You are on page 1of 28

This article was downloaded by: [Memorial University of Newfoundland]

On: 27 November 2014, At: 05:58


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Population Studies: A Journal of Demography


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/rpst20

Demographic transition in sub-Saharan Africa: How


big will the economic dividend be?
a a
Robert Eastwood & Michael Lipton
a
University of Sussex
Published online: 19 Feb 2011.

To cite this article: Robert Eastwood & Michael Lipton (2011) Demographic transition in sub-Saharan Africa: How big will
the economic dividend be?, Population Studies: A Journal of Demography, 65:1, 9-35, DOI: 10.1080/00324728.2010.547946

To link to this article: http://dx.doi.org/10.1080/00324728.2010.547946

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Population Studies, Vol. 65, No. 1, 2011, pp. 935

Demographic transition in sub-Saharan Africa: How


big will the economic dividend be?

Robert Eastwood and Michael Lipton


University of Sussex

In mid-demographic-transition, many Asian countries enjoyed a large demographic ‘dividend’: extra


economic growth owing to falling dependant/workforce ratios, or slower natural increase, or both. We

estimate the dividend, 1985 2025, in sub-Saharan Africa and its populous countries. Dependency and
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

natural increase peaked around 1985, 20 years after Asia. The UN projects an acceleration of the
subsequent slow falls but disregards slowish declines in young-age mortality and thus, we argue,
overestimates future fertility decline. Even if one accepts their projection, arithmetical and econometric
evidence suggests an annual, if not total, dividend well below Asia’s. The dividend arises more from falling
dependency than reduced natural increase, and could be increased by accelerating the fertility decline (e.g.,
by reducing young-age mortality) or by employing a larger workforce productively. Any dividend from
transition apart, low saving in much of Africa (unlike Asia) means that, given likely natural increase,
current consumption per person is unsustainable because it depletes capital per person.

Keywords: sub-Saharan Africa; demographic dividend; demographic transition; dependency burden;


economic development; fertility decline; labour force; natural increase; population projection;
sustainable development

[Submitted January 2010; Final version accepted August 2010]

1. Introduction: The background ratio in both regions exhibit hump-shaped trajec-


tories (Figures 1 and 2), SSA’s levels lie markedly
Sub-Saharan Africa (SSA) has experienced, and is above Asia’s. This matters, because the pace of
projected to experience, rapid population growth* development may depend not only on the trajec-
from 183 million in 1950 to 863 million in 2010 and tories of these variables but also on their levels; for
1,753 million in 2050. Yet both the rate of natural instance, high young-age dependency may slow
increase (crude birth rate minus crude death rate, development by reducing savings.
that is, population growth net of migration) and the
dependency ratio peaked around 1985. Projected
falls in these variables in the period 19852025 Demographic transition
exceed, by about a third and a quarter respectively,
the preceding rises from 1950. This mirrors a Demographers (e.g., Coale 1973; Montgomery 2009)
comparable trajectory in much of Asia, where the identify four stages, of which only the middle two are
peak was some 20 years earlier, and where the fall in part of the demographic transition as such: a pre-
the dependency ratio has been linked to a large modern equilibrium, with crude birth and death rates
‘demographic dividend’ in the form of more rapid (CBR and CDR) both around 3545; an urbanizing/
economic growth (Bloom and Williamson 1998; industrializing stage, with CBR unchanged but CDR
Bloom et al. 2000). falling towards 15; a mature industrializing stage,
This paper asks whether such a dividend can be with both rates falling towards 10; and post-industrial
expected in SSA during the period 19852025, as equilibrium, with the rates roughly equal at 10 or
natural increase slows and the dependency ratio fewer. The recent emphasis given by economists to
falls. At the outset, one important difference must be the effects of changes in the dependency ratio*and
noted. While natural increase and the dependency the balance within it between young and old

ISSN 0032-4728 print/ISSN 1477-4747 online # 2011 Population Investigation Committee


DOI: 10.1080/00324728.2010.547946
10 Robert Eastwood and Michael Lipton

dependants*has led some to break the transition This view suggests that (i) factors other than falling
into three phases (Lee 2003, Figure 6) as follows: child mortality cause fertility decline, notably the
Phase 1 (c.25 years): rising dependency ratio driven (pre-work) education of females and their wages
by rising young-age dependency; Phase 2 (c.40 (reflecting the opportunity cost of motherhood); and
years): falling dependency ratio driven by falling (ii) that without such factors, child mortality de-
young-age dependency; Phase 3 (c.50 years): rising cline*which on its own raises natural increase*may
dependency ratio driven by rising old-age depen- fail to induce enough fertility decline to reduce
dency. We adopt this scheme, noting that the start of natural increase, let alone reduce it to a post-
Phase 2, defined by peak dependency, also approxi- industrial equilibrium of zero (Preston 1978; Doepke
mately coincides with peak natural increase in Asia 2005). Whether falling young-age mortality raises or
and SSA (Figures 1 and 2). We concentrate on the lowers the expected (or average) number of surviving
possible dividend that SSA may reap in Phase 2. children is unclear. The reduced expected cost of
Crucial for the start, and speed, of Phase 2*that is, rearing a new-born to adulthood induces couples to
for the move to falling dependency and natural plan for more surviving children (Tzannatos and
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

increase*is the shift to falling total fertility (TFR) Symons 1989; Birchenall and Soares 2009). Working
and hence birth rates. Malthus (1824), using regional the other way is the insurance motive: if couples wish
census data from Switzerland and Norway, suggested to have a given chance of attaining a fixed number
that ‘extreme healthiness’ accelerated ‘prudential (e.g., 2) or more adult offspring, then a reduction in
checks’. In today’s language, he meant that lower risk will cause them to plan for fewer surviving
mortality (we might add: especially among children) children on average (for example, with zero risk,
induced behavioural change that reduced TFR and they would plan to have exactly 2). Econometric
hence CBR. That remained the standard view of attempts to resolve such issues have been bedevilled
demographic transition theory at its formulation by the need to control for other determinants of
(Thompson 1929) and during its modern develop- fertility, such as the education of females, and for the
ment (Coale and Hoover 1958, pp. 1213). Falling endogeneity of infant mortality (either because IMR
child mortality (CMR, used throughout the paper to and TFR are jointly determined by other variables,
mean deaths of children under age 5 per thousand or because less frequent births change the risk of
live-born children) may quantifiably affect three of child mortality). Nevertheless, recent literature using
the proximate determinants of total fertility (Bon- best-practice econometric techniques concludes that
gaarts 1978, 1982). They are: post-partum infecund- a fall in young-age mortality is the main driver of
ability, use of contraception, and proportion married TFR decline (Section 2).
(Preston 1978, but see Montgomery and Cohen 1998).
This view of fertility transition has been challenged
on the ground and the evidence that marital age and Impact of Phase 2 of demographic transition on
fertility are determined jointly with other variables, growth of income per person: Accounting
notably the contribution of women to the labour
force, within some kind of household optimizing Projection of the impact of Phase 2 of the demo-
framework (Becker 1960, 1981; Schultz 1981, 2008). graphic transition in less developed countries was

3
90

80 2
Per cent

Per cent

70

1
60

50
0
1950 1970 1990 2010 2030 2050 1950 1970 1990 2010 2030 2050
Year Year

Dependency Dependency Natural Natural


ratio: Asia ratio: SSA increase: Asia increase: SSA

Figure 1 Dependency ratios: Asia and SSA Figure 2 Natural increase: Asia and SSA
Source: UN (2009a) Source: UN (2009a)
Demographic transition in sub-Saharan Africa 11

pioneered in 1957, for India (Coale and Hoover to be some 23 times as large as those from falling
1958). Losses in Phase 1, and especially gains in dilution of capital.
Phase 2, seemed small because both peaks and This age-structure accounting, however, has se-
subsequent falls in the dependency ratio were vere limitations. For it to be the whole story, there
under-projected. For the period 19862006, the ratio must be no effects of Phase 2 on the time path of
was projected to fall only from 52 to 50. This output per worker. However, there may be positive
followed from the projection that TFR would halve effects either from age structure itself, if for instance
from 1956 to 1986, and stay at the 1986 rate until a fall in youth dependency raises the savings rate, or
2006 (ibid. pp. 233, 322). In fact, after 1956, India’s from slower natural increase, since this implies that
TFR fell more slowly than projected: it took until given savings are spread less thinly, an effect that
2000 to halve. Therefore India’s dependency peak tends to raise capital per worker and therefore
was higher than expected82 in 1965and there was output per worker. The first effect fits into the age-
a greater subsequent dependency declineto 73 in structure hypothesis, simply implying that the effects
1985 and 60 in 2005 (UN 2009a)*so that much are more than arithmetical, while the second does
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

larger economic effects from age-structure change not. Similarly, for the natural-increase accounting to
now seemed plausible. Kelley and Schmidt (1995) be the whole story, we must assume that sustain-
and Bloom and Williamson (1998), disaggregating ability can be identified with constant capital per
fertility and mortality, claimed that the ‘Asian tigers’, person, and thus must neglect technological advance,
and to some extent South Asia too, had gained a in particular the possibility that age-structure change
large demographic dividend as a result of age- might stimulate such advance. The natural-increase
structure change during Phase 2 and, moreover, accounting also has a normative character, calculat-
that this was the total effect of demographic change. ing extra savings ‘needed’ for sustainability under
To set this age-structure hypothesis in context, an given circumstances, but silent on whether transition
accounting approach is useful. Figures 1 and 2 show might make these savings either more or less likely
two features of Phase 2 of the transition: age- to materialize.
structure change and falling natural increase. Each To assess the effect of demographic change on
can be held to yield growth benefits from a simple growth in output per person, there is an alternative
accounting standpoint. For age structure, a falling to the accounting approach: econometric estimation.
dependency ratio implies arithmetically that a given The literature on this subject, reviewed in Section 4,
rate of growth of output per worker translates into a strongly supports the age-structure hypothesis. Not
faster rate of growth of output per person. For only does falling dependency increase growth arith-
natural increase, the slower it is, the less an economy metically, as described above, but the effect*that is,
needs to save in order to maintain total (natural plus the demographic dividend*is more than arithmeti-
reproducible) capital per person, so the higher is cal. Moreover, no link from natural increase to
sustainable consumption per person. In an economy growth is found. These findings cannot, however,
where natural increase is falling through time, be taken as final because of doubts over the
therefore, the amount that has to be saved in order robustness of the econometric techniques employed.
to equip extra workers is also falling, so that the Overshadowing the demographic dividend in
amount left over and available for (sustainable) Phase 2 in SSA, and highly pertinent to any
consumption is growing through time. To express comparison with Asia, are the consequences of
this somewhat differently, any natural increase SSA’s combination, throughout the transition, of
means that a given stock of capital is ‘diluted’ by much more rapid natural increase (Figure 2) with
having to be shared among more workers, with the much lower savings rates. In Asia, not only did the
consequence that falling natural increase delivers a slowing of natural increase raise sustainable con-
demographic dividend via reduced dilution. sumption per head during the period 19652005, but
Accounting thus allows us to calculate two types of savings were high enough for the path of develop-
demographic dividend, one from improved age ment itself to be broadly sustainable. This is very
structure and one from slower natural increase. unlikely to be the case in SSA over the period 1985
These dividends are not strictly commensurate, 2025 (Section 4). Slower natural increase will indeed
because one comprises extra growth of output per raise sustainable consumption per head by reducing
person, and the other extra growth of sustainable the savings that would be needed to sustain capital
consumption per person. Nevertheless we estimate per person, but this prospect pales into insignificance
and compare the dividends for a number of countries beside the likelihood that savings will fall far short of
in Section 4: in SSA the age-structure gains are likely the necessary level.
12 Robert Eastwood and Michael Lipton

Outline and contributions of this paper from demographic transition as such are the effects
of continuing natural increase, when set in the
Section 2 states the paths of the main demographic context of national savings rates. We estimate these
variables during the three phases of transition. We underlying rates of capital dilution for populous SSA
analyse natural increase rather than the rate of countries, with comparative estimates for some
growth of population, because the causes and con- Asian countries. Section 5 presents our conclusions.
sequences of migration would require a major The main contribution of the paper is to bring
digression. We then enquire whether UN middle together the elements needed for an assessment of
projections, used in this paper, correctly predict the demographic dividend in SSA, reviewing past
Phase 2 of the transition. Being global and univari- estimates and presenting some new calculations.
ate, they do not allow for region-specific impacts on Since all this must rest on demographic projections,
fertility of causal variables, such as literacy of we first set UN middle projections in the context of
females and young-age mortality. In particular, some determinants of fertility, and of the extent to
where the latter is higher, or falling more slowly, which policy might modify them. Given the demo-
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

than the global norm, these projections may over- graphy, we calculate two sorts of accounting divi-
state future fertility decline, and therefore reduc- dend for the region and its populous countries, and
tions in both the dependency ratio and the rate of assess the influential view that the dividend arises
natural increase. entirely from changes in dependency. We show that,
Section 3 tracks transition data in SSA, its regions, under current policies, the adverse effects of ongoing
and its most populous countries. In most cases, rapid natural increase in much of SSA exceed likely
natural increase and dependency peaked (bringing gains from its reduction during the demographic
in Phase 2) at higher levels than in Asia, and about transition.
20 years later, around 1985. This sets the stage for
larger total declines in dependency and natural
increase in Phase 2. However, UN middle projec- 2. Three phases of demographic transition
tions to 2030 show annual declines lower in most of
SSA than comparable falls 20 years earlier in Asia, Defining the phases
and even those projections look optimistic. But the
future can be changed. Parts of Asia and Africa In the stylized ‘demographic equilibrium’ before
experienced fast reductions in young-age mortality, transition, births equal deaths, typically at high
and measures were undertaken to transmit these levels. Age-specific mortality and fertility rates are
rapidly into less pro-natalist behaviour. It is techni- trendless, as is age structure, and therefore the
cally possible for much of SSA to do the same. Yet dependency ratio. Phase 1 begins with a sustained
total fertility decline has been sluggish in some large fall in the crude death rate. This raises the rate of
countries of West and Middle Africa, and has stalled natural increase. Because the reduced deaths (e.g.,
in many, but not all, groups and regions of East from malnutrition, dysentery, and malaria) are of a
Africa (Ezeh et al. 2009). Groups and areas slow to much larger proportion of under-15s than of persons
reduce natural increase come to represent growing aged 1564, the dependency ratio rises substantially.
proportions of populations. Most groups and regions Towards the end of Phase 1, total fertility starts to
will need faster reduction of child mortality, plus fall. In Phase 2, more and more ‘survivors’ (bene-
measures influencing fertility decisions more di- ficiaries of earlier falls in under-5 mortality) reach
rectly, for Africa as a whole to reap a large age 15. Therefore, even though young-age mortality
demographic dividend. is still falling somewhat, dependency ratios fall.
Section 4 moves to estimates of the demographic Meanwhile the fall in TFR continues. ‘Momentum’,
dividend. Giving appropriate Asian comparisons, we from rising numbers of women ‘survivors’ reaching
evaluate econometric evidence as it affects SSA’s childbearing age, slows the consequent falls in CBR,
dividend so far, and present alternative accounting but after some years these falls accelerate, leading to
estimates for the period 19852025. As is evident a steady slowing of natural increase. In Phase 3, as
from Figures 1 and 2, both the dependency ratio and ‘survivors’ reach age 65, old-age dependency starts
natural increase in SSA reach a higher peak and to rise, and natural increase has little further to fall,
then decline more slowly in SSA than in Asia. having approached zero.
Therefore, the accounting dividends for SSA are All three phases, like the preceding equilibrium,
larger in total but smaller per year. Yet, as noted are stylized (Table 1). Fluctuations and crucial
above, arguably more important than the dividends simultaneities, notably between the start of falls in
Demographic transition in sub-Saharan Africa 13

