You are on page 1of 22

Department of Economics, Delhi School of Economics, University of Delhi

Economic Approaches to Education and Manpower Planning


Author(s): Amartya K. Sen
Source: Indian Economic Review, New Series, Vol. 1, No. 1 (APRIL 1966), pp. 1-21
Published by: Department of Economics, Delhi School of Economics, University of Delhi
Stable URL: http://www.jstor.org/stable/41969624 .
Accessed: 31/07/2014 05:27
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Department of Economics, Delhi School of Economics, University of Delhi is collaborating with JSTOR to
digitize, preserve and extend access to Indian Economic Review.

http://www.jstor.org

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

Economic
to

Approaches

Education

Manpower

and
Planning*

relation between theory and statistics is an intricate one.


THE Theories, to be sound empirically,must be based on data, but
collected and compiled
statistics,in their turn, cannot be efficiently
a new field like that
In
notion
of
some
a
without
backgroundtheory.
of economicsof educationand that of manpower planning, such backgroundtheoriescan prove to be the biggest stumblingblock. A number
of approaches have cropped up in the field,and statisticalwork seem
to be proceeding in a variety of directions. It is not altogetherclear,
however,whetherall theseapproachesare particularlypromising. Since
the qualityand the amountof the data leave a lot to be desired, rejection
all the more necessary
or verificationis not very easy. It is, therefore,
that the underlyinglogic of the background theories should be very
carefullyexamined, to make sure what exact economic assumptions
theyimply,and to what extentthese are plausible relationsto expect.
In this paper threeof the more dominantapproaches to the problem
willbe examined:
(i) The Fixed RequirementsApproach, e.g., that of Professors
Tinbergen,Correa, and Bos ;x
(ii) The Income Shares Approach of the Marginalist School, e.g.,
thatof Mr. Denison ;2
in June1964at the Asian Institute
Thisformed
delivered
thebasisofa lecture
ofEconomicDevelopment
oftheUnitedNations.
andPlanning
to AccelereofEducation
1. J.Tinbergen
andH. Correa,"Quantitative
Adaptation
and H. C, Bos,"A Planning
ted Growth,"Kyklos,yL'(1962),4; J.Tinbergen
in O.E C.D.,
Model of EducationRequirements
ofEconomicDevelopment,"
TheResidual
FactorandEconomic
Growth
(Paris,1964).
Statesandthe
in the United
2. E. F. Denison.The Sourcesof Economic
Growth
theContribution
Alternatives
Us(New York,1962);Denison,"Measuring
Before
of Education(and the 'residual')to EconomicGrowth",O.E.C.D., The
Residual
FactorandEconomic
Growth
(Paris,1964). Theapproachwas pioneered
of
seetheSupplement
Schultz
anda number
of Chicagoeconomits;
byProfessor
the Journal
of PoliticalEconomy
, Vol. 70, No. 5, Part2 (October1962);
seealsoG. S. Becker,
HumanCapital(NewYork,1964).

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

. fc. SSf

(iii) The Human Resouree Index Approach of Professors Harbison


and Myers.8
In the light of our assessmentof these approaches some suggestions
would be made on what otherdata to collect,and what furtherconsiderations to bring into our analysis of the economics of education and
manpowerplanning.
The Fixed Requirements
Approach
The firstapproach consistsessentiallyin postulating fixed relations
between stocks of educated manpowerand unitsof nationalincome,or
itssectoral components. It assumes away implicitlythe possibility of
substitutionbetween educational resources and productive factors of
otherkinds. Theorizationof thiskind is implicitin the often repeated
casual statementabout educationalplanning,viz., educational expansion
to economicexpansion. ProfessorsTinbergen,
mustgo up proportionately
Correa and Bos have formulatedthe implications of this approach
rigorouslyin two models,whichare veryworthstudying.
Both in the Tinbergen-Correapaper (1962), as well as in the
Tinbergen-Bospaper (1963), there are two scarce resources,viz., "the
total stockof people with a secondary education," and "those with a
third-leveleducation," indicated respectivelyby N2,, and N3,, fortime
period t. In the 1962 paper, the need for N2 is takento be simply
proportionalto the nationalincome. The same forthat part of N3 which
is used forproducingnationalincome; but there are two otherparts of
N3 employedin teachingpeople at the secondaryand the third level, and
are also takento be proportional (given teacher-pupil
theserequirements
ratios). The stock of secondary educated manpower and third-level
educated manpower this period depends on the stocklast period,plus
new entrantsthisperiod afterfinishing
education last period, minus the
part of the last period'sstock eliminatedby death and retirement.This
"drop out" is takento be a fixed proportion of the last period's stock.
These are the main featuresof the 1962 paper, and the rest consistsin the
working out of the implicationsof all this and of a fewotherminor
of alternativerates of growth
assumptions. The educationrequirements
are examinedwithinthis framework.
In the 1963 paper, Tinbergen and Bos develop thismodel further,
introducingsome changes,withoutaltering the basic approach. Instead
of assumingthat all the studentseventuallyjoin the labour force, a fixed
3. F. Harbison
andC.A. Myers,
andEconomicGrowth
Education,
Manpower
(New
York,1964).

