Escolar Documentos
Profissional Documentos
Cultura Documentos
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=popcouncil.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
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.
Population Council is collaborating with JSTOR to digitize, preserve and extend access to Population and
Development Review.
http://www.jstor.org
Age Structure Dynamics
in Asia and Dependence
on Foreign Capital
MATTHEW
HIGGINS
JEFFREY G. WILLIAMSON
Does Asian history suggest that heavy child dependency burdens tend
to create heavy foreign capital dependency? As Asian countries aged and
their child dependency burdens fell, did dependence on foreign capital tend
to diminish? And will Asian foreign capital dependency diminish still fur-
ther as dependency burdens continue to fall in the next century?
Coale and Hoover showed in 1958 that the sharp decline in infant mortal-
ity initiated in the 1930s implied a sizable future increase in the Asian youth
dependency rate, especially if reinforced by the persistence of high birth
rates. They were, of course, correct. With only two precocious exceptions,
Sri Lanka (1955-59) and Japan (1950-54), Asia reached peak youth de-
pendency rates in the 1960s and 1970s.2 The modal country peaked in the
1960s. The evidence on changing Asian dependency rates in the second
half of this century is summarized in Table 1.
Elderly dependency rates have been comparatively stable in Asia dur-
ing this century. Japan, the region's demographic and economic leader,
began to see a rise in its share of persons aged 65 and older beginning in
the 1970s, while a rising trend emerged among the more advanced Asian
tigers-Hong Kong, Singapore, South Korea, Taiwan-by the 1980s. As
these countries enter the next century, however, the trend toward rising
elderly shares is expected to show a marked acceleration.
Youth dependency rates were much higher in developing Asia than
they were in the developed countries. While the OECD 'young" share-
those aged less than 15-averaged about 26 percent during the baby boom
in the 1950S,3 the peak rates in Asia were in many cases 20 percentage
points higher, two of the most extreme examples being from the area about
which Coale and Hoover were writing in 1958-Bangladesh and Pakistan
(both about 46 percent). At their respective peaks, the youth dependency
burden in Asia was far higher than in Europe and North America. It was
also far higher than in Japan at its peak (34.7 percent in 1950-54). Fur-
thermore, what limited data we have confirm that the increase in Asian
youth dependency rates has occurred largely in the second half of the twen-
tieth century, with youth dependency remaining fairly stable at high lev-
els prior to World War II (Higgins and Williamson 1996: Table 2). Asian
demographic history has been in dynamic transition ever since.
Asian dependency rates were not exceptional, however, when com-
pared to the experience of newly settled regions in the late nineteenth cen-
tury. Alan Taylor and one of the present authors (Taylor and Williamson
264 AGE STRUCTURE DYNAMICS IN ASIA
NOTE:'A to peak' refersto the percentagepoint changebetween the earliestdate,usually 1950-54, and the
peak.The figurein parenthesesrefersto changefrom 1931 or 1941 to the peak,where availablefromHiggins
and Williamson(1996:Table2). Theage sharesfor multiyearperiodsare averages,derivedfromquadrennial
datavia linearinterpolation.
SOURCE: UnitedNations(1992).
MATTHEW HIGGINS / JEFFREY G. WILLIAMSON 265
1994: 352-353) document very high youth dependency rates in the coun-
tries of the New World when they started modern economic growth; for
example, 56 percent for Canada in 1851 and 45.2 percent for Argentina in
1869.4 Furthermore, those youth dependency burdens correlate well with
heavy foreign capital dependency in both cases. The capital flows from ma-
ture Europe to the adolescent New World were enormous, with Argen-
tina, Australia, and Canada largely dependent on foreign capital to fulfill
their accumulation requirements. The same seems to have been true of
Asia. Is the correlation spurious?
Table 1 also documents the large decline in the youth dependency
rate in Asia from its peak. Some of the declines from peak have been spec-
tacular. The biggest have occurred among the old and new tigers in East
and Southeast Asia: for example, South Korea, -18.0 percentage points;
Japan, -16.7. The smallest declines have occurred among the economic
slow growers of South Asia: for example, India, -4.1 percentage points;
Pakistan, -0.4. With the exception of Sri Lanka, the intermediate cases are
all in Southeast Asia. This decline in youth dependency was compressed
within two or three decades, only half the time it took most late-nine-
teenth-century industrializing countries to record less-spectacular reduc-
tions (Taylor and Williamson 1994: Appendix Table Al). Even by the stan-
dards of history, then, Asian dependency rate changes in the late twentieth
century have been revolutionary.
How does Asia look today? Table 2 reports the evidence for 1990-92. This
time the figures are given for the prime labor force age group (those aged
25-59).
The range across Asia is very large, almost 21 percentage points be-
tween rich Singapore and poor Bangladesh. The rich city-states of Hong
Kong and Singapore are in the economically most favorable situation, with
just over half of their populations in prime ages. Japan, South Korea, and
Taiwan are close behind. At the other extreme, Pakistan and Bangladesh
are the least favored, with less than a third of their populations in prime
ages. Nepal and the Philippines slightly exceed one-third, and the remain-
ing countries fall somewhere in between. This wide variation in Asian de-
pendency burdens may have inhibited economic convergence in the region.
Equally important, there has been no dependency rate convergence
in Asia over the past three decades. Compared with the richer countries in
East Asia, the poorer countries in South Asia already had higher depen-
dency burdens and lower prime age shares in the 1960s, but the demo-
graphic distance between the two regions grew even wider by the 1990s.
266 AGE STRUCTURE DYNAMICS IN ASIA
NOTE: 'A 2025-1990" refers to the percentage point change between 2025 and 1990-92 in the percent aged
25-59 years.
SOURCE:United Nations (1991).
