain amount of circularity,since the selected for describing socio-economicdevelopment are
taxonomic distance is made up of distances on dimensions the continuous variables most closely associated with the
represented by all the candidate variables. A better ex- categorialvariable whether the country falls within the
ample may be chosen from principal componentsanalysis denomination L.D.C.or M.D.C.It is very difficult to deal
where the criterion of selecting the most representative adequately with a mixed bag ofcontinuous,discrete and
variables is the degree of explained total variance. categorialvariables simultaneously.Most times the distinc-
If we opt for an endogenous criterionwe may apply tion between those types is clear-cut.Sometimesthe line
a method developed in some earlier studies carried out of division is blurred,just as a continuousstream may have
under the Uneco project and which relates to the con- successive cascadingjumps. Another example is the con-
cept of maximum information capacityof a predict0r.l tinuous flow of cars in a city and the interruptionsof that
Alternatively one may as well ayply a better-known flow by traffic lights,converting it from a continuousto
method: that of examining the statistical significance a quantum form. Luckily the problem of classificationis
of regression function parameters in a multidimensional not intractable enough to suggest the analogy with the
case. For the application of this method which falls contemporaiieous wave theory and corpuscular theory of
within a closely-relatedfamily of multiple regression light which testified differently to its continuousor tQ its
analytical methods,the use of the very handy and effec- discrete nature. Neither will it be necessary to carry the
tive stepwise regression and checking procedure is analogy further still with the electron,which partakes of
strongly advocated. In any reasonablewell-equipped both natures.
computational centre there are ready-madeprogrammes When our list contains all the types then one has either
available for these operations. It should however be to convert by appropriate scaling continuousvariables
remembered that when choosing this method the check- into discrete variables,or vice-versa.
ing of some assuptions should be carried out beforehand. In tilt: first case we would have to stop speaking ofthe
These assumptions are implied in the method with re- variables and of their numerical values. We would have to
gard to the multidimensional random variable distribu- translate them into adequate probabilities. It means,for
tion. This checking is usually a rather dull exercise. instance,that the expected value E (X)would have to be
replaced by np, the standard deviation S (X)by (np (1 -p))
the correlation coefficient r (X,Y)by the coefficient of
3.CHARACTER (OR TYPE) OF VARIABLES stochasticaldependence:
There are two other points which should be emphasized
when analysing the different methods for selecting from d2 =
1-xyf(Pij-Pi x P.)J
the total set of candidate variables the best combination 1
of these. The first point is mainly theoretical and relates 1- .
min (s,t)
to the nature of the variables. All social indicators can be
looked upon as continuous variables. This statement has
(where p-.= probabilitiescorresponding to
?1
(Y,5); p- p.-
1 J
far-reachingmethodological consequences. It means marginal probability distributions;s,t - number of rows
among other things that we are justified to apply such and columnsrespectively in the probability contingency
basic statistical parameters as the mean value,the standard table). It means also that instead of applying a conceptu-
deviation or the coefficient of correlation. It also means ally and numerically easy taxonomic method or the less
that if we do not involve discrete or categorial variables, easy but well-knownprincipal componentsmethod,we
then for clustering countriesand variables we can use would have,for clustering and selecting the most appro-
either the Wroclaw taxonomic method or the principal priate variables,to devise some special methods to deal
componentsmethod to mention but two of them. The with discrete variables and categories. One of these
use of the majority of socio-economicvariables selected methods is the Analyse des correspondanceswhich can
for analysing the statistical aspects of socio-economicde- be regarded as an adaptation of the classical principal
velopment rests upon the assumption of their continuous componentsmethod to cover a discrete case.
character. However,the range of variables normally appli- If instead of converting continuous variables into dis-
cable to the description of countries may include those we crete ones we operated in the opposite direction,we
might describe as discrete or categorial. Some of these could expect at least two advantages:
(a) Since the number of discrete variables and categories
variables may be,for instance,as following: the number
of ethnic groups living in a country with a plural society, in the list of socio-economicindicators should be by
or the number of languagesor dialects spoken in that far smaller than the number of continuous variables,
country. We have here examples of a discrete variable. the amount of computational effort to convert dis-
crete and categorialvariables into the continuoustype
The variables may also be categorialones,e.g.when they
can assume the form of a question like Does the country
belong to the UN?, which is illustrative of a categorial 1. Studies VI and XVI: O n the optimal choice of predictors;
approximative methods of selection of an optimal set of
variable capable of having only the two dichotomized predictors,by Z.Hellwig. (Studies and working papers pre-
values yes or no. GNP,life expectancy.birth and pared within the Unesco Social Science project on h u m a n
death rates which are among the favourite variables resources indicators).