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

EDUCATIONAND MANPOWER
PLANNIN

proportionof "wastage" (e.g. due to failureto pass or to continue study,


due to seekingeducationforpurely cultural reasons withoutwanting to
use it) is introduced. The possibilityof an initialsurplusof some kindof
educated stockis also considered. Manpower requirementsare related
to sectoralcomponentsof nationalincome,and not to aggregatednational
incomealtogether. And insteadof takinga proportionalrelationbetween
educated manpower and the product, a not-necessarily-proportional,
butunique,relationbetweenthe two is assumed. This last relaxation of
the assumptionof the firstmodel has to be examined carefully, We may
now, lor example,postulatethat when output goes up 10 per cent, the
stockof secondaryeducated manpowerrequirement
goes up by 5 per cent.
But whilethis relationis not proportional,it is unique, in the sense that
a 10 per centincreasein outputrequiresa fixed amount of expansionof
the educated stock, no matter what else we do in the economy. We
cannotcut down the requirementof educated stock by choosing other
techniquesof productionrequiring,say, less of this kind of education.
In fact,what characterisesthesemodelsis thisnotionof fixedrequirement
of educationalstockforproducinga givensize of nationalincome and its
sectoralcomponents.
Any model of this kind requires some simplification,and such
theoriesshouldnot be taken to task for assumingthe realityto be a little
simplerthan it in factis, in orderto achieve practicalusability. What is
worrying,therefore,is not the existence of simplifying
assumptionsas
us
on
the
but
those
ones
that
to
seem
such,
put
specific
wrong track
altogether. I have had occasion to discussthese models elsewhere,4and
here I shall only verybrieflystatewhat appears to me to be the three
chiefdrawbacksof thisapproach, which, somewhat unfairlyto Correa
and Bos, I shall call the Tinbergenapproach.
The major difficulty
withthe approach lies in its basic assumption
of the completefixityof the educational requirementsof producing a
givensize and composition of national income. There are possibilities
of substitution
which are completelyoverlooked. We cannot say with
any degreeof confidencethatan extra $ 1 millionof national income in
a certainsectorrequiresan extrabundleof, say, 100 secondary educated
men ; the requirements
of productionare not fixed in this way. There
are possibilities of substitutionbetween differentlevels of skill and
education, and also between human resources and physical ones in
choosingbetweenalternativemethodsof increasingnational income from
: Some
ofEconomic
ModelofEducation
4. "A Planning
Development
Requirements
FactorandEconomic
Growth.
inO.E.C.D., TheResidual
Comments,"

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

A. K. SEN

is not supported
any givensector. The fixityof educationalreqiiirement
or
economic
argument, by any thing more
by any very convincing
causal
observations.
two
one
or
than
fairly
A second featureof the Tinbergenapproach is its complete concentration on formal education without any weight being placed on the
processof learningwhile at work. This is probably derived from an
analogy withphysicalcapital, but unlike machines men do learn from
experience. The rental of machinerytends to decline uniformlywith
age, but preciselythe oppositeis trueof human beings for quite a while
aftertheyjoin work5. Justby countingthenumberof secondary-educated
men we may not get much idea of their usefulness to the industries.
This questionof learningis an importantone.
A thirdweaknessof the Tinbergenapproach is its assumption that
drop-out due to death and retirementis a given proportion of the
last period's stock of educated manpower. This requires the assumption of the forceof mortalityand retirementbeing independentof age,
and is taken as a straightforward
analogy with the assumptionof the
so-called radio-active depreciation used in the context of physical
capital.6 I doubt that this is a good assumption for physical capital,
but forhumanbeingsthis is hardlytolerable. The chance of retiringor
dyingnextyear is not the same at the ages of 20, 40, 60, 100, 200.
is easilyremediable. It is the firstthat mainly
This last difficulty
limitsthe Fixed Requirements Approach to educational planning, and
adds to it. Altogetherit is not at all obvious why
the second difficulty
thisapproach should be taken as even a first step in the field, though
there is no doubt that models of this kind are sometimesimplicity
assumed in casual statementsof political leaders. We owe, I think, a
debt of gratitudeto ProfessorTinbergenand others for formulatingthe
modelsrigorously,whichhas allowed us to perceive its limitationswhich
are not so obvious in the casually made statements. Their works have
cleared the groundforfurtherstudies.
II.

The Income Share Approach

This approach is based on applying the marginal productivity


theoryof distributionto the economicsof education. If it is true that
and
"Educationand Income/'Reviewof Economics
5. See H. S. Houthakker,
HumanCapital(NewYork,1964).
1959; G. S. Becker,
Statistics
, February
"ParableandRealismmCapitalTheory: The Surrogate
6. See PaulA. Samuelson,
Studies
Reviewof Economic
"7Yre
Production
, June1962,p. 197; N.
Function,
ibid, p. 177.
"A NewModelofEconomicGrowth,"
KaldorandJ.A. Mirrlees,