East Asia
1950-54 18.0
1955-59 13.9 18.8 -4.9
1960-64 18.3 23.5 -4.6
1965-69 24.0 25.1 -3.4
1970-74 29.0 28.5 -1.4
1975-79 29.7 30.0 -0.7
1980-84 28.6 29.1 0.1
1985-89 34.0 28.3 5.3
1990-92 35.0 30.8 2.4
South Asia
1950-54 15.2 11.0 3.1
1955-59 10.4 12.8 -3.2
1960-64 9.1 14.0 -6.8
1965-69 9.5 13.4 -5.9
1970-74 8.4 12.8 -4.4
1975-79 11.0 16.4 -5.4
1980-84 10.7 20.4 -9.8
1985-89 10.0 18.8 -8.2
1990-92 9.5 19.7 -8.6
Southeast Asia
1950-54 13.2 13.9 -2.0
1955-59 15.5 14.1 1.4
1960-64 15.4 15.6 -0.3
1965-69 16.2 19.0 -2.8
1970-74 22.3 26.5 -4.2
1975-79 26.8 29.7 -2.9
1980-84 27.2 32.4 -5.3
1985-89 28.3 28.2 0.1
1990-92 32.6 34.6 -2.0
NOTES: All entries are unweighted country averages. Accordingly, the identity (savings -
investment = current account balance) is only approximate. For population-weighted
averages, see Higgins and Williamson (1996: Table 5).
the current account balance (CAB) or, equivalently, the gap, country by
country, between domestic savings and domestic investment. Negative num-
bers imply net inflows, that is to say, foreign capital dependency.
These unweighted regional averages show that the investment rate
has risen markedly since the early 1950s, gaining an impressive 12.8 per-
268 AGE STRUCTURE DYNAMICS IN ASIA
dependency during the postwar era. First, the variable rate-of-growth ef-
fect model describes only the steady-state relationship between dependency
and saving rates-a shortcoming derived from its life-cycle ancestry.7 Yet
the rapid pace of demographic change in Asia over the past half-century
surely supports the presumption that the observed surge in the saving rate
since the 1960s reflects out-of-steady-state behavior. Second, the variable
rate-of-growth effect model focuses exclusively on the link between de-
pendency and saving rates, ignoring the determinants of investment de-
mand (but see Auerbach and Kotlikoff 1992). Yet saving is determined in-
dependently of investment only under perfect capital mobility, that is, in a
small, open economy facing an exogenous world interest rate. In any other
setting, the observed saving rate depends on both domestic savings supply
and investment demand (e.g., Feldstein and Horioka 1980; Obstfeld 1986;
Feldstein and Bacchetta 1991; Frankel 1991). By abstracting from demo-
graphic influences on investment demand, the variable rate-of-growth ef-
fect model provides no guidance concerning the effects of demographic
change on the residual of greatest interest in this article, net capital flows.
To address these issues requires a dynamic model, one capable of in-
corporating demographic effects on both savings supply and investment
demand. Fortunately, the variable rate-of-growth effect model is easily ex-
tended to accommodate these requirements. Higgins and Williamson (1996)
show that the variable rate-of-growth effect model can be subsumed un-
der the textbook neoclassical growth model (e.g., Blanchard and Fischer
1988), inhabited by an overlapping generations population: the former
model is, in essence, the open-economy, steady-state version of the latter.
In the extended model (outlined in the Appendix), saving remains moti-
vated by life-cycle considerations, while investment demand is driven by
the need to equip a growing labor force. For an open economy, the gap
between savings supply and investment demand is reflected in net capital
flows (the current account balance); for a closed economy, the balance be-
tween these forces determines the equilibrium (i.e., observed) saving rate.
We rely on simulations to study the economy's dynamic evolution in
response to empirically plausible patterns of demographic change. (Ana-
lytical solutions are generally unavailable, even given simple functional
forms.) A simulated "demographic transition,' in which fertility rises for
several generations before falling to a new, lower steady-state level, repre-
sents an attractive experiment, because such a pattern roughly describes
the demographic history of many Asian countries over the postwar era.
The simulation results highlight an important qualitative implication
of standard growth models: the demographic 'center of gravity' for invest-
ment demand should be earlier in the age distribution than that for sav-
ings supply. In particular, investment demand should be most closely re-
lated to the share of young (through its connection with labor force growth),
MATTHEW HIGGINS / JEFFREY G. WILLIAMSON 271
while savings supply should be most closely related to share of mature adults
(through its connection with retirement needs). The divergence between
these centers of gravity implies that the effects of demographic change on
savings and net capital flows will depend on the economy's degree of open-
ness to capital flows. For a financially open economy, a shift in the popula-
tion age distribution toward younger ages should produce a tendency to-
ward current account deficits: saving falls due to the increased dependency
burden, even as investment rises in response to higher labor force growth.
As the age distribution shifts toward the center, savings supply should in-
crease, even as investment demand slackens, pushing the current account
into surplus. In contrast, for a closed economy, the younger center of gravity
for investment would cause the observed (i.e., equilibrium)saving rate to ap-
pear positively related to relativelyyoung aspects of the age distribution.
It is not immediately clear whether the open- or closed-economy as-
sumption provides the better guide in analyzing the observed dynamics of
savings, investment, and capital flows. The fact that many Asian countries
have, since the 1950s, relied so heavily on capital imports might appear to
create a strong prima facie case against the closed-economy assumption.
The real issue, however, is whether investment is constrained by domestic
savings (Williamson 1993). If the economy faces a binding constraint on
capital inflows, equilibrium in the domestic capital market will depend on
both local savings supply and investment demand, and demographic de-
pendency will alter the market outcome in a way qualitatively similar to a
closed economy. For example, an increase in fertility under such condi-
tions would lower the supply of savings at a given interest rate, leading to
an equal decline in investment, leaving the volume of capital inflows un-
changed. Under weaker restrictions on foreign mobility, dependency rate
changes will affect equilibrium in the domestic capital market in a way
that would display features of both the closed- and open-economy models.