13
would be much smaller than if we proceeded the other large number it follows that unity is partitioned into many
way round. sub-intervalsthe mean length of which should tend to-
The continuous case is much simpler from the con- wards zero. Hence if we agree to accept a given number
ceptual,and much better known,from the theoreti- po with 0<po<l as a threshold value and retain all those
cal,point of view. It means that in this case we are
not only better equipped with powerful tools of
5
variables correspondingto weightsp. arranged in a
J
decreasing order which fulfil the condition:
statistical analysis but we have also many ready-made
programmes if we choose to use a computer. E p j 2 1 - p ~ ;no\< n
There are different methods of converting discrete and and if we take into account on the one hand that po is an
categorial variables into continuousones. For discrete arbitrary value which can vary within a pretty wide inter-
variables one of the usual techniquesis linear interpolation. val and on the other hand that we can never measure the
Recently a new technique has been successfully applied, relative importance of variables with a high degree of
the so-calledlinear randomizationprocedure. To convert accuracy,then we should adopt a practical rule that all
categorial variables into continuousones the best-known retained variables are of the same high relative importance
and widely recommendedmethod is the so-calleddummy and should be given the same weights,say p,whereas all
variables procedure. rejected variables being of minor importance have their
weights,say q ,although equal as well but smaller than
4.THE PROBLEM OF WEIGHTS weights p ascribed to the retained variables (see fig. 1).
Hence the problem of selecting the most appropriate
The modern techniqueof describing some objects of variables boils down to the equivalent problem of weight-
scientificinvestigation by means of vectors in an ing which in turn consists in finding such a number
N-dimensionalspace immediately brings up the question no = f(p,) that no)p 1 - p, ,where of course
of the relative importance of particular vector components. no p + (n-no) q = 1
Experience or intuition,even in the absence of knowledge
of a particular field,would suggest that the relative im- since
portance of particular variables,whether descriptive or p = n1, pj and q = - 1 1 Pj
explanatory,cannot be uniform. Data in raw form would n-no j>no
jSn0
at the very start cloud any quick evidence of that relative We have attempted to show that the problem of measuring
importance. The common practice before examining by the relative importance of variables (or components of
the usual tools of statisticalanalysis the relative import- vectors by means of which we define the object of our in-
ance of these variables is to standardize the latter. The vestigation) is equivalent to the weighting problem and this
standardizing procedures are conventional. These do not can always be expressed in terms of reducing the initial
constitute a difficult task. They however cover a wide tentative list of variables to a much smaller size.
range of practices. The common standardizing practice This reduced set of variables should contain only the
of replacing xi by (xi-%)/ s, where 5 is the standard most important or the most representative ones among
deviation,introducesalso the element of weighting. the original set. The problem of finding this reduced or
While eliminating the problem of units,it also favours the compact set of variables is at the very heart of cross-
variable with a smaller standard deviation. This first bias sectionalanalysis and constitutes a primordial element in
left aside,there comes the real problem of ascribing the whole theory of internationalcomparison in the socio-
weights to the variables. This cannot be satisfactorily economic field.
solved unless there is agreement upon the most appro-
priate criterion for evaluating the relative importance of
particular variables. This brings us to the examination of
the different criteria one may consider for differentiating
between the variables and for situating them on the various
steps of a hierarchical ladder. As however we need the
same criteria to select the most meaningful variables and
point out the most important among the selected ones,
we may conclude that the problem of selecting the vari- 1 2 3...n0.........j ............ n-1 n
ables and that of assessing their relative importance by
means of a given system of weights are mutually equivalent Whether the objective be theory-buildingaround causal
and show the two different sides of the same coin i.e. that hypotheses or,more soberly,the testing of empirical
before one starts speakingabout weighting variables one models,the selection of a smallnumber of relevant
has to determine the most adequate criterion for compar- variables from the forest of those for which data are col-
ing and selecting these variables. lected has paramount importance. Both thiorie and
There is also another important fact about weighting empirie are required; the latter is usually within better
which is worth remembering.Weights are non-negative reach of analysts. The reason is that work usually starts
numbers: pl,p2 ...pn such that >: p. = 1. If _n is a
J
with data that are available. Other variables might be
j=1
14
more relevant which are unfortunately not collected.The where i,j = 1,2,...and d.. 0)
1J
>
testing oftheoretical or hypothetical constructs would Now the smallest element has to be found in each
require that data be collected on variables which are row of the matrix D. If we denote these elements
absent from international compendia,national reportsor by dithen di = m i n d..
administrative files. The problem of selection on the 4
contrary implies a reduction of the number of variables j
forwhich data are collected internationallyaiid are there- After having found diin the ith row,one finds the
number of the column which forms with the given
fore available. This type of analysis therefore suits
row a cross-cellcontaining the value di. In this way
empirical investigations better. Put differently empirical one gets n pairs of indices:
studies would come earlier in the day than theory building
and testing. The importance of selecting a set of core (l,Jl), (2,J2),......(njjn)
variables to describe or predict some phenomenon,like Let us now define a few terms. The terminology
for example,socio-economicdevelopment is for the reasons will be very simple. Each of our pairs of indices
given above very important. No wonder therefore that so will be called a link whereas i and jiin (i, ji) are
much attention has already been given to the problem and nodes. The first node is the begirinirzg of the link
so much effort has gone into the search for satisfactory which ends with the second node. The links may
solutions. Earlier studies carried out within the Unesco therefore be called oriented links. When we draw
project , research studies carried out by various institutes a graph all nodes will be represented by points and
and agencies2havedealt with the methodology of indica- all links by arrows. The number of arrows pointing
tor selection and weighting. They have shown most of at a given node will be called the power of that node
them varying degrees of sophistication. If+ it is suggested The node with the highest power is next found. If
above,selectionand weighting are looked at as being in- there are more than one with the highest power the
timately connected,we could propose a rather simplified choice will fall on the one to which corresponds the
approach to the whole problem. smallest value of - n
This is of course a conclusion the importance of which d.=l C d..
i n j - 1 1J
should not be underestimated. After all the most sophi-
sticated contemporary statistical methods such as Fishers Now all nodes which are beginnings of links point-
test of overall significance of regression parameters, step- ing towards the node of the highest power are to be
wise regression,principal components method,factor linked with this node. All nodes attached to the
analysis,the method of maximum capacity of informa- node of the highest power can be now looked upon
tion,la mkthode danalysedes correspondancesare all as the ends of some other nodes which are in to be
dealing with the same bunch of questions: how to deter- linked with all these ends. By continuing this pro-
mine a compact set of variables,how to split the set of cess ofjoining together all the nodes related to the
variables into some more homogeneous groups and how to node of the highest power,one obtains a certain
find the best representativesfor particular groups. family of nodes which can be referred to as the
In the next paragraph a very simple and convenient first concentrationof nodes,or as the first
method will be presented which could supply some cluster of points.
answers to these questions. The method is graph-based Next,one finds all the clusters which correspond to
and does not therefore require going through complex the nodes with powers below the highest one,in a
computationalprocesses which are unavoidable when decreasing order of magnitude.
utilizing other techniquese.g. the principal components The node with the highest power in a cluster is
method. termed the centre of the cluster. If in a cluster
nodes of the highest order number more than one,
we select the one to which corresponds the highest
5. VARIABLES CLUSTERING BY MEANS value of d.. The centre of a cluster can be looked
OF THE TAXONOMIC METHOD J
upon as the representative of variables belonging to
To make the method more easily understandable we shall this cluster. The set of representativesof all clusters
in the course of its presentation illustrateit with numeri- will be referred to as a set of core variables.
cal examples. The range of variation of the parameters dishould
now be divided into three parts:
(i) The algorithm starts with a correlationmatrix
R (n x n) where n stands forthe number of vari-
ables (see Table 13). 1. Studies VI,VII,IX (Unesco mimeographed working documents).
(ii) In the second step the following transformationis 2. A valuable example of selecting key indicators from a wide
range of thesc is given in a research report of the United Nations
i~ I 1
effected d.. = 1- r
ij
by means of which the matrix R is mapped into a
Research Institute for Social Development: Report No. 70.10
Geneva 1970: Contents and measurement of socio-economic
new matrix D (see Table 2). This matrix is also of development: A n empirical enquiry.
3. N.V. Sovani, A.M. Subramanian: A n Index of Socio-Economic
(n x n )type,symetricaland with elements d- Development of Nations, UNRISD, 1969.
1J

5
6. CONCLUSION

where -
d--
1
n
i= 1
di ; Sd [5
= -
i=l
(di-d)*
I
The selectionmethod which has been proposed,while
being a variation on the intercorrelationtheme,has the
virtue of simplicity and provides a clear picture,through
the taxonomic graph,of the relationshipsand distances
All nodes which are beginnings of some links the length between the variables. Furthermore a novel idea has
d. of which belongs to the first interval can be removed found expression in the elaboration of the method: it
3 refers to the problems of selection and weighting as close
from the graph and the correspondingvariables may be
called redundant. Nodes which are beginnings of some correlates requiring a unique solution.
other links the length diof which falls in the third inter- The clusteringtechniquewhichhas been described here
hasidentified clustersofvariables and clustercentres. Each
val can also be eliminated and the respective variables centre could be made representative of the whole group.
may be referred to as irrelevant. All remaining variables Expanding the number of the more representativevariables
would form the compact set. All the steps which lead would mean a movement in concentric circles from the
to the obtention of particular clusters,the core variables centre towards the periphery so as to include the number
as well as the compact set of variables are summarized of variables which would be retained for any specific
and illustrated in the diagrams which follow. purpose. The two exercises which have been carried out
The procedure which has been described is simple and have led to taxonomic graphswhich are reproduced on
liable to easy interpretation. It is equally applicable to the following pages. It may be that in each case the
variables which are continuous,discrete or categorial.If original number of variables has been somewhattoo
however one is put off by,or is suspiciousof,its great limited to derive many useful conclusionsfrom the
simplicity,there are more sophisticatedmethods avail- clustering. Most of the groupingsappear to be consistent
able,like for example,the principal componentsmethod, with what is already known or suspected of the relative
which will provide good solutionsto the problem as it movement of variables as development proceeds.
has been posed above,but at much greater cost.