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

EDUCATION
AND MANPOWER
PLANNING

marginalproductivitydeterminesthe prices of factors of production,


thenwe do not need to measurethe marginal productivitydirectly,and
all we have to do is to look at the size of rental of the relevant factors
of production. If a worker earns $100 more witheducationof typex,
when otherthingsare given, then it can be taken, according to this
approach, that his marginal productivityis $100 higheras a resultof
this education. This provides an extermlytemptingway of by-passing
the complicatedproblemsof estimatingeducational productivity,and it
is not surprisingthat the method has been grabbedwithconsiderable
eagerness. The approach has been used by T. W. Schultz,G. S. Becker,
T. Mincerand others in a varietyof studiesin the contextof specific
and somewhatmore heroically for calculating the sources
estimations,7
of economic growth in the United States by ProfessorKendrick8and
Mr. Denison.9 For a glimpseat how all thisis done, we may look at
Mr. Denison's method.
Mr. Denison startswith Professor Houthakker's figureson mean
incomesearned in the United States before tax classified into groups
accordingto age and years of schooling,obtained from the 1950 Census
of Population. From this a set of typical differentialsaccordingto
schoolingare obtained formales of the same age. But since differences
in schooling are correlated partly with differencesin ability, energy,
motivation,and otherfactors that will also influenceproductivityand
of the reported income
earnings,Mr. Denison assumes that "three-fifths
differentialrepresentdifferences
in incomesfromworkdue to differences
in educationas distinguishedfrom associated characteristics".10As the
nextstep thesedifferentials
"due to" educational change are applied to
distribution
of males by years of school completedat various past dates,
and therebythe rise in averageincome due to increase in the number of
school years completedis estimated. Since therehas also been a rise in
thenumberof school days per year,some further
adjustmentsare needed.
It is assumed that a given percentage increase in the number of days
spentin school per year has the same effectas an equal precentage
increasein the numberof years spentin school.
7. See particularly,
theSupplement
on "Investment
in HumanBeings",of The
Journal
ofPoliticalEconomy
, October1962,LXX,No. 5, Part2.
8. J.W. Kendrick,
Trendsin the UnitedStates
Productivity
, NationalBureauof
Economic
Research(Princeton,
1961).
9. E.F. Denison,TheSources
Growth
intheUnited
Statesandthe AlterofEconomic
natives
forEconomic
before
Us,Committee
Development
(NewYork,1962).
10. Denison,
p. 69,

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

A. K. SEN

All these calculations done, Denison findsthat from1929 to 1957


education"raised the averagequalityof labour by 29.6 per cent, or at an
averageannual rate of 0.93 per cent."11 The nextquestionis how much
effectdid thisrisein the qualityof thelabour forcehave on the national
use of marginalproductivity
output. Here Denison make some further
of
labour
in the national income
share
Since
the
"average
theory.
over thisperiod" was 73 per cent,it is concluded that the elesticityof
outputwithrespectto labour is 0.73.12 Thus, Denison concludes, that
the averageannual rate of riseof 0.93 per centof the quality of labour
force,would have increased the national product by 0.68 percentage
pointsper year (being 73 per cent of 0.93). And this amounts to saying
that of the total growth of 2.33 per cent per year in the UnitedStates,
in education (since 0.68
as much as 23 per cent was due to improvement
is 23 per cent of 2.93). The figureis even more impressiveifwe look at
product per capita, and it turns out fromthis calculationthatno less
than 42 per cent of the growthrate in the product per person employed
education.
was due to the contributionof further
This in nut-shellis Denison's approach in measuring the contributionof educationto past growth. He goes on to use thesame approach
to measuringthe likelycontributionof educationto futuregrowth. One
will
gets an estimate of how any scheme of educationalimprovement
and
and dependingon our values
affectthe national productivity,
goals
we can accordinglyselectan appropriateeducationalpolicy.
One can make severalcriticismsof the Denison approach. Some of
these concern his specific assumptions,while othersare commonto the
approach of gettingfiguresof productivityindirectlyfrom the earning
figures,involving crucial use of the marginal productivitytheory
11. Denison,p. 73
butcanbespelledouthere.LetY=output,w=wage
is wellknown,
12. Theargument
shareof wagesin nationaloutput.Now,we
m=the
and
rate,L=employment,
that: w= -0Y
themarginal
knowfrom
assumption
productivity
0L
w.L 8Y/ Y
of output
of elasticity
whichis the definition
m=
Therefore,
to labour. Henceif73 percentis theshareof wagesin national
withrespect
a 1 percentrisein the labourforcewill increasethe nationaloutput
income,
thatthisresultwill be true
be remembered
by0.73percent. Itmust,however,
ofthisforbigchangescan
and
small changes, application
onlyformarginally
of substitution
of
unit
elasticity
be madeonlywith the further
assumption
function.
ofproduction,
production
i.e.,ofa Cobb-Douglas
between
thefactors

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

EDUCATIONAND MANPOWER
PLANNING

postulates. Here we may mentionmainly the more general objections,


and formore specificdifficulties
about the Denison approach the reader
could be referredto ProfessorMoses Abramovitz'scriticalevaluation.18
The fulfilment
of the marginalproductivity
theoryrequiresa variety
of assumptions,manyof whichhave been widelydiscussedfromthe point
view of theirrealism. Some of us continueto remain very skeptical of
the assumptions of the theory,14
withoutnecessarilybeingenamouredof
themorewidelydiscussedalternativemodels, e.g., that of Mr. Kaldor.15
This whole area of economicsseems to be in such a bad shape that any
restrictive
big policy decisionbased on theextremely
assumptionsof these
theoriescan be viewedwithconsiderablesuspicion. In theseapplications
of the marginal productivitytheory, the temptationis providedby the
factthat it short-circuits
the need .for an independent measurementof
and insteadone can happilyuse the
productivity,which is verydifficult,
income data, which is very easy to obtain. But precisely because
is so difficult
to measure, we cannot easily verify
marginalproductivity
the theoryby actual statisticalobservations,so that the entire approach
is based on a giganticassumptionrather than on empirical verification.
Indeed in so far as the underlyingassumptionsof the theory have been
put to test, e.g., the assumptions of perfect competition, of perfect
foresight,of profitlessequilibrium the results have been far from
encouraging,and the continued popularity of the theory in these
estimatescan perhaps largely be explained by the simplification
of these calculations that one heroic assumption like this allows
one to make. I am not altogetheropposed to yielding to temptations,
but the ease with which economists like Denison make the validityof
marginal productivitytheory almost a part of a latter-day Gospel,
distractsfromthe highquality of theirworksand thepainstakingtrouble
whichtheytake over the otherassumptionsthey make.
There is, in fact,more difficultyin Denison's approach than the
simple assumption of marginalproductivity
theory. (In thisrespectwe
have been a littleunfairin lumpingall the uses of this theory together.)
Denison also assumesthe existenceof economies of scale, and as much as
17 per cent of the growth of per capita income in the United States
13. "EconomicGrowthin the UnitedStates", The American
Economic
Review
September
1962,Vol.LII, No.4.
14. Someofmyownreservations
are statedin mynote,"Neo-Classical
and NeoTheoriesofDistribution,"
Economic
March1963.
Keynesian
Record,
15. N. Kaldor,"Alternative
Theories
ofDistribution",
Reviewof Economic
Studies,
1955-56.

