THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 321
A SYSTEMIC VIEW OF THE POPULATION OF CHINA, TODAY AND TO-MORROW
André LAMBERT € Louis LOHLE-TART
A.D.R.A.S.S. (Ottignies, Belgium)
ABSTRACT
The authors present models, written in Dynamo language, used for reconstitu-
ting the demographic evolution of China between 1953 and 1978, and for dynamic
simulations of the Chinese population. These models incorporate standard fea~
tures of demographic projections but also a highly disaggregated sub-system of
the female fertility, by age of women and number of children ever born. The
spontaneous evolution of population is easily analyzed; any number of changes
in fertility and/or mortality can be simulated, together with their consequen~
ces on population figures.
The paper presents simulations of modifications in the trends of mortality and
fertility. The effects of such modifications are discussed.
The authors show the importance of existing stocks of people (by sex, age and
number of children) for understanding and predicting the evolution, "natural"
or consecutive to governmental actions, in population parameters; the evolu-
tion can be accelerated or delayed due to structural constraints.
The short- and long-term effects of the possible evolutions in demographic
parameters on the several stocks (population structures) are discussed, in the
perspective of their usefulness for Gecision-makers. The disaggregation of the
model allows to make sectorial projections for each simulated hypothesis.
The feasability of a stabilized population in China by the year 2000 is also
discussed.
A RECONSTITUTION OF THE RECENT EVOLUTION OF THE CHINESE POPULATION
Position of the Problem
Population dynamics can essentially be thought in terms of many disaggregated
stocks of people, the so-called structures, determined by flows, internal to
the system (ageing) or between the system and the "outer world" (births,
deaths and migrations). System Dynamics tools are thus perfectly suited to
demographic modelling.
The strong inertia of population structures is well known : the evolution of a
population is largely determined by its current distribution of people by sex
and age. For instance, the number of births is a function of the number of wo-
men in childbearing ages, and these births will determine the number of women
reaching childbearing ages twenty years later. Relevant simulations of the
future population growth are thus only possible on the basis of very detailed
data about present structures; however, this is precisely the kind of data
322 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
usually unavailable in most countries of the world, ‘includind China.
But this inertia has also a reverse consequence : whenever series of data,
even incomplete, are available on structures, total population and events
(births, deaths), it is possible to guess a range of likely values for un-
known details - and the more available are data, the narrower is the range of
guesses.
Nevertheless, the inertia of structures do not last forever; on the contrary,
a given structure submitted to constants laws of evolution will result in some
"final" convergent structure depending only on these laws (strong ergodicity
theorem). In other words, this property of convergence could be used for gues-
sing a structure whenever there are some evidences concerning its evolution :
for instance an arbitrary initial age structure has only to be simulated until
convergence, subject to measured or estimated parameters of fertility and mor-
tality, for getting the most likely initial structure corresponding to these
parameters.
Another characteristic of population dynamics is very relevant to modellers :
detailed rates of events are not random. For instance, the death rate at a gi-
ven age is correlated with the rates of next ages; each correlation is rather
loose, but there are only a few contrasted patterns of rates across a variety
of ages. And the same holds for fertility. Consequently, even with a poor in-
formation about deaths and births, it is possible to derive good estimates at
a detailed level from these data and some mathematical model, requesting only
a few parameters : model life tables (distributed between four or five "fami-
lies"), functions of fertility.
Reconstituting the Population
The exercise of reconstituting the Chinese population between 1953 and 1978
has been described elsewhere from the point of view of demographers (Lambert,
1981). We will only briefly summarize the process, stressing some points rele-
vant to modelling and simulation techniques.
Basic data were very poor : a census realized in 1953 (without a detailed pu-
blished age structure) and scattered information on the number of births and
deaths, mainly taken from official statements and articles in the People's
Daily or in some periodicals.
Needed data were considerably more detailed : an initial structure of popula-
tion by age (in single years) and sex, and yearly series of rates of mortality
and fertility, at least for each five-year age group.
The first step, getting an age structure, is straightforward by using the
classical approach of a set of polynomials applied to five-year age groups and
giving single-year groups (Sprague's multipliers).
The estimates of mortality and fertility are obtained by trial and error,
through an iterative process. Initially, several sources of knowledge (inclu-
ding clinical data and ethnological information) give an idea of the pattern
of the phenomena : global shape of the curves (for instance, relative impor-
tance of the infant and juvenile mortality), indications on quantifiable pa-
rameters (for instance, brackets for the life expectancy or the mean age at
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 323
childbearing).