16
z
m
N
N
N
3
N
0
N
2
00
3
o 2
s5
W
3
w In
2i
n
3
0
W b,
m m d 0 0 d m W 3 N
3
';? 9 \9 Y T ? b S T 3
2
3
3
2
m
00
l-
Q
m
Tt
m
N
3 3
17
a
n
c
--
3
I1
.I
:3
a
II
5-
W
43
0
0
18
w
I
n
-s
&I
e,
+J
U
\
k
0
(1
20
Part 1 1 1

A Method of Establishing a List of


Development I ndicators
by Branislav lvanovic
Statistics Division
United Nations Conference o n Trade and Development

I. INTRODUCTION and a more complete school education. Lastly,where


there is a high degree of concentration,one rarely en-
The choice of social and economic development indicators countersunfavourable cases such as those ofmothers who
is certainly one of the most important problems in the are too young or too old,children born of a second mar-
project for development evaluation and forecasting during riage,etc. In short,the more concentrated the births are
the Second Development Decade: the more complete the in relation to the age of the mother,the more ordered is
information about national development levels contained the family life and the more effective the familysactivity
in the indicators selected,the better the results. from the social and economic points of view.
There are a certain number of indicatorswhich we al- Quantitative analysis shows that this indicatorgives
ways think of as providing the most reliable information more information about a countryslevel of development
about national social and economic development and than certain traditional indicators such as per capita ex-
which are the most frequently used in evaluating a countrys ports of manufactured goods,average expectation of life,
development level,for example,the per capita GDP,the number of literates,the contribution ofthe manufacturing
number of illiterates,the percentage of the population industries to the GDP,the number of school enrolments,
not employed in agriculture or the share of the GDP etc.
accounted for by industrialproducts. The conclusion could be that to arrive at an exhaustive
However,the lists of development indicators used in list of development indicators,those which have previously
various national or internationalresearch institutions been neglected or overlooked should be added to those
are not always identical,and there is constant controversy appearing in all the lists used hitherto.
about the value and importance of one or other of these This would make a very long list,which we shall call
well-establishedindicators. In actual fact,it may be the maximum list ofsocial and economic development
wondered whether there are any objective reasons why indicators. The statistics established for these indicators
they should all be considered as social and economic de- would give a very complete picture of all aspects of de-
velopment indicators and whether others,so far over- velopment in the countries concerned.
looked or neglected,might not be more suitable than However,if we examine all the statistical publications
certain ofthe ones traditionally used. of the United Nations StatisticalOffice,regional economic
We have thus endeavoured to find new development commissions,and various Specialized Agencies of the
indicators,and in so doing we have defined and examined United Nations,as well as national publications and other
the indicatorof concentration of births in relation to the official or semi-officialdocuments,we see that the choice
age of the mother.I This indicator provides informa- of development indicatorsis greatly limited by the absence
tion about the family situation in a given country and is of statistical data about many developing countries.
obviously also an important development indicator.First
of all,a high degree of concentrationof births in relation
1. In a country j, the measurement of this indicator is given by
to the mothersage in a particular country signifiesthat
the number of children per family is limited,which en-
sures a higher standard ofliving in that country. Next,
if the area of concentrationis situated between the
mothers ages of twenty and thirty,this means that the where n represents the number of age groups, x.. the number
U
parents are of an age which makes it possible for the of births in which the mothers age is in the i-th group, and
children to have a good family education,prolonged X.the total number of births in the country j during the year
J
economic protection,favourable physical development in question.