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

A. K. SEN

by himto such economies. This raises some


during 1929-57 is attributed
First
of
were
these economies internal or external? If
all,
questions.
not have been paid according to their
the
factors
could
internal,
respectivemarginalproducts,becausethatwould have exhaustedmorethan
the total output. On the otherhand, if these economies were external
thenfactorrewardcan equal privatemarginalproduct,but will fall short
of the appropriatesocial marginalproduct. Thus one would be misledby
takingthe additionalearningsdue to education as indicatorsof additional
contributionof education to national (as opposed to private)productivity.
No matterwhich assumptionis made, once increasing returnsto scale
is introduced into the picture,the rationaleof the incomeapproach to
productivity
gets a severe jolt. Thus, the approach here is not only
to
the
open
charge of being unrealisticin assumingthe validityof the
marginalproductivitytheory, it is also open to the charge of being
internallyinconsistent.
Two otherquestions can be raised here. All these works draw
heavilyon an articleby ProfessorRobert Solow,16 who seemed to have
meantthe paper ratherless seriouslythan the restof the world has taken
it. And as Solow has pointedout in his later works,17 lookingmerely at
income shares tends to under-estimate the contributionof capital
accumulation to economicgrowth,sincenew techniquesand thefruitsof
progressof knowledgecan be absorbed mainly throughthe vehicle of
in
of grosscapital investment
capital investment. Thus the contribution
the determination
of growthis muchhigherin Solow's laterestimatesthan
in thatof Denison.
A second pointto note is thatwhen "learning by doing" is introduced, we have a divergencebetweenprivateand social product,which
has been analysedby ProfessorArrow,18and this type of consideration
is also leftout by Denison's approach. But even apart fromtheserather
subtle but importantcomplications, the basic approach of avoiding
estimatesby lookingat income shares seems to be dubious,
productivity
when
the existenceof economies of large scale are assumed
particularly
simultaneously. We have concentratedmainlyon Denison's works. In
Production
Function",The[Reviewof
16. "TechnicalChangeand the Aggregate
Economic
andStatistics,
August1957.
S. Karlin,and P. Suppes
andTechnical
17. Investment
Progress",in K. J.Arrow,
in
Social
Sciences
Calif.,1960; "TechniMathematical
Methods
(Stanford,
(eds.),
Economic
The
American
and
Economic
Formation
Growth",
cal Progress,
Capital
Proceedings,
May 1962.
Review,
of Learningby Doing , The Reviewof Economic
18. 'TheEconomicImplications
June1962.
Studies,

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

PLANNING
EDUCATIONAND MANPOWER

some respectsthe less ambitiousworks of Schultz, Becker, Mincer, and


othershave been muchmore convincing. The Income Share Approach
has contributedmuchto the study of economics of education, but its
limitationsdo not seem to be adequatelyemphasizedin some of the more
publicizedworksof thisschool.
III.

Human Resourceindicators

In contrastto theheroicattemptsof theFixed RequirementSchool and


of the Income share approach, the recentattemptby ProfessorsHarbison
and Myersto constructand use some simple quantitativeindicators of
human resource developmentmustappear to be a pedestrianexcercise.19
But partlyforthe same reason,theirattemptis likely to mislead people
less about the statewe have reached in the complicated field of educationalplanning. Theirapproach consistsof compilinga very valuable
collection of data on educationaland relatedfields,and in tryingto see
whethersome indicesof humanresourcedevelopmentcan be formedfrom
these data, and whethersome use can be made of these indices. Since I
shall claim below thatin certainrespectsHarbison and Myersmisinterpret
the significance
of theirown indices,I should like to make it very clear
that
I do regardtheirpainstakingworkto be a path-breaking
straightaway
one in the fieldof internationalcomparisons.
With a great deal of research Professors Harbison and Myers
have compiledinternational
data on the followingseven "human resource
indicators".
1. Numberof teachers(firstand second levels) per 10,000 population.
2. Engineersand scientistsper 10,000 population.
3. Physiciansand dentistsper 10,000population.
4. Pupils enrolledat first-level
(primary)educationas a precentage of
the estimatedpopulationaged 5 to 14 inclusive.
5. The adjusted school enrolmentratiosfor first and second levels
combined.
6. Pupils enrolledat second-level(secondary)educationas a precentage
of the population estimatedaged 15 to 19 inclusive, adjusted for
length of schooling.
7. Enrolmentin third-levelhighereducation as a percentage of the
age group 20 to 24.
19. F. Harbison
andC.A. Myers,Education,
and Economic
Growth
Manpower
(New
York,1964).