For the first year of simulation, the level of the selected curves of morta~
lity and fertility is found by searching for parameters enabling to reproduce
correctly the total number of events. For instance, the nature of mortality in
China is nearest to the "B" family of the OECD model life tables; taking into
account the size and structure of the 1953 population, a total around 9.8
millions of deaths according official sources can only be obtained with a life
expectancy at birth of 47.7 years; this figure is only a synthetic estimate,
based on a complete set of rates of mortality by age. The same is true for
fertility, where a complete set of 36 age-specific rates is derived from a
Pearson's gamma function based in fact on only two parameters, a mean age at
childbearing (around 27 years) and an average number of children per woman at
the level of 4.50. This estimate results in turn of the trial and error proce-
dure, using the mean age estimate and aiming at a correct reproduction of the
total number of births.
The detailed population of the next year is easily deduced from the initial
structure and the natural movement estimated by the above-mentionned method.
The simulation continues in this way, year after year, in the best case. In
many cases, however, even data on births and deaths are lacking or unreliable,
and the trial and error procedure is made more complex by the fact that inter-
mediate trends should also be guessed on the mere base of qualitative informa-
tion (such as the awareness of poor harvests, and so on). Actually, the re-
constitution of a likely evolution of the Chinese population between 1953 and
1978 requested above one hundred simulations before getting a "path" compati-
ble. with every reliable known information.
From a technical point of view, let us mention that this project was executed
with a model written in Dynamo III/370 (running on an IBM 370/168 and later on
an IBM 4381). While the basic tools of demographic projections were included
in the model, it implements some typical Dynamo features. For instance, we de-
vised a mean for "un-smoothing" the trends in mortality. Given the reference
model life table, an approximate level of mortality at both ends of the simu-
lated period can be computed, giving a trend, translated ina number of years
needed for the transition. Application of this smooth trend is obvious : it
only needs a linear interpolation between transforms of the limit life tables,
driven by the time. However, the actual evolution of mortality was rather cha-
otic; so, in order to follow the very large fluctuations observed, the model
goes on interpolating between tables, but using a variable number of steps
between ends. In other words, worsening conditions are translated as a longer
period of transition, and the situation of the simulated year respectively by
reference to the true calendar and to the size of the period of transition
make the interpclation leaping backwards. The set of “instantaneous durations"
was of course designed by trial and error too.
Some Results
The results presented below (Table 1) are only synthetic indicators based on
the functionning of the model and compared to the available pieces of infor-
mation, Let us keep in mind that the true "results" do not appear here, since
they are in fact the thousands of detailed figures reconstituted by the model
and dynamically related, year after year. It should be noticed that no simu-
lated data differs from real ones by over one percent, except for the so-cal-
$24 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY, CHINA
led "black years", for which there are no reliable estimates. The cumulated
numbers of events along both series converge about within 0.5 per thousand,
confirming that the reconstitution is plausible, even if there is no proof
things happened exactly in this way.
In the Table 1, we present, for selected years, numbers of deaths and births
given by Chinese sources ("observed") and computed in the model, along with
the computed total population at end of year and two key demographic indices
from the model.
Table 1. Synthetic Results of a Reconstitution of the Chinese Population,
1953-1978. Selected Years.
Numbers (in millions) of Total |child. Life
Popul. per Expec-
Births Deaths in Woman tancy
Year | observed model observed model millions|(units) | (years)
1953 21.3 21.29 9.8 9.89 583 4.50 47.7
1958 18.5 18.61 (13.5) 14.01 649 3.65 40.6
1963 29.0 - 28.93 6.5 6.51 674 5.30 60.9
1968 28.5 28.71 7.6 7.56 718 4.63 61.3
1973 25.0 25.03 6.3 6.35 880 3.59 66.3
1978 17.5 17.49 6.4 6.15 952 2.35 68.2
A sharp decline in both fertility and mortality is clear from these figures.
Consequences on age structure are predictible; for instance, the share of the
population under 20 years fell from 48.7% in 1953 to 44.3% in 1978. A conti-
nued trend towards low fertility, an important political goal for Chinese of-
ficials, will still have more important consequences, with repercussions on
the number of people in reproductive ages, delayed by one generation. Conver-
sely, a rise in fertility among age groups not yet reduced in size could ge-
nerate an oscillating system where waves of large birth cohorts would alterna-
te with smaller ones and lead to irregularities difficult to manage for any
social system. For instance, the number of children reaching school ages can
vary by a factor of 2 over a short period of time : the number of births drop-
ped from 21.5 millions in -1957 to 13 millions in 1960 but was back over 29
millions in 1963, only three years later.