21
The absence or poor quality of statistical data hinders the same quantity of information about the countrysde-
the introduction of a large number of development in- velopment level.
dicators such as those providing information about pro- Generally speaking,development indicatorsare not in-
ductivity in industry or agriculture,standard of living, dependent of each other, so that information provided by
infrastructure,industrial and educational cadres etc. one indicator is partially contained in the information al-
The introduction of even a single indicator into a ready provided by others. Taking the whole set of indi-
countrys statistical system frequently involves setting cators selected, therefore,it could be that certain quanti-
up a new department with suitably qualified staff and ties of information are repeated two or more times. It is
adequate equipment for the compilation,processing and obvious that these duplicated items would have to be sub-
publication of the data. These data should be of a high tracted from the total volume of information so as to
quality and comparable with data from other countries. leave only the basic information contained in this set of
The introduction of new items into our list of develop- indicators.
ment indicators consequently requires time,new institu- Third conclusion: The total amount of information
tions and great financial efforts. It goes without saying (withoutduplication)given by a set of indicators is
that this problem is even more serious in developing generally less than the sum of the quantities of informa-
countries. tion contained individually in each indicator of that set.
If we are to keep track of the development of coun- Let us take the example of two completely inter-
tries during the Second Decade with anything like suf- dependent indicators. If we arrange the countries in
ficient accuracy to reach our objectives,some new items question according to the values corresponding to either
ought unquestionably to be included in the list of develop- indicator,we shall obtain two identical lists. This means
ment indicators. The question is, however,which should that both indicators contain the same information about
be chosen from the many already suggested. the countriesdevelopment level and that consequently
The first problem to be solved in carrying out the pro- only one of the indicators need be considered,the other
ject for the evaluation and forecasting of development being superfluous.
during the Second Decade will thus be to establish,on a Fourth conclusion: Despite the fact that two given
scientific basis, an optimum list of a limited number of indicators may be very important from the point of
development indicators. view of the information they provide, separately,about
a country 3 development level, the contribution of one
2. CRITERION FOR SELECTING SOCIAL of them is insignificantif there is a high degree of cor-
AND ECONOMIC DEVELOPMENT INDICATORS relation between the two.
Fifth conclusion: Indicators producing information
Various statistical indicatorsmay be used to estimate and which is completely contained in the overall information
keep track of the social and economic development of a provided by the indicators already taken into account
country. Each of these indicators provides information should be discarded.
about one aspect of the level of development,and informa- It follows that a relatively small number of carefully
tion about this aspect is not,generally speaking,identical chosen indicators could contain a greater sum total of
with other information.The indicator percentage of the information than a large number of badly chosen
population not employed in agriculture,for example, indicators.
gives us information about a countryslevel of develop- Consequently,to obtain as complete an idea as possible
ment which is not identical with the information given by of the development level of the countries under observa-
the indicator per capita energy consumption. A choice tion,it is not enough to increase the number of indicators.
must thus be made and a whole set of indicatorshas to be To increase the total amount of information,the import-
prepared in order to give a more complete idea of the level ance of each indicator,expressed in the form of a quantity
of development of the countries concerned. of information,and duplications of information,expressed
First conclusion: Increasing the number of indicators in the form of interdependence between indicators,must
also increases the total amount of information about the be taken into account.
countrys level of development. The optimum selection of a limited number of develop-
The various development indicators do not contain the ment indicators should give a maximum total amount of
same volume of information about the development level information (without duplication), whilst the sum total of
of a country,i.e.they are not all of equal importance in duplications should be reduced to a minimum.
relation to development. The per capita GDP indicator The more information an indicator provides which is
thus provides more information about the social and not already contained in the sum of the information pro-
economic development level of a country than,for example, vided by more important indicators,the greater that in-
the average life expectancy indicator. If the quantity of dicators contribution to the evaluation of the countrys
information contained in development indicators is development level will be. Nevertheless,even an indicator
measurable,it should be possible to classify them in order which makes quite a considerable contribution will not be
of importance. very useful if its value is almost constant in all the coun-
Second conclusion: Sets of indicators of the same size tries observed. The indicator percentage of illiterates
(same number of indicators)do not, in general,contain over ten years of age in the population, for example,is
22
of no significance in the rich countries where it has be- Developmentindicators could thus be classified ob-
come stabilized at between l and 3 per cent. On the jectively according to the correlation between the various
other hand,in the poor countries this value varies between indicatorsactually used and the overall index containing
5 and 99 per cent,so that this same indicator provides a the maximum amount of information. We cannot cal-
high degree of discriminationin the evaluation of the culate the correlation coefficients directly,however,
development level ofdeveloping countries. The opposite since before we can do so we need the series of I-distances,
would be true if we took contribution ofindustrial pro- to calculate which we need the indicator classification
duction to the GDPas the development indicator. which we are in the process of establishing.
Sixth conclusion: A good development indicator should As we are thus in a vicious circle,we have to see
be discriminatory in all the countries under observation. whether it is at least possible to resolve the problem of
W e may therefore complete the criterion for selecting classifying development indicators by an approximate
a development indicator for a given set ofcountriesas method.
follows: It should be observed,first of all,that there are two
The greater UFIindicatorsdiscriminatory capacity in possible extreme cases of our problem. In the first,all
all the countries observed and the greater the quantity of indicatorsare independentand the problem of the same
information it provis7swhich is not contained in the sum quantitiesof informationbeing duplicated no longer
of the informatioil giwii by more important indicators al- occurs. In the second case,all the indicators selected
ready taken into condcration, the greater its contribu- are completely dependent on one of them,i.e. the quanti-
tion to an evaluation of the countries development level. ties of informationof all the indicatorsare contained in
the information of the dominant indicator. In both
3. OBJECTIVE EVALUATION OF THE DEGREE cases,it is possible to establish the series of I-distances
OF IMPORTANCE OF A DEVELOPMENT without any difficulty and subsequently to order the
INDICATOR indicators by their degree of importance.
In practice,however,one is usually in a position mid-
The degree of importance or significance of a development way between these two extremes. One may say,never-
indicator may be thought of as the quantity of informa- theless,that one is always either in a position in which
tion about a countrysdevelopment level it contains. one indicator contains most of the sum total of informa-
This quantity of information cannot,of course,be ex- tion about the countriesdevelopment level,or in a posi-
pressed in absolute terms. What we are able to determine tion in which it is impossible to say that one indicatoris
is more a relative quantity,dependent on which countries obviously dominant in relation to the rest.
are being observed and which indicatorsare selected. The first position is closer to the extreme case of the
Ifa given indicator is completely dependent on the total dependence of all indicators on one dominant in-
development level of a country,we shallhave the same dicator than to the extreme case of the total independence
classificationofcountries if we list them according to of all indicators. This is why,as a first approximation,
their value in terms of this indicator,or according to their we shall take the dominant indicator as containing the
development level. In this case,we may say that all the sum total of all the informationin all the indicators be-
informationabout the development level of countries is ing studied. In other words,we shall assume for the
contained in this single indicator,since one only has to time being that the development level correspondsto the
list these countries by their values according to this in- value of the dominant indicator for the country in
dicator to obtain at the same time the list by levels of question.
development. Ifwe calculate the correlation matrix and list the given
If dependence is partial,the information contained in indicatorsin accordance with the order of magnitude of
the given indicatoris only part of the total information their coefficients of correlationwith the dominant indica-
about the development level of the countries. tor,we shall obtain a provisional list of development in-
Consequently,we may say that the degree of import- dicatorsby degree of importance. This we shall call the
ance of a development indicator corresponds to the first-stageclassification.
degree of dependence between that indicator and the Using for the time being,the indicators arranged in
development level. this order,we are able to calculate the I-distancevalues
Ifwe take the I-distanceD1 as the relative measure for all the countries under consideration. As this
of a countryslevel of development and ifwe express I-distancebased on all the indicators under consideration
the degree of dependence between indicator Xiand the contains more informationabout the development level
development level by the simple correlationcoefficient of the countries than the dominant indicator,we shall,
ri,the value ri will express the degree of importance of in the second stage,determine the order of the indicators
according to the magnitude of their coefficients of cor-
Xi. relation with the Idistance of the first stage. This will
Listing the given indicatorsaccording to the order of be the second-stageclassification.
magnitude of their coefficients of normal correlation
with the I-distance,we obtain the classificationof de- 1.See Research Memorandum No. 41 of5 November 1970
velopment indicatorsin order of importance. (UNCTAD).

23
If we calculate the series of I-distancesaccording to 7 Concentration of births in relation 0.77
the second-stageclassification and if we calculate the to the age of the mother
coefficients of correlation between the indicators under 8 Exports of manufactured goods per 0.76
consideration and the I-distancethus obtained,we arrive head of population
at the third-stageclassification of indicators. 9 Average expectation of life 0.71
This iterative process is continued until such time as 10 Literacy rate 0.69
the classification of indicatorsis identical with that of the 11 Share of manufacturingindustries 0.66
previous stage. The classificationis then definitive since in the GDP
it remainsinvariable in all succeeding stages. 12 Rate of school enrolments 0.64
In the second case (without a dominant indicator) one
is closer to the extreme case of the complete independence
of all the indicators than to the extreme case of the com- 4.QUANTITATIVE ASSESSMENT OF THE
plete dependence of all the indicatorson the dominant CONTRIBUTION OF AN INDICATOR TO THE
one. We thus assume,as a first approximation,that all EVALUATION O F A COUNTRYS
the indicators are independent and that there is conse- DEVELOPMENT LEVEL
quently no duplication of information.
The I-distanceis reduced in this case to FrBchets 4.1 Statistical parameters
distance F,the calculation of which requires no order
of indicators since they enter into the F formula on an {
Let X = X1, ..., %} represent the set of n indicators
equal footing. being considered for possible use in evaluating the devel-
If we calculate the values of the F distance for all opment level of all the countries under observation,
countries under consideration and if we subsequently
determine the Coefficients of correlation between the in-
{
represented by P = P1,...,Pm 1 *