Suppose, for example, that potential borrowers must pay a risk premium
in order to attract foreign capital, and that the premium rises with the de-
sired inflow. A dependency-rate-induced fall in the domestic supply of sav-
ings at a given interest rate would then be only partly offset by increased
capital inflows, forcing up the interest rate and causing a decline in invest-
ment. A positive link between youth dependency and capital dependency
would be observed, but it would be weaker than the one generated under
perfect world capital markets.
Although the model described here represents a generalization of the
existing macroeconomic models of the savings/demography nexus, it re-
mains subject to several obvious limitations. Because parents are treated as
having no bequest motive, the model abstracts from the effects of demo-
graphic change on the transmission of wealth across generations. Fertility
itself is treated as exogenous, so that the model abstracts from the chang-
272 AGE STRUCTURE DYNAMICS IN ASIA
ing costs and benefits of childrearing over the course of development. (We
return to this topic in the concluding section. The exogeneity assumption
implies, of course, that we sidestep the issue of whether childrearingin poor
countries is motivated partly by the need for old-age support.) Investment is
driven by demand for social overhead and residentialas well as business capi-
tal, and thus the model may be unduly reductionistin emphasizing the con-
nection between investment and laborforce growth. We are aware of no trac-
table model that accommodatesthese complexities,but they may be important
nonetheless. Our empirical analysis (in common with previous cross-coun-
try studies) should be regardedas capturing the effect of demographic forces
driving accumulation, rather than as a test of a particular causal channel.
Empirical results
NOTES:
1 See Appendixforvariabledefinitions,variablesused as instruments,data
sources,and detailsof econometricmethod.
2 Methodof estimation:three-stageleast squares(3SLS).
3 Totalnumberof observations= 916. System-weightedR2= .915.
4 Estimatesfor constanttermand dummyvariablesnot reported.
5 Absolutet-statisticsare in parentheses.The F-statisticsat the bottomof the
tableindicatethe joint significanceof the two demographicvariables.
6 Testfor equalityof the demographicvariablesacrossthe two equations: F2,,.
= 4.43, with marginalsignificancelevel of .012. Thistest is equivalentto a test
of the significanceof the demographicvariablesin explainingmovementsin
the currentaccount.
2.5
-2.5 Xav'mgs
-5 / ~~~~~~~~~~~~~Investment
\
-5
-7.5
-10
0-4 10-14 20-25 30-35 40-44 50-54 60-64 70+
Age
-1
-2
0-4 10-14 20-25 30-35 40-44 50-54 60-64 70+
Age
midlife. But they reach a peak rather early in Asia, at ages 35-39, declin-
ing sharply thereafter so as to become negative by ages 55-59. Yet, the
rather young "center of gravity" found for the saving rate is what might
have been expected if Asia has been only imperfectly integrated into the
world capital market.'0
The implicit age distribution coefficients for the investment equation
appear at first glance to be quite similar to those for savings. To bring the
differences into relief, the bottom half of Figure 1 plots the implicit age
distribution coefficients for the current account balance. The coefficients
are clearly negative for the early portion of life (up to age 39), indicating
that the young-adult-induced increase in investment demand (transmit-
ted via both employment and infrastructure needs) outweighs its induced
increase in savings supply. This implies that relatively young countries pass
through a long period of foreign capital dependency which includes peri-
ods of child, adolescent, and young adult gluts. The coefficients turn posi-
tive after age 40 as the induced fall in investment demand outstrips the
induced fall in savings." According to these estimates, relatively young
countries (like those in Asia) are net capital importers and relatively old
countries are net capital exporters: if global capital markets let it happen,
capital in the late twentieth century tends to move between countries in
the manner of an intergenerational transfer.
The results just described will be used in the rest of the article to construct
estimates of the impact of the changing dependency burden on Asian sav-
ings, investment, and net capital dependency. The estimates refer to de-
mographically induced deviations of, say, a country's saving rate from the
country average for the full sample period. We also construct projections
of the effects of expected changes in the dependency burden on savings,
investment, and capital flows, using the UN's estimates of population age
shares for 2010 and 2025. The same procedure, of course, is followed for
investment and the current account balance.
South Asia
1950-54 1.79 1.80 -0.02
1955-59 0.97 1.14 -0.17
1960-64 -0.21 0.09 -0.30
1965-69 -1.07 -0.73 -0.34
1970-74 -1.51 -1.21 -0.30
1975-79 -1.16 -1.05 -0.11
1980-84 -0.49 -0.61 0.12
1985-89 0.20 -0.16 0.36
1990-92 1.49 0.73 0.76
Southeast Asia
1950-54 -0.14 -0.53 0.39
1955-59 -1.32 -0.98 -0.34
1960-64 -2.54 -1.68 -0.88
1965-69 -3.41 -2.28 -1.13
1970-74 -3.64 -2.42 -1.21
1975-79 -2.00 -1.32 -0.68
1980-84 0.88 0.63 0.25
1985-89 4.27 3.05 1.22
1990-92 7.90 5.54 2.36
the risingdependency rate served by itself to lower the saving rate in EastAsia
by 6.1 percentage points, to lower it by 3.5 percentage points in Southeast
Asia, and to lower it by 3.3 percentage points in South Asia.
Although the dependency rate literature rarely notes it, these demo-
graphic variables also had their impact on investment demand, and in the
278 AGE STRUCTURE DYNAMICS IN ASIA
for both countries dependency effects alone would have produced a small
change in the opposite direction. Sri Lanka experienced a marked rise in
foreign capital dependency between 1950-54 and 1990-92 (perhaps due
to the depressing effects on saving of civil strife and the attendant military
expenditure), in spite of improved dependency rate conditions. Only for
Bangladesh do demographic forces appear to have dominated, with the
rising youth dependency burden accounting for the bulk of the 2.9 per-
centage point decline in the CAB balance between 1970-74 and 1990-92.