Advanced Modelling of Fertility
Whenever. the synthetic indices of fertility change strikingly, it means that
the reproductive behaviour of people is modified, especially the parity, i. e.
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 325
the rank of birth : proportions of women having one, two, etc.. children also
change, so do the probabilities of getting an additional child for every
actual number of children. Beside this technical consideration, the offspring
of each individual woman has for long been a sensitive issue for the Chinese
decision-makers. Consequently, it seemed relevant to try to simulate not only
the global fertility, but also to estimate proportions of women by parity and
the evolution of this particular structure. Since, obviously, the parity is a
function of the duration of "exposure to the risk of childbearing", this dis-
tribution has to be simulated by age. An advanced version of the initial model
enabled us to guess the structure of Chinese women distributed by age and pa-
rity (limited to the fifth rank) at the time of the 1953 Census and to follow
its evolution along the whole period of simulation. The exercise was conducted
by the same trial and error process used previously and it would be tedious to
describe it at length. It could be felt more hypothetical than the global
exercise, e. g. because of a greater sensitiveness of fertility curves by pa-
rity to departures of the model parameters (themselves less known) from the
real ones. However, this disaggregated model is also submitted to additional
constraints : the parameters computed and tested in the global model have to
be reproduced by the synthesis of the disaggregated model. For instance, the
mean age at childbearing for the whole female population is checked against
the weighted average of mean ages for each parity group. With this proviso in
mind, the hypothetic structure by parity we derived for the Chinese population
of 1978 should be understood as one possible scenario, but a very likely one
because alternative scenarios respecting the same constraints are not evident.
We will not present here results of the advanced model, which are only rele-
vant for demographers (its features are included in the further discussion).
A MODEL FOR THE FUTURE
The Model
The first model, presented above, served as a tool for initializing another
model, dedicated to forecasting. The reason for building a different model is
simple : we desired to use it on micro-computers but the severe limitations on
the size of micro-computerized models do not allow to implement it. We rewrote
thus totally the model, dropping all the complex systems used in the reconsti-
tution of the population and keeping only the resulting structures. We were so
in a position to write a simplified model in the Professional Dynamo version 2
plus language (running on an IBM PC/AT, it uses only 57K of core). Of course,
only simple scenarios can be simulated, but they are demonstrative enough of
the abilities of this technique as an helping tool for decision-makers.
The actual model embodies mainly a population module where the total popula-
tion is distributed by sex and single years of age; women in reproductive ages
are also distributed by parity; the flows of individuals are controlled by
sets of age-specific rates of mortality and fertility; there is no. provision
for migrations. Of course, any change in the parameters governing the flows
can be simulated, and the output of population in terms of school enrolment,
working force, and so on, is available. For the present paper, we only used
hypothetical coefficients for the sake of the demonstration, but any set of
realistic coefficients can easily be substituted for meeting real conditions.
326 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
4
Inertia or The Weight of the Past Times
We have shown the erratic and typically non-linear pattern of evolution of
both mortality and fertility in the past thirty years. An easy demonstration
of the inertia of demographic phenomena lies in an “everything constant"
scenario. In this case, we first simulated the next fifteen years under the
simple hypothesis of constant mortality and fertility patterns, i. e. a life
expectancy at birth of 68 years and a final descent of 2.35 children per wo-
man. Table 2 shows, at selected years, the evolution in the number of births
and deaths. Births reach a maximum during the nineties, due to the coming in
age of the large births cohorts of the sixties, and then begin to fall; deaths
are steadily growing, due to both the ageing of the population and the abso-
lute growth of its size; for the same reason, in spite of an actually constant
mortality, the crude death rate grows. However, with the constant hypothesis,
these two opposite trends in numbers of deaths and births never converge. The
annual growth rate of the whole population, around 1% in 2001, would still
decline but would stabilize around 0.5% and stay indefinitely at this level.
Table 2. Total Population, Births and Deaths, China, 1986-2001,
Selected Years. Constant Hypothesis (figures in millions).