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

10

A. K. SEN

In additiontwo indicatorsof the "orientationof higher education"


are used :
1. The percentage of students enrolled in scientificand technical
facultiesin a recentyear.
2. The percentageof studentsenrolledin faculties in humanities,fine
artsand law in the same year.
These lattertwo indicescould be compiledfor manyof the seventyfivecountriestaken up in thisstudy,but not for all of them. Harbison
and Myershave a lot of interesting
things to say in detail about the educational policiesof thesecountries,but what perhaps stand out as their
chiefcontributionto this fieldis theiruse of what theycall the"composite
index" of "levels of human resource development." This series is
constructedfromitems6 and 7 of the firstlist, i.e., enrolmentratiosat
the second and the third level. The composite index is simply the
arithmeticaltotal of (i) enrolment at second level of education as a
percentage of the age-group 15 to 19, adjusted forlengthof schooling,
and (ii) enrolmentat the thirdlevel of education as a percentageof the
age-group20 to 24 multiplied by a weight of 5. If we referto (i) as s,
and (ii) by t, the human resourceindex,is :
I = s+5.t
Then all countries in the world are classified into four groups
("underdeveloped","partiallydeveloped", "semiadvanced",and "advanced"), the resultsof whichare quoted in Table I below. We also mention
the per capita G.N.P. of thesecountries.
Table I
Harbison-and-Myers'Rankingof Countriesby CompositeIndex
of Human Resource Development and their
GNP Per Capita
CompositeIndex
Level I :

"Underdeveloped"

Niger
Ethiopia
Nyasaland
Somalia
Afghanistan

0.3
0.7
1.2
1.6
1.9

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

N.A.
55
60
50
50

EDUCATIONAND MANPOWER
PLANNING

11

CompositeIndex
Saudi Arabia
Tanganyika
ivory Coast
N. Rhodesia
Congo
Liberia
Kenya
Nigeria
Haiti

1.9
2.2
2.6
2.9
3.6
4.1
4.7
5.0
5.3
5.5
5.5
7.5

170
61
N.A.
150
92
100
87
78
105
N.A.
64
60

"PartiallyDeveloped"
Guatemala
10.7
Indonesia
10.7
10.8
Libya
Burma
14.2
Dominican Republic
14.5
Bolivia
14.8
Tunisia
15.2
Iran
17.3
China (Mainland)
19.5
Brazil
20.9
Colombia
22.6
22.7
Paraguay
Ghana
23.2
23.6
Malaya
Lebanon
24.3
Ecuador
24.4
Pakistan
25.2
Jamaica
26.8
27.2
Turkey
Peru
30.2
31.2
Iraq

189
131
60
57
239
99
173
108
73
293
263
114
172
356
362
189
70
316
220
179
156

Senegal
Uganda
Sudan
Level II :

Level III i "Semiadvanced"


Mexico

33.0

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

262

A K. SEN

CompositeIndex

GNP Per Capita


(U.S. dollars)

Thailand
India
Cuba
Spain
South Africa
Egypt
Portugal
Costa Rica
Venezuela
Taiwan
Greece
Chile
Hungary
South Korea
Italy
Yugoslavia
Poland
Czechoslovakia
Uruguay
Norway
Level IV : "Advanced"

35.1
35.2
35.5
39.6
40.0
40.1
40.8
47.3
47.7
48.4
48.5
51.2
53.9
55.0
56.8
60.3
66.5
68.9
69.8
73.8

96
73
431
293
395
142
224
357
648
161
340
379
490
144
516
265
475
680
478
1,130

Denmark
Sweden
Argentina
Israel
West Germany
Finland
U.S.S.R.
Canada
France
Japan
United Kingdom
Belgium
Netherlands
Australia
New Zealand
United States

77.1
79.2
82.0
84.9
85.8
88.7
92.9
101.6
107.8
111.4
121.6
123.6
133.7
137.7
147.3
261.3

1,057
1,380
490
726
927
794
600
1,947
943
306
1,189
1,196
836
1,316
1,310
2,577

Source - Harbison and Myers,Table 5, pp. 45-48.

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

13

EDUCATIONAND MANPOWER
PLANNING

Harbison and Myers find that the "composite index" is highiy


correlated withGNP per capita, thoughof course individualexceptions
are clearly visible in the table. The correlationcoefficientturns out to
be .888, whichis quite remarkablefor such a large numberof countries.
It is, however,somewhat curiousthatwithineach level of human resource
development,the correlation coefficientsare a great deal lower, viz..025 for Level I, .373 for Level II, .574 for Level III, and .692 for
Level IV. The relation is particularlyweak for lower levels, and get
strongeras we move to higher groups. This indicates some cause for
considerablecautionin interpreting
the data, but it also seemsto imply
that the relation is a strongerone over big changes than over narrow
variationswithineach group,particularlythe low groups.
Anotherinteresting
feature to note is thatthe average value of the
index risesmuch fasterthan GNP per capita between Levels I and II,
somewhat fasterbetween Levels II and III, and slower between Levels
III and IV. The average value of the composite index and of the GNP
per capita are shownin Table II, below.
Table II
ArithmeticMeans of "Composite Index" of Human Resource
Developmentand of GNP Per Capita (Level-wise)
CompositeIndex
Level I
(17 countries)
Level II
(21 countries)
Level III
(21 countries)
Level IV
(16 countries)

GNP Per Capita ($)

64

21

182

50

380

115

1,100

Source : Harbison and Myers,Table 2, p. 38.


Harbison and Myersinterpretthisto mean that "the highest rates
of humandevelopmentshould be made by countriesin Levels I and II."20
We do not have the opportunityhere of going into the detailsof
the analysis of Harbison and Myers, but it is necessaryto have a quick
20. Harbison
andMyers,
p. 37.