Table 1 Three phases of demographic transition: trends in key variables

Phase 1 Phase 2 Phase 3

Crude death rate Fall Fall Slow fall


Young-age mortality Fall, becoming rapid Fall Slow fall
Total fertility No trend or slow rise, then fall Fall Fall
Crude birth rate No trend or slow rise Fall Slow fall
Natural increase Rapid rise Fall Fall, then no trend
Dependency ratio Rapid rise Rapid fall Rise

natural increase and dependency, are neglected. As (Shapiro and Tambashe 2002; cf. Garenne and Joseph
we shall see, this is not too damaging for assessing 2002). The delay in rural changes can retard national
transition in most parts of Africa. All three phases trends in total fertility, natural increase, and depen-
are visible in Figures 1 and 2 for Asia, while only dency during Phase 2, but we do not look at sub-
Phase 1 and a prolonged Phase 2 are visible for SSA national trends, except where overlap of phases
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

as a whole; however, Phase 3 starts much sooner in within a country has major policy importance.
some countries, notably South Africa.

UN estimates of the past and projections of the


Diversity within Asia and sub-Saharan Africa future: impact of young-age mortality on fertility

Though it is of interest to compare SSA with Asia, To estimate demographic dividends in Section 4, we
both are regionally diverse, demographically (Table used past estimates and middle projections from the
2 below) and economically. In Africa, Phase 2 has UN (2009a), based on censuses and DHSs. The last
progressed fastest in the Southern region, with census round was 200001. All the countries in
Eastern SSA slower and West and Central SSA Tables A1A3 except China, Sudan, and Thailand
slowest. South-Central Asia’s transition has lagged also have a post-2000 DHS (1989). In SSA, demo-
well behind East Asia’s, with West and South East graphic estimates vary from relatively good (Kenya)
Asia in between. Tables A1A3 show some of the to very rough-and-ready (DR Congo), with past
country diversity, with data for India, China, Indo- mortality less disputed (Hill and Amouzou 2006)
nesia, and Thailand, and for the 16 most populous than past total fertility; the UN fertility estimates
countries (out of 51) in SSA. These housed 764 appear to be in the middle range (Bulatao 2006, p.
million people in 2005, 77 per cent of SSA’s 62). However, UN middle projections of total
population. Just four countries*Nigeria, Ethiopia, fertility in many countries are probably biased
DR Congo, and South Africa*contained 42 per downwards. The UN model assumes that a country’s
cent. future total fertility depends only on: (i) its TFR in
Diversity within countries also affects transition. the latest year that it was measured (typically in
Each phase arrives earlier, and proceeds faster, for 200008 for the most recent projections); (ii) past
urban, richer, more economically integrated, and less trends in TFR in other countries after reaching that
gender-unequal groups and places. Thus child mor- country’s latest measured TFR, except that (iii) if
tality, total fertility, and hence natural increase and TFR ‘has stalled or where there is no evidence of
dependency all fell later and more slowly, and are [its] decline’, it is ‘projected to remain constant for
now much higher, in North-Central India, above all [510] years’ after which ‘the model projection takes
in Bihar, than in Southern India, especially Kerala over’ and TFR falls toward 1.85 (UN 2009b).
(Cassen et al. 2004). Arguably Bihar has emerged UN projections are therefore univariate, that is,
from Phase 1 as slowly and recently as Middle Africa, they predict a country’s TFR by past TFRs alone,
while Kerala, like Southern Africa, is in Phase 3. For and global, that is, based on past TFR paths
13 SSA countries in the late 1990s, between the two worldwide (except when a national TFR is assumed
most recent Demographic and Health Surveys briefly stationary, as in (iii) above). Such projec-
(DHSs) ‘unweighted average rural TFR declines by tions cannot allow for possible country-specific
about 0.3’ but by 0.9 in urban areas ‘[usually] with determinants of future TFR such as its child
fertility initially declining in urban areas [only], then mortality, education and earnings of females, and
. . . in both settings but more rapidly in urban places, access to contraception. If these are worse, or
and finally . . . more in rural than in urban areas’ slower-improving, than the global norm, the UN
14 Robert Eastwood and Michael Lipton

model will project implausibly rapid acceleration of similar hierarchy of country disadvantage (IDRC
that country’s fertility decline. For example, com- 2002). Countries with higher IMR tend to have
pared with global experience of Phases 1 and 2, even higher post-1995 child mortality. The UN
infant (and, by implication, child) mortality in (2009a) estimates that, for every 100 infant deaths,
Nigeria is far higher, and has fallen far more slowly there are 163 child deaths (this includes infant
(Table A2). If these drive subsequent total fertility, deaths) in SSA (over 170 in Angola, Mozambique,
its fall in Nigeria will not accelerate steeply, as the and Nigeria), but fewer where IMR rates are lower
UN projects: in Nigeria from 198590 to 200005 (Asia 138, China 129, Thailand 152) (UN 2009a; see
TFR fell by only 0.320.39 per quinquennium, but also Hill and Amouzou 2006 for West Africa).
from 200005 to 202530 the UN middle projection Countries with high infant mortality show even
is a fall of 0.60 per quinquennium (UN 2009a). more child deaths (per hundred infant deaths) in
Where young-age mortality has fallen more slowly, local life tables, which are more reliable: the
or stalled at higher levels, than elsewhere, such average figures from local surveys are Ethiopia
fertility projections are too high (UN high projec- 182, Gambia 251, Guinea-Bissau 203, Ghana 170,
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

tions do not help, because they rely on the same Senegal 153, Zambia 173 (IDRC 2002).
global univariate model). But how important is the
mortality-to-fertility effect?
Recent literature, using best-practice econo-
metrics to reduce problems of reverse and joint

3. Transitions, 1920/50 2030: African
dependency and natural increase in Asian
causation, confirms earlier demographic transition context
theory: ‘the biggest driver quantitatively [of TFR
decline] remains child survival’ (Conley et al. 2007). Phase 1
In India in 198191, a fall in child mortality of 50
was estimated to have reduced total fertility by 0.20 Mortality decline. This led to rising natural in-
(out of a total fall in TFR of 0.7), slightly more than crease in most of Africa and Asia by 1930. In the
the response to higher literacy of females (Drèze 1930s population rose annually by 0.6 per cent in
and Murthi 2001; Murthi 2002). Turning to multi- China, 1.4 per cent in India, 1.5 per cent in
country panels, Lorentzen et al. (2008) imply that a Indonesia, and over 1 per cent in Congo, Kenya,
fall of 50 in infant mortality cuts total fertility by Nigeria, and South Africa (Visaria and Visaria 1982,
0.73. Angeles (2010, Table 3) estimates an immedi- p. 488; Fage et al. 1986, p. 49; Dawson 1987; Brass
ate fall of 0.21 in total fertility owing to a fall of 50 and Jolly 1993, p. 29; Iliffe 1995; Populstat n.d.-a;
in CMR; he indicates that the fall in fertility is n.d.-b; Wikipedia n.d.).
much greater if one allows for the lagged effect. Thus by 195055, when the UN’s data series
Conley et al. (2007, pp. 2930) find that a fall of 50 begins (2009a), Phase 1 had been under way for
in CMR cuts TFR by between 0.5 and 0.9; the some time. From 195055 infant mortality declined
effect is ‘robust across all specifications, and holds if through Phase 1, into Phase 2, in all regions, but
CMR is replaced with IMR’. Group disaggregations much more slowly in SSA, so that by 200005 it was
support such effects. Ruralurban gaps in fertility still 95, compared to 47 in Asia. Southern Africa did
(and its decline) among the main childbearing age better than the rest of SSA until 198590, after which
groups, especially those aged 2024, are signifi- infant mortality rose. Taken on its own, the relatively
cantly wider where rural-urban child mortality gaps slow decline of IMR in SSA should have slowed
are greater (Shapiro and Tambashe 2002, Table 3). natural increase relative to the position in Asia. That
The relative impacts of infant, child, and under- this did not happen was because the path of fertility
15 mortality on fertility decisions are unclear. was so different in the two regions.
There are few comparable data based on reliable
life tables for under-15 mortality, yet it is the whole Fertility trajectories. In Asia, TFR fell dramatically,
age span 015 that affects both rearing costs and from 5.7 to 3.5, between 195055 and 198590; in
the prospects that children will continue the family SSA it hardly fell at all, with an Asian-scale fall in
line. Some local findings are based on CMR, but Southern Africa and a modest fall in East Africa
because worldwide data acceptable to the UN offset by rises in Middle and West Africa (Cohen
(2009a) are recent and few (1995, 2000, and 2005 1993, 1998; Table A1). Even by 200005, TFR was
only), reliance on IMR data for time-series estima- above 195055 levels in Burkina Faso and DR
tion is inevitable. In any event, IMR, CMR, and Congo and almost the same in Uganda. Owing to
what we know of under-15 mortality suggest a population momentum, the TFR downturn in late
Demographic transition in sub-Saharan Africa 15

Phase 1 is translated into a fall in CBR and natural of SSA. In four more, it peaked in 1980 (and in
increase only after a time lag. In many cases (India, Angola in 1995) at 13 points above the 1985 level.
Ghana, Kenya, Sri Lanka, Southern Africa) the fall Ethiopia, Uganda, and DR Congo peaked in 1995*
in TFR was fast enough for CBR to start to fall, at 57 points above the 1985 dependency level, not
albeit slowly, almost at once. In East and South East much compared with the 21- to 25-point (UN-
Asia, the fall in TFR was precipitate and led to early projected) decline for 19952025. Only South Africa
and sharp falls in CBR. is an outlier, with peak dependency early: 84 in 1970.
Despite variations, it is reasonable to see 1985 as
Dependency and natural increase outcomes. Old- SSA’s switch point from Phase 1 to Phase 2.
age dependency in most of Asia and Africa was very
stable throughout Phases 1 and 2. Hence the depen-
dency ratio was largely determined by youth depen- Phase 2
dency and its changes. Asia’s dependency ratio rose
from 67 in 1950 to a peak of 80 in 15 years, in 1965; Natural increase. In which countries of SSA is the
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

SSA’s rose from 82 in 1950 to a peak of 94 in 35 years. Phase 2 slowdown in natural increase comparable to
There was substantial country variation (Table A3): the earlier slowdown in Asian countries? Thirty
at the extremes, Sudan’s dependency ratio rose years after the country peak, natural increase had
negligibly while Kenya’s rose from 78 in 1950 to fallen by 1.7 percentage points in China, 0.8 in India
a peak of 113, owing to very high, and at first rising, and Indonesia, and 1.5 in Thailand. In only six
total fertility. SSA’s natural increase in 195055 was populous SSA countries did natural increase peak by
0.3 per cent per year above Asia’s, and peaked at 0.6 1980, so we can track 30 years of its fall. This was of
per cent above the Asian peak, but 20 years later. ‘Asian’ size in South Africa (1.03), Côte d’Ivoire
(0.98), Ghana (0.76), and Sudan (0.70), but only 0.11
Dating the peak. How uniform across countries was in Uganda and 0.20 in Tanzania.
the 1985 switch from Phase 1 to Phase 2 in SSA? In In the other ten populous countries of SSA,
the quinquennium 198085, natural increase peaked natural increase peaked after 1980. Thus 30-year
in Cameroon, Kenya, and Sudan; Nigeria, South trajectories for Phase 2 are partly projections. As
Africa, and four other ‘populous’ countries were late as 200510, natural increase in four of the six
within 0.1 per cent per year of the peak, and DR most populous countries*DR Congo, Ethiopia,
Congo and two others within 0.25 per cent per year. Kenya, and Nigeria*ranged from 2.36 to 2.79 per
The only populous outliers, where natural increase cent, well above India’s 2.21 per cent peak that
peaked well after the period 198085, were Ethiopia occurred 35 years previously. On the UN middle
(2.64 per cent per year 198085; peak 3.01 per cent projection, these four countries will slow their
199095), Madagascar (2.57 per cent; 3.05 per cent natural increase after 2010 more than did India after
199095), and Burkina Faso (2.99 per cent; 3.47 per 1975. However, so far, unlike India 1535 years ago,
cent 200510). As for dependency, it peaked in 1985 change in these countries has been sluggish: little
in Nigeria, Sudan, and four other populous countries decline in total fertility in DR Congo, stalled decline

Table 2 Natural increase and dependency ratio, 195055 to peak rate, by region, sub-Saharan Africa and Asia

Natural Natural
increase 1950 Natural increase peak Dependency Dependency Dependency ratio
Region 55 increase peak years ratio 1950 ratio peak peak year