The poor countries of South Asia have carried greater dependency bur-
dens, implying higher net domestic savings shortfalls and greater foreign
capital dependency. The best way to illustrate that fact is by recourse to
the counterfactual. Suppose the fast-transition countries of East Asia had
carried the heavier demographic burdens of South Asia.'2 And suppose
South Asia had carried the lighter demographic burdens of the fast-transi-
tion countries of East Asia.
The results of these thought experiments appear in Table 6 for three
points in time: early in the demographic transition, 1955-59; the middle,
1975-79; and toward the end, 1990-92. The table first reports the share of
the current account balance in GDP that was recorded under the actual
dependency burden. The table also reports the share that would have pre-
vailed had the country carried the dependency burden which character-
ized other parts of Asia ('counterfactual'), and the difference between the
two. (The shares are calculated following equation (6) in the Appendix.)
Even though national accounts data may make it impossible to document
the actual share for some periods and countries, the difference (col. 3) can
always be calculated given the estimated within-sample beta coefficients
and the dependency burdens. Panel A asks how East Asia would have be-
haved given South Asian demographics. Panel B turns the question around,
asking how South Asia would have behaved given the demographics of
East Asia.
To illustrate the first question, consider Panel B in Table 6: while Pa-
kistan was a significant net capital importer in 1990-92 (-6.9 percent of
GDP), it would have been a net capital exporter (+3.9 of GDP) had it been
favored with the lighter demographic burden of the fast-transition coun-
tries during the same period. The switch between the actual and counter-
factual demographic regimes implies a 10.8 percentage point change. This
figure is huge: if the dependency rate can really be taken as exogenous,
and if the coefficients estimated above are even close to the mark, it sug-
gests that much of South Asia's heavy dependence on foreign capital has
been due to the fact that the dependency burden has been far higher there
than in East Asian countries. And the 1990-92 period is no outlier:
MATTHEW HIGGINS / JEFFREY G. WILLIAMSON 281
Bangladesh Myanmar
1990-92 -8.22 1985-89 -3.39*
2005 -6.56 2005 -1.26
2015 -3.11 2015 +1.35
2025 +1.40 2025 +4.80
China Pakistan
1990-92 +2.29 1990-92 -6.92
2005 +5.93 2005 -4.67
2015 +9.15 2015 -1.22
2025 +11.21 2025 +3.02
India Philippines
1985-89 -3.81* 1990-92 -2.33
2005 -1.51 2005 +1.28
2015 +1.08 2015 +4.67
2025 +4.05 2025 +8.18
Indonesia Singapore
1990-92 -3.45 1990-92 +9.61
2005 +0.93 2005 +14.93
2015 +4.03 2015 +18.04
2025 +6.97 2025 +19.13
Japan Sri Lanka
1990-92 +2.46 1990-92 -10.74
2005 +4.36 2005 -6.33
2015 +4.81 2015 -3.57
2025 +4.51 2025 -1.61
Korea Thailand
1985-89 +3.18* 1990-92 -7.66
2005 +8.82 2005 -2.31
2015 +11.68 2015 +1.07
2025 +13.86 2025 +4.02
Malaysia
1990-92 -6.20
2005 -1.67
2015 +2.18
2025 +5.10
NOTES:Thefiguresdenotedby an asteriskmean that the CABsharefor 1990-92 cannotbe documentedin our source;we use
1985-89 instead.The CABsharefor Nepalis not even availablefor 1985-89, and thus was excludedfromthis table.HongKong
and Taiwanare not includedin these projections.
SOURCE: See discussionin text.
to the impact of future demographic events. But some loud and interesting
suggestions emerge from the table.
284 AGE STRUCTURE DYNAMICS IN ASIA
Population aging in Japan will lower the saving rate, of course, but it
will also lower the investment rate, and by a bit more. Thus, if demographic
forces are permitted to have their way, Japan will not become a nation of
old people unable to find the resources to export capital to the poorer parts
of Asia. Actually, it looks as though the CAB share will rise by about 2
percentage points over its level in 1990-92. But what distinguishes Japan
is that the rise is so modest: it is likely to be dramatic elsewhere in Asia.
South Asia will graduate from its current heavy dependency on for-
eign capital to complete independence by 2025. The CAB in Bangladesh
will switch from -8.2 percent of GDP in 1990-92 to +1.4 percent in 2025,
a total change of 9.6 percentage points, almost identical with the impact of
the fall in the dependency rate recorded by Japan between 1950-54 and
1990-92. India will undergo a similar trend, although the total impact will
not be as great (a 7.9 percentage point drop in the CAB share). Pakistan
and Sri Lanka will closely replicate the case of Bangladesh. The experience
in East Asia will be similar (Japan aside), with the CAB share rising by
about 10 percentage points for China and Korea.
Nevertheless, the movement toward net capital export positions will
be most pronounced in Southeast Asia. Between 1990-92 and 2025, the
CAB share will rise by 10.4 percentage points in Indonesia, by 11.3 in Ma-
laysia, by 10.5 in the Philippines,by 9.5 in Singapore,and by 11.7 in Thailand.
If demographic forces are allowed to have their way, capital flows over
Asian borders will be very different three decades from now.
burden and its persistence or decline. As to the future, aging will not di-
minish Japan's capacity to export capital in the next century, but little of it
will go to the rest of Asia, at least if dependency rates are allowed to have
their way. Rather, the rest of Asia will switch to capital export positions,
even in South Asia.