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

14

A. K. SEN

look at the meaningand interpretationof their"composite index." An


obvious question to ask is whyshould we choose the weight of 5 to put
enrolmentratio with the second-level enrolment
togetherthethird-level
ratio; why not some otherweight,say, 3 or 7 or anything else ? The
authorsdo not defend theirsystem of weighting,but we can thinkof a
number of alternativeways of proceeding in this question of weightselection. Since Harbison and Myersseem to regard this index to be
some kindof a guide to human resourcedevelopmenton the production
side (an assumption that we shall examine presently), one possible
method will be to compare the productivitiesof the two groups of
educatedpeople at the margin. But the productivityfiguresdo not exist
in thisform, and in any case one would expect thisratioto varygreatly
between the countries. In so far as there are earning figures,the
differentialseemsto be a greatdeal less than 5 in most countries.21We
in view
need not take the earningfiguresas good guidesto productivity
of our earlierdiscussionbut it is not easy to thinkof some otherbasis
forjustifyingthe weightof 5.
In fact, the exercise implies an economicproblemof interpretato get at than a purelystatistical
tionthat may be somewhatmoredifficult
the
correlationbetween the index
problemof what weightingmaximizes
and GNP per capita. It is this last problemthat Dr. R Sundrum has
posed in an unpublished note, and it turns out that a weight of 5.9
maximizes the correlation coefficientbetween the combined indexand
the GNP per capita.22 I would guess that considerationsof this kind
of the
probablydid influenceHarbison and Myersin theirdetermination
correlation
coefficient
is
because
the
of
5,
corresponding
quite
weight
close to the maximum. In fact,the resultis ratherinsensitiveto changes
in the weight, as the variables involved move more or less in the same
will naturallybe relatively
way,23and near the maximum the sensitivity
small.
But even if we take a purelystatisticalexplanationof the determination of the weight, there is an economic question involved in
interpretingthe indexitself. To make any use of it we must attribute
indicate
thattheratio ofmeanincomeof collegegraduates
21. Denison'sestimates
Table8, p. 68),
intheU.S.A.is around1.68( TheSources,
tohigh-school-finishers
a figure
near5.
nowhere
"Note onHarbisonandMyers'CompositeIndexof Human
22 R. M. Sundrum,
AsianInstitute
Resource
mimeographed,
Development",
(1964).
which
seemto movemoreorlessinthesame direction,
23. In factall theindicators
of the composite
ofthe "usefulness"
HarbisonandMyerstakeas demonstration
indexas a discriminator
(p. 37).

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

PLANNING
EDUCATIONAND MANPOWER

15

some clear economic meaningto the index. And here I find myself in
some considerabledisagreementwithHarbison and Myers. For themthe
index seems to stand for some kind of a guide to the stock of human
resources, whereasas far as I can see it tellsus nothingmore thanthe
rateat whichpeople are makingadditions to the stock. What proportion
of people today is receivingeducation of a given typeneed not bear any
close (not to mentionproportional)relationbetween the stockof people
with this kind of education. Yet this is precisely what the authors
implicitlyassume in their interpretationof the indices. To give an
instance, the authors indicate that "for example,a particular country
may have a low composite index and a much higher GNP per capita
because of richnaturalresources."24 But the compositeindexis not one
of thestockof educated manpower,and thereis no argumentfortreating
it as such. On the production side, the "composite index" tells us
nothingwhateverabout withGNP per capita to expect.
It seemsto me that the best way of explainingthe relationbetween
the "composite index" and the GNP per capita that Harbison and Myers
observe, is to take it as a consumption relationship. Depending on
current prosperity, decisions may be taken by individuals and the
governmentson how much to spend on education. We can expect the
flowof currenteducation will depend verymuch more on thiskind of
GNP per capita consideration,than GNP per capita itselfwill depend on
the flow of current education,on the production side. However, this
pointabout consumptionrelations should not be made too much of, as
even here the relationshipmay be farfroma rigid one, and the high
correlationbetween the index and GNP per capita, on which all this
speculation is based, is not freefrom, as we have mentionedbefore, a
certain amount of suspicious features. However, we can have some
discussion on the consumptionversus the production aspects,even on
the basis of moreor less pure theory, and it can be pointed out that
if we expect any relation betweenthe proportions of students schooling
and the GNP per capita on the productionside, it should be betweenthe
proportions now and the increase in productionsome years later. The
relationshipmustbe one of a stock-flowkind witha timelag, ratherthan
of a flow-flow
kind withouta timelag.
It might be thought, however, that there may be such a close
relationbetweeneducationalstockand the flow of students today, that
one can be taken roughly as an index of theother. We have not got
much data to confirmor contradictthe picture, though theoretically
24. HarbisonandMyers,
p. 41.

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

A. K. SEN

there is no convincingreason why this should be the case. Some


internationaldata on the stock of educatedpeople relatedto GNP can
be found in an unpublished paper of Mr. L. J. Emmerij.25 These
concern9 countriesonly,but it is interesting
to compare the ordering of
the educational flow ratios and the educationalstockratiosforthese 9
countries. The flowratiosare all computedfrom Harbison and Myers's
Table 5, using columnson "Per Capita GNP, US Dollars," "Enrolment
Ratios: 2d Level Adj.", and "Enrolment Ratios: 3d Level Unadj."
Harbisonand Myers's"compositeindex" is of course made out of these
last two columns. The number of secondaryeducated men and that of
educatedmen each per unitof GNP, i.e., the stockfigures,are
third-level
taken simply from Emmerij's calculations. Table III gives the comparisonof theorderingsof the secondarystockand flow ratios and Table
IV thatof third-levelstockand flowratios.
Table III
Orderingsof SecondaryEducation Stock and Flow
Per Unit of GNP