Assuming that statistics are available for each indicator


dicators observed and the F distance,this gives us the of X and for each country of P,if the value of the indica-
first-stage classification of development indicators by
their degree of importance.
3
tor for country P.is expressed by the formula x..
J 4
The same iterative method as in the first case is sub- these statistics may be presented in the form of the matrix
sequently used to arrive at the definitive list of indicators. M = [x.. 1,i = 1, ...,n,j = 1, ...,m.
4
It should be noted that the selection of social and Calculation of the statistical parameters of the indica-
economic development indicators is an instance of the tors X, requires the values x.. to be weighted. Designating
1J
first case,since the indicator per capita GDP is quite the matrix of the weighting coefficientsby the formula
clearly dominant. N =[ n.. 1,the arithmetic mean will be defined by the
If the development indicators observed {Xl,X2,..., 4
are arranged according to the definitive classification and
if the I-distanceD for each country is calculated accord-
formula: - -2
xi =
1 m
nij xij , i = 1 ,....,n
ing to this classification,the quantitative measurement of C nij j=1
j
the importance of the indicator will be expressed by
the formula: Ii = r(%; D). and the standard deviations by:
Their order of magnitude corresponds to the definitive
classification,namely:
r(X1;D)2 r(X2; D)>, ... >/ 4%; D).
Taking the developing countries and restricting the Using the covariance wicof the indicators and Xc,
development indicators to 12, the order of importance of it is possible to calculate normal correlation coefficients:
these indicators in 1968 was as follows:
Degree of Wic .
Order Indicator ric = - i,c=l, ...,n ,
importance ai ac
1 Per capita GDP 0.97 partial correlation coefficients:
2
3
Per capita energy consumption
GDP of the population employed
0.95
0.81
-
ric ris rcs
in agriculture
4 Number of doctors per 100,000 0.81
ric--= i
inhabitants fori<c, i, c = l , ...,nands#i,c,
5 Number of newspapers printed per 0.80 and multiple correlation coefficients:
1,000inhabitants
6 Percentage of the population not 0.79
employed in agriculture

24
where R representsthe determinant of the correlation k i- 1
matrix and RI the minor of R corresponding to the
element r 1. (4.3.1) C (1,2,...,k) = 1 CDi (P)
i =I
n
j =1
(1-rji.12...j-1)

4.2 The discriminatory capacity where CDi(P)representsthe coefficientof discrimination


of an indicator of indicator Xiin all the countries P and rjie12 ...j-l the
As a measurement of the discriminatory capacity of coefficientof partial correlation between indicatorsXi
the indicatorX,in the countries P as a whole,we suggest and X,.
the coefficient of discrimination which can be expressed This formula is increasingand homogeneousin relation
algebraically in the following form: to the discriminatory capacities of the indicatorsselected.
Each capacity is weighted so that the higher the indicator's
rank and the smaller the partial correlationsbetween it
and the indicators preceding it, the greater the importance
of its contribution in the sum total of information.
or If all indicatorsare mutually independent,the total
amount of information becomes the sum of the dis-
criminatory capacities of all the indicators selected,namely:
k
(4.3.2) C (1,2,...,k)= C CDi (P) ,
It should be noted that this coefficientdoes not have i=l
the same significance as the coefficient of variation which
measures the relative mean dispersion of Xiin P. and if a functional-lineardependence exists between them,
the total amount of information is reduced to the dis-
Thus in our example,which illustratesthree distribu- criminatory capacity of the first indicator,namely to:
tions in which the discriminatory capacity obviously
varies (with respective coefficients of discrimination of (4.3.3) C (1J ...,k= ) CD1 (P).
3.22,3.00and 2.67), the dispersion is practically con- If the set of indicators selected can be divided into two
stant (2.88,2.86 and 2.83). mutually independent groups,the total amount of informa-
tion will be the sum of the total information correspond-
A . A X1* ing to the two groups.
WWW It should be emphasized again that in accordance with
.. x2
the analytical formula for the total amount of informa-
tion (4.3.1), the discriminatory capacities of all the in-
~

.e
dicatorsrepresent additive quantitiesand that the quantity
of information from indicator Xialready contained in the
..
e. x3

information from the indicators preceding it in order of


importance may be subtracted by using the weighting
I 1 I I I I I I I I I 1 1 1 1 I X * factor
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

4.3 Total information of a group


ofdevelopment indicators
4.4 Optimum list of a limited number
of development indicators
To determine the total quantity of information con-
tained in the set of k indicators selected XI, x2,...xk, The fundamental question involved in the choice of a
we must follow the criterion concerning the choice of list of development indicatorsis: if we have a maximum
development indicators formulated in section 2,accord- list of n indicators,how can it be reduced to a list of k
ing to which we must take into account each indicator's <
indicators (k n) such that this set of k indicatorsshall
degree of importance,the duplicationsof information contain a total quantity of informationabout the develop-
contained in the various indicators,and each indicator's ment level of the countries under observation greater
discriminatory capacity. than any other set of k indicators selected from the n in-
Given that the indicators XI,X2,...xk are already dicatorsof the maximum list?
arranged in order of their importance,the overall con- Let us assume that the indicators{XI, X2,...,Xn}
tribution of this set of indicators to the evaluation of the are already arranged according to the definitive classifica-
development level of the countries under observation tion
is expressed by the formula:

25
and that (4.4.3)

is an ordered sub-setof k indicatorsin which

Lastly,if the k indicators are already selected,either


The total contribution of this set of indicators to the by establishing them in advance or by identifying them
evaluation of the development level of the countries is, by the formulae (4.4.2) or (4.4.3), it remains to identify
according to formula (4.3.1), as follows: the indicator xk+l which would produce the maximum
increment in the total amount of information C (1,2, ...,k).
This will be the indicator X m if

The optimum list of k indicators will thus consist of


the group 1 x,, xp,...x k 1
which maximizes formula (4.4.1). In other words,the
optimum list will consist of the indicators Xa,Xp , ...x k ,
if I forik+1 E 1 -
k+l, k+2, ...,n
1 .