We have made numerous assumptions along the way: that world sav-
ings supplies do not alter incentives for capital flows over Asian borders;
that world capital markets are relatively open; and that the demographic
transition is exogenous to accumulation performance. We have discussed
the world capital market assumptions at length, but have only alluded to
the exogeneity of the demographic transition.
We have taken dependency rates as exogenous, as if the transition
can be viewed solely as a response to the import of foreign health technol-
ogy in the 1950s, a shock that sharply lowers mortality, especially infant
mortality, over a decade or two, thus inducing a half-century evolution in
the age structure and in dependency rates. But that is much too simple, of
course. With a lag, fertilityrates have also fallen, and these have been driven,
among other things, by economic success (and their fall limited by eco-
nomic failure). Additional reductions in infant mortality have also been
assured by economic success (or limited by economic failure). This article
ignores this important connection, and this may affect our interpretation
of the past.
In making projections about the future, we acknowledged that our
assumptions about integrated world capital markets may be made irrel-
evant by a retreat back to pre-1970 autarky-if not by all of Asia, then by
some important parts like China. Furthermore, we have ignored the possi-
bility of any change in global capital supply that might influence incen-
tives for capital flows over Asian borders.
Subject to these important qualifications, we conclude that the de-
pendency rate hypothesis is alive and well in Asia. It has played an important
role in the past half-century and probably will do the same over the next.
Appendix
Themodel in brief
The model allows for three periodsof life-youth, the prime ages, and old age.
The populationat time t consistsof N,, prime-ageadults in the labor force, N2,
retiredelderly,and No, dependentyoung. The populationof dependentyoung is
determinedby No',= nN,N1,that is, each prime-ageadulthas n, survivingoffspring.
We assumethat prime-ageadultscareaboutthe welfareof their dependent
offspring,but that elderly adults have no bequest motive. Prime-ageadults are
286 AGE STRUCTURE DYNAMICS IN ASIA
endowed with one unit of time, which is inelastically supplied to the labor force.
Labor income is divided among current consumption, child support, and savings
for old age. The lifetime budget constraint of a representative prime-age adult at
time t can then be written as:
C2,t + +n
Wt = Cl,t1 + r + ntco, (1)
where C1,and C2t+l refer, respectively, to consumption during the prime years and
retirement, CO,is consumption per dependent offspring, and rt+,is the interest rate
on assets acquired at t and held until t+1.
Preferences are described by an additively separable utility function of the
form:
the steady-state investment rate, with i'(n*) = g k*lf(k*). It can be shown that this
investment-promoting effect dominates even if saving rises, at least in our model,
so an increase in fertility always lowers the steady-state CAB (Higgins and
Williamson 1996).
For the closed economy, savings supply and investment demand must be
equal. Using this equilibrium condition, we can show that higher fertility brings
an increase in the saving rate when:
where XL is labor's income share and ekn is the elasticity of steady-state capital
intensity with respect to population growth. This elasticity is negative for 1/0 2 1
(that is, as long as higher interest rates do not lower savings), so that higher fertil-
ity lowers the steady-state saving rate. The model thus allows for the possibility
that the saving (and investment) rate may fall with an increase in fertility, thus
generalizing the basic insight of the variable rate-of-growth effect model to a closed-
economy setting.
sit = P + PS + P5RPIt
+p2Z.t+ p3GROWTH.t+p4Z,tGROWTH.t +u (4)
where si, is the national savings or investment rate in country i at year t, Z, refers
to a vector of demographic variables, GROWTH, is the growth rate of national in-
come, RPI,t represents the relative price of investment goods, and ui, is a random
disturbance term.
Z,. is constructed by using a quadratic polynomial to represent 15 popula-
tion age shares: 0-4, 5-9,...,65-69, and 70+. This technique for incorporating
demographic information into macroeconomic equations was introduced by Fair
and Dominguez (1991). The technique has the advantage of capturing the infor-
mation contained in the entire age distribution while maintaining a parsimonious
parameterization (see Higgins and Williamson 1996: Appendix, for additional de-
tails). In particular, it allows us to derive implicit estimates for all 15 age shares
while estimating only two underlying parameters.14
The estimated models are of the "fixed-effects"variety: the intercept term is
allowed to vary across countries, but the slope coefficients are treated as common
to all countries. A lagged dependent variable (LDV) specification was chosen in
order to control for temporal persistence in national savings or investment rates.
The models are estimated using instrumental variables techniques to control
for the fact that some of our explanatory variables, particularly income growth
and the relative price of investment goods, should themselves be endogenous. Two-
stage least squares (2SLS) is used to construct initial estimates for national sav-
ings, investment, and the current balance. The 2SLS estimates provide the basis
for various specification tests, reported fully in Higgins and Williamson (1996)."
288 AGE STRUCTURE DYNAMICS IN ASIA
We then rely on 3SLS to construct our benchmark estimates, treating the savings
and investment equations as elements of a simultaneous system.'6 This procedure
implicitly yields estimates of demographic effects on capital flows via the identity:
National Savings = Investment + Current Account Balance. The statistical signifi-
cance of these demographic effects can be tested by imposing the restriction that
the coefficients of Zi and Z2 are equal for the savings and investment equations.
The national savings, investment, and current account shares are based on
national accounts data taken from International Financial Statistics (International
Monetary Fund 1995). The relative price of investment goods (RPI)and the growth
rate of aggregate GDP (GROWTH)are derived from the Penn World Tables, Mark 5.6
(Heston and Summers 1995). Demographic data are from the United Nations
(1992). Taiwan represents the sole exception to the above, with national accounts
and demographic data derived from the Statistical Yearbook of the Republic of China
(Republic of China, various years). Annual population age shares are derived from
quadrennial data via linear interpolation. As for variables used as instruments,
real output per worker and per capita are measured at purchasing power parity
from the Penn World Tables. The Penn World Tables is also the source of data on
labor force growth and the purchasing power parity for GDP. Exports plus imports
as a share of GDP is based on national accounts data.