Total
Year Population Births Deaths
1978 952 17.5 6.4
1986 1044 21.3 7.7
1989 1087 22.4 8.4
“1992 1136 25.7 9.0
1995 1185 25.7 9.8
1998 1231 24.2 10.4
2001 1270 22.5 11.0
The parameters of fertility and mortality used in the constant hypothesis are
not especially low; however they represent a sharp decline by reference to the
high levels which determined the Chinese population structure for centuries,
until the early sixties. That means a drastic change in the determinants of
the structures, reflected by a slow but perceptible drift in the age distri-
bution of the population. Table 3 shows the repartition between conventional
age groups of "young", "adult" and "old" people. From this table, it is clear
that the ageing of the total population is mainly due to the diminution of the
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 327
Table 3. Relative Structures (per thousand people) of the Chinese Population,
1986-2001, Selected Years. Constant Hypothesis.
Proportion of Age Groups
Dependency
Young Adult Old Ratio
Year | (under 20) (20-59) (60 and over) (units)
1978 443 484 713 1.065
1986 398 516 86 +937
1989 373 537 90 864
1992 353 552 9s .810
1995 342 559 99 «790
1998 343 554 103 -806
2001 346 548 106 825
Table 4. Absolute Structures (in millions) of the Chinese Population,
1986-2001, Selected Years. Constant Hypothesis.
Size of Age Groups
Young Adult old
Year (under 20) (20-59) (60 and over)
1978 421.7 460.8 69.5
1986 415.7 539.4 89.6
1989 405.7 583.5 98.3
1992 400.6 627.3 107.7
1995 405.7 661.9 417.4
1998 422.4 681.6 127.0
2001 438.8 695.7 135.3
328 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM: DYNAMICS SOCITY. CHINA
share of younger people, rather than to the growth of the share of older
people. This double move is made apparent by the dependency ratio, a conven-
tional measure of the burden on people at "active" ages (i. e. adults) due to
people in “inactive” ages : the dependency ratio declines very fast to a mini-
mum in the mid-nineties.
However, we should be aware that relative structures visualize the evolution
of population, but do not directly reflect what is most meaningful from the
point of view of decision makers, i. e. the absolute size of the age groups
(population in school ages, in active ages, and so on). Indeed, we can observe
from Table 4 that the overall growth of the population overcompensates the
fall in the proportion of youngsters from the mid-nineties. For instance,
China will reach its ever-biggest number of young people around the year 2000,
while their share in the population would be near its historical minimum.
Going into Uncertainties : Changes in Mortality?
According to an hypothesis of decreasing mortality, we simulated the evolution
of population if the observed trend of improved conditions of mortality would
prevail, reaching in 2001 the level of life expectancy to birth of 74.0 year
(i. e. the level of mortality observed in some countries, like Sweden or the
Netherlands, around 1985). Though such an evolution is quite important, the
impact on the total population is not very significant, as shown by Table 5,
compared with Tables 2 and 3 : only an additional 24 millions people, or an
increase under 2 per cent, and the changes in the age groups distribution is
still less noticeable. This is another way to demonstrate the inertia of
structures inherited from the past. In fact, even when the overall mortality
is yet rather low, progresses'do first benefit to infant, younger children and
Table 5. Total Size (in millions) and Relative Structures (per thousand
people) of the Chinese Population, 1986-2001, Selected Years.
Hypothesis of Decreasing Mortality.
Proportion of Age Groups
Total Young Adult Old
Year | Population (under 20) (20-59) (60 and over)
1986 1048 : 398 516 86
1989 1093 373 536 91
; 4992 1144 353 551 9,
1995 1197 343 557 100
“1998 1247 344 551 105
2001 1294 347 545 108 ;
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 329
older people. The impact of such a progress on the adult population is only
noticeable later, when the additional surviving children come in age.
Uncertainties under Control : Fertility Changes and Population Policies
A well-known objective of Chinese authorities is to reach as soon as possible
a zero population growth (ZPG). The present situation is rather favourable,
with a very low fertility by reference to Third World countries. But, as shown
above, population will still grow at the rate of one per cent per year at the
beginning of the 2ist Century, and later continue indefinitely to grow at a
lower pace.
In order to get a first idea of what would be possible, we have simulated some
rough scenarios of population policy, namely the "no third child" and "no se-
cond child" policies. For this present paper, we limited ourselves to simulate
an instantaneous change, i. e. the hypothesis where the policy is fully imple-
mented - and fully effective - starting in 1986 (it is of course technically
easy to simulate both a steady evolution over years and any proportion of
failure of the policy). That means that, from 1986, the simulated reproductive
behaviour of childless women is unchanged, as is the behaviour of mothers of
first parity in the "no third child" scenario; consequently, since no every
woman is married or mother, the synthetic indices of "number of children per
woman" would respectively amount to 1.7 and to 0.9. As a reference, let us say
that 1.7 children per woman is the approximate level of fertility in France in
the early eighties (one of the highest levels in Europe), and that 0.9 chil-
dren per women is the level reached in the mid-eighties by a few big European
cities like Hamburg or Torino.