SSA 2.15 2.89 198085 82 94 1985


East 2.29 2.99 198590 85 97 1980
Middle 1.94 3.07 198595 82 99 1995
South 2.29 2.52 197075 74 85 196570
West 2.02 2.87 198085 81 94 198590
Asia 1.88 2.38 196570 67 80 1965
East 1.81 2.42 196570 63 76 1965
(China) 1.87 2.60 196570 61 80 1965
S. Central 1.84 2.43 197580 70 83 1965
S. East 2.08 2.52 196070 73 88 1970
West 2.52 2.71 196065 74 89 1965

Source: UN (2009a).
16 Robert Eastwood and Michael Lipton

in Kenya (i.e., no convergence to any global path of Actual and projected fertility change: an
fertility decline), and slow young-age mortality Afro-Asian parallel?
decline in DR Congo, Ethiopia, and Nigeria, that
is, a weakness in the main driver of fertility decline. A parallel between South-Central Asia and SSA
Modelled, not empirical, age-specific death rates in makes it useful to compare their fertility behaviour.
many African countries, including DR Congo, Ken- In South-Central Asia, total fertility started to fall
ya, and Nigeria, make it precarious to project significantly from 196570, and in most of SSA from
demographic change from past birth and death rates 198590. In SSA it then fell by 16 per cent in 15
(IDRC 2002, chap. 6). years. The UN (2009a) projects the fall to accelerate
In several African countries, HIV/AIDS clouds to 22 per cent in the subsequent 15 years, cutting
any projections for natural increase. In South Africa, TFR to 4.2 in the period 201520. Indeed, the Phase
the crude death rate troughed at 8.8 in the years 2 path of total fertility in SSA has quite precisely
199095, but then rose to 15.1 in 200510, auto- paralleled that of South-Central Asia 20 years ear-
matically reducing annual natural increase by 0.63 lier. In each quinquennium, the absolute falls in TFR
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

per cent. Plainly, this 0.63 per cent is not part of


in South-Central Asia from 196570 (5.86) to 1980
transition. In UN 2009b, ‘the impact of HIV/AIDS
85 (4.88) are almost identical to those in SSA from
on mortality is modelled explicitly . . . where HIV
198590 (6.42) to 200005 (5.41). Why, then, might
prevalence among persons aged 15 to 49 was [over] 1
one doubt the UN middle projection that this
per cent during 19802007’, using adult HIV pre-
parallelism will continue?
valence rates from DHSs, and evidence on spread,
First, global projections of a decline in TFR to the
treatment, and survival prospects. Although best-
same endpoint in all countries rest, implicitly, on the
practice methods were used, projections based on
assumption that replacement fertility is much
such evidence are inevitably speculative. The impact
of HIV/AIDS on natural increase is further compli- the same everywhere (2.1 lifetime births per woman).
cated by the effects on fertility, and interactions Yet it ranges ‘from less than 2.1 to nearly 3.5 . . . due
between HIV-related fertility and mortality change. almost entirely to cross-country differences in
mortality . . . the current replacement TFR for [East
Dependency ratio. So far, SSA’s Phase 2 fall in Africa] is 2.94’ (Espenshade et al. 2003, p. 575). Given
dependency is even slower, relative to Asia’s, than high projected mortality, UN fertility projections
was the Phase 1 rise. Among the populous countries, therefore imply that Ethiopia, Madagascar,
only in Kenya and South Africa has dependency Mozambique, Kenya, and Sudan reduce TFR
fallen almost as fast in Phase 2 as it did in China, implausibly close to (current) replacement levels by
India, and Indonesia. In the 20 years following its 203035.
1985 peak, the all-SSA dependency ratio fell only Second, one proximate determinant of fertility,
from 94 to 86. However, the UN middle projection is contraceptive prevalence, lags well behind levels
an accelerated fall to 65 by 2030. In most of Africa, seen at a comparable point in transition in Asia.
fertility trends have been sluggish. Remarkable For example, among women aged 1549, prevalence
acceleration seems required to drive the projected was over 35 per cent in India in the early 1980s (Ross
declines in dependency. That is especially the case in et al. 2004; Visaria 2009, p. 188), but only 23 per cent
countries heavily affected by HIV/AIDS, which in SSA in 200308 (UNICEF 2010b: DR Congo 21,
‘hollows out’ the age group 3050. Thus from 2000
Ethiopia 15, Kenya 39, Nigeria 15, South Africa 60).
to 2005, despite ongoing falls in total fertility, the
Contraceptive use (given availability, which is not
dependency ratio actually rose in Malawi from 96 to
considered here) depends on demand. This, like age
99, and in Mozambique from 88 to 90 (Tables A1
of marriage and frequency of cohabitation, depends
and A3).
partly on political attitudes and cultural norms.
Fertility. Among populous countries of SSA, only Many allege that these are more pro-natalist in
South Africa provides a precedent for an absolute SSA than elsewhere, which may help to account for
fall in total fertility, from its peak, comparable to cross-country variation in desired number of chil-
Asia’s. It has also fallen by over two points in Côte dren (Westoff 2010). Yet such attitudes and norms
d’Ivoire, Ghana, and Kenya (where it has stalled: vary greatly across SSA, just as they do across Asia,
Ezeh et al. 2009). Only in Ghana, South Africa, and seem unlikely to explain why contraception (and
Cameroon, and Sudan is TFR now below 5. probably other proximate determinants of fertility
Demographic transition in sub-Saharan Africa 17

decline) lags well behind Asian experience in most projected to fall to 3.7. There is no sign of the sharp
of the populous countries of SSA. fall in total fertility without which Middle Africa will
Might the apparent reluctance to adopt contra- not approach UN middle projections of declining
ception (and, more generally, weaker prospects for dependency. For these to happen, big changes in
fertility decline than in Asia) arise because women’s policy or institutions seem needed.
advancement in SSA lags behind Asia’s of 2025
years previously? It is well established that later
marriage, lower marital fertility, and the use of Improving the complex of young-age mortality
contraception are advanced by better opportunities and fertility in SSA: Prospects and policies
for women. However, the main proxy for these, adult
literacy of females, is not worse in most of Africa
The mortality prospect. To a considerable extent,
than in India 2025 years earlier: India’s 1981 rate of
fertility decline depends on previous falls in young-
30 per cent (RGCC 2001) was exceeded in 200306
age mortality. Here too, most of SSA has lagged far
by many populous African countries, including, of
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

behind Asia, yet the UN (2009a) projects rapid catch-


the few with reliable DHS data, Nigeria (55 per
up ‘on the basis of [regional] models of change of life
cent) and Uganda (49 per cent), though not Burkina expectancy [with] smaller gains the [higher] the life
Faso (12 per cent) (Huebler 2008). expectancy reached’ (UN 2009b). In other words,
Third, we saw that by Asian standards most of regional and national age-specific mortality*like
Africa has seen slow, or (in several countries most fertility*is projected on a univariate model impos-
affected by HIV/AIDS) reversed, falls in young-age ing convergence to global trends: where mortality
mortality. To the extent that such falls drive fertility remains high, accelerating decline is projected.
decline, the UN’s projection model will overstate it Asia cut IMR by 51 per cent in 196595; the
in much of Africa. projection for 19952025 is 46 per cent. Yet for
Finally, within-country variation in Africa impedes the same periods SSA achieved only 27 per cent; the
rapid national fertility decline arithmetically. As in projection is 41 per cent. Some countries would
India (Cassen et al. 2004, pp. 98, 105106), total achieve projected falls merely by resuming past falls
fertility is increasingly influenced by regions (and in IMR (as in Kenya) or at least consolidation or
rural sectors) where it is persistently high. So, big modest improvement (Burkina Faso, Cameroon,
falls in TFR, and therefore in dependency and Côte d’Ivoire, Malawi, Sudan), but most country
natural increase in Phase 2, increasingly depend on projections seem unduly optimistic. Infant mortality
progress among SSA’s poor, rural people. Yet a 2002 in DR Congo, having fallen 10 per cent (from 143)
study, using the most recent DHSs for 24 SSA from 1965 to 1995, is projected to fall three times as
countries, found that average rural TFR was 6.4 fast from 1995 to 2025, and in Nigeria the rate of
compared to an urban average of 4.6; for the 13 decline is projected to double. Does recent evidence
countries with repeat surveys, all but two showed a support such accelerations?
widening ruralurban gap (Shapiro and Tambashe During the years 19902000, 17 countries in SSA
2002). Urbanrural and overall inequality is greater had two DHSs. Infant mortality rose in nine (sig-
in most African countries than in most of Asia nificant at 5 per cent in four), with significant falls only
(Eastwood and Lipton 2004), thus impeding the in Malawi, Niger, and Nigeria (Mahy 2003, p. 31).
education of females, contraceptive availability, and South Africa (among others) has slipped back, mainly
child nutrition and preventive medicine in lagging as a result of HIV/AIDS. In the twenty-first century,
areas. Assuming roughly similar prospects for urba- Kenya reversed a similar setback: IMR decreased
nization, this leaves most of Africa worse placed from 119 in 1969 to 66 in 1989, increased to 77 in 1999
than was Asia to provide preconditions for reduced and 2003, but then resumed its decline to 52 in 2008
TFR among their rural majorities. 2009 (KNBS and ICF Macro 2010, pp. xv, 3). As
In Middle Africa, the slow pace of past fertility regards child mortality, UNICEF (2010a) shows it
decline makes UN middle projections of depen- falling at 1.8 per cent per year in 200005, and at the
dency especially vulnerable. In the period 195095 same rate in 200508 (to 144). Other careful recent
the ratio rose, to peak at 99. In the years 19952005 reviews of censuses and quality-controlled micro-data
it fell to 95. Yet it is projected to fall sharply to 63 by also find slow, non-accelerating falls in CMR (and
202530 because by then TFR, which was 6.0 in IMR) (UNICEF, WHO, World Bank and UN Popula-
195055 and even higher, at 6.2, in 200005, is tion Division 2007; UN 2009a (revised 1 August 2010);
18 Robert Eastwood and Michael Lipton

Macro International Inc., for DHSs, 2010; UNICEF are mutual causes, and need joint solutions. Data
2010b). A dissenting view, apparently based on the ‘from all 52 countries [with DHS surveys] between
use*and trust*of a very wide variety of data sources, 2000 and 2005 [except] India’ show that ‘if all women
is Rajaratnam et al. (2010). would wait at least 24 months to conceive again,
Whatever the past trends, Asian and some African under-five deaths would fall by 13 per cent[, and if] 36
experiences show that rapid improvements in child months [by] 25 per cent’ (Rutstein 2008, pp. 14, 70
mortality are feasible (Table A2; Conley et al. 2007). 1). The processes of lowering under-5 mortality and
In SSA, most child mortality is due to the interaction fertility in Africa are well under way, even if mostly
of malnutrition with gastro-intestinal infections and slower and less visible than in Asia at a comparable
malaria, and in some countries to HIV/AIDS. For stage of transition. Africans are used, therefore, to
the 60 per cent of SSA population at high malaria mortality declines, and to reviewing the conse-
risk, P. falciparum alone directly causes 28 per cent quences for their fertility decisions. It is reasonable
of all child deaths. Finding and delivering a vaccine to expect a shorter lag between falling young-age
against all forms of malaria would cut child mortality mortality and the fertility response*and hence a
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

by over a third. Even with today’s medical knowl- quicker, more vigorous Phase 2*if couples know
edge, West and Middle African infant mortality can that access to contraception, improved education and
be pushed far below its 200005 level of 100115 prospects for women, and measures to reduce child
(Snow and Omumbo 2006, p. 202), as Ghana’s rate mortality are being tackled jointly. Put another way,
(70) shows. In Benin between the 2000 and 2006 these three (expensive) objectives are complemen-
DHS rounds, ‘increased possession of bed nets led to tary: spending on each is usually more cost-effective
a decrease of about 21 percent in what [under-5] if accompanied by spending on the others. Putting
mortality would have been without the increased family planning in the context of child health has
possession’ (Rutstein et al. 2009, Abstract and pp. much more than rhetorical importance.
167). SSA has seen huge rises in public health We now set aside the above concerns. We assume
expenditure to combat malaria, tuberculosis, and that middle UN projections of falls in dependency
HIV/AIDS. However, this is mostly aid-financed; a and natural increase, for SSA and each region and
lasting impact requires embedding the largely pre- populous country, are correct*or will be made so by
ventive measures in national politico-administrative changing policies and institutions. What demo-
systems. This may not occur swiftly. South-Central graphic dividend can be expected from 1985 (the
Asia last saw today’s rate of West African infant posited start of Phase 2) to 2025, via lower depen-
mortality around 1980, and took 25 years to get it dency ratios and slower natural increase?
down to a still-high rate of 60. As noted earlier, even
maintaining past rates of fertility decline in most
African countries (let alone accelerating them on the 4. Economic effects of demographic changes
Asian pattern) will require falls in fertility among
areas and groups where it is particularly high. To Causal channels from demography to output per
facilitate this, measures to reduce young-age head
mortality*principally, affordable access to preven-
tive and primary health care*must increasingly be Dilution of natural capital. If the constraint repre-
spread to poorer rural and slum areas and groups. sented by natural resources, such as (quality-
adjusted) land or water, is binding, then a rise in
A paradox. There is a paradox about expecting, let the number of people will reduce steady-state GDP
alone requiring, such a policy to strengthen Phase 2 per head. Niger, whose population is projected by
of the transition. Especially where child mortality is the UN to grow from 13 million in 2005 to 58 million
still very high, effective action to cut it*though in 2050, is an example of an agrarian economy with a
needed for big medium-term cuts in natural increase fast-growing population, where such a mechanism
and dependency*will for some years increase them. may seem potentially important. Another example
Can this time lag be reduced? Thompson (1959) appears to be Rwanda (André and Platteau 1998).
thought 25 years was the normal lag in Asia. The raw Also, a sufficient rise in population (density) might
data, and the econometric literature reviewed above, trigger progressive environmental degradation, in
hint at a shorter lag, but do not resolve the issue. We which case there would be a link to the growth rate
can see big country differences ex post, and can be as well as the level of GDP per head. An example
sure that policy matters. What is crucial is integrated is the account of environmental collapse on Easter
policy on young-age mortality and total fertility. They Island by Diamond (2005, p. 118), which he
Demographic transition in sub-Saharan Africa 19