The current account balance (CAB) is measured as the sum of the trade bal-
ance and net factor income. National savings is measured by adding gross invest-
ment to this total. Our measures of both national savings and current account
balance exclude private and official transfers and, thus, are not strictly correct.
However, these definitions were chosen because data concerning transfers come
from balance-of-payments sources generally unavailable before 1970. The differ-
ence in measured savings or current account shares is almost always quite small,
and appears not to affect our results.
The demographic effect on the saving rate for country i at time t is calculated as:
where PZ3and PZ2 are the estimated coefficients for Zi and Z2, ZTIand Z27are the
country i averages for these variables, and IJDvisthe estimated coefficient of the
lagged dependent variable. To capture the long-run effect of changes in the demo-
graphic variables, 1 -PLDVappears in the denominator.
Similarly, the effect of changes in country i's age structure on its saving rate
between, say, 1990 and 2025 is calculated as:
Because these calculations refer to the long-run or total effects, they may
slightly overstate the effects of changes in the demographic variables be-
tween any two given years.
MATTHEW HIGGINS / JEFFREY G. WILLIAMSON 289
Notes
The authors acknowledge the research assis- growth rate, and thus does not allow one to
tance of Ben Dennis and Suny Lay and help- determine whether the (negative) depen-
ful comments by David Bloom, Allen Kelley, dency or (positive) rate-of-growth effect on
Peter Lindert, Andrew Mason, Jonathan savings is the stronger.
Morduch, Alan Taylor, Peter Timmer, and 6 Continuing skepticism is illustrated by
participants at the Conference on East and Hammer (1986) and Kelley (1988). But see
Southeast Asian Economic Change in the also the recent supporting evidence reported
Long Run, Honolulu, Hawaii (11 April in Mason (1988); Masson (1990); Webb and
1996), where an earlier version of this pa- Zia (1990); Williamson (1993); Kang (1994);
per was presented. The views expressed in Taylor (1995); Kelley and Schmidt (1995,
this article are those of the authors and do 1996); Higgins and Williamson (1996);
not necessarily reflect those of the Federal Higgins (1997); and Lee, Mason, and Miller
Reserve Bank of New York or the Federal
(1997).
Reserve System.
7 The variable rate-of-growth effect
1 The term dependency should be inter- model assumes steady-state population and
preted in this artide as "reliance."We do not productivity growth (Mason 1987). Rather
intend to imply that reliance on foreign capi- inconsistently, the youth dependency rate is
tal as a source of investment funds need have allowed to vary, although it is simply an in-
negative welfare consequences. Indeed, in the creasing function of the steady-state popu-
simple theoreticalmodel outlined below, equi- lation growth. We show below that the
librium net capital flows are welfare-improv- model's qualitative implications remain un-
ing for both borrowing and lending nations. changed when the correspondence between
2 We define the youth dependency rate population growth and youth dependency is
as the share of the population aged 14 years acknowledged.
and younger within the total population. 8 Initial specificationsincluded the prod-
Some other work defines the youth depen- ucts of the age-distribution variables and na-
dency rate as the share 14 and younger di- tional income growth (that is, GROWTH,-
vided by the share 15 and older. Zz,).
However, these interactive terms proved to
3 We refer to 18 OECD members, ex- be statistically insignificant at the 10 percent
cluding Australia and Japan: Mitchell (1978, level for the saving, investment, and CAB
1983, 1992). equations. Although the variable rate-of-
4 Fogel (1991) shows that these large growth effect model plausibly implies that in-
mid-century New World dependency rates teractive terms should be important, the
were driven by labor supply responses in la- finding that they are not such in late-twen-
bor-scarce areas of recent settlement-early tieth-century Asia supports similar evidence
marriage, many births within marriage, and from other times and places (Taylor and
high infant survival in an environment of Wiliamson 1994; Taylor 1995; Higgins 1997).
good-quality nutrition. In contrast, the Asian 9 These basic results appear insensitive
experience in this century was driven by ex- to plausible modifications in the empirical
ogenous declines in mortality, especially in- specification. For example, the estimated de-
fant and early childhood mortality, associ- mographic effects are qualitatively similar
ated with the diffusion of modern health and remain statistically significant if we rely
technologies from center to periphery. Yet on an AR(1) specificationto deal with the is-
the result was the same: high youth depen- sue of temporal persistence. Adding a time
dency rates. trend or year-specificdummy variablesto the
5 Tobin (1967) also describes a second model also has little effect on the scale or mag-
model that allows for dependency effects on nitude of the estimated demographiceffects.
savings. As observed by Mason (1988), how- 10 Savings supply and investment de-
ever, this model assumes a fixed population mand are separately identified in the empiri-
290 AGE STRUCTURE DYNAMICS IN ASIA
cal models developed here only to the ex- mial specification against one in which the
tent that countriescan borrowand lend on the age shares are allowed to enter in an unre-
internationalcapitalmarketwithout constraint stricted manner. Simple Chow tests, applied
and at a given world interest rate. In the ab- to each equation separately, indicate that a
sence of perfect capitalmobility, the estimates third or higher-order polynomial is not re-
for savings will reflect a mix of the separate quired to capture the relationship between
demographic influences on both savings and accumulation and age structure.