It will not be easy to determine this optimum list, 5. CONCLUSION


even using a computer,if the maximum list contains a
large number of indicators. We have tried in this study to outline an objective pro-
In practice,however,high-levelauthorities such as the cedure for establishing an optimum list of development
United Nations General Assembly,UNCTAD's Trade and indicatorswhich would provide statistics of use in evaluat-
Development Board,ECOSOC,etc., which decide on and ing and forecasting development in the Second Develop-
adopt a definitive list of development indicators,could ment Decade.
insist on a certain number of indicatorsbeing included The first step is to establish the maximum list of
in advance and leave it to the experts to add others to development indicatorsby combining the various lists
make up an optimum list. This list would perhaps no used hitherto with other indicatorswhich have been
neglected or overlooked in the past.
longer be the best possible,but from the technical point
The procedure then consists of making a quantitative
of view it would be much easier to carry out calculations.
evaluation of the discriminatory capacity and importance
Let X1,X2, ...,Xp be the p indicators decided in ad- of each indicator in this maximum list.
vance and let X. ,...,X, be the (k-p)indica- The total amount of information of a group of indi-
'p+175p+2 k cators is defined as the sum of the coefficients of the dis-
tors to be chosen from the (n-p)indicators of the maxi- criminatory capacity of all the indicatorsin this group
m u m list. We can order the set formed by the union of whose weighting coefficients correspond to the import-
any combination of (k-p)indicatorswith the p indicators ance of the indicatorsand do away with duplications of
decided in advance,i.e. the same quantities of information. The optimum short
<]1, 2,...)P/U I ip+l, ip+2, ik 1> list of development indicators will then be formed by the
indicatorswhich provide the maximum quantity of total
information.
The order of these indices will be established according Considering the absence of statistics in certain cases
to the degree ofimportance of the corresponding indicators. or the poor quality of the statistics provided by the
The optimum list will now be made up of the p indica- large number of indicators in a maximum list, and the
tors decided in advance and the (k-p)indicators expense and time involved in introducing new indicators
1 xu, x,Uw,..-,
xk 1 for which
into the statistical system of developing countries,this
procedure makes it possible to draw up a rational plan
for determining a limited number of indicatorspro-
viding the maximum amount of total information.

26
Annex

ProgressReport I on the Methodology of Human Re- Type Production Function,by H.V.Muhsam,Hebrew


sourcesIndicators. (Unesco/COM/(SHC)CS/ 194/5). University,Jerusalem,Israel.
Study 11. An Analytical Approach to Educational Attain- Study XII. The Use of Taxonomic Measures in Target
ment Indicators Based on the Analysis of School En- Setting Based on InternationalComparisons,by
rolment by Grade. Zygmunt Gostkowski.
Study 111. Procedure of Evaluating High-LevelManpower Study XIII. Human Resources and Economic Growth:
Data and Typology of Countriesby means of the The Case of Mexico,by Tord Hoivik,International
Taxonomic Method,by Zdzislaw Hellwig. Peace Research Institute,Oslo.
Study IV.On the Definition and Theory of Development Study XIV. Human Resources and Socio-Economic
with a View to the Application of Rank Order Indi- Development: I. Theory,Methods and Data;
cators in the Elaboration of a Composite Index of 11. The Case of Japan,by Johan Galtung.
Human Resources (Paper prepared for the Unesco Study XV. Human Resources and Socio-Economic
Expert Meeting on the Methodology of Human Re- Development: The Case of Venezuela,by Kristin
sources Indicators,held in Warsaw 11-16December Tornes,Internaitonal Peace Research Institute,Oslo.
1967. The paper can also be identified as PRIO- Study XVI. Approximative Methods of Selection of
publication No.6-4from the InternationalPeace an Optimal Set of Predictors,by Zdzislaw Hellwig.
Research Institute,Oslo), by Johan Galtung &Tord Study XVII. Mathematical Methods in the Unesco
Hoivik. Project on Human Resources Indicators:
Study V. Discussion of methodological guidelines, (i) An evaluation
problem areas and difficulties connected with the (ii) Alternative Analytical Methods.
e!aboration of coherent systems of indices of human by Ludovic Lebart (SHC/WS/219).
resources,by ZygmuntGostkowski. Study XVIII.Human Resources and Economic Develop-
Study VI. On the Optimal Choice of Predictors,by ment. Some Problemsof Measurement by Prof.
Zdzislaw Hellwig. Wilfred Beckerman (University College,London).
Study VII. On the Problem of Weighting in Internation- Study XIX. Distance-basedanalysis,numerical taxonomy
al Comparisons,by Zdzislaw Hellwig. and classification of countriesaccording to selected
Study VIII. Indices of Development for Selected Latin areas ofsocio-economicdevelopment,by Serge
American and Middle African Countries. An Experi- Fanchette,Methods and Analysis Division,Unesco.
mental Exercise using the Hellwig Taxonomic Method, Study XX. Diachronic Analysis of Relationships between
by Frederick H.Harbison,Joan Maruhnic,Jane R. Human Resources Componentsand the Rate of
Resnick. Economic Growth in Selected Countries (COM/WS/13 1)
Study IX. Alternative Combinations of Human Re- by Johan Galtung.
sources Components as Predictors of Economic Study XXI. The Selectionof a Set of Core Indicators
Growth in Latin America, by The Higher School of of Socio-EconomicDevelopment,by Zdzislaw Hellwig
Economics,Wroclaw,Poland. and Serge Fanchette.
Study X. Concerning Data Availability and Conceptual Study XXII. The Quest for an Employment Strategy in
Validity,Selected Latin American and African Coun- Developing Countries and its Relationship to the Work
tries, by Frederick H. Harbison,Joan Maruhnic and on Human Resources Indicators,by H.W.Singer
Jane R. Resnick. (Institute of Development Studies,The University of
Study XI. The Assessment of the Validity of Human Sussex,UK).
Resources Indicatorsby means of a Cobb-Douglas
27
Study XXIII. Synchronic and Diachronic Approaches wealth in Kenya,by H.W. Singer and Stuart D.
in the Unesco Project on Human Resources Indicators. Reynolds (The Instituteof Development Studies,
Wroclaw Taxonomy and Bivariate Diachronic Analysis University of Sussex,UK).
-
(SHC/WS/209), by Serge Fanchette Methods and Study XXVII.Distributionalpatterns of development
Analysis Division,Unesco. and welfare: A case study of Zambia,by Charles
Study XXIV. A method of establishing a list of develop- Elliott (Overseas Development Group,School of
ment indicators,by B. Ivanovic (Director of the Development Studies,University of East Anglia).
Office of Statisticsof UNCTAD). Study XXVIII.Typologicalstudy using the Wroclaw
Study XXV. Grouping and ranking of 30 countriesof Taxonomic Method (A study of regional disparities
Sub-SaharanAfrica. Two distance-basedmethods in Venezuela), by JosC F.Silvio-Pomenta(Centro de
compared,by B. Ivanovicand Serge Fanchette. Estudios del Desarrollo(CENDES), Universidad
Study XXVI. Aspects of the distribution of income and Central de Venezuela.

28
UNESCO PUBLICATIONS: NATIONAL DISTRIBUTORS

Argentina Editorial Losada, S.A., Alsina I 131. BUENOSAI=.


