The World's Changing Human Capital Stock: Multi-
State Population Projections by Educational
Attainment
Wolfgang Lutz
International Institute for Applied Systems Analysis, Laxenburg, Austria
Anne Goujon
International Institute for Applied Systems Analysis, Laxenburg, Austria
RP-01-11
June 2001
Reprinted from Population and Development Review 27 (2): 323-339 (June
2001)
International Institute for Applied Systems Analysis • Schlossplatz 1 • A-2361 Laxenburg • Austria
Tel: (+43 2236) 807 • Fax: (+43 2236) 71313 • E-mail: publications@iiasa.ac.at • Web: www.iiasa.ac .at
llASA Reprints make research conducted at the International Institute for Applied Systems Analysis
more accessible to a wider audience. They reprint independently reviewed articles that have been
previously published in journals. Views or opinions expressed herein do not necessarily represent those
of the Institute, its National Member Organizations, or other organizations supporting the work .
Reprinted with permission from Population and Development Review 27(2): 323-339 (June 2001)
Copyright © 2001
All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopy, recording, or any information storage or retrieval
system, without permission in writing from the copyright holder.
Vienna
Yearbook
of
Population
Research
2003,
pp.
19-33
The World's Changing Human Capital
Stock: Multi-State Population
Projections
by
Educational Attainment*
Wolfgang Lutz and Anne Goujon
19
EDUCATION
IS
GENERALLY
assumed
to
have far-reaching beneficial consequences.
At
the indi-
vidual level more education tends
to
imply better health, wider economic opportunities, and
greater autonomy, especially
for
women (Federici et
al.
1993; Jejeebhoy 1995).
At
the aggre-
gate level the educational composition
of
the population has long been considered a key
factor
in economic, institutional, and social development (Bellew et al. 1992; Benavot 1989; Hadden
and
London 1996)
and
in the rate
of
technological progress (Grossman and Helpman 1991;
Romer 1992). The extensive theoretical
and
empirical literature on the relationship between
human capital formation and various aspects
of
development is not reviewed here.
1
instead,
this research note demonstrates the feasibility
of
multi-state cohort component population pro-
jections
for
groups defined by different educational attainment. With the increasing impor-
tance
of
education in a knowledge-based economy, this approach can make a contribution not
only
to
the
field
of
demography, but also to long-range economic planning.
The multi-state approach
The increasing awareness over the past decade
of
the importance
of
human capital in
development has stimulated attempts to estimate and project the educational composition
of
the population. Most empirical studies have approximated educational stocks in terms
of
enrollment ratios or illiteracy rates (Mankiw et al. 1992; Romer 1989). What is needed,
however,
is
a complete matrix
of
the composition
of
the population
by
age, sex, and levels
of
educational attainment for different points in time. Many attempts to measure human
capital stock have failed to meet this goal because
of
problems with country-level data and
the lack
of
appropriate demographic methodologies (Ahuja and Filmer 1995; Barro and
Lee 1993; Dubey and King 1994; Kyriacou 1991; Nehru et al. 1993; Psacharopoulos and
Arrigada 1986, 1992). Ahuja and Filmer (1995) came nearest to our approach by taking
existing
UN population projections and superimposing onto them an educational distribu-
tion estimated for two broad age groups (ages 6-24 and
25
and older) from given sets
of
enrollment ratios and UNESCO projections. Using this approach they projected the educa-
tional composition (for four educational groups) for
71
developing countries. Apart from
the lack
of
more detailed information by age, this approach was also
of
a static nature: it
An extensive international bibliography is given in Brock and Cammish ( 1997).
* Reprint from Population and Development Review, 27 (2), pp. 323-339 (June 2001). New York.
NY:
The Population Council. Reprinted with the permission
of
the Population Council.
20
The World's Changing Human Capital Stock
did not allow the educational composition
of
the population to influence fertility despite
the marked educational fertility differentials found in most developing countries.
In this study we apply the demographic methodology
of
multi-state population projec-
tion to the task. This method is based on a multi-dimensional expansion
of
the life table
(increment-decrement tables) and
of
the cohort-component projection method developed
at the International Institute for Applied
Systems Analysis (IIASA) during the 1970s (Keyfitz
1985; Rogers 1975). The multi-state model divides the population by age and sex into
"states." Originally states were conceived
of
as geographic units, with the movements
between the states being migration streams. But a
"state" can also reflect any other clearly
defined subgroup
of
the population, for example groups with different educational attain-
ment, with the movements then becoming educational transition rates. The projection
of
human capital stocks by age and sex
is
an ideal example for the application
of
the multi-
dimensional cohort-component model, because education tends to be acquired at younger
ages and then simply moves along cohort lines. Change in the educational composition
of
the total population (aged
15
and older) is typically caused by the depletion (through mor-
tality)
of
less-educated older cohorts and the entry
of
more-educated younger cohorts.