Countries

Venezuela
France
Italy
Canada
Brazil
U.S.A.
Costa Rica
Colombia
India

Numberof RatioofSecondSecond-level levelEnrolment Orderingof


EducatedMen Ratio to GNP
Emmerij's
Per Unitof
Per Capita
Stock Ratios
GNP {Emmerij)
(H-M)
1.5
3.3
5
5
5.7
6.3
7.5
8.2
10.7

4.04
6.24
6.74
2.80
4.40
3.70
8.07
5.17
33.15

1
2
3/4
3/4
5
6
7
8
9

Orderingof
H-M Flow
Ratios
3
6
7
1
4
2
8
5
9

toMr.Emmerij
to allowme to quote the data in my note,
25. I amverygrateful
:
"A PlanningModel of EducationRequirements
of EconomicDevelapment
Some Comments,"
figures
givethedata
O.E.C.D.,op. cit.,TableI. Emmerij's
numbersof decimal points dependingon the variations
up to different
in the sourcesand the quality of the data. We have multipliedthe
the
ratiosby100foreasy comparability,
withoutof coursechanging
Emmerij
are:
forthe different
relativeproportions.
countries
The yearsof reference
Venezuela(1950),France(1954;,Italy(1951), Canada (1951),Brazil (1950),
U.S.A.(1950),CostaRica(1950),Colombia(1950),andIndia(1955).

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

EDUCATIONAND MAMPOWER
PLANNING

17

In spite of some similaritiesthat the orderings are completely


differentin the two cases should be obvious by glancingat the table.
Also the quantitativeratiosare significantly
different.
Tablb IV
Orderingsof Third-levelEducation Stock and Flow
Per Unit of GNP

Countries

Venezuela
France
Colombia
Brazil
Canada
U.S.A.
Costa Rica
Italy
India

Numberof Ratio of ThirdThird-level levelEnrolment Orderingof


EducatedMen Ratio to GNP
Emmerij's
Per Unitof
Per Capita
Stock Ratios
GNP {Emmerij)
(H-M)
0.45
0.9
1.0
1.1
1.2
1.4
2.0
2.2
2.6

0 66
1.04
0.68
0.55
0.48
1.29
1.04
0.85
3.01

1
2
3
4
5
6
7
8
9

Orderingof
H-M Flow
Ratios
3
6/7
4
2
1
8
6/7
5
9

Here again the orderings are very differentbetween stocksand


of educationto GNP
flows,so thatany generalizationsabout contribution
cannot easily be based on theflow figures. Also no immediateconclusion
can be drawnabout whichcountries"may need to" make "major effort
in investmentin the developmentof human resources"(p. 37), on the
basis of the flowdata alone.
The point that is being consideredherecannot be treatedas one of
only academic interest. For example, let us examine Harbison and
Myers's method of constructingthe indexonlyfromsecond-levelenrolmentratios, leaving out primaryeducation. The rationale for this is
based on the relativelylower correlation coefficientbetween primary
enrolmentratio and GNP per capita, which is only .668 for these
countries. This allows Harbison and Myers to confirmthat an "index
based on higher levels of education correlates more significantly
with
measuresof economic developmentthan one based on the lowest level
of education."26 This may well be the case, but thisargumentcertainly
andMyers,
26. Harbison
p. 40,

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

18

Ar K. SEN

cannotbe derivedfromthe low correlation with the primary enrolment


ratio. One has to look at some otherindex,such as the proportionof
firstlevel educated men in total population to see whether primary
education correlates significantlywith GNP per capita or not. Indeed
it is possible to argue thatin mostcountriesprimaryeducationhas become such an importantminimalrequirementof civilized existence that
they try to make the enrolmentratio as highas possible,withoutundue
considerationof nationalincome. (The limitationsthat are there arise
mostly from non-income considerations.) But thiswill affectthe stock
of primaryeducated persons only in the rather long run,27and its
consequences on the national income, which may be more significant
thanotherkindsof education for a number of countries,will be seen
only in the long run. Thus the low correlationbetweenprimaryenrolmentratio and GNP per capita tells us very litlle about the importance
of primaryeducation as an indicatorof humanresource development.
And it is in fieldslike thisthat thereis a genuinechance of basing policy
on wrongconclusionsdrawn fromthe H-M indicesand the corresponding
correlationcoefficients.
Finally, in some of the statementsof Harbison and Myersthey
give the impressionas if theyhave at the back of their mind a model of
educational planning almost like that of Tinbergenand Correa. Take
theirconclusion,quoted a little bit eailier, that "it is clear that the
highest rates of human resource developmentshould be made by the
countriesin Levels I and II."28 This is based on the simple ground that
the composite index rises relativeto GNP per capita a greatdeal faster
betweenLevel I and Level II, and between Level II and III, compared
with that between Levels III and IV. Quite apart fromthe factthat
these data give no indicationwhatever about the amount of eifort the
Level IV countries should spend on this,comparedwithLevels I and II
countries,even the comparison with Level III is not quite legitimate
unlessone makes theassumptionthat the pattern of relationshipbetween
national income and educational resources is a unique one, so thata
movementfromLevel I GNP to Level II GNP cannot take place without
accumulatingthe amount of average educational stock as in Level II ;
ofprimary
educational
flowaffects
27. It mustalso be notedthatanyadjustment
moreslowlythanadjustment
insecondor third
thetotalsizeofthestockrather
of lifeaftercompleting
level flows.This is becausethe expectation
primary
secondor thirdlevel
is considerably
education
longerthanthataftercompleting
education.
andMyers
28. Harbison
, p. 37.