describes as ‘the clearest example of a society that note output by Y, population by N, the population of
destroyed itself by overexploiting its own resources’. working age by WA, and the percentage growth rate
The population is believed to have declined by the of any variable x by g(x). Then by definition:
end of the eighteenth century by some 70 per cent
gðY=NÞ ¼ gðY=WAÞ þ gðWA=NÞ (1)
from the peak reached in the period 14001600;
deforestation of the island reached a peak in about Assume, just in this paragraph, that the given
1400 and was complete by the end of the 1600s. demographic change has no impact on g(Y/WA).
Then, if the demographic transition causes the
Increasing returns to population via the rate of working-age share (WA/N) to grow by an extra 1
technical advance or the sharing of infrastructure per cent (e.g., at 1 per cent instead of 0 per cent per
overheads. Extra people may raise long-run in- year), there will be an equal rise of 1 per cent in the
come per person, via higher density. This may rate of growth of output per person. We refer to this
accelerate research and invention (Kremer 1993; as the arithmetic age-structure dividend. It can be
Simon 1996): ‘exogenous’ advances can be shared, linked directly to the change in dependency, since
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

and incentives to invent and innovate are increased g(WA/N)  g(1 dependency ratio). A falling
by the prospect of returns in larger markets. Higher dependency ratio implies a rising working-age share
density may induce land-saving agricultural innova- and, arithmetically, a faster rate of growth of output
tions (Boserup 2005). In SSA it has sometimes per person.
stimulated environmental improvement (Tiffen Going beyond this is the age-structure hypothesis,
et al. 1994; see, however, Zaal and Oostendorp which in what we call its strong form says that all the
2002), some forms of mechanization (Pingali et al. effect of demographic change (of any kind) on
1987), and rural roads in isolated regions (Tiffen et output per person comes through changes in the
al. 1994, pp. 102104). However, evidence is lacking working-age share (equivalently, through the depen-
that such mechanisms are widespread in SSA. dency ratio): whether the magnitude of this effect is
only arithmetic, as above, or larger (or smaller) than
Dilution of reproducible capital (Solow model). In that is to be determined econometrically. The weak
economic models that allow only for reproducible form of the hypothesis says that all the effect comes
capital, rises in the number of people lead only through the age structure, allowing for the possibility
temporarily to capital dilution (falls in capital per that, for instance, the effects of young and old
head) and lower output per head. To see this in an dependency might be different. What the age-
idealized way, consider a once-for-all instantaneous structure hypothesis excludes are any effects of
10 per cent rise in population (and workforce). population growth per se on output per person.
Capital per head falls by 10 per cent at once, but To explore this hypothesis from the standpoint of
the extra workers are productive, generating extra a model in which all capital is reproducible, it is
savings, which cause capital per head to begin to rise helpful to decompose the sources of growth of
again. This compensating capital deepening is, how- output per person using an aggregate production
ever, self-limiting, because*given diminishing function.
returns*rises in capital per head raise the propor- Let Y T.F(K, L), where T.F(K, L) is a constant-
tion of savings per head that must be devoted to returns-to-scale production function with capital K,
replacing worn-out machines (and equipping new employed labour L, and technology parameter T;
workers if there is an underlying rate of population technological advance is represented by a rise in T.
growth greater than zero). Therefore, in the long run, Denote the labour force by LF. Then we can
capital per head and output per head return to their decompose g(Y/N), in a slightly different way from
initial equilibrium values. Similar reasoning implies (1), as follows:
that, as opposed to a one-off rise in population, a
sustained rise in its rate of growth does permanently gðY=NÞ ¼ gðY=LÞ þ gðL=NÞ (2)
lower capital per head and output per head. In ¼ ½gðTÞ þ a:gðK=LÞ
practice the distinction made here is less sharp than þ ½gðL=LFÞ þ gðLF=WAÞ þ gðWA=NÞ (3)
it appears, since simulations suggest that ‘tempora-
rily’ may mean many ‘for decades’ (Sato 1963). where a 0 is another technological parameter: the
elasticity of output per head with respect to capital
Age-structure effects. The fall in the dependency per head.
ratio in Phase 2 of the demographic transition yields Equation (2) breaks down growth in GDP per
a straightforward economic benefit as follows. De- head into growth in labour productivity (Y/L) and
20 Robert Eastwood and Michael Lipton

growth in the ratio of employment to population (L/ necessarily the same, in the rate of growth of the
N). Equation (3) breaks down each of these compo- working-age population. From (6) we see that in this
nents. Labour productivity growth comes either case, the net effects via (i) productivity growth
from technological advance (g(T)) or capital dee- (technological advance increased capital per
pening (g(K/L)). The ratio of employment to worker) and (ii) unemployment and participation
population can rise from a fall in the unemployment rates, have to add to zero. (i) and (ii) correspond to
rate (L/LF rises), a rise in the participation rate (LF/ the two square-bracketed terms in both (6) and (7).
WA), or a rise in the working-age share of the Might the effects via each of (i) and (ii) be zero?
population (WA/N). For simplicity, this assumes that Using (7), for (b), this would mean that employment
all labour force participants are of working age. growth was just keeping up with faster growth in the
To bring out the meaning of the strong form of the working-age population (Dg(L/WA)  0). Then for
age-structure hypothesis, consider any demographic (i), if there were no effect on technological advance,
change whatsoever and use ‘D’ to represent its there would have to be faster capital accumulation,
effects on any variable. Then the hypothesis can be financed either by higher domestic savings or capital
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

written in proportional form as inflow, so that faster employment growth was not
reducing capital per worker. The effects could add
DgðY=NÞ ¼ cDgðWA=NÞ (4)
up to zero in other ways: for instance, there might be
where g is a parameter, inserted to leave open the some capital dilution as a result of insufficient
question whether the effects of changes in the savings that was being offset by faster (Boserupian)
working-age share are just arithmetical (g1) or technological advance. Or else females’ labour
more or less than that. For instance, as above, if the market participation might be raised: if the unem-
working-age share has been constant (g(WA/N)  0) ployment rate was unaffected, then, from (6), the
and the demographic transition causes it to start contribution from (ii) would be positive. Summing
rising at 1 per cent per year, then if g 1 the growth up, for growth of output per person not to fall as a
of GDP/head is raised by 1 per cent per year. result of faster population growth, something has to
Pursuing this further, we can write (3) in ‘changes’ offset the extra capital dilution that this brings
form thus: about.
Nevertheless, the econometric literature has
DgðY=NÞ ¼ ½DgðTÞ þ a:DgðK=LÞ claimed not merely that the age-structure hypothesis
þ ½DgðL=LFÞ þ DgðLF=WAÞ is true, but that the effects are greater than
þ DgðWA=NÞ (5) arithmetical: g 1. As noted above, there are
mechanisms that might account for this. A fall in
which implies, after substitution of (4) and rearran- the dependency ratio*especially the fall in the
gement: young dependency ratio that a fall in fertility will
½DgðTÞ þ a:DgðK=LÞ þ ½DgðL=LFÞ þ DgðLF=WAÞ bring about*may well induce more participation of
females in the labour force and raised savings. As
¼ ðc  1Þ:DgðWA=NÞ:
regards savings, if the population is considered as
(6) consisting of dissaving dependents at both ends of
the age spectrum plus saving workers in the middle,
Equivalently
then the early consequences of a fall in fertility will
½DgðTÞ þ a:DgðK=LÞ þ ½DgðL=WAÞ be to raise savings ratios, by reducing the weight of
¼ ðc  1Þ:DgðWA=NÞ: (7) young dissavers in the population.
If falls in fertility simultaneously raise females’
Continuing with the special case g1, implying that participation and savings, then an appealing scenario
the right-hand sides of (6) and (7) are zero, the presents itself: at the same time as a rise in females’
strong form of the age-structure hypothesis is then participation is swelling the labour force, both
equivalent to the claim that the terms on the left- financial and labour-market conditions will be tend-
hand sides add to zero (with g1 they would have to ing to favour the investment that will facilitate
add to more than zero, if the demographic change gainful employment of the extra labour. The extent
was improving the age structure, i.e., Dg(WA/ to which raised domestic savings are a necessary
N) 0). ingredient of this naturally depends on how impor-
For example, suppose there is a rise in the rate of tant are barriers to the international mobility of
population growth accompanied by a rise, not capital.
Demographic transition in sub-Saharan Africa 21

What can the data tell us about the consequences technology gap. From this comes the idea that a
of demographic change for SSA development? given country at a given time has a steady-state
equilibrium value of (labour) productivity, defined
Evidence from cross-country regressions. The as what productivity would be if the gaps were closed,
mainstream economics literature on the conse- and that productivity growth depends partly on the
quences of demographic change for development gap between actual and steady-state productivity. In
has been dominated by cross-country regression spite of its name, the convergence framework does
analysis. Cross-country growth regressions, founded not imply that convergence will necessarily be
observed, since international gaps in steady-state
in the ‘convergence’ framework described below,
productivities could exceed those in actual produc-
have sought to quantify both capital-dilution and
tivities, perhaps because of low savings rates or rapid
age-structure effects of demographic change on the
natural increase in some poor countries (on diver-
growth rate of output per head. Some of this work
gence in practice, see Pritchett 1997). Evidence on
has used a pure cross section of countries, with the
this for SSA is explored in the last part of this section.
dependent variable being, say, their growth of output
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

A demography-oriented elaboration of the con-


per head in the period 196590; other work has also
vergence framework is as follows:
sought to exploit within-country variation over time,

and has used panel data, with the dependent variable gðY=LÞ ¼ dððY=LÞ  ðY=LÞÞ: (8)
being growth over several 5-year or 10-year periods.
The use of the cross-country regression method has This is the ‘convergence’ hypothesis: the rate of
not been confined to growth regressions. The finding growth of productivity, g(Y/L), depends on how far
by some of large age-structure effects has motivated productivity falls short of its ‘steady-state’ value, (Y/
others to shed light on the mechanisms, looking L)*. d is the speed of convergence, assumed constant
especially for age-structure effects on savings. Cross- across time and space.
country regression methods have also been used to 
ðY=LÞ ¼ Xb (9)
look for effects on poverty, conditional on growth
(Eastwood and Lipton 1999). where b is a vector of coefficients and X is a vector
One reservation should be noted. The research of variables that are held to influence steady-state
aims to give evidence of causal links from demo- productivity. Those employed by Barro (1997) in-
graphy to development, but ultimately what is clude education (of males, secondary and tertiary),
exhibited is correlation. Correlation between X and health (e.g., life expectancy), the share of govern-
Y may reflect causation from X to Y, but also ment consumption in GDP, and proxies for the rule
possible are reverse causation from Y to X and of law and the robustness of democratic institutions.
incidental association, reflecting causal links from Demographic variables may also be included. The
some omitted variable to both Y and X. In practice, Solow model implies, as noted above, that faster
as shown below, these possibilities are very hard to population growth lowers (Y/L)*; higher population
exclude. This has led to scepticism among econo- density might raise it.
mists about the usefulness of the method.
gðY=NÞ gðY=LÞ þ gðLÞ  gðNÞ
Cross-country growth econometrics has, overwhel-
mingly, been based on the so-called ‘convergence’ or ¼ gðY=LÞ þ gðWAÞ  gðNÞ (10)
‘technology-gap’ framework (Barro 1991, 1997;
The first part of (10) is an identity; the second part
Kelley and Schmidt 2005, 2007). The underlying
replaces g(L) with g(WA), by assuming that
idea is that economic growth is partly a matter of
unemployment and participation rates are constant
‘catch-up’. Even if a less developed country has low
through time.
capital per worker, it is rising through time*and
taking productivity up with it*for two reasons. First, gðY=NÞ ¼ / þ dXb  dðY=LÞ þ c1 gðWAÞ
domestic savings, plus inflow of capital from abroad,  c2 gðNÞ þ e (11)
is more than enough merely to equip new entrants to
the labour force with as much capital as existing Equation (11), used by Bloom and Williamson
workers have. Second, the country may be using (1998) and Bloom and Sachs (1998), is derived by
technology well inside the global technology frontier, eliminating (Y/L)* and g(Y/L) between (8) and (10)
and deriving some growth from closing this and making three amendments: (i) a constant f is
22 Robert Eastwood and Michael Lipton

added to allow for labour-augmenting technical geography and health, and g(L) and g(N) are also
progress at the same rate in all countries; (ii) included. Geography is proxied by the percentage of
coefficients g1 and g2 are inserted, allowing the land in the tropics and coastal and inland population
age-structure hypothesis to be tested rather than densities, and health by life expectancy at birth
imposed; (iii) an error term o is added. in 1965. Expanding the X vector in this way works,
In Bloom and Williamson and Bloom and Sachs in that the ‘Africa effect’, as measured by the dummy,
the dependent variable is the annual growth rate of is eliminated. Once the hypothesis that g1 g2 is
GDP per head over the period 196590, and the accepted, the regression is re-run to impose this and
sample of countries is all those (developed and less the estimated age-structure effect is represented by a
developed) for which data can be obtained. Bloom gamma of between 1.5 and 3.5, with the higher of
and Williamson ask how far the relatively rapid these figures being obtained if the sample is restricted
growth in East Asia is traceable to the demographic to African countries (Bloom and Sachs 1998, Table 6).
transition in that region; Bloom and Sachs ask As in Bloom and Williamson, g can be interpreted
whether relatively slow growth in Africa can be straightforwardly as a multiplier to be applied to the
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

explained by a combination of geographic and ‘arithmetic’ dividend in (10), potentially explicable