investment-a lesson made clear by the 15 We conduct specification tests to ad-
Higgins-Williamson(1996) simulationmodel. dress the following issues. Serial correlation:
In this setting, an increase in the share of Durbin's h-test for serial correlation in an
young adults, who presumably save little, LDVspecification is applied to each equation
might lead to an increase in the equilibrium separately. The presence of serial correlation
quantityof savingsby causingan outwardshift is rejected at the 5 percent level for each
in the investment demand schedule. Similarly, equation. Overidentifying Hansen's
restrictions:
an increase in the share of the middle-aged J-test is applied to each equation separately,
might actually reduce savings if any outward using 2SLS. For each equation, we fail to re-
shift in savings supply is more than offset by ject at the 10 percent level the null hypoth-
an inward shift in investment demand. esis that the six overidentifying restrictions
11 Caution is warranted regarding the are valid (see Davidson and MacKinnon
CAB coefficents corresponding to elderly de- 1993: 232-237). Consistency of OLS: The
pendency. Because there has been relatively Durbin-Wu-Hausman test is applied to each
little variation in our Asian sample in the eld- equation separately, using 2SLS. The null
erly dependency share, these coefficients are hypothesis that OLS is consistent is rejected
likely to be less precisely estimated than at the one percent level for the investment
those corresponding to the youth or prime- equation and CAB equations, and narrowly
age shares. escapes rejection at the 10 percent level for
12 We omit Japanfrom the fast-transition the savings equation(see Davidson and
group in order to focus the counterfactualex- MacKinnon 1993: 237-242). Fixed-effectsvs.
ercise on countries that attained their peak random-effectsand common-intercept specifica-
youth dependency rates in the early 1960s or tions:Hausman's test for the consistency of
later. Thus, the group includes China, Hong the random-effects estimator is applied to
Kong, Korea, Singapore,and Taiwan. each equation separately. We reject the con-
sistency of the random-effects specification
13 For simplicity, we abstract from de- at the one percent level for the savings and
preciation. The more general expression investment equations (see Greene 1993:
would be: f'(k*) = r* + 8. Introducing depre- 479-480). Simple Chow tests reject the null
ciation, as is done in the simulation experi- alternative of a common intercept for all
ment reported in Higgins and Williamson countries at the one percent level for both
(1996), does not affect the model's qualita- equations.
tive implications.
16 Because the initial 2SLS estimates
14 This representation of demographic produced nearly identical coefficients for the
effects imposes some structure on the data, lagged dependent variable, this term was
requiring the relationship between savings constrained to be equal in generating the
(or investment) and the population age 3SLS estimates. Relaxing this constraint has
shares to vary smoothly and to exhibit a no appreciable effect on the coefficient esti-
single global maximum or minimum. We are mates for the other variables.
unable to reject these restrictions at even the
25 percent level when testing the polyno-
MATTHEW HIGGINS / JEPPREY G. WILLIAMSON 291
References
Auerbach, Alan and Laurence Kotlikoff. 1992. "The impact of the demographic transition
on capital formation," Scandinavian Journal of Economics 94, no. 2: 281-295.
Barro, Robert. 1992. "World interest rates and investment," Scandinavian Journal of Econom-
ics 94, no. 2: 323-342.
Barro, R. and J. Lee. 1994. Data Set for a Panel of 138 Countries. Available on diskette from R.
Barro, HarvardUniversity, or via FTPfrom the NBER.
Blanchard, Oliver and Stanley Fischer. 1988. Lectureson Macroeconomics.
Cambridge, MA:
MITPress.
Borsch-Supan, Axel. 1996. 'The impact of population ageing on savings, investment and
growth in the OECD area," in Future Global Capital Shortages: Real Threat or Pure Fiction?
Paris: OECD.
Cass, David. 1965. 'Optimum growth in an aggregative model of capital accumulation,"
Review of Economic Studies 32: 233-240.
Coale, Ansley J. and Edgar M. Hoover. 1958. Population Growth and Economic Development in
Low-IncomeCountries.Princeton: Princeton University Press.
Collins, Susan. 1991. 'Saving behavior in ten developing countries," in Douglas Bernheim
and John Shoven (eds.), National Saving and Economic Performance, pp. 349-373. Chi-
cago: University of Chicago Press.
Cutler, David, James Poterba, Louise Scheiner, and Lawrence Summers. 1990. 'An aging
population: Opportunity or challenge?" Brookings Papers on Economic Activity, No. 1.
Davidson, Russell and James MacKinnon. 1993. Estimation and Inference in Econometrics. New
York: Oxford University Press.
Deaton, Angus and Christina Paxson. 1995. 'Saving, inequality and aging: An East Asia
perspective," Asia-Pacific Economic Review 1, no. 1: 7-19.
Edelstein, Michael. 1982. OverseasInvestment in the Age of High Imperialism. London: Muthuen.
Fair, Ray and Kathryn Dominguez. 1991. "Effectsof the changing U.S. age distribution on
macroeconomic equations," American Economic Review 81, no. 5: 1276-1294.
Feldstein, Martin and C. Horioka. 1980. 'Domestic saving and international capital flows,
Economic Journal 90 (June): 314-329.
Feldstein, Martin and P. Bacchetta. 1991. 'National saving and international investment,"
in Douglas Bernheim and John Shoven (eds.), National Saving and Economic Performance,
pp. 201-224. Chicago: University of Chicago Press.
Fogel, Robert. 1991. 'The conquest of high mortality and hunger in Europe and America:
Timing and mechanism," in PatriceHigonnet, David Landes, and Henry Rosovsky (eds.),
Favoritesof Fortune,pp. 33-72. Cambridge, MA: HarvardUniversity Press.
Frankel, Jeffrey. 1991. 'Quantifying international capital mobility in the 1980's," in Dou-
glas Bernheim and John Shoven (eds.), National Saving and Economic Performance, pp.
227-265. Chicago: University of Chicago Press.
Fry, Maxwell and Andrew Mason. 1982. 'The variable rate-of-growth effect in the life-
cycle saving model: Children, capital inflows, interest and growth in a new specifica-
tion of the life-cycle model applied to seven Asian developing countries," Economic
Inquiry 20 (July): 426-442.