We simulated also the impact of these changes under different hypotheses of
mortality. Table 6 presents only results under the hypothesis of decreasing
mortality used for Table 5 (to reach a life expectancy of 74 years by the year
2000 is a realistic assumption). In this Table, we show for the same selected
years what would be the total population of China if both scenarios are imple-
mented in 1986.
The "no third child" scenario leads the population to a sustained, while
lower, growth; the ZPG target would only by reached by the year 2023, with a
total population of 1.297 billion people.
The 2PG target will be reached in 2001 with the "no second child" scenario :
from 2002, we would observe a negative growth. Let us stress two conclusions
from this finding. First, the ZPG is indeed possible at dawn of the Century.
But the objective can only be met with untenable hypotheses : it seems some-
what irrealistic to believe in the possibility of a radical disappearance,
effective right now, of any birth among women having yet at least one child,
and of holding this constraint for years. The ZPG cannot be reached with less
stringent demands on the population, except if an important worsening in the
mortality conditions happens : but this is clearly neither desirable nor like-
ly. A second conclusion we lack of place for documenting at long is the long-
term consequences of such a fast change. Indeed, the growth rate will not be
stabilized at its zero level, but become increasingly negative. A true conti-
nued ZPG would impose to decision-makers to react in the other direction by
stimulating fertility at that time; however, due to the inertia of population
Phenomena, the trend would be reversed by coming back in the area of positive
380 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA
Table 6. Total Population (in millions) and Population Growth (per thousand),
China, 1986-2001, Selected Years. Hypothesis of Decreasing Mortality;
Two Scenarios of Decreasing Fertility.
Scenarios
"no third child" "no second child"
Total Population Total Population
Year Population Growth Population Growth
1986 1048 9.1 1048 2.3
1989 1078 10.3 1056 2.8
1992 1112 10.4 1064 2.5
1995 1146 9.1 1071 1.4
1998 1177 8.0 1075, 1.2
2001 1204 6.5 1077 0.6
Table 7. Relative Structures (per thousand), China, 1986-2001, Selected Years.
Hypothesis of Decreasing Mortality; Two Scenarios of Decreasing
Fertility.
Scenarios
"no third child" "no second child"
Age Groups Dependency Age Groups Dependency
Year| 0-19 20-59 60-+ Ratio 0-19 20-59 60-+ Ratio
1986) 398 516 86 |, «£939 398 516 86 «939
1989; 365 543 92 841 344 555 94 +803
1992) 334 567 99 ~ 764 305 592 103 688
1995; 314 581 105 +720 266 622 112 -608
1998) 305 584 144 741 239 640 121 563
2001) 297 586 117 +707 215 655 131 528
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 331
growth, and so on, generating oscillations more and more difficult to manage.
Table 7 compares the evolution of the relative structures of the Chinese popu-
lation under both scenarios. The reason of a further negative growth or oscil-
lations if the system has to be influenced again by policies are clear from
the dramatic fall in the proportion of youngsters, especially in the "no se~
cond child" scenario : this age group is only renewed (in proportion) by 54
per cent on this fifteen year period. That means that ageing adults will no
more be replaced by young adults, and the fathers and mothers of to-morrow
will be less numerous, bearing thus less children if fertility does not chan-
ge. At the same time, the increasing number of older people will produce a
greater number of deaths; the negative growth can thus only be more and more
important for years.
The structure of population at the time of reaching ZPG is noticeably diffe-
rent :
0-19 20-59 60-+ Dependency Ratio
"no third child" (ZPG in 2023) 231 591 178 693
"no second child" (ZPG in 2002) 206 660 134 ~515
The apparent paradox of getting an older population with a higher fertility is
another illustration of the inertia of demographix phenomena. Indeed, during
the twenty-one more years between 2002 and 2023, the faster growing population
has time enough for the ageing of the large birth cohorts before 1963, while
the decreasing fertility renews more slowly the stocks of adults. Moreover,
the youngsters of 2023 are mainly the offspring of smaller birth cohorts, born
in the eighties and nineties. On the whole, thus, the ZPG is reached with a
significantly less favourable dependency ratio due to the progressive deple-
tion of the adults stocks.