health variables, included in vector X, together via the savings and participation effects noted earlier.
with the lack of a demographic transition. In Table 3 we present estimates of this ‘arithmetic’
The first empirical question is whether age- dividend for SSA and its 16 most populous countries,
structure effects and capital-dilution effects are with some comparative data for Asia. Projected
both discernible. The strong version of the age- dependency ratios are UN ‘medium variant’ ones
structure hypothesis is represented by g1 g2 in (so that projected falls may be exaggerated, as
(11), since equal rises in the rates of population and
argued earlier) and the dividends are calculated
labour force growth then have no net effects on
using equation (2) above with gamma equal to unity.
growth in GDP per head. If ‘Solow-model’ capital
A comparison of SSA 19852025 with Asia 1965
dilution matters as well, then (g1 g2) should be
2005 shows a projected SSA arithmetic dividend of
significantly positive. Most studies employing cross-
0.32 per cent p.a., compared with 0.41 per cent for
country growth regressions do not have age-structure
Asia and 0.52 per cent for East Asia. South Africa’s
effects as a prime focus, so they suppress g(L) (in
early transition is visible*it is projected to enjoy a
effect imposing g1 0) and include g(N) to test for
dividend equal to that of Asia over the period 1965
capital dilution; generally speaking significant capi-
2005. Among the populous SSA countries, those
tal-dilution effects are not found (Kelley and
with the largest projected dividends are those where
Schmidt 1995). Bloom and Sachs and Bloom and
Phase 2 of the demographic transition is well
Williamson similarly find no evidence of capital
dilution, but they do find a more-than-arithmetical established*Cameroon, Ghana, Kenya, and Sudan
age-structure effect. Thus the hypothesis that g1 g2 are all projected to enjoy a dividend of 0.4 per cent
is not rejected, and when g1 g2 g is imposed, the or more. Clearly, if gamma really were as high as 3.5
estimate of g is significantly greater than unity. for African countries, then in these countries at least,
Bloom and Williamson attribute about a third of the age-structure demographic dividend in prospect
the ‘economic miracle’ in East Asia to the age- would be substantial (e.g., 1.9 per cent p.a. for Kenya
structure demographic dividend that East Asian over the period 19852025).
countries enjoyed during the period 196590. By including population densities in X, Bloom and
In their paper on Africa, Bloom and Sachs begin Sachs allow for both the positive effects via increas-
with a version of (11) which excludes the demographic ing returns and the negative ones via the dilution of
variables, so that only initial GDP per worker, three X natural capital discussed earlier. While, in the model,
variables (trade openness, a proxy for the quality of X variables are supposed to affect growth only via
institutions, and the central government deficit), and the steady-state level of GDP per worker, (Y/L)*,
an Africa dummy are included. One might expect the and thereby the productivity ‘gap’, their estimating
savings rate to be included as an X variable, but it is equation would evidently also pick up unmodelled
excluded on the ground that it should be considered country-specific effects on the rate of technical
endogenous, that is, dependent on the other X advance (i.e., on f). They find a significantly positive
variables. The coefficient on the Africa dummy coefficient on coastal density and an insignificantly
indicates an unexplained shortfall of growth in Africa negative one on interior density, results which (at
of about 2.2 per cent per annum (p.a.). Then the X face value) give evidence for ‘increasing returns’, but
vector is expanded to include variables representing not ‘natural capital dilution’.
Demographic transition in sub-Saharan Africa 23

Table 3 Age-structure dividends (g 1), countries of sub-Saharan Africa, and regions and selected countries of Asia,
selected years

Dependency ratios Dividends

1950 1965 1985 2005 2025 19652005 % per year 19852025 % per year

SSA 82 88 94 86 71 0.03 0.32


E SSA 85 92 96 90 74 0.03 0.30
Ethiopia 89 85 90 92 70 0.09 0.28
Kenya 78 108 112 83 71 0.32 0.54
Madagascar 71 92 94 90 67 0.03 0.37
Malawi 95 93 101 99 86 0.08 0.19
Mozambique 80 84 92 90 73 0.08 0.26
Sudan 89 89 91 79 58 0.14 0.47
Tanzania 93 93 96 91 82 0.03 0.19
Uganda 85 97 102 108 88 0.14 0.18
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

M SSA 82 86 96 95 75 0.12 0.28


Angola 79 92 99 95 76 0.04 0.31
Cameroon 76 82 96 83 66 0.01 0.42
DRC 90 88 96 101 78 0.17 0.24
S SSA 74 85 80 58 54 0.39 0.39
South Africa 73 84 77 56 53 0.41 0.36
W SSA 81 85 94 86 70 0.01 0.33
Burkina Faso 75 81 100 92 79 0.15 0.28
Côte d’Ivoire 83 85 93 82 66 0.04 0.38
Ghana 91 89 93 76 62 0.18 0.44
Nigeria 81 87 95 86 67 0.01 0.39
China 61 80 56 42 46 0.59 0.17
India 68 82 73 60 47 0.32 0.41
Indonesia 76 81 72 51 43 0.45 0.46
Thailand 83 93 63 43 48 0.75 0.24
Asia 67 80 67 53 48 0.41 0.30
E Asia 63 76 55 43 47 0.52 0.13
SC Asia 70 83 77 61 49 0.32 0.43
SE Asia 73 87 74 53 46 0.50 0.44
W Asia 74 89 81 61 50 0.40 0.47

Source: UN (2009a).

How credible are these cross-country findings? regressions cannot tell us how much of the negative
Unfortunately, such models are subject to important association between these two variables is attribu-
econometric reservations. Consistency of the para- table to reverse causation from growth of GDP per
meter estimates requires that the error term in (11) head to population growth.
is independent of the included regressors. There are One way of alleviating such endogeneity problems
reasons to doubt this independence here (Udry has been to use a panel of data, allowing a particular
1998). First, the error term amounts to the total application of the method of instrumental variables,
effect on the growth of GDP per head of all relevant described below. Kelley and Schmidt (2005, 2007)
variables that are not included as regressors. It is use the same data period as Bloom and Williamson
implausible that such variables are uncorrelated with and Bloom and Sachs but split the data into three
the level of GDP per head in 1965, which is one of decades plus a quinquennium (for 199095). Their
the regressors. There will surely be omitted serially estimating equation is the same as (11), although its
correlated variables that have affected national conceptualization is a little different. Their X vector
growth rates both before and after 1965, creating includes a set of politico-economic variables (de-
an incidental association between error term and rived from Barro’s work, noted earlier), alongside
regressor. Problems of endogeneity will also affect four demographic variables: youth and old-age
other regressors. In particular, causation is likely to dependency ratios, population and population den-
run in both directions between growth in GDP sity. The authors argue persuasively that the right
per head and population growth; cross-section way to model age-structure effects is to isolate the
24 Robert Eastwood and Michael Lipton

arithmetical effect in the term [g(W) g(N)], while that Z is (i) highly correlated with X, and (ii)
hoping to pick up effects on savings*effects that independent of the error term o. The normal
may, as they note, affect not only the steady state but practice, followed by Kelley and Schmidt, is to use
the speed of convergence towards it*via disaggre- lagged Xs as instruments, but this is controversial,
gated dependency ratios. They therefore test the since it has to assume that*while the included
age-structure hypothesis in its weak form. The causal variables, the Xs, are serially correlated, as
inclusion of population and population density may be expected with slow-moving X variables*the
means that they are also taking account of the omitted causal variables, whose aggregate influence
possibilities of increasing returns or dilution of comprise o, are not (Weil 2007, p. 1271). It is difficult
natural capital. to see how this assumption could ever be convin-
Kelley and Schmidt’s findings, like those of Bloom cingly defended, since, whatever properties lead
and Sachs and Bloom and Williamson, fit their model some variables to be included as regressors and
well if judged by the correspondence between the others to be excluded, that of being serially corre-
signs of estimated coefficients and theoretical ex- lated or not can hardly be one of them. Further, the
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

pectation (Kelley and Schmidt 2005, Table 1). There assumption cannot be tested, in view of the small
is support for the hypothesis that g1 g2 1 once the number of time periods*typically four or five.
demographic X variables are included in the model. The argument for age-structure effects that are
Of these four variables, only the young-age depen- more than arithmetical rests in part on the hypoth-
dency ratio is significant, but its estimated effects are esis that a fall in the dependency ratio raises savings.
very large compared to the arithmetic dividends What is the evidence for this? For East Asia, there is
reported in Table 3. Those dividends were calculated certainly a strong negative association between
against a benchmark of unchanged age structure. On young-age dependency and aggregate savings as a
the same basis, the projected decline in the youth share of income. Between the mid-1960s and the
dependency ratio in SSA from 88 in 1985 to 64 in early 1990s, young-age dependency rates in some
2025 would, using the parameter estimate given in East Asian countries fell from about 40 per cent to
Kelley and Schmidt (2007, Table 3.1), raise average 25 per cent or less, while the savings ratio rose from
growth over the years 19852025 by 0.59 per cent about 20 per cent to about 35 per cent.
p.a., nearly double the arithmetic dividend. The Empirical work to test whether this association
corresponding figure for Asia 19652005 is 1.14 per can be interpreted as a causal consequence of life-
cent p.a., nearly 3 times the arithmetic dividend. cycle savings has two strands. The first consists of
It is worth noting some contrasts between the panel studies on country-level data using the method
Bloom and Sachs and Kelley and Schmidt ap- described above. Thus Higgins and Williamson
proaches. While it is true that some of the variation (1997), on the basis of a panel of annual observations
across time in the data is exploited by Kelley and over 40 years on 16 Asian countries, claim that
Schmidt, they do include dummy variables for each nearly all (about 13 percentage points) of the rise in
time period other than the first, and, as the authors the savings rate can be attributed to the life-cycle
acknowledge, these dummies swallow up most of the mechanism via reduced dependency burdens. How-
aggregate variation in growth from decade to decade ever, it must be questionable whether year-on-year
(Kelley and Schmidt 2005, p. 294 and Table 2). As variations in noisy age structure and savings rates
regards variation across space, Kelley and Schmidt can be informative, especially given what are likely
include regional dummy variables and, given the to be complex dynamics. In a reconsideration of this
limited range of their included regressors, are not work Schultz (2004) suggests an effect that is only
able to eliminate a strong ‘Africa effect’: African one-tenth as great and not statistically significant.
growth, ceteris paribus, is estimated to be between A second strand of research (Deaton and Paxson
0.7 per cent and 1.3 per cent p.a. slower than growth 1997, 2000; Lee et al. 2000) uses household data to
in Asia, Europe, and Latin America (Kelley and identify age and cohort effects on savings, using
Schmidt 2005, Table 1, col. 5). simulation methods to estimate the effects of demo-
The use of the instrumental-variables method in graphic change on aggregate savings. This research is
the estimation of the model is designed to circum- based on the life-cycle model of savings, according to
vent endogeneity issues of the kind discussed earlier. which individuals follow a consistent pattern of
Space precludes extensive discussion (for an intro- smoothing consumption through the life span, per-
duction, see Angrist and Krueger 2001), but the iods of dissaving early and late in life being balanced
essential requirement is that one can identify an by a hump of saving in between. Demographic
‘instrument’ Z for an endogenous regressor X, such change affects the aggregate savings rate in this
Demographic transition in sub-Saharan Africa 25

model via changes in age structure. In particular, the effects of HIV on development in South Africa.
fertility decline should raise the savings rate by Such models involve strong simplifying assumptions
reducing the proportion of young dissavers in the and have attracted powerful critiques (e.g., Bell et al.
population. Deaton and Paxson (2000), using Tai- 2006, who emphasize adverse effects of HIV on
wanese data for the period 197695, test the life- human capital accumulation via the diversion of
cycle hypothesis rather than imposing it, and their public resources from education towards care of the
critical finding is that the hypothesis in its simplest sick, and the effects of both death of a parent and
form fails, a result confirmed in their work on other reduced child life expectancy on private incentives
countries (Deaton and Paxson 1997). Successive to educate children).
cohorts of Taiwanese individuals at every age are Going further away from estimation than this is
found to save higher fractions of their permanent recent work (World Bank 2006) that eschews the
incomes, a result which the life-cycle model would strong assumptions required for cross-country
have to attribute to a trend of strengthening of the econometric analysis in favour of strong theoretical
bequest motive. The household saving rate in Taiwan assumptions and country-by-country calculation.
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

rose from about 10 per cent in 1970 to about 30 per The primary aim of this work is to investigate at
cent in 1990, but the authors can attribute, at most, country level whether savings are consistent with
only one-fifth of this rise to the effect of the sustainability, defined as non-decreasing wealth per
demographic transition (Deaton and Paxson 2000, head. It is assumed that output per head is a function
p. 167). of wealth per head, with no exogenous technical
Summing up, what does the cross-country econo- change. Slower population growth, ceteris paribus,
metric literature claim about the importance in lowers the savings share consistent with sustainabil-
practice of the four causal channels identified ear- ity and therefore raises sustainable consumption per
lier? Population density is included in some of the head*a demographic dividend defined slightly dif-
research, acting as an imperfect proxy for the ferently from that found in the age-structure litera-
dilution of natural capital or increasing returns or ture. For a given deceleration of population growth
both, but does not provide strong evidence for either over, say, a 40-year interval, we can represent the
causal channel, although the finding by Bloom and gain in consumption per head in the form of an
Sachs that coastal density raises growth per head annual growth rate, making it broadly (but not
provides some weak support for increasing returns. exactly) commensurate with the growth in output
The strongest claim in the literature relates to the per head identified in the age-structure approach
relative importance of the dilution of reproducible and shown in Table 3.
capital and benefits from improved age structure, Before estimating such ‘reduced-dilution’ divi-
coming out in favour of the latter and arguing that dends, we consider the World Bank’s estimates of
greater-than-arithmetical age-structure effects arise, dilution itself. These are interesting in themselves
in particular, because reduced youth dependency and offer an important contrast between SSA and
raises savings. However, neither conclusion is firmly Asia. The algebra is as follows. With W denoting
established. Both rest on a controversial macro- wealth and other notation as before (Y now stands
econometric method, and investigations of the sav- for income rather than GDP), we can write the
ings channel using microeconomic data suggest for growth rate of W/N as
East Asia that most of its rise in savings cannot be
attributed to the demographic transition. gðW=NÞ gðWÞ  gðNÞ DW=W  gðNÞ (12)