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

PLANNING
EDUCATIONAND MANPOWER

19

and similarlybetweenII and III. Withoutassuming somethingapproximating a unique relation between education and income such recommendationscannot be made.
I mustconfessthatI findmyselfverymuch more in sympathywith
the Harbison-Myersmethod of collecting and presenting information
and constructing
indices,than withthe use they make of these indices.
Particularlythe problem of stock and flow,that of the consumptionand
the productionside, the question of weighting,the non-sequituron the
importanceof primaryeducation, and the implicitassumption(in some
of the statements)of an approximateunique relation between income and
education,should worryus a greatdeal.
Some General Conclusions
The main object of this paper was to present, compare and
criticallyevaluate the various approaches to the problemof education
and manpowerplanningthat are being widely used at the moment. It
has not, of course, been possible to cover all the approaches to the
those workswhichare based on skilful application
question,particularly
of commonsenseto the available data, e.g , thatof JohnVaizey.29 But
the evaluationof themore formal approaches does leave one with the
impressionthat while a great deal has been achieved,the subjecthas
perhaps sufferedmore than a little from trying to arrive at quick
withratherstrong,
conclusions,based on coveringup gaps of information
and not very easily supportable,assumptions. The Fixed Requirement
Approach is perhapsthe worstoffenderin this, though the heroic uses
of the Income Share Approach also leave one acutelyuneasy. While it
is truethattime,tide and policy decisions wait for no one, I am not
convinced that the need forenlightenedpolicy is best servedby patching
up our gaps of knowledge by a set of assumptions that are neither
intuitivelyvery plausible, nor are verifiedby data (in some cases are not
even verifiable).
Leavingout the quick conclusionsthat Harbison and Myers draw,
their approach of constructingrelevant indices to see whatuse can be
a veryuseful one. On
made of them,mustbe regardedas fundamentally
and
the basis of our review of Harbison
Myers'sparticularmethods,
however,it may be suggestedthat as a guide to theproductionside of the
picture, their "composite index" is not quite the rightthingto look for,
thoughits value in otherrespectsis not to be denied. What we need
on the stocksof educatedpeople of different
clearly is more information
29. JohnVaizey,Economics
1962).
ofEducation
(London,

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

20

A. K. SEN

kinds,forwhichthe enrolmentratios are very poor guides. For most


do not exist. But lack of information
countries these information
is not
an immutablelaw of nature,and part of thejob of the economists workthe
ing in this field is to point out preciselywhat kind of information
various countriesshould collect. It will be not only useful, but essential
for educational planning, to have informationon the numberof people
witha giventypeof educationclassifiedby age, occupation and earnings.
Even theircrude numberswill be moreusefulto correlatewithGNP than
enrolmentratios. With informationon age, we can do one better,by
tryingto bringin the forces of learning by doing beyond the level of
formaleducation,whichhas been so farmore or less leftout of orthodox
discussions on the economics of education.80 Earnings would provide
on the value of a workerwitha given kind of education to
information
the enterpriseemploying him, but in usingthese the difficulties
arising
betweenprivateand social productivities,fromimperfromthe difference
fect foresight,from the existenceof monopolies, and fromthe presence
of conventionalwages not determinedby market forces, must be kept
in mind.
The data on educationalstock withfurther
on age and
information
be
can
collected
the
with
occupation
fairly easily
periodical censuses.81
Informationon earningsis, however,rather more difficultto come by,
partly because the rates of taxation depend on this. Here the method
of sample surveyscan be of veryconsiderableuse, maintaininganonymity
of informationto reduce the risk of deliberate misreporting. Sample
surveysshould also be useful at the other end of the picture, viz., in
findingout how the enterpriseregardsthe usefulnessof variouskindsof
skill,dependingon formal education, on-the-job training,and learning
fromexperience.
Finally,the effectof primaryeducationon the peasants* attitude to
modern methods of production and on various social and economic
problemsin the ruralareas will be extremelyimportantto study in the
contextof underdevelopedcountries. This typeofconsiderationis usually
on
30. On learningby doing,seeK. J. Arrow, "The EconomicImplications
, XXIX (3),No. 80 (June1962). On
Learningby Doing, Reviewof Economic
of this processon skill,productivity
someoftheconsequences
and earnings
ofworkers,
seeA. K. Sen,"Education,
by Doing and Productivity",
Learning
Asian Institutefor EconomicDevelopment
and Planning
mimeographed,
(1964).
fromothersources. See Pitamber
can also be collected
Pantand
31. Information
Educated
inIndiv
Persons
T. P. Chaudhry,
, Perspective
PlanningDivision,PlanNewDefti,1959.
ningCommission,

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

EDUCATIONAND MANPOWER
PLANNING

21

leftout of discussionsbecause of the difficulty


in gettingprecise information on this,thoughthisgap in our knowledgeis sometimesmade to look
unimportanton the basis of some misleading argumentson correlation
coefficients,
e.g., thatof Harbison and Myers, whichwe discussedearlier.
Rational discussions on educational policy requirethatwe look closely
into this side of the picturein the underdevelopedcountries.
We startedby lookingat the economic approaches now being used
for educational and manpower planning, and we have ended up by
research. If this
drawing up somethinglike a big agenda forfurther
seemsto implya certainscepticismabout thecurrentstateof this fieldof
economics,that is preciselywhat is intended.
Delhi School of Economics
UNIVERSITY
OF DELHI

Amartya K. Sen

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:27:54 AM


All use subject to JSTOR Terms and Conditions

You might also like