A Brief Look at Population and Schooling
These comments on relative structures are still more meaningful when the ab-
solute figures are considered. Table 8 shows the total number of young people
in the Chinese population according to both scenarios, contrasted with an
hypothesis of constant fertility at the 1978 level (i. e. 2.35 children per
woman). These figures differs slightly from Table 2 because they were computed
here under the hypothesis of decreasing mortality. From the point of view of
social investments, it is clear from Table 8 that the "no third child" scena-
rio is more favourable than the other two. Indeed, we can observe a steady
decrease in the number of youngsters, not’ as dramatic as in the "no second
child" scenario, and smooth by reference to the constant hypothesis.
These differences are meaningful for social organization : the planification
of schooling and the entry in the labour force is obviously made easier when
changes of size of the concerned populations are smooth, not too fast, and
oriented in a constant direction. For instance, such changes make possible to
plan improvements in quality or coverage of schooling, synchronized with the
evolution of populations.
For exemplifying this dynamic interaction between population and social
332 THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA,
Table 8. Population under 20 (in millions), China, 1986-2001, Selected Years;
Hypothesis of Decreasing Mortality: Three Scenarios of Fertility.
Fertility Hypothesis
(number of children per woman)
Year 2.35 1.7 0.9
1986 417.2 417.2 417.2
1989 408.1 393.1 370.6
1992 404.2 372.4 324.4
1995 410.7 360.1 285.0
1998 429.2 358.6 257.0
2001 448.5 357.9 231.0
poiicy, we simulated an arbitrary school system (but any other initial and
simulated conditions can be used). In this exercise, we hypothetized that 95
per cent of the children enter school at 6 and dropouts occur so that 80 per
cent of pupils complete their sixth year and 30 per cent their twelfth year,
and we consider only the first twelve years of school. The “improved school-
ing" scenario implies that every child is admitted at school, with respective
final rates of completion of six and twelve years fixed to 90% and 60%. Our
simulation also supposed that five years are needed for fully implementing the
total coverage of the population and the lower level of dropouts after six
years, while the improvement of the schooling between six and twelve years is
only achieved in fifteen years. The simulated reform is supposed to occur in
1986. Let us mention, from the technical point of view, that dropouts are
either simulated as an instant departure from school nor as a linear inter-
‘polation of proportions attending school, but by computing (in the model)
linearized conditional probabilities, following the life table technique.
able 9 presents the comparison between both scenarios for the case of a "no
third child" policy in a context of decreasing mortality. Under initial con-
ditions, the school population culminated to 211 millions pupils in 1981, then
rapidly declining until 1986. One can observe that school attendance continues
falling until the nineties; but in the "improved schooling" scenario, the
school population in 2001 is back to its level of 1985. Contrastingly, the
"constant schooling" scenario gives a loss above 16% of the total school popu-
lation in 2001 (the loss is above 26% by comparison to the peak.year 1981),
meaning some "over-equipment" of schools (by reference to the initial situa-
tion).
Comparable simulations are possible for manpower planning, independently or in
relation with alternate schooling policies, and, in this case, the qualitative
THE 1987 INTERNATIONAL CONFERENCE OF THE SYSTEM DYNAMICS SOCITY. CHINA 333
Table 9. School Population (in millions), China, 1986-2001, Selected Years;
Two Hypothetic Scenarios of Schooling under Decreasing Mortality and
Fertility Conditions.
Schooling
Year Constant Improved
1986 185.2 185.2
1989 166.7 171.5
1992 156.7 173.4
1995 150.8 174.8
1998 152.1 , 181.1
2001 : 155.5 187.5
composition of the labour force, as a consequence of the evolution in learn-
ing, can also be considered.
CONCLUSIONS
In this paper, we tried to show some applications of dynamic modelling useful
for decision-makers in global situations where classical demo-economic ap-
proaches are limited by an apparent lack of suitable data and by the complexi-
ty of many non-linear dynamic interrelations. As a matter of fact, population
and population-related problems are typically fit to a System Dynamics ap-
proach, due to their fundamental stocks-and-flows nature.
For the sake Gf demonstration, however, we limited. ourselves to a few simple,
or even simplistic, scenarios. It should nevertheless be stressed that our
presentation is far from exhausting the possibilities of the models, even with
the restricted size of micro-computers (for instance, another national model
we build has no less than 820 variables modifiables for getting a variety of
scenarios). Under this respect, System Dynamics modelling for population stu-
dies give a matchless flexibility and ease of use to researchers and planners.
REFERENCES
Lambert, A. (1981), TI opulatii ina, Cabay, Louvain-la-Neuve (Belgium)
coll. Working Papers of the Département de Démographie.