But the change in wealth, DW, is equal to adjusted


net savings, that is, the product of the net savings
Capital stocks, savings, and sustainability
rate and income, with an adjustment, explained
below. To keep W/N constant, the required savings
One response to doubts over the use of aggregate
rate, denoted sR, can therefore be obtained by
econometric methods to identify links from demo-
setting equation (12) equal to zero, substituting
graphy to development has been to use relationships
DW sRY, and rearranging:
estimated using household data as inputs to macro-
economic simulation models, with parameters sR ¼ ðW=YÞgðNÞ (13)
mostly obtained by calibration rather than estima-
tion. For example, Young (2005) uses micro-level Then the ‘savings gap’ is defined as the gap between
estimates of the effects of HIV infection on fertility this ‘required’ savings rate (to keep consumption per
as an input to a long-run macro simulation model of person just sustainable, that is, without depleting
26 Robert Eastwood and Michael Lipton

capital per person) and the adjusted net savings rate. from net saving conventionally defined (i.e., gross
Table 4 shows estimated savings gaps for the 11 out saving minus the depreciation of reproducible capi-
of the 16 most populous countries in SSA for which tal) by adding education expenditure and deducting
the data are available. What is striking is how large the consumption of natural capital, comprising
these gaps are*over 10 per cent of income for all petroleum resources, mineral resources, and forest
but South Africa, Kenya, and Ghana*in spite of the resources (excluding soil degradation and the deple-
fact that gross savings rates in most of these tion of fish stocks). For some SSA countries the
countries already exceed 10 per cent. consumption of natural capital is important, as
To grasp these results, note two features of the indicated by the difference between columns 1 and
World Bank’s method. The first is the measurement of 2 in Table 4: among our eleven countries, oil
national wealth, which comprises ‘tangible wealth’ depletion has a very large impact for Nigeria, and
and ‘intangible wealth’ (for details on ‘intangible a significant one for Cameroon, while net deforesta-
wealth’, see World Bank 2006, p. XIV). What matters tion looms large for Ethiopia.
here is ‘tangible wealth’, W, defined to include not The idea behind these calculations is that repro-
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

only reproducible capital but also land and natural ducible capital can substitute for natural capital ($
resources, defined as arable land, pasture, timber and for $ at the margin), for example, enough extra
non-timber forest resources, protected areas, and roads, factories, or irrigation can make up for
subsoil resources (fish stocks and subsoil water are excessive depletion of mineral resources or the
excluded). Reproducible capital averages only 29 per crowding of land resources that population growth
cent of tangible wealth in the eleven large SSA entails. However, population growth rates of 2 per
countries considered, so the dilution of wealth result- cent per year or more require very high savings rates
ing from population growth is, on average, over 3 if wealth per capita is not to fall. Table 4 shows that
times as great as would be the case if only reproducible in Burkina Faso, where depletion of exhaustible
capital were included. As equation (13) shows, a high resources is not a significant factor and where
value of W/Y (i.e., low capital productivity) raises the adjusted saving is over 6 per cent of national income,
required savings rate: inclusion of natural capital in the required savings share is 27.5 per cent and so
the analysis tends to raise W/Y in SSA relative to Asia, there is a savings gap of 21 per cent. Since raising the
where natural capital is generally a lower fraction of savings share by an amount of this order is hard to
wealth, and so contributes to the relatively high contemplate, the implication might appear to be that
required savings rates for SSA that are shown in a marked fall in the rate of population growth is a
Table 4. sine qua non of a transition to sustainability. This
The second feature of the World Bank’s method is argument cannot be transplanted without caveats to
the definition of adjusted net saving. This is obtained countries, such as Nigeria, where the counterpart of

Table 4 Savings gaps in sub-Saharan Africa and Asia, 2000

Net saving Adjusted net Reproducible Tangible Natural Required net Savings
% of saving % of capital share of wealth to increase, % saving % of gap % of
income income tangible wealth income ratio 200005 income income

Burkina Faso 4.0 6.3 0.4 8.87 3.1 27.5 21.2


Cameroon 5.7 1.3 0.27 11.83 2.33 27.6 28.9
Côte d’Ivoire 0.7 0.9 0.24 6.59 2.52 16.6 17.5
Ethiopia 4.5 4.0 0.18 9.63 2.71 26.1 30.1
Ghana 8.4 6.5 0.34 7.93 2.29 18.2 11.7
Kenya 5.7 11.6 0.39 6.52 2.59 16.9 5.3
Madagascar 1.7 3.5 0.19 8.47 2.85 24.1 20.6
Malawi 3.8 0.9 0.41 8.19 2.91 23.8 24.7
Mozambique 3.8 7.6 0.31 7.88 2.66 21.0 13.4
Nigeria 17.3 32.5 0.14 15.85 2.44 38.7 71.2
South Africa 2.4 8.7 0.68 3.76 1.43 5.4 3.3
China 29.8 28.0 0.55 6.14 0.74 4.5 23.4
India 14.6 14.1 0.45 6.47 1.65 10.7 3.4
Indonesia 15.4 3.1 0.42 9.08 1.41 12.8 9.7
Thailand 15.9 17.6 0.32 5.81 0.68 4.0 13.7

Source: World Bank (2006), UN (2009a).


Demographic transition in sub-Saharan Africa 27

a high savings gap is rapid oil depletion. To the happened in Asia in the 1950s and 1960s (Lipton
extent that oil or other mineral rents are accruing to 1977, chap. 10).
foreigners or national elites, it may be that the gap How robust are these results? Two reservations
between consumption and sustainable consumption may be noted, pulling in opposite directions. First,
for the mass of the population is much smaller than technological advance*whether exogenous or via
the calculation would indicate. Working the other the closing of an international technology gap*is
way, some countries have high annual rates of excluded. Otherwise, constant wealth per head
depletion of soil nutrients (and fish stocks), which would allow rising, not constant, consumption per
are probably due in part to population increase, but head, undermining the sustainability calculations
not accounted for here (Haileselassie et al. 2005; given here (Weitzman and Löfgren 1997). Second,
Henao and Baanante 2006). natural and reproducible capital are assumed per-
What are the results of similar calculations for fectly substitutable. If they are not, then $ for $
Asian countries? If it could be shown that Asia in substitution of natural capital with reproducible
1965 or 1985 had similar savings gaps to SSA in 2000, capital will fail to sustain output and, as a result,
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

then the credibility of this exercise would be wea- the numbers in Table 4 will underestimate savings
kened. Estimates of pre-2000 wealth levels are not gaps (compare Weil and Wilde 2009).
available, so that equation (13) cannot be applied We now turn to the effects of the fall in natural
historically, but for 2000, gaps can be calculated increase during Phase 2. For SSA, we take the
for nine populous Asian countries (China, India, projected change in annual natural increase between
Indonesia (Table 4), Korea, Malaysia, Nepal, Philip- a reference period representing the end of Phase 1,
pines, Sri Lanka, and Thailand). In all of these taken as 197085 and the first 40 years of Phase 2,
countries with the exception of Indonesia (an oil taken as 19852025. For comparisons with Asia, the
producer to which arguments parallel to those for two periods chosen are 195065 and 19652005. Falls
Nigeria apply) the gaps are negative: net savings in natural increase between the two periods (col-
rates are high and easily sufficient to outweigh umns 2 and 3 of Table 5) imply from equation (13)
dilution. Estimated net savings rates over the period that required savings rates fall, raising sustainable
19702000 averaged about 20 per cent for East Asia consumption per head by given amounts (col. 5).
and the Pacific, little more than zero for SSA (World These rises are quite substantial, 5.8 per cent for
Bank 2006, Figure 3.3, p. 41). This gap may be Ghana and 6.8 per cent for Kenya, for example, but
exaggerated owing to the under-recording of non- when they are converted to rates of increase over 40
monetized savings in SSA, as appears to have years, so as to make them roughly comparable with

Table 5 Reduced dilution dividends: selected countries of sub-Saharan Africa (197085 to 19852025) and Asia (195065
to 19652005)

Natural Tangible
increase: SSA Natural increase: wealth: Change in sustainable Change in sustainable
197085 Asia SSA 19852025 multiple of consumption/head: share consumption/head:
195065 Asia 19652005 income of national income annual change, per cent

Burkina Faso 2.73 3.03 8.87 2.7 0.07


Cameroon 2.84 2.31 11.83 6.2 0.15
Côte d’Ivoire 3.54 2.50 6.59 6.8 0.17
Ethiopia 2.63 2.63 9.63 0.0 0.00
Ghana 3.01 2.28 7.93 5.8 0.14
Kenya 3.72 2.68 6.52 6.8 0.16
Madagascar 2.72 2.69 8.47 0.2 0.01
Malawi 3.31 2.92 8.19 3.2 0.08
Mozambique 2.45 2.25 7.88 1.5 0.04
Nigeria 2.71 2.29 15.85 6.8 0.16
South Africa 2.42 1.16 3.76 4.7 0.12
China 1.83 1.52 6.14 1.9 0.05
India 1.93 2.07 6.47 0.9 0.02
Indonesia 1.99 1.91 9.08 0.8 0.02
Thailand 2.97 1.70 5.81 7.4 0.18

Source: World Bank (2006), UN (2009a).


28 Robert Eastwood and Michael Lipton

the age-structure dividends in Table 3, much smaller case for policy to cut young-age mortality, alongside
numbers result. fertility, in rural areas.
Thus the message of Table 5 is simple. Whether or We then turned to the economic consequences of
not natural increase is acting as a drag on SSA the demographic transition, specifying the channels
development, as the World Bank research suggests, through which demographic change might have
changes in it seem in most cases to have a small macro-economic consequences. Phase 2 of the
projected effect on sustainable consumption per transition entails falls in both dependency and
head. The projected SSA age-structure dividends natural increase, each of which may deliver a
over the period 19852025 in Table 3 range from demographic dividend. In each case, accounting
0.18 per cent p.a. to 0.54 per cent p.a., while those can be used to calculate an ‘arithmetic’ dividend
from reducing the dilution of capital in Table 5 range which, under strong assumptions specific to that
from minus 0.07 per cent p.a. to plus 0.17 per cent case, constitutes the total dividend. For dependency
p.a. (negative numbers indicate rises in natural falls, the arithmetic dividend for SSA over the
increase in countries that remain in Phase 1 for period 19852025, defined in terms of the growth
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

much of the period). Estimates for Asian countries* of GDP per person, is 0.32 per cent per year. For
inaccurate to the extent that they are based on falls in natural increase, data gaps prevent us from
tangible wealthincome ratios for 2000*are simi-
calculating a figure for SSA as a whole, but for
larly small in comparison with the estimated (arith-
eleven of its populous countries the median arith-
metic) age-structure dividends. The small numbers
metic dividend*here defined in terms of growth of
in Table 5 are, of course, an artefact of the chosen
sustainable consumption per person*is 0.12 per
counterfactual. If fertility were to decline so much
cent per year.
faster than shown in UN middle projections that
An influential body of econometric work has
Phase 2 ended with the rate of natural increase equal
supported an ‘age-structure hypothesis’ that all of
or close to zero, then dividends 5 or 6 times larger
the macro-economic effects of the transition are
would, on average, be the result.
mediated through age-structure change*that natur-
al increase as such is irrelevant. In its strong form,
the hypothesis claims that the working-age share
5. Conclusions
(equivalently, the dependency ratio) is a sufficient
statistic for the age structure; the econometrics then
We first reviewed the demographic transition in
suggests a growth dividend 1.53.5 times greater
SSA. In Phase 1, compared with Asia, dependency
than the arithmetic dividend. In its weak form, the
and natural increase peaked some 20 years later in
hypothesis allows for differential effects from young-
SSA*and higher, by about 14 points and 0.6 per
age and old-age dependency, and strong effects from
cent, respectively. UN middle projections suggest
that in Phase 2 both will fall more slowly in SSA young-age dependency, perhaps via savings beha-
than they did in Asia, and these projections depend viour, are found. Reservations about these results
on accelerated falls in African total fertility. This is arise, however, both for technical reasons and
driven mainly by earlier falls in young-age mortality, because of doubts about whether findings that
but this continues to fall relatively slowly in SSA. depend a good deal on Asian experience can be
Since that is not taken into account in UN projec- projected onto SSA’s future.
tions, these probably overstate future TFR declines. On the latter point, the achievement of even the
Yet there is sufficient national diversity to suggest arithmetic age-structure dividend requires that de-
possible policy-driven acceleration of falls in young- mographic transition does not reduce labour pro-
age mortality, though the immediate effect would be ductivity, in spite of a continuing rise in the labour
to raise both dependency and natural increase. Even force. That this was achieved in Phase 2 in much of
in countries where total fertility has fallen overall, it Asia was due to the sharp rise in the savings rate,
has remained stubbornly high in some regions, and only about a fifth of which can be convincingly
in rural areas. Despite urbanization, for some time attributed to dependency falls, to improved educa-
the demographics of such areas will increasingly tion of the workforce, and to rapid, labour-intensive
dominate national outcomes. This strengthens the agrotechnical progress on smallholdings, especially
Demographic transition in sub-Saharan Africa 29

in the period 196585. Such achievements have interaction among determinants of labour-productivity,
proved more elusive in Africa. suggest an efficient route to higher demographic
Falling natural increase delivers a demographic dividend: could complementary policies cut young-age
dividend in terms of sustainable consumption, by mortality, raise availability of contraception, improve
reducing the proportion of income that has to be the education and earning prospects of girls, and
saved to prevent capital per person from falling. absorb extra workers productively?
Borrowing data and methods from the World Bank,
we have shown that falls in natural increase during
Phase 2 yield rather small dividends of this form. Far Notes
more significant, and in sharp contrast to Asia, is the
alarming implication that continuing high levels of 1 Robert Eastwood and Michael Lipton are Senior
natural increase, in combination with low savings Lecturer and Research Professor in the Department of
rates and low capital productivity, render current Economics at the University of Sussex. Eastwood is
consumption per person in most of SSA unsustain- corresponding author: email r.k.eastwood@sussex.ac.uk
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