Goldberger,Arthur. 1973. 'Dependency rates and savings rates: Further comment," Ameri-
can Economic Review 63, no. 1: 232-233.
Greene, William. 1993. EconometricAnalysis. New York: Macmillan.
Hammer, Jeffrey. 1986. 'Population growth and savings in LDCs:A survey," WorldDevelop-
ment 14, no. 5: 579-591.
Haveman, Robert and BarbaraWolfe. 1995. 'The determinants of children's attainments: A
review of methods and findings," Journal of Economic Literature 33, no. 4: 1829-1878.
292 AGE STRUCTURE DYNAMICS IN ASIA
Heston, Alan and Robert Summers. 1995. Penn WorldTables(Mark5.6), on diskette. Avail-
able from the National Bureau of Economic Research, Cambridge, MA.
Higgins, Matthew. 1994. "The demographic determinants of savings, investment and inter-
national capital flows," Ph.D. dissertation, HarvardUniversity.
. 1997. 'Demography, national savings and international capital flows," International
Economic Review, forthcoming.
Higgins, Matthew and Jeffrey G. Williamson. 1996. "Asian demography and foreign capital
dependence," NBERWorkingPaperNo. 5560, National Bureau of Economic Research,
Cambridge, MA.
Hsiao, Cheng. 1986. Analysisof Panel Data. Cambridge:CambridgeUniversity Press.
International Monetary Fund. 1995. InternationalFinancialStatistics,on magnetic tape. Wash-
ington, D.C.: International Monetary Fund.
Judson, Ruth and Anne Owen. 1997. 'Estimating dynamic panel data models: A practical
guide for macroeconomists," Finance and Economics Discussion Series, Federal Reserve
Board, Washington, D.C.
Kang, Kenneth. 1994. 'Why did Koreans save so little and why do they now save so much?"
International Economic Journal 8, no. 4: 99-111.
Kelley, Allen C. 1988. 'Economic consequences of population change in the third world,"
Journal of Economic Literature 26, no. 4: 1685-1728.
Kelley, Allen C. and Robert M. Schmidt. 1995. 'Aggregate population and economic growth
correlations: The role of the components of demographic change," Demography32:
543-555.
. 1996. 'Saving, dependency and development," Journal of Population Economics 9:
365-386.
Lee, Ronald, Andrew Mason, and Tim Miller. 1997. 'Saving, population and economic
growth in East Asia," paper presented at the Conference on Population and the East
Asian Economic Miracle, East-West Center, Honolulu (7-10 January).
Leff, Nathaniel. 1969. 'Dependency rates and savings rates," AmericanEconomicReview59,
no. 5: 886-896.
Mason, Andrew. 1987. 'National saving rates and population growth: A new model and
new evidence," in D. Gale Johnson and Ronald D. Lee (eds.), PopulationGrowthand
Economic Development: Issues and Evidence, pp. 523-560. Madison: University of Wis-
consin Press.
. 1988. 'Saving, economic growth, and demographic change," Populationand Devel-
opment Review 14, no. 1: 113-144.
Masson, Paul. 1990. 'Long-term macroeconomic effects of aging populations," Financeand
Development 27: 6-9.
Mitchell, B. R. 1978. European Historical Statistics, 1750-1975. New York: Facts on File.
1983. International Historical Statistics: The Americas and Australasia. Detroit: Gale Re-
search.
. 1992. International Historical Statistics: Europe 1750-1988, 3rd ed. New York: Stockton
Press.
Obstfeld, Maurice. 1986. 'Capital mobility in the world economy: Theory and measure-
ment," Carnegie-RochesterConferenceSeries on Public Policy 24: 55-104.
Phelps, E. S. 1968. 'Population increase," CanadianJournalof Economics1, no. 3: 497-518.
Ram, Rati. 1982. 'Dependency rates and aggregate savings: A new international cross-sec-
tion study," American Economic Review 72, no. 3: 537-544.
Republic of China. Statistical Yearbookof the Republic of China, various issues. Taipei: Republic
of China.
Sachs, J. and A. Warner. 1995. 'Economic reform and the process of global integration,"
Brookings Papers on Economic Activity, No. 1.
Solow, Robert. 1956. 'A contribution to the theory of economic growth," QuarterlyJournal
of Economics 70, no. 1: 65-94.
MATTHEW HIGGINS / JEFFREY G. WILLIAMSON 293
Taylor, Alan. 1995. "Debt, dependence and the demographic transition: Latin America in
to the next century," WorldDevelopment23, no. 5: 869-879.
Taylor, Alan and Jeffrey G. Williamson. 1994. "Capitalflows to the New World as an inter-
generational transfer," Journal of Political Economy 102, no. 2: 348-369.
Tobin, James. 1967. 'Life-cycle savings and balanced economic growth," in William Fellner
(ed.), Ten Essays in the Tradition of Irving Fischer. New York: Wiley Press.
United Nations. 1992. The Sex and Age Distribution of the World Populations: The 1992 Revision.
New York.
Urga, Giovanni. 1992. "The econometrics of panel data: A selective introduction," Oxford
University Institute of Economics and Statistics discussion paper.
Webb, Steven and Heidi Zia. 1990. 'Lower birth rates = higher saving in LDCs,"Financeand
Development 27: 12-14.
Williamson, Jeffrey G. 1979. 'Why do Koreans save 'so little'?" Journalof DevelopmentEco-
nomics6, no. 3: 343-362.
.1993. 'Human capital deepening, inequality, and demographic events along the
Asia Pacific Rim," in Gavin Jones, Naohiro Ogawa, and Jeffrey G. Williamson (eds.),
Human Resources and Development Along the Asia-Pacific Rim, pp. 129-158. Oxford: Ox-
ford University Press.
- 1996. 'Globalization, convergence, and history," Journalof EconomicHistory56, no.
2: 277-306.