You are on page 1of 23

World Development Vol. 28, No. 2, pp.

197±219, 2000
Ó 2000 Elsevier Science Ltd. All rights reserved
Printed in Great Britain 0305-750X/00/$ - see front matter
PII: S0305-750X(99)00131-X

Economic Growth and Human Development

Yale University, New Haven, Connecticut, USA

University of Oxford, UK


United Nations Development Program, New York, USA
Summary. Ð The connections between economic growth (EG) and human development (HD) form
two chains. Crosscountry regressions show a signi®cant relationship in both directions, with public
expenditures on health and education, notably female, especially important in the chain from EG to
HD; and the investment rate and income distribution signi®cant in the HD to EG chain. This gives
rise to virtuous or vicious cycles, with good or bad performance on HD and EG reinforcing each
other. Evidence over time has strong sequencing implications: countries initially favoring economic
growth lapse into the vicious category, while those with good HD and poor EG sometimes move
into the virtuous category. Where choice is necessary human development should be given
sequencing priority. Ó 2000 Elsevier Science Ltd. All rights reserved.

Key words Ð human development, economic growth, income distribution, poverty, health,

1. INTRODUCTION provides the resources to permit sustained

improvements in HD. On the other, improve-
Human development has recently been ments in the quality of the labor force are an
advanced as the ultimate objective of human important contributor to EG. Yet, while this
activity in place of economic growth. 1 Its two-way relationship between HD and EG may
intellectual antecedents may be traced to the now be widely accepted, the speci®c factors
earlier basic needs approach of the ILO and the linking them have not been systematically
World Bank, as well as Sen's concept of capa- explored. Nor has the question of priorities in
bilities. 2 Human development has been de®ned the phasing of policy. The purpose of this paper
as enlarging people's choices in a way which is to sharpen understanding of the two-way
enables them to lead longer, healthier and fuller links between HD and EG at both theoretical
lives. 3 The de®nition of HD as ``enlarging and empirical levels. This in turn permits us to
people's choices'' is very broad. For the analyze priorities in the phasing of policy and
purpose of exploring the links between HD and to examine the usual assumption that EG must
EG theoretically, and especially empirically, we precede progress on HD. 4
need to narrow it down. We shall consider the Section 2 identi®es the major links between
HD of a country as consisting of the health and EG and HD. Section 3 presents some empirical
education of its people, recognizing that this is
very much a reductionist interpretation.
Clearly, there exists a strong connection * We are grateful for very helpful comments on an
between economic growth (EG) and human earlier draft from three anonymous referees. Final
development (HD). On the one hand, EG revision accepted: 19 July 1999.