able. Matters are worse if man-made capital cannot 2 The authors thank three reviewers, Guenther Fink, and
readily be substituted for natural capital, or if an Bruno Schoumaker for helpful comments on an earlier
environmental resource is progressively degraded draft and Kenneth Hill for sending us the IDEC Life
(perhaps below a critical threshold) because it is Tables.
used by more and more people.
One escape from such predicaments could be via
References
inflows of savings from abroad, attracted by the high
marginal returns to capital that one might expect to André, C. and J. Platteau. 1998. Land relations under
be available in capital-scarce economies. But recent unbearable stress: Rwanda caught in the Malthusian
work that takes proper account of natural capital trap, Journal of Economic Behavior and Organisation
suggests that the marginal product of reproducible 34(1): 147.
capital is not high in SSA, which may help to account Angeles, L. 2010. Demographic transitions: analyzing the
both for low private capital inflows and low domestic effects of mortality on fertility, Journal of Population
savings rates (Caselli and Feyrer 2007). If this is Economics 23(1): 99120.
correct, then the prospects for stemming capital Angrist, J. D. and A. B. Krueger. 2001. Instrumental
dilution in SSA in the coming decades will depend variables and the search for identification: from supply
on technological advances raising the marginal and demand to natural experiments, Journal of Eco-
product of capital, unless greater falls in natural nomic Perspectives 15(4): 6985.
increase than currently projected can be achieved. Barro, R. 1991. Economic growth in a cross section of
Other things equal, it will be advances in the least countries, Quarterly Journal of Economics 106(2): 407
capital-intensive sectors (e.g., agriculture) where the 444.
greatest potential benefits are likely to lie, simply Barro, R. 1997. Determinants of Economic Growth: A
because a given investment will be complemented Cross-Country Empirical Study. Cambridge, MA: MIT
with more extra labour in such sectors. Such Press.
advances are not impossible: capital pessimism was Becker, G. 1960. An economic analysis of fertility, in G.
also a feature of debates on Asian economic devel- Becker (ed.), Demographic and Economic Change in
opment in the 1950s and 1960s (e.g., Higgins 1959), Developed Countries. Princeton, NJ: Princeton Univer-
and a series of key technological advances in sity Press and NBER.
agriculture (the Green Revolution) helped to pre- Becker, G. 1981. A Treatise on the Family. Cambridge, MA:
vent its predictions from being realized. Harvard University Press.
Finally, neither arithmetic nor econometrics Bell, C., S. Devarajan, and H. Gersbach. 2006. The long-
determines the future of demographic variables any- run economic costs of AIDS: a model with an applica-
where, nor the resulting economic gains in Phase 2. tion to South Africa, World Bank Economic Review
We have hinted at relevant policy choices, but a 20(1): 5589.
general issue arises. Might mutual causation (e.g., Birchenall, J. and R. Soares. 2009. Altruism, fertility, and
between reduction in fertility and in mortality), and the value of children: health policy evaluation and
30 Robert Eastwood and Michael Lipton

intergenerational welfare, Journal of Public Economics Dawson, M. 1987. Health, nutrition and population in
93(12): 280295. central Kenya, 18901945, in D. Cordell and J. Gregory
Bloom, D. and J. Sachs. 1998. Geography, demography, (eds.), African Population and Capitalism: Historical
and economic growth in Africa, Brookings Papers on Perspectives. Boulder, CO: Westview, pp. 201217.
Economic Activity 29(2): 207296. Deaton, A. and C. Paxson. 1997. The effects of economic
Bloom, D. and J. Williamson. 1998. Demographic transi- and population growth on national saving and inequal-
tions and economic miracles in emerging Asia, World ity, Demography 34(1): 97114.
Bank Economic Review 12(3): 419456. Deaton, A. and C. Paxson. 2000. Growth, demographic
Bloom, D., D. Canning, and P. Malaney. 2000. Population structure, and national saving in Taiwan, Population
dynamics and economic growth in Asia, Population and and Development Review 26(Suppl.): 141173.
Development Review 26(Suppl.): 257290. Diamond, J. 2005. Collapse: How Societies Choose to Fail
Bongaarts, J. 1978. A framework for analysing the or Survive. London: Allen Lane.
proximate determinants of fertility, Population and Doepke, M. 2005. Child mortality and fertility decline:
Development Review 4(1): 105132. does the BarroBecker model fit the facts? Journal of
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

Bongaarts, J. 1982. The fertility-inhibiting effects of the Population Economics 18(2): 337366.
intermediate fertility variables, Studies in Family Plan- Drèze, J. and M. Murthi. 2001. Fertility, education, and
ning 13(67): 178189. development: evidence from India, Population and
Boserup, E. 2005. Conditions of Agricultural Growth: The Development Review 27(1): 3363.
Economics of Agrarian Change under Population Eastwood, R. and M. Lipton. 1999. The impact of changes
Pressure. New Brunswick: Transaction (1st edn 1965. in human fertility on poverty, Journal of Development
London: Allen and Unwin). Studies 36(1): 130.
Brass, W. and C. Jolly. 1993. Population Dynamics in Eastwood, R. and M. Lipton. 2004. The ruralurban
Kenya. Washington, DC: Commission on Behavioral dimension of inequality change, in G. A. Cornia (ed.),
Sciences and Education, National Research Council. Inequality, Growth and Poverty in an Era of Liberali-
Bulatao, R. 2006. Population and mortality after AIDS, in zation and Globalization. Oxford: Oxford University
D. Jamison, R. Feachem, M. Makgoba, E. Bos, F. Press, pp. 112141.
Baingana, K. Hofman, and K. Rogo (eds.), Disease Espenshade, T., J. Guzman, and C. Westoff. 2003. The
and Mortality in Sub-Saharan Africa, 2nd edn. Wa- surprising global variation in replacement fertility,
shington, DC: World Bank, pp. 5374. Population Research and Policy Review 22(56):
Caselli, F. and J. Feyrer. 2007. The marginal product of 575583.
capital, Quarterly Journal of Economics 122(2): 535 Ezeh, S., B. Mberu, and J. Emina. 2009. Stall in fertility
568. decline in Eastern African countries: regional analysis
Cassen, R., T. Dyson, and L. Visaria (Eds) 2004. Twenty- of patterns, determinants and implications, Philosophi-
First Century India: Population, Economy, Human cal Transactions of the Royal Society B 364(1532):
Development and the Environment. Oxford: Oxford 29913007.
University Press. Fage, J., A. Roberts, and R. Oliver. 1986. Cambridge
Coale, A. and E. Hoover. 1958. Population Growth and Economic History of Africa Vol. 7:. Cambridge: Cam-
Economic Development in Low Income Countries. bridge University Press.
Princeton, NJ: Princeton University Press. Garenne, M. and V. Joseph. 2002. The timing of the
Coale, A. 1973. The demographic transition, in Proceed- fertility transition in sub-Saharan Africa, World Devel-
ings of International Population Conference, Liège, opment 30(10): 18351843.
1973, Volume 1. Liège: International Union for the Haileselassie, A., J. Priess, E. Veldtkamp, D. Teketay, and
Scientific Study of Population, pp. 5372. J.-P. Lessche. 2005. Assessment of soil nutrient deple-
Cohen, B. 1993. Fertility levels, differentials, and trends, in tion and its spatial variability on smallholders mixed
K. Foote, K. Hill, and L. Martin (eds.), Demographic farming systems in Ethiopia using partial versus full
Change in Sub-Saharan Africa. Washington, DC: Na- nutrient balances, Agriculture, Ecosystems and Envir-
tional Academy Press, pp. 867. onment 108(1): 116.
Cohen, B. 1998. The emerging fertility transition in sub- Henao, Julio and C. Baanante. 2006. Agricultural Produc-
Saharan Africa, World Development 26(8): 14311461. tion and Soil Nutrient Mining in Africa: Implications for
Conley, D., G. McCord, and J. Sachs. 2007. Africa’s lagging Resource Conservation and Policy Development. Mus-
demographic transition: evidence from exogenous im- cle Shoals, AL: International Fertilizer Development
pacts of malaria ecology and agricultural technology. Corporation.
Working Paper 12892, National Bureau of Economic Higgins, B. 1959. Economic Development: Principles,
Research, Cambridge, MA. Problems and Policies. New York: Norton.
Demographic transition in sub-Saharan Africa 31

Higgins, M. and J. Williamson. 1997. Age structure Malthus, T. R. 1824. Population (Encyclopedia Britannica
dynamics in Asia and dependence on foreign capital, (4th6th edn)), in T. Huxley and J. Huxley (Eds) Three
Population and Development Review 23(2): 261293. Essays on Population. New York: Mentor.
Hill, K. and A. Amouzou. 2006. Trends in child mortality, Montgomery, K. 2009. The demographic transition. Avail-
19602000, in D. Jamison, R. Feachem, M. Makgoba, E. able: http://www.marathon.uwc.edu/geography/Demotr
Bos, F. Baingana, K. Hofman, and K. Rogo (eds.), ans/demtran.htm
Disease and Mortality in Sub-Saharan Africa, 2nd edn. Montgomery, M. and B. Cohen (Eds) 1998. From Death to
Washington, DC: World Bank, pp. 1530. Birth: Mortality Decline and Reproductive Change.
Huebler, F. 2008. Adult literacy in sub-Saharan Africa Washington, DC: Committee on Population, National
from Demographic and Health Surveys 20036. Avail- Research Council.
able: http://huebler.blogspot.com/2008/05/literacy.html Murthi, M. 2002. Fertility change in Asia and Africa,
IDRC (International Development Research Centre). World Development 30(10): 17691778.
2002. Population and Health in Developing Countries: Pingali, P., Y. Bigot, and H. Binswanger. 1987. Agricultural
Population, Health and Survival at INDEPTH Sites. Mechanization and the Evolution of Farming Systems in
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

Available: http://www.idrc.ca/en/ev-43000-201-1-DO_ Sub-Saharan Africa. Washington, DC: World Bank.


TOPIC.html Populstat. N.d.-a. http://www.populstat.info/Asia/chinac.
Iliffe, J. 1995. Africans: The History of a Continent. htm
Cambridge: Cambridge University Press. Populstat. N.d.-b. http://www.populstat.info/Africa/safri
Kelley, A. and R. Schmidt. 1995. Aggregate population cac.htm
and economic growth correlations: the role of the Preston, S. 1978. The Effects of Infant and Child Mortality
components of demographic change, Demography on Fertility. New York: Academic Press.
32(4): 543555. Pritchett, L. 1997. Divergence big time, Journal of
Kelley, A. and R. Schmidt. 2005. Evolution of recent Economic Perspectives 11(3): 317.
economic-demographic modeling: a synthesis, Journal Rajaratnam, J., J. Marcus, A. Flaxman, H. Wang, A. Levin-
of Population Economics 18(2): 275300. Rector, L. Dwyer, M. Costa, A. Lopez, and C. Murray.
Kelley, A. and R. Schmidt. 2007. A century of demo- 2010. Neonatal, postneonatal, childhood, and under-5
graphic change and economic growth: the Asian mortality for 187 countries, 19702010: a systematic
experience in regional and temporal perspective, in A. analysis of progress towards Millennium Development
Mason and M. Yamaguchi (eds.), Population Change, Goal 4, Lancet 375, pp. 19882008. Available: http://
Labor Markets and Sustainable Growth: Towards a www.thelancet.com/journals/lancet/article/PIIS0140-
New Economic Paradigm. Amsterdam: Elsevier, pp. 6736(10)60703-9/fulltext
3974. RGCC (Registrar General & Census Commissioner).
KNBS (Kenya National Bureau of Statistics) and ICF 2001. India, Provisional Population Totals, Census of
Macro. 2010. Kenya Demographic and Health Survey India 2001, Series 1. Delhi: Controller of Publications.
200809. Calverton, MD: KNBS and ICF Macro. Ross, R., E. Abel and K. Abel. 2004. Planteaus during the
Kremer, M. 1993. Population growth and technological rise of contraceptive prevalence. Policy Working Paper
change: one million BC to 1990, Quarterly Journal of Series No. 10. Washington DC: USAID.
Economics 108(3): 681716. Rutstein, S. 2008. Further evidence of the effects of
Lee, R., A. Mason, and T. Miller. 2000. Life cycle saving preceding birth intervals on neonatal, infant, and
and the demographic transition: the case of Taiwan, under-five-years mortality and nutritional status in
Population and Development Review 26(Suppl.): 194 developing countries: evidence from the demographic
219. and health surveys. DHS Working Paper 41, US
Lee, R. 2003. The demographic transition: three centuries Agency for International Development, Washington,
of fundamental change, Journal of Economic Perspec- DC.
tives 17(4): 167190. Rutstein, S., M. Ayad, R. Ren, and R. Hong. 2009.
Lipton, M. 1977. Why Poor People Stay Poor: Urban Bias Changing Health Conditions and the Decline of Infant
and World Development. London: Temple Smith. and Child Mortality in Benin. Calverton, MD: ICF
Lorentzen, P., J. McMillan, and R. Wacziarg. 2008. Death Macro, for Republic of Benin and US Agency for
and development, Journal of Economic Growth 13(2): International Development.
81124. Sato, R. 1963. Fiscal policy in a neo-classical growth
Mahy, M. 2003. Childhood Mortality in the Developing model: an analysis of time required for equilibrating
World: A Review of Evidence from the Demographic adjustment, Review of Economic Studies 30(1): 1623.
and Health Surveys, DHS Comparative Reports No. 4. Schultz, P. 1981. Economics of Population. Reading, MA:
Calverton, MD: ORC Macro. Addison Wesley.
32 Robert Eastwood and Michael Lipton

Schultz, P. 2004. Demographic determinants of savings: UNICEF. 2010b. The State of the World’s Children.
estimating and interpreting the aggregate association in Updated at: http://data.un.org/Data.aspx?d SOWC&
Asia. Discussion Paper 901, Economic Growth Center, finID%3A34
Yale. UNICEF, WHO, The World Bank, and UN Population
Schultz, P. 2008. Population policies, fertility, women’s Division. 2007. Levels and Trends of Child Mortality in
human capital and child quality, in P. Schultz and J. 2006: Estimates Developed by the Inter-Agency Group
Strauss (eds.), Handbook of Development Economics for Child Mortality Estimation. New York: UNICEF.
Vol. 4. Amsterdam: Elsevier, pp. 32493303. Visaria, L. and P. Visaria. 1982. Population, in D. Kumar
Shapiro, D. and B. Tambashe. 2002. Fertility transition in (Ed.) Cambridge Economic History of India, Vol. II:
urban and rural sub-Saharan Africa: preliminary evi- c.1757c.1970. Cambridge: Cambridge University Press.
dence of a three-stage process, Journal of African Visaria, P. 2009. Demographic aspects of development: the
Policy Studies 8(23): 103127. Indian experience, in U. Kapila (Ed.) Indian Economy
Simon, J. 1996. The Ultimate Resource II. Princeton, NJ: since Independence, 19th edn 20082009. New Delhi:
Princeton University Press. Academic Foundation, pp. 181206.
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