Figure 1 shows the specific structure
of
the multi-state model chosen for this study.
It
subdivides the population into four distinct groups according to educational attainment.
Each subpopulation
is
further stratified by age (five-year age groups) and sex, and can be
represented through a separate population pyramid. The key parameters
of
the model are
three sets
of
age- and sex-specific educational transition rates, that is, the age-specific
probabilities for young men or women to move, for example, from the category
of
primary
educational attainment to that
of
secondary attainment. Although this model can handle
transitions at any age, for example as a result
of
adult education ca;npaigns, in reality
transitions at older ages are very rare. Transitions here are concentrated in the age range
below 25 years depending on the kind
of
transition. Another important feature gives this
model a dynamic element: it considers different fertility rates for different educational
groups. Hence, even with constant status-specific fertility, a change in the relative size
of
the educational subpopulation results in changes in the fertility rate
of
the total population.
This feature significantly influences the results discussed in the data section below. Mi-
gration and mortality are only considered by age and sex in this application. While for
international migration we assumed a hypothetical distribution by education, we did not
feel in the position to do so for mortality.
2
As more empirical information becomes avail-
able, empirically founded educational differentials may also be assumed for these two
components
of
change.
Social science tells us that not only fertility rates, but also school enrollment rates
(flows) tend to depend on the educational composition (stocks). There
is
much evidence
of
intergenerational transmission
of
education. For this reason we also run some special sce-
narios that take account
of
such feedbacks.
It
is useful here to distinguish between first-
order feedbacks that represent compositional effects resulting from educational fertility
differentials, and second-order feedbacks that represent behavioral responses to changing
stocks. Although there is little reliable empirical evidence to model such responses at the
Data are available from surveys on infant mortality
by
education
of
the mother, but neither this
information
nor
the information about surviving relatives
is
sufficient to
study
the educational
level
of
the person who dies. As discussed in Lutz et al. ( 1999), mortality differentials have
significant effects only for the size
of
the elderly population.
They
are less relevant for the study
of
the working-age population.
Wolfgang Lutz
and
Anne Goujon
Figure
I:
Structure
of
the
multi-state
population
projection
model
by
level
of
education
Population
with no
education by
age
and
sex
Population
with primary
education by
age
and
sex
Population
with
secondary
education by
age
and
sex
~
-
.
..............
...
Educational transition rates by age and sex
Ferti
li
ty by age and e
du
cation
of
mother
In-
and
out
-migration by age, sex, and education
Mortality by age
and
sex
Population
with
tertiary
education by
age and sex
Mortality
21
macro level,
we
will
de
scribe an experimental scenario incorporating such second-order
feedbacks.
3
It
is evident that
under
such a setup, educational projections
cannot
be
simply superim-
posed onto given population projections, as has been done in previous studies.
Population
projections must
be
carried out as an integral
part
of
the exercise, since alternative educa-
tional scenarios will result in alternative fertility trends.
The
educational projections pre-
sented here, therefore, expand on the demographic assumptions
of
IIASA's
newest and
still unpublished world population projections, which will document
our
more extensive
projections (Lutz
200
I).
Our
projections pick up methodological recommendations made
in
Frontiers
of
Population Forecasting (Lutz
et
al. 1999).
The
substantive fertility, mor-
tality, and migration assumptions are largely based
on
recommendations in the National
Academy
of
Sciences report entitled Be
yo
nd
Six Billion (Bongaarts and Bulatao 2000) and
are specified in detail in Lutz
(2001 ).
Population projections
by
level
of
education are a logical next step in improving popu-
lation forecasts and making them more relevant. As discussed in Lutz
et
a
l.
( 1999) adding
education to
age
and sex as
an
explicit demographic dimension in population forecasting
also affects the demographic output parameters themselves because a significant source
of
so-far-unobserved heterogeneity is being observed and explicitly endogenized.
It
may,
therefore, be considered an improvement, even
of
the purely demographic output parame-
ters
of
the projection. More importantly, however, the future educational composition
of
the population is
of
interest in its own right.
4
We
thank an
anonymous
referee for encouraging us to present such a scenario.
For
a detailed listing
of
countries
by
region, see Lutz ( 1996).