Australia Publications :Educational Supplies Pty. Ltd., Box 33, Post Office BROOKVALE 2100, N.S.W. Periodicals :
Dominie Pty. Ltd., Box 33, Post Office, BROOKVALE 2100,N.S.W.Sub-agent: United Nations Association
of Australia (VictorianDivision), 5th Floor, 134-136Flinders St.. MELBOURNE 3000.
Austria Verlag Georg Fromme & Co., Arbeitergasse 1-7,1051 WIEN.
Belgium Jean D e Lannoy, 112, rue du Trsne, BRUXFLLES5. CCP 708 23.
Bolivia Libreria Universitaria.Universidad San Francisco Xavier. apartado 212.SUCRE.
Brazil Fundacgo G e t W o Vargas, Serviso de Publicasiies, caka postal ZIIZO Praia de Botafogo 183,RIO DE JANEIRO
G.B.
Bulgaria Hemus, Kantora Literatura, bd. Rousky 6,SOFIIA.
Burma Trade Corporation n.O (9),550-552 Merchant Street, RANGOON.
Cameroon L e Secrktaire gknkral de la Commission nationale de la Rkpublique fkdkrale du Cameroun pour IUnesco,
B.P. 1061,YAOUND6.
Canada Information Canada. OTTAWA (Ont.); Bookshops: 640 Ouest, rue Ste Catherine, MONTR~AL I I I (Que.);
1683 Burrington St.HALIFAX(N.S.); 393 Portage Ave., WINNIPEG (Manitoba), 171 rueslater,OTTAWA (Ont.);
221 Yonye St, TORONTO (Ont.); 800 Granville St, VANCOUVER (B.C.).
Chile Editorial Universitaria,S.A., casilla IOZZO,SANTIAGO.
Colombia Libreria Buccholz Galeria, avenida Jimknez de Quesada 8-40,apartado a6reo 49-56,BOGOTA; Distrilibros
Ltda., Pi0 Alfonso Garcia, carrera 4, non 36-119y 36-125,CARTAGENA; J. Germin Rodriguez N.,Calle
17,6-59,apartado nacional 83, GIRARDOT (Cundinarnarca). Editorial Losada Ltda., Calle 18A. n.O 7-37.
apartado akreo 5829,apartado nacional 931.BOGOTA. Sub-depots :Edificio L a Ceiba, Oficina 804 MEDELLfN;
Calle 37. n.O 14-73Oficina, 305, BUCARAMANGA; Edificio Zaccour, Oficina 736. CALI.
Congo Librairie populaire. B.P. 577,BRAZZAVILLE.
Costa Ricn Libreria Trejos S.A.,apartado 1313.SAN Jose.Telkfonos: 2285 y 3200.
Cuba Distribuidora Nacional de Publicaciones,Neptuno 674,LA HABANA.
CYPNS M A M ,Archbishop Makarios 3rd Avenue, P.O.Box 1722, NICOSIA.
Czechoslovakia SNTL,Spalena 51.PRAHAI (Permanent display) ; Zahranicni literatura, I I Soukenicka,PRAHA I.
For Slovakia only :Alfa Verlag. Publishers.Hurhanovo nam. 6,893 31 BRATISLAVA.
Dahomey Librairie nationale,B.P. 294,PORTONovo.
Denmark Ejnar Munksgaard Ltd.. 6 Nsrregade. 1165KUBENHAVN K
Egypt Librairie Kasr El Nil 38, rue K a s r El Nil.LE CAIRE.National Centre for Unesco Publications I Tlaat
Harb Street,Tahrir Square. CAIRO.
Ethiopia National Commission for Unesco, P.O. Box 2996,ADDIS ABARA.
Finland Akateeminen Kirjakauppa, z Keskuskatu,H ~ I N K I .
France Librairie de IUnesco,7 place de Fontenoy, 75700 PARIS. CCP 12598-48.
French West Indies Librairie Au Boul Mich. L Rue Perrinon and 66 Avenue du Parquet, 972 FORT-DE-FRANCE (Martinique).
German Dem. Rep. Deutscher Bucb-Export und -Import G m b H , Leninstrasse 16, 701 LEIPZIG.
Fed. Rep. of Germany Verlag Dokumentation, Postfach 148, Jaiserstrasse 13. 8023 MLNCHEN-PULLACH. The Courier (German
edition only): Bahrenfelder Chaussee 160 HAMBURG-BAHRENFELD. CCP 27 66 50. For scientific maps only:
GEO C E N T R E , D - 7 STUTTGART 80, Postfach 800830.
Ghana Presbyterian Bookshop Depot Ltd., P.O. Box 195, ACCRA; Ghana Book Suppliers Ltd., P.O. Box 7869,
ACCRA;T h e University Bookshop of Ghana, ACCRA; T h e University Bookshop of Cape Coast; T h e Uni-
versity Bookshop of Legon, P.O. Box I, LEGON.
Greece Anglo-Hellenic Agency, 5 Koumpari Street, ATHINAI138.
Hong K o n g Swindon Book Co., 13-15 Lock Road, KOWLOON.
Hungary Akadhiai Konyvesbolt Vdci U 22, BUDAPE~T V. A.K.V. Konyvtarosok Boltja. NkpkGztPrsasig utja 16,
BUDAPFST VI.
Iceland Snaebjorn Jonsson &Co. H.F.. Hafnarstracti 9. REYKJAVIK.
India Orient Longman Ltd.; Nicol Road, Ballard Estate, BOMBAYI; 17 Chittaranjan Avenue, CALCUTTA 13;
36A Anna Salai, Mount Road, MADRAS 2 ; B-3/7Asaf Ali Road, NEW DELHI I. Sub-depots: Oxford Book &
Stationery Co., 17Park Street,CALCUTTA 16;and Scindia House, NEWDELHI; Publications Section, Ministry
of Education and Social Welfare, 72 Theatre Communication Building, Connaught Place, NEW DELHI I.
Indonesia Indira P.T.,JI. Dr.Sam Ratulangie 37, JAKARTA.
Iran Commission nationale iranienne pour IUnesco, avenue Iranchahr Cbomali no 300. B.P. 1533. T ~ H ~ R A N .
Kharazmie Publishing and Distribution Co., 229 Daneshgahe Street, Shah Avenue, P.O. Box 14l1486,
TBHBRAN.
Iraq McKenzies Bookshop, AI-Rashid Street, BAGHDAD; University Bookstore, University of Baghdad, P.O.
BOX75.BAGHDAD.
Ireland T h e National Press, 2 Wellington Road, Ballsbridge. DUBLIN 4.
Israel Emanuel Brown, formerly Blumsteins Bookstores: 35 Allenby Road and 48 Nachlat Benjamin Street, TEL
Avrv; 9 Shlomzion Hamalka Street. JERUSALEM.
Italy L I C O S A (Libreria Commissionaria Sansoni S.p.A.). via Lamarmora 45. casella postale 552, 50121 FIBENZE.
Jamaica Sangsters Book Stores Ltd., P.O. Box 366, I O I Water Lane, KINGSTON.
Japan Maruzen Co. Ltd.. P.O.Box 5050. Tokyo International,TOKYO.
Kenya T h e E S A Ltd.. P.O.BOX 30167,NAIROBI.