Snow, R. and J. Omumbo. 2006. Malaria, in D. Jamison, R. Weil, D. 2007. Accounting for the effect of health on
Feachem, M. Makgoba, E. Bos, F. Baingana, K. Hof- economic growth, Quarterly Journal of Economics
man, and K. Rogo (eds.), Disease and Mortality in Sub- 122(3): 12651306.
Saharan Africa, 2nd edn. Washington, DC: World Weil, D. and J. Wilde. 2009. How relevant is Malthus for
Bank, pp. 195214. economic development today? American Economic
Thompson, W. 1929. Population, American Journal of Review 99(2): 255260.
Sociology 34(6): 959975. Weitzman, M. and K.-G. Löfgren. 1997. On the welfare
Thompson, W. 1959. Population and Progress in Asia. New significance of green accounting as taught by parable,
York: McGraw-Hill. Journal of Environmental Economics and Management
Tiffen, M., M. Mortimore, and F. Gichuki. 1994. More 32(2): 139153.
People, Less Erosion: Environmental Recovery in Westoff, C. F. 2010. Desired Number of Children 2000
Kenya. Chichester: Wiley. 2010, DHS Comparative Report No. 25. Calverton,
Tzannatos, Z. and J. Symons. 1989. An economic approach
MD: ICF Macro.
to fertility in Britain since 1860, Journal of Population
Wikipedia. N.d. http://en.wikipedia.org/wiki/DutchEastIn
Economics 2(2): 121138.
dies
Udry, C. 1998. Comment on Bloom and Sachs, Brookings
World Bank. 2006. Where is the Wealth of Nations?
Papers on Economic Activity 29(2): 281286.
Measuring Capital for the 21st Century. Washington,
UN. 2009a. World Population Prospects: The 2008 Re-
DC: World Bank.
vision*Population Database. Available: http://esa.un.
Young, A. 2005. The gift of the dying: the tragedy of AIDS
org/unpp/index.asp?panel2
and the welfare of future African generations, Quar-
UN. 2009b. Assumptions Underlying the Results of the 2008
terly Journal of Economics 120(2): 423466.
Revision of World Population Prospects. Available:
Zaal, F. and R. Oostendorp. 2002. Explaining a miracle:
http://esa.un.org/unpp/assumptions.html
intensification and the transition towards sustainable
UNICEF (United Nations Children’s Fund). 2010a. Sta-
small-scale agriculture in dryland Machakos and Kitui
tistics by Area: Monitoring Child Survival and Health.
Districts, Kenya, World Development 30(7): 12711287.
Available: http://www.childinfo.org/mortality.html
APPENDIX
Table A1 Total fertility rate and crude birth rate in countries of sub-Saharan Africa and regions of Asia, selected years

TFR5055 TFR6570 TFR8590 TFR0005 TFR0510 TFR2530 CBR5055 CBR6570 CBR8590 CBR0005 CBR0510 CBR2530

SSA 6.57 6.70 6.42 5.41 5.08 3.40 47.8 47.3 45.0 40.2 38.6 28.2
E SSA 7.03 7.10 6.67 5.60 5.30 3.47 49.7 48.6 46.0 41.0 39.6 28.7
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

Ethiopia 7.17 6.87 7.06 5.87 5.38 3.25 49.4 47.7 48.0 40.6 38.6 27.0
Kenya 7.48 8.11 6.54 5.00 4.96 3.22 51.2 50.6 45.1 38.8 39.0 27.3
Madagascar 7.30 7.30 6.30 5.28 4.78 3.06 49.0 47.9 45.0 39.3 36.2 26.1
Malawi 6.78 7.20 7.16 6.03 5.59 3.64 52.3 55.8 50.9 43.7 40.5 30.6
Mozambique 6.60 6.60 6.33 5.52 5.11 3.26 49.3 48.0 43.6 43.5 39.5 28.5
Sudan 6.65 6.60 6.08 4.82 4.23 2.75 46.4 46.2 41.8 34.5 31.6 22.9
Tanzania 6.74 6.79 6.36 5.66 5.58 3.80 49.0 48.7 44.9 41.8 41.6 30.8
Uganda 6.90 7.12 7.10 6.70 6.38 4.22 51.3 49.0 49.4 47.4 46.3 34.3
M SSA 6.03 6.31 6.72 6.22 5.67 3.68 46.8 47.1 48.2 45.5 42.7 31.1
Angola 7.00 7.40 7.20 6.63 5.79 3.83 54.0 53.2 52.8 47.6 42.9 32.1
Cameroon 5.68 6.10 6.10 4.92 4.67 3.02 43.4 44.6 42.9 37.8 37.1 26.0
DRC 5.98 6.15 6.98 6.70 6.07 3.81 47.2 47.5 50.2 48.3 45.1 32.3
S SSA 6.22 5.77 4.16 2.90 2.64 2.15 43.4 38.9 32.2 24.9 22.8 18.3
South Africa 6.50 5.70 4.00 2.80 2.55 2.10 43.3 38.2 31.1 24.2 22.1 17.8
W SSA 6.42 6.71 6.68 5.59 5.27 3.47 47.4 48.0 46.2 41.2 40.0 28.5
Burkina Faso 6.10 6.56 6.94 6.14 5.94 4.05 47.2 47.5 48.0 45.0 47.8 32.7

Demographic transition in sub-Saharan Africa


Côte d’Ivoire 6.77 7.83 6.61 5.05 4.65 3.01 51.2 53.0 42.8 37.1 35.1 25.5
Ghana 6.43 6.95 5.88 4.54 4.31 3.10 44.6 47.5 40.4 33.7 32.5 24.5
Nigeria 6.55 6.55 6.76 5.67 5.32 3.27 48.0 47.0 46.9 41.8 40.1 27.2
China 6.11 5.94 2.63 1.77 1.77 1.85 43.8 36.9 23.7 14.0 13.5 10.6
India 5.91 5.64 4.15 3.11 2.76 1.96 43.4 38.8 32.5 25.4 23.0 15.5
Indonesia 5.49 5.57 3.40 2.38 2.19 1.85 42.7 42.1 27.4 20.7 18.8 13.7
Thailand 6.35 5.99 2.30 1.81 1.81 1.85 43.6 39.6 20.8 15.3 14.6 12.2
Asia 5.73 5.46 3.47 2.50 2.35 2.01 42.3 37.7 28.4 20.3 19.0 14.2
E Asia 5.42 5.01 2.52 1.71 1.72 1.81 40.7 34.6 22.2 13.4 13.0 10.2
SC Asia 6.04 5.86 4.48 3.17 2.82 2.10 43.3 40.0 34.3 26.0 23.7 16.6
SE Asia 5.98 6.03 3.63 2.48 2.32 2.02 44.1 41.3 29.2 21.0 19.5 14.6
W Asia 6.33 5.95 4.46 3.18 2.95 2.29 47.2 40.7 32.6 25.2 23.9 17.5

Source: UN (2009a).

33
34
Table A2 Crude death rate, infant mortality rate, HIV prevalence, countries of sub-Saharan Africa and regions of Asia, selected years

Robert Eastwood and Michael Lipton


CDR5055 CDR6570 CDR8590 CDR0005 CDR0510 CDR2530 HIV1549 IMR5055 IMR6570 IMR8590 IMR0005 IMR0510 IMR2530

SSA 26.3 21.3 16.4 14.9 13.9 9.7 5.0 174.3 142.8 112.0 94.8 88.6 61.1
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

E SSA 26.8 20.4 16.1 14.3 12.7 8.5 178.0 137.4 108.7 85.3 76.3 49.0
Ethiopia 29.9 21.8 19.1 13.5 12.0 8.2 2.1 199.3 147.9 125.8 88 79.1 50.2
Kenya 23.6 16.5 10.0 12.9 11.7 8.5 7.8 147.1 104.4 67.2 70.5 63.9 43.8
Madagascar 27.5 22.1 16.4 10.8 9.3 6.5 0.1 180.9 143.2 109.9 74.7 65.2 40.2
Malawi 30.9 25.5 18.6 14.6 12.4 8.4 11.9 212.0 195.0 142.4 96.5 83.7 57.0
Mozambique 32.8 25.6 21.0 16.9 16.1 11.8 12.5 220.2 172.1 143.0 102.9 90.1 53.4
Sudan 22.8 19.2 14.4 11 10.3 8.0 1.4 152.0 129.5 99.1 73.3 69.1 43.5
Tanzania 22.4 18.9 14.7 13.3 11.5 7.7 6.2 153.3 128.1 99.8 74 64.8 38.5
Uganda 24.5 17.4 16.1 15.2 12.9 7.8 5.4 160.5 116.5 103.6 82.3 74.0 50.5
M SSA 27.4 22.2 17.5 17.1 16.4 11.5 182.1 149.6 116.4 112.7 112.2 82.6
Angola 35.9 28.7 23.2 19.6 17.1 11.5 2.1 230.5 186.0 150.5 132.9 117.5 79.8
Cameroon 24.9 20.1 13.0 14.5 14.3 10.5 5.1 169.5 133.0 83.4 89.4 86.9 61.2
DRC 25.4 21.2 18.1 17.4 17.2 11.9 1.3 166.5 143.3 117.8 112 116.8 86.5
S SSA 20.5 14.9 8.8 13.5 14.8 14.1 104.3 88.6 55.0 60.3 49.4 26.9
South Africa 20.3 14.7 8.7 9.9 15.1 14.4 18.1 96.1 83.6 52.3 59.1 49.1 26.7
W SSA 27.2 23.7 18.3 15.4 14.4 9.9 185.3 158.1 124.6 103.5 97.5 69.5
Burkina Faso 29.0 23.7 18.1 14 13.1 8.7 1.6 219.9 162.3 111.7 86.8 80.1 64.7
Côte d’Ivoire 26.6 20.9 11.4 11.9 10.9 7.6 3.9 192.1 155.3 102.9 86.8 86.8 61.2
Ghana 20.1 17.4 11.9 10.8 11.2 8.6 1.9 139.9 117.1 83.3 70.4 73.4 48.0
Nigeria 27.7 24.2 20.0 17.4 16.5 11.7 3.1 185.7 162.1 132.3 113.8 109.4 77.7
China 25.1 10.9 7.5 6.6 7.0 9.1 0.1 195.0 80.8 31.4 25.6 22.9 14.9
India 25.5 17.3 11.1 8.9 8.5 8.0 0.3 163.7 129.9 87.8 61.8 54.6 38.3
Indonesia 26.1 18.6 9.2 6.6 6.3 7.0 0.2 201.2 144.1 69.8 34.2 26.6 13.1
Thailand 14.9 10.8 5.9 8.5 8.9 10.2 1.4 94.2 69.9 26.0 9.6 6.8 5.6
Asia 23.5 13.9 9.0 7.4 7.4 8.0 175.0 107.0 64.0 47 41.5 28.1
E Asia 22.6 10.4 7.3 6.8 7.2 9.4 0.1 181.7 76.7 29.7 24.2 21.8 14.5
SC Asia 24.9 17.4 11.1 8.6 8.1 7.6 0.3 168.2 133.1 89.6 63.5 56.3 39.8
SE Asia 23.3 16.1 8.6 6.7 6.5 7.0 0.3 165.7 117.3 60.7 33.5 28.3 15.4
W Asia 22.0 13.8 7.9 5.5 5.5 5.4 n/a 192.2 129.5 59.8 34 29.9 17.1

Source: UN (2009a).
Table A3 Rate of natural increase and dependency ratio, countries of sub-Saharan Africa and regions of Asia, selected years

NGR 195055 NGR 196570 NGR 198590 NGR 200005 NGR 200510 NGR 202530 DpR 1950 DpR 1965 DpR 1985 DpR 2000 DpR 2005 DpR 2025

SSA 2.15 2.60 2.86 2.53 2.47 1.85 82 88 93 89 86 71


E SSA 2.29 2.82 2.99 2.67 2.69 2.02 85 92 96 93 90 74
Ethiopia 1.95 2.59 2.89 2.71 2.66 1.88 89 85 90 95 92 70
Downloaded by [Memorial University of Newfoundland] at 05:58 27 November 2014

Kenya 2.76 3.41 3.51 2.59 2.73 1.88 77 108 111 89 83 71


Madagascar 2.15 2.58 2.86 2.85 2.69 1.96 70 92 93 93 90 67
Malawi 2.14 3.03 3.23 2.91 2.81 2.22 95 93 101 96 99 79
Mozambique 1.65 2.24 2.26 2.66 2.34 1.67 80 84 91 88 90 73
Sudan 2.36 2.7 2.74 2.35 2.13 1.49 89 88 90 83 79 58
Tanzania 2.66 2.98 3.02 2.85 3.01 2.31 93 93 96 91 91 82
Uganda 2.68 3.16 3.33 3.22 3.34 2.65 85 96 101 109 108 88
M SSA 1.94 2.49 3.07 2.84 2.63 1.96 81 86 95 97 95 75
Angola 1.81 2.45 2.96 2.8 2.58 2.06 79 92 98 98 95 76
Cameroon 1.85 2.45 2.99 2.33 2.28 1.55 75 82 95 87 83 66
DRC 2.18 2.63 3.21 3.09 2.79 2.04 89 88 96 103 101 78
S SSA 2.29 2.40 2.34 1.14 0.80 0.42 73 85 79 62 58 54
South Africa 2.3 2.35 2.24 1.43 0.7 0.34 72 83 77 59 56 53
W SSA 2.02 2.43 2.79 2.58 2.56 1.86 81 85 94 88 86 70
Burkina Faso 1.82 2.38 2.99 3.1 3.47 2.4 74 80 99 96 92 79
Côte d’Ivoire 2.46 3.21 3.14 2.52 2.42 1.79 83 84 92 81 82 66

Demographic transition in sub-Saharan Africa


Ghana 2.45 3.01 2.85 2.29 2.13 1.59 90 89 92 81 76 62
Nigeria 2.03 2.28 2.69 2.44 2.36 1.55 80 86 94 88 86 67
China 1.87 2.6 1.62 0.74 0.65 0.15 61 80 55 48 42 46
India 1.79 2.15 2.14 1.65 1.45 0.75 68 81 73 65 60 47
Indonesia 1.66 2.35 1.82 1.41 1.25 0.67 75 81 72 54 51 43
Thailand 2.87 2.88 1.49 0.68 0.57 0.2 82 92 61 46 43 48
Asia 1.88 2.38 1.94 1.29 1.16 0.62 67 80 66 57 53 48
E Asia 1.81 2.42 1.49 0.66 0.58 0.08 63 75 53 48 43 47
SC Asia 1.84 2.26 2.32 1.74 1.56 0.9 69 82 77 67 61 49
SE Asia 2.08 2.52 2.06 1.43 1.3 0.76 73 87 74 58 53 46
W Asia 2.52 2.69 2.47 1.97 1.84 1.21 74 88 80 67 61 50

Source: UN (2009a).

35

You might also like