crosscountry evidence on these links. Section 4 tion of income and its change over time. 6 The
develops a typology of country cases, some way in which growth translates into income
representing the mutual enhancement of HD distribution and poverty reduction depends on
and EG and some demonstrating asymmetric the nature of the growth processÐin particular,
performance. The ®nal section investigates the the extent to which it is based on the generation
movement of countries from one category to of employment and on increasing rural
another and re¯ects on the implications for incomes, e.g., if the output mix is labor inten-
policy. sive and rural incomes rise rapidly income
distribution is more likely to improve and
poverty reduction to occur than if growth is
2. THE TWO CHAINS urban biased and capital intensive. 7
Expenditure on HD-related items is strongly
We view HD as the central objective of a€ected by the rate of poverty reduction. Not
human activity and economic growth as surprisingly, if poor households receive extra
potentially a very important instrument for income, they increase their food expenditure and
advancing it. At the same time, achievements in calorie consumption signi®cantly. 8 Empirical
HD themselves can make a critical contribution evidenceÐfor example, for Bolivia, Brazil,
to economic growth. There are thus two Chile, C^ ote d'Ivoire, Ghana, India, Indonesia,
distinct causal chains to be examined: one runs Pakistan, Philippines, Malaysia, Nicaragua and
from EG to HD, as the resources from national PeruÐalso indicates the positive e€ects of
income are allocated to activities contributing family income change on child schooling. 9 For
to HD; the other runs from HD to EG, indi- example, a crosscountry study of the determi-
cating how, in addition to being an end in itself, nants of secondary education found that more
HD helps increase national income. The two egalitarian countries had higher secondary
chains are pictured in Figure 1. enrollment rates. 10 One estimate suggests that if
the distribution of income in Brazil were as equal
(a) Chain A: from EG to HD as Malaysia's, school enrollments among poor
children would be 40% higher. 11 While the
GNP contributes to HD mainly through evidence on the relations between income and
household and government activity; civil soci- health is less extensive, studies in Brazil, Chile,
ety, e.g., through community organizations and C^ote d'Ivoire and Nicaragua suggest that
other nongovernmental organizations (NGOs), household income also has a signi®cant e€ect on
also plays a role. The same level of GNP can the demand for health, 12 some showing a much
lead to very di€erent performance on HD higher relative response for low than for high-
according to the allocation of GNP among and income households. 13
within these institutions and variations in their Where women control cash income, it
behavior. appears that expenditure patterns are geared
Households' propensity to spend their after- relatively more toward HD inputs, such as food
tax income on items which contribute most and education. For example, among Gambian
directly to the promotion of HD in poor coun- households, the larger the proportion of food
tries, e.g., food, potable water, education and under womenÕs control the larger household
health, varies, depending on such factors as the calorie consumption. 14 Similarly, in the Phil-
level and distribution of income across house- ippines it has been shown that consumption of
holds as well as on who controls the allocation calories and proteins increases with the share of
of expenditure within households. In general, income accruing directly to women. 15 In the
poor households spend a higher proportion of C^ote d'Ivoire, an increase in womenÕs share of
their incomes on HD items than those with cash income was associated with signi®cantly
higher incomes, and similar results ¯ow from higher spending on food and reduced spending
greater female control over household income. on alcohol and cigarettes. 16
When levels of poverty in a country are high, Turning to the government, the allocation of
either because per capita income is low or badly resources to improving HD is a function of
distributed, the expenditure of many house- total public sector expenditure, of how much of
holds on HD is bound to be low. While evi- this ¯ows to the HD sectors, and of the way in
dence 5 indicates that, in general, poverty is which it is allocated within these sectors. This
reduced with economic growth, the extent of can be expressed in the form of three ratios: 17
the reduction varies greatly with the distribu- the public expenditure ratio, de®ned as the

Figure 1. The HD-GNP cycle.

proportion of GNP spent by the various levels across countries in each of these ratios, which
of government; the HD-allocation ratio, means that the same level of GNP may be
de®ned as the proportion of total government associated with very di€erent levels of govern-
expenditure going to the HD-sectors; and, ment spending on HD priorities. 18
®nally, the HD priority ratio, de®ned as the The underlying determinants of these three
proportion of total HD-sector expenditure ratios are complex, but include the following: (i)
going to ``priority areas.'' Within the HD-sec- the tax capacity of the system; (ii) the strength
tors, some expenditures are clearly much more of the demand for military expenditure and for
productive in terms of achieving advances in other non-HD priorities of the government; (iii)
HD than others; for example, basic education, the varying interplay between bureaucratic
especially at an early stage of development, is forces, vested interests and popular pressures. It
generally recognized to have a larger impact on should be noted that all three ratios are a€ected
HD than tertiary education. But the precise by the extent of decentralization of government:
de®nition of what constitutes a ``priority area'' country evidence suggests that real decentral-
will inevitably vary according to a country's ization, i.e. devolution, tends to increase the
stage of development, rendering this third ratio total revenue available; it often raises the HD-
more arbitrary and dicult to measure than the allocation ratio; and it almost always improves
other two. There are very large variations the HD-priority ratios. 19

The signi®cance of public expenditure choi- e€ectiveness of these expenditures in raising

ces for improving HD is illustrated by a HD levels, i.e. what type of provision is most
comparison between Kenya and Malawi. In the productive at what level of development, and
1980s, a similar proportion of national income how di€erent combinations e€ect a change in
went to public expenditure (27% in Kenya; 30% HD. We shall call this link in the chain the
in Malawi) but Kenya had a signi®cantly higher ``HD improvement function'' (HDIF); it
social allocation ratio (47% compared to 35%) resembles a production function in that it
and social priority ratio (34% compared to relates the inputs into HD, such as public
14%) so that the proportion of GDP going expenditure on health services or water, to the
directly to HD-improving priorities in Kenya HD objective of achieving better health. An
was over three times that of Malawi (5.1% illustration of the type of relationship repre-
compared to 1.5%). 20 sented in the HDIF is whether and how far the
The importance of economic growth for provision of safe water is complementary to or
raising resources to promote HD is illustrated substitutes for education in contributing to
by a comparison between Botswana and improvements in health. Another is what
Sudan. In 1970 public expenditure on health combination of family planning expenditure,
and education per person was similar in the two improved nutrition and immunisation levels
countries ($96 in 1987 prices in Sudan, and $65 most e€ectively yield improvements in infant
in Botswana). During 1970±92 per capita mortality. The relationships embodied in the
expenditure increased more than seven times in HDIF are complex, depending on individual
Botswana, while remaining practically and community behavior and local knowledge
unchanged in Sudan. Hence, by 1992, Sudanese about relevant technologies. Moreover, the
expenditure was less than a quarter of that in relevant arguments of the HDIF are bound to
Botswana ($114 versus $466). This was not due di€er at di€erent stages of development.
to di€erences in the public expenditure ratio, Some aspects of the HDIF have been greatly
which was higher in Sudan during the 1970s illuminated by detailed empirical work. 22 An
and equal to Botswana's during the 1980s, but example of the potential usefulness of an
to the much faster income growth in Botswana. improved understanding of this production
With a similar share of Botswana's GNP going function is the abundant evidence that educa-
to health and education, real expenditure rose tion, especially female, tends to improve infant
much faster there. survival and nutrition. 23 Another study, for
Finally, NGO or other civil society activity, Brazil, showed that an increase in the nonlabor
on which information is more scattered, is income of women increased the probability of
typically heavily oriented toward HD objectives child survival by 20 times that of a comparable
(e.g., generating incomes for the poor and on increase in the nonlabor income of men. 24
schools, nutrition and health projects). Research in Ghana has shown that in rural
Resources are derived from private donations areas the provision of basic health services,
and government, in each case including both including adequate drugs, increases child health
foreign and domestic sources. There are and survival signi®cantly, while the evidence is
considerable variations in the extent, vitality less clear on urban services. 25
and e€ectiveness of NGO activities across Conditions liable to bring about an
countries, depending on their history, culture, improvement in the eciency of the HDIF
tax laws and actual or perceived government include better information about available
de®ciencies in providing services. In most technologies and the appropriate combinations
contexts, NGOs play a supplemental or even of inputs; the generation of new and more
marginal role, but in a few areasÐe.g., BRAC e€ective technologies; and the enhancement of
and the Grameen Bank in Bangladesh, the the motivation to make use of available
Harambee Schools in Kenya and the options, e.g., to send children to schools and
``Comedores Populares'' in PeruÐthey appear clinics. Such motivation appears to be in¯u-
to represent a major source of HD-enhance- enced by education levels, the power structure
ment. 21 within the family, incentives and the magnitude
Expenditures on HD inputs are clearly not of the opportunity costs.
objectives in themselves, but rather constitute It should be clear from this discussion of the
instruments for achieving advances in various various links in the ED-HD chain that, in
dimensions of basic well-being. A further general, we would expect important causal
important link in Chain A, therefore, is the connections to exist between the economy and

HD achievements, but these connections are important ingredient in a system's capacity to

not automatic: the strength of the links in Chain borrow foreign technology e€ectively.
A varies according to a large range of factors, Speci®cally, (i) health, primary and second-
including the structure of the economy, the ary education and nutrition raise the produc-
distribution of assets, and the policy choices tivity of workers, rural and urban; (ii)
made. The richness of the so-called social secondary education, including vocational,
capital 26 of a society presumably also a€ects facilitates the acquisition of skills and mana-
these choices and the strength of links at each gerial capacity; (iii) tertiary education supports
stage; when people act together to promote the development of basic science, the appro-
their well-being, when public morality is high, priate selection of technology imports and the
when the community monitors malfeasance, domestic adaptation and development of tech-
and when it participates extensively in public nologies; (iv) secondary and tertiary education
life, ceteris paribus, we can expect all the links also represent critical elements in the develop-
in Chain A to be stronger. ment of key institutions, of government, the
In summary, we would hypothesize from this law, the ®nancial system, among others, all
review of the links in Chain A that the essential for economic growth.
connection between GNP to HD is likely to be Empirical evidence at both micro and macro
stronger: levels lends support to the importance of these
A.iÐthe lower the proportion of the popula- relationships.
tion below the poverty line; for a given level At a micro-level, numerous studies indicate
of GNP per capita, this means the more that increases in earnings are associated with
equally income is distributed; additional years of education, with the rate of
A.iiÐthe more income households allocate return varying with the level of education. 28
to HD at a given income level; this may be While social rates of return (which deduct the
related both to the level of female education publicly ®nanced costs of education but make
and to female control over income within the no other adjustments to the private returns) are
household. below the private, they are still typically greater
A.iiiÐthe higher the proportion of GNP de- than returns to most physical investment. 29
voted to priority social expenditure by the These rate of return estimates are sometimes
government; interpreted as indicating the magnitude of the
A.ivÐthe more e€ective the contribution of impact of education on productivity; but this
social capital, including community organi- view has also been challenged by the argument
zations and other NGOs; that education performs a signalling function,
A.vÐthe more ecient the HDIF. distinguishing among people of di€erent innate
We shall test some of these hypotheses; others ability, rather than itself raising productivity.
we do not test, either because of conceptual But, investigations in rural Pakistan, and in
problems or data constraints, notably A.iv. and urban Kenya and Tanzania, di€erentiating
A.v. between additional earnings due to cognitive
achievement and those due simply to schooling,
(b) Chain B: from HD to EG showed that cognitive achievement accounted
for a high proportion of the extra earnings. 30
Turning our attention to the second chain, In agriculture, evidence suggests positive
from HD to EG, ample evidence suggests that e€ects of education on productivity among
as people become healthier, better nourished farmers using modern technologies, but less
and educated they contribute more to economic impact, as might be expected, among those
growth, although some important dimensions using traditional methods. 31 In Thailand,
of HD, such as making the lives of the termi- farmers with four or more years of schooling
nally ill tolerable, do not directly lead to were three times more likely to adopt fertilizer
enhanced productivity. 27 Higher levels of HD, and other modern inputs than less educated
in addition to being an end in themselves, a€ect farmers. 32 Similarly, in Nepal, the completion
the economy through enhancing people's of at least seven years of schooling increased
capabilities and consequently their creativity productivity in wheat by over a quarter, and in
and productivity. Clearly, the health and edu- rice by 13%. 33
cation of a population are among the main Education is also an important contributor
determinants of the composition and growth to technological capability and technical
of output and exports, and constitute an change in industry. Statistical analysis of the

clothing and engineering industries in Sri much more emphasized in the development
Lanka, to cite just one example, showed that literature.
the skill and education levels of workers and From a macro-perspective, the ``new growth
entrepreneurs were positively related to the rate theories'' aim to endogenize technical progress
of technical change of the ®rm. 34 Education by incorporating some of these same e€ects,
alone, of course, cannot transform an econ- emphasizing education as well as learning and
omy. The quantity and quality of investment, research and development (R&D). According
domestic and foreign, together with the overall to Lucas (1988), for example, the higher the
policy environment, form other important level of education of the workforce the higher
determinants of economic performance. Yet the the overall productivity of capital because the
level of human development has a bearing on more educated are more likely to innovate, and
these factors too. The quality of policy-making thus a€ect everyone's productivity. In other
and of investment decisions is likely to be models a similar externality is generated as the
in¯uenced by the education of both policy- increased education of individuals raises not
makers and managers; moreover, the volume of only their own productivity but also that of
both domestic and foreign investment will others with whom they interact, so that total
probably be larger when a system's human productivity increases as the average level of
capital supply is more plentiful. education rises. 42 A complementary view is
Improved health and nutrition have been that technical progress depends on the level of
shown to have direct e€ects on labor produc- R&D in an economy. By investing labor and
tivity, especially among poorer individuals. 35 capital in R&D a ®rm is able to improve not
A range of labor productivity gains has been only its own pro®tability but also the produc-
observed associated with calorie increases in tivity of the ®rms which consume its output.
poor countries, 36 including studies of farmers Again, education plays a key role, both in
in Sierra Leone, sugar cane workers in Guate- contributing to R&D and via interactive
mala, and road construction workers in learning. 43 A number of empirical studies have
Kenya. 37 In these cases productivity enhance- shown the positive e€ect of education on
ment appears to follow fairly immediately as economic growth at a macro level, with its size
current intakes of calories or micro-nutrients varying according to the measure of education
are increased. In other cases the e€ects are and the particular macro growth model adop-
medium-run (as re¯ected in weight) or long-run ted. 44
(as re¯ected in height), based on evidence from The impact of education on the nature and
Bangladesh, Brazil, India, Sri Lanka, and the growth of exports, which, in turn, a€ect the
Philippines. 38 A longitudinal study of a sample aggregate growth rate, is another way in which
of children in Chile concluded that providing human development in¯uences macro perfor-
nutritional supplements to children to prevent mance. The education and skills of a devel-
malnutrition would generate bene®ts six to oping country's labor force in¯uence the
eight times the cost of the intervention in terms nature of its factor endowment and conse-
of additional productivity. 39 A similar study of quently the composition of its trade. It has
Cali, Colombia found that a health and nutri- been argued that even ``unskilled'' workers in
tion program increased the lifetime earnings of a modern factory normally need the literacy,
individuals to from 2.5 to 8.9 times those of an numeracy and discipline which are acquired in
illiterate worker. 40 primary and lower secondary school. 45
It is dicult to capture the e€ects of ill-health Theoretical models incorporating skills and
(other than malnutrition) as there are few learning as important determinants of
accurate estimates of the incidence of illness. comparative advantage have led to modi®ca-
But studies in both Ghana and C^ ote d'Ivoire tions of the simple two factor Heckscher±
show the negative impact of morbidity, with Ohlin model, helping to explain the Leontief
men who reported that their activities had been Paradox and the dramatic success in the
curtailed by illness having lower hourly wage manufactured export growth of some devel-
rates, reduced hours of work and a smaller oping countries, notably those of East Asia. 46
probability of being in the labor force. 41 In Investigations have shown a signi®cant posi-
some contexts the evidence even indicates tive correlation between the growth of manu-
larger productivity e€ects arising from health factured exports and the growth of GDP,
and nutrition than from formal schooling, although serious identi®cation problems
although the impact of education has been remain. 47

There is also a positive feedback from institutional framework. As pointed out earlier,
improved education to greater income equality both domestic investment and direct foreign
which, in turn, is likely to favor higher rates of investment are in¯uenced by a country's HD
growth. As education becomes more broadly levelÐparticularly the education and skill levels
based, low-income people are better able to of the workforce. 53 But di€erences in
seek out economic opportunities. For example, economic growth across societies are again not
a study of the relation between schooling, only due to the level of inputs but also by how
income inequality and poverty in 18 countries e€ectively they are used and for what purpose.
of Latin America in the 1980s found that one- One important in¯uence is the incentives set by
quarter of the variation in workers' incomes economic policy. Another derives from people's
was accounted for by variations in schooling motives and behavior patterns.
attainment; it concludes that ``clearly education Income distribution again appears to be
is the variable with the strongest impact on important in Chain B, as it was in Chain A.
income equality.'' 48 Another study suggested Recent empirical evidence suggests that the
that a 1% increase in the labor force with at distribution of assets and income has an e€ect
least secondary education would increase the on economic growth, with a more equal
share of income of the bottom 40 and 60% by distribution favoring higher rates of growth. 54
between 6 and 15% respectively. 49 An investi- Among the explanations put forward for this,
gation of the determinants of income distribu- one is derived from the relationships in Chain
tion in 36 countries found secondary A, i.e., that a more equal distribution of income
enrollment rates to be signi®cant. 50 implies better nutrition and a stronger demand
Finally, education may a€ect per capita for education and hence raises labor produc-
income growth via its impact on the denomi- tivity. Others derive from political economy
nator, i.e. population growth. For example, a considerations and relate to Chain B, e.g., that
study of 14 African countries for the mid-1980s an unequal distribution of income may be
showed a negative correlation between female associated with greater political and economic
schooling and fertility in almost all countries, instability, more likely to interrupt economic
with primary education having a negative progress. 55
impact in about half the countries and no In summary, the connection between HD
signi®cant e€ects in the other half, while and GNP, represented by Chain B, is likely to
secondary education invariably reduced fertil- be stronger:
ity. 51 The three success countries in terms of B.iÐthe higher the investment rate;
reduced fertility, Kenya, Botswana and Zimba- B.iiÐthe more equal the distribution of in-
bwe, had the highest levels of female schooling come;
as well as the lowest child mortality rates. 52 B.iiiÐthe more appropriate the economic
As in Chain A, the strength of the various policy setting.
links in Chain B varies considerably and there Because of diculties in de®ning and measur-
is no automatic connection between an ing hypothesis B.iii we shall con®ne our tests to
improved level of HD and increases in per B.i and B.ii.
capita GNP. It is not enough to create a larger
pool of educated people; there must also be
opportunities for them to be productively 3. EMPIRICAL FINDINGS
employed or it might simply increase the
number of educated unemployed. Relevant to The discussion above led us to a set of
the demand-side are the savings and investment hypotheses about the links between HD and
rates, technology choice and the overall policy EG for both causal chainsÐfrom EG to HD
setting. (Chain A) and from HD to EG (Chain B),
Although high rates of saving and investment some of which we test with crosscountry
by themselves, as shown by past East European regressions for 1960±92. Our sample consists of
experience, do not guarantee high levels of 35 to 76 developing countries, according to the
sustained growth, normally a positive rela- availability of data for particular variables. We
tionship between investment and growth generally use lags of the original variables as
prevails, with its strength depending on such instruments to reduce the simultaneity bias that
factors as the policy environment, the quantity would have resulted from applying ordinary
and quality of human resources, the availability least squares (OLS). Lagged values are
of technology choices and the ¯exibility of the reasonable candidates as instruments since the

correlation between the residuals in the two ment of life expectancy. We attribute this to the
periods analyzed is not substantial. impact on household behavior of female
income, knowledge and control within the
(a) Chain A household. Moreover, it should be noted that
when this variable is added, social expenditure
For Chain A the variable chosen as a proxy becomes less signi®cant, which suggests that
for achievement in human development was life much of the impact of social expenditure
expectancy shortfall reduction, 1970±92, from a appears to occur through its e€ect on female
maximum of 85 years. 56 education. According to equations (4), (5) and
The explanatory variables selected were: (6), a 1% increase in the female primary gross
Ðlagged GDP per capita growth rate (for enrollment rate is estimated to reduce the life
1960±70) as a measure of overall EG; expectancy shortfall by 0.1%.
Ðsocial expenditure (de®ned as public The income distribution variables run coun-
expenditure on education and health) as a ter to our expectations, i.e. a more equal
percentage of GDP for the whole period distribution does not seem to advance human
(1970±92), as well as lagged (1970±80). The development.
hypothesis advanced earlier was that HD Both the African and Latin American
improvement would depend in part on the dummies are negative and signi®cant through-
proportion of GDP devoted to social expen- out, as we might have expected, given that the
diture; priority social expenditure data were comparator is highly successful East Asia. In
viewed as somewhat too arbitrary for each case the coecient is quite small. The
deployment here; regional dummies include a number of region-
Ðseveral measures of income distribution, speci®c features, including the level of per
i.e. income share of the bottom 20% or capita income.
40%, 1960±92, and the ratio of income share
of the top to the bottom quintile, 1960±92; (b) Chain B
Ðfemale primary school gross enrollment
rate in 1965; this is a proxy for the change For Chain B, the dependent variable chosen
in the stock of educated females. The latter was GDP per capita growth, 1970±92.
is likely to be associated with greater female The explanatory variables selected were:
control over household expenditure, thus Ðlog GDP per capita in 1960, to test for
tending to improve the HDIF; convergence of income levels as countries
Ðregional dummies, with East Asia allo- approach high income levels;
cated a zero value. Ðinitial levels of HD, using three di€erent
Table 1 summarizes the results for Chain A. measures, log life expectancy in 1962, adult
GDP per capita growth proved signi®cant and literacy 1970±72, and a combined index of
quite strong in all of the equations, with higher life expectancy and literacy for 1970
growth of per capita income leading to better (HDI*); 57
HD performance. According to equation (1), a Ðchanges in HD over time for two of the
one percentage point increase in the average measures of HD, the change in the log of life
growth rate of GDP per capita is estimated to expectancy, 1962±82; and HDI* shortfall
reduce life expectancy shortfall by more than reduction 1960±80; 58
three percentage points over the period. The Ðgross domestic investment as a % of GDP
social expenditure ratio also proved signi®- for the period as a whole (1960±92);
cantly positive in all but one case. For every Ðincome distribution, lagged, (1960±70),
percentage point increase in the average share using three alternative measures, the ratio
of GDP invested in health and education, when of the income share of the top to bottom
lagged, the life expectancy shortfall decreases quintiles, the income share of the bottom
by about 1.75 percentage points (equations (1) 40% of the population, and the income share
and (3)). of the bottom 20%;
An interesting ®nding is the mechanism Ðregional dummies, as in Chain A.
through which the social expenditure ratio The results are summarized in Table 2. Our
seems to a€ect human development. As measures of the initial level of human devel-
regressions (4), (5) and (6) indicate, the female opment invariably proved signi®cant, although
primary enrollment rate for 1965 has a signi®- with low coecients; i.e. the initial level of life
cant but small impact on the rate of improve- expectancy in equations (10)±(12), as well as
Table 1. Chain AÐordinary least squares regressions dependent variable: life expectancy shortfall reduction 1970±92
Variable 1 2 3 4 5 6 7 8 9
GDP/n Growth rate 1960±70 3.25 1.66 3.19 1.27 2.65 2.70 2.10 2.14 2.31
(4.09)a (3.61) (4.06) (2.56) (3.17) (3.17) (2.82) (2.52) (2.95)
Social Expenditures as a % of GDP ± 1.03 ± ± ± ± ± ± ±
1970±92 (2.84)
Social Expenditures as a % of GDP 1.73 ± 1.75 ± 1.36 1.51 1.28 1.54 1.36
1970±80 (2.37) (2.56) (1.64) (1.74) (1.91) (1.76) (1.91)
Income share of bottom 40% 1960±92 ÿ0:004 ± ± ÿ0:004 ± ÿ0:002 ± ± ±
(ÿ0.93) (ÿ1.29) (ÿ0.45)
Income share of bottom 20% 1960±92 ± ± ± ± ± ± ± ÿ1:66 ÿ1:88
(ÿ1.76) (ÿ2.26)
Ratio of income share top to bottom ± ± 0.002 ± 0.002 ± 0.005 ± ±
20% 1960±92 (1.19) (1.01) (3.02)
Female primary gross enrollment rate ± ± ± 0.12 0.10 0.10 ± 0.05 ±
1965 (2.59) (1.94) (1.85) (0.71)

Latin America dummy ± ± ± ± ± ± ÿ0:08 ÿ0:08 ÿ0:07
(ÿ2.35) (ÿ1.98) (ÿ2.