Khmer Republic Librarie Albert Portail, 14.avenue Boulloche. PHNOM-PENH.
Republic of Korea Korean National Commission for Unesco. P.O. Box Central 64,SEOUL.
Kuwait T h e Kuwait Bookshop Co.Ltd., P.O. Box 2942, KUWAIT.
Liberia Cole &Yancy Bookshops Ltd., P.O. Box 286, MONROVIA.
Libya Agency for Development of Publication and Distributlon,P.O. Uox 34-35,TRIPOLI.
Luxembourg Librairie Paul Bruck, 22 Grande-Rue, LUXEMBOURG.
Madagascar Cornmission Nationale de la RBpublique Malpache, MinistAre de 1Education Nationale, TANANARIVB.
Malaysia Federal Publications Sdn. Bhd.. Balai Berita, 31 Jalan Riong. KUALA LUMPUFL
Malta Sapienzas Library, 26 Kingsway, VALLETTA.
Mauritius Nalanda Co. Ltd, 30 Bourbon Street, PORT-LOUIS.
Mexico C I L A (Centra Interamericano de Libros Acadhicos), Sullivan 31 bis, MfXlCO 4. D.F.
Monaco British Library. 30. boulevard des Moulins, MONTE-CARLO.
Netherlands N.V. Martinus Nijhoff, Lange Voorhout 9. s-GRAVENHAGE; Systemen Keesing, Ruysdaelstraat 71-75>
AMSTERDAM.
Netherlands Antilles G.C.T.Van Dorp &Co. (Ned. Ant.) N.V., WILLEMSTAD (Curacao,N.A.).
New Caledonia Reprex S.A.R.L., B.P. 1572, N O U M ~ A .
New Zealand Government Printing Office, Government Bookshops: Rutland Street, P.O. Box 5344, AUCKLAND; 130
Oxford Terrace. P.O. Box 1721,CHRISTCHURCH; Alma Street, P.O. Box 857, HAMILTON; Princes Street,
P.O. Box I 104,DUNEDIN; Mulgrave Street, Private Bag, WELLINGTON.
Niger Librairie Manclert, B.P. 868, NIAMEY.
Nigeria T h e University Bookshop of Ife; T h e University Bookshop of Ibadan,P.O. Box 286,IBADAN; T h e University
of Nsukka; T h e University Bookshop of Lagos; T h e Ahmadu Bello University Bookshop of Zaria.
Norway All publications: Johan Grundt Tanum, Karl Johans gate 41!43. OSLO I. TheCmUieronly :A/SNarveqns
Litteraturtjeneste. Box 6125,OSLO 6.
Pakistan T h e West-Pak Publishing Co. Ltd., Unesco Publications House, P.O. Box 374, G.P.O., LAHORE.
Showrooms :Urdu Bazaar, LAHORE and 57-58 Murree Highway, G/6-1,ISLAMABAD. Pakistan Publications
Bookshop: Sarwar Road, RAWALPINDI; Mina Book Agency, 65 Shahrah Quaid-e-azam, P.O. Box 729
pi3.
Peru The Courier only :Editorial Losada Peruana, apartado 472, LIMA.Otk publications: Distribuidora Inca
S.A., Emilio Althaus 470. Lince, casilla 3115.LIMA.
Philippines T h e M o d e m Book Co., 926 Rizal Avenue, P.O. Box 632. MANILA.
Poland Osrodek Rozpowzechniania Wydawnictw Naukowych PAN,Palac Kultury i Nauki, WARSZAWA.
Portugal Dias &Andrade Ltda., Libraria Portugal, rua o Carmo 70, LISBOA.
Southern Rhodesia Textbook Sales PVT) Ltd., 67 Union Avenue, SALISBURY.
Romania I.C.E. LIBRI. Calea Victoriei nr. 126, P.O. Box 134-135 BUCURESTI.SubscriptMns to p ~ r i ~ d ti ~ ~ l ~
Rompresfilatelia. Calea Victoriei nr. 29, BUCURESTI.
Senegal La Maison du Livre, 13, avenue Roume, B.P. 20-60,DAKAR; Librairie Clairafrique, B.P. zoos, DAKAR;
Librairie LeSenegal. B.P. 1594,DAKAR.
Singapore Federal Publications Sdn Bhd., Times House, River Valley Road, SINGAPORB 9
South Africa Van Schaiks Bookstore (Pty.) Ltd., Lihri Building, Church Street,P.O.Box 7.24. PRBTORIA.
Spain AI1 publications :Editiones Iberoamericanas, S.A., calle de OEate 15, MADRID 20; Distribuci6n de Publi-
caciones del Consejo Superior de Investigaciones Cientificas, Vitrubio 16,MADRID 6 ; Libreria del Consejo
Superior de Investigaciones Cientificas, Epigciacas 15. BARCELONA. For TheCourieronly :Ediciones Liber,
apartado 17.ONDdRROA (Viscaya).
Sri Lanka Lake House Bookshop, Sir Chittampalam Gardiner Mawata, P.O.Box 244, C O L o m o 2.
Sudan AI Bashir Bookshop,P.O. Box I I 18,KHARTOUM.
Sweden All publications: A/BC.E. Fritzes Kungl. Hovbokhandel, Fredsgatan 2, Box 16356,10327 STOCKHOLM16.
For TheCourier: Svenska FN-Forbundet, SkolgrPnd 2, Box 150 so. s-Io4 65 STOCKHOLM.
Switzerland Europa Verlag. Ramistrasse 5, ZURICH; Librairie Payot, 6,rue Grenus, 1211 GENBVE XI.
Tanzania D a r es Salaam Bookshop,P.O. Box 9030, DARES SALAAM.
Thailand Suksapan Panit, Mansion 9. Rajdamnem Avenue, BANGKOK.
Togo Librairie evangPlique, B.P. 378; LoMB; Librairie du Bon Pasteur, B.P. 1164, LoMB; Librairie modeme,
B.P.777, LoM~.
Turkev Librairie Hachette. 469 Istiklal Caddesi, Beyoglu, ISTANBUL.
Uganda Bookshop, P.O.Box 145.KAMPALA.
Mezhdunarodnaja Kniga, MOSKVA, G-zoo.
H.M.Stationery Office, P.O. Box 569, LONDONSEIgNH; Government bookshops: London, Belfast, Bir-
mingham, Bristol, Cardiff, Edinburgh, Manchester.
United States Unesco Puhlicstions Center, P.O. Box 433. N e w York, N.Y. 10016.
U p p e r Volta Librairie Attie B.P.64. Ouagadougou, Librairie catholique JeunessedAfrique, OUAGADOUGOU.
Venezuela Libreria Historia, Monjas a Padre Sierra, Edificio Oeste 2. n.O 6 (frente a1 Capitolio), apartadode correo8
7320-101,CARACAS.
Yugoslavia Jugoslovenska Knjiga. Terazije 27. BEOGW. Drzavna Zalozba Slovenije Mestni Trg. 7.6, LJUBLJANA.
Zaire L a Librairie,Institut national detudes politiquen, B.P. 2307, KINSHASA:
Commiaaion nationale de la Rdpu-
blique du ZaIre Dour IUnenco, Ministere de IMucntion nationale. KINSFIMA.

ISBN 92-3-1
01 1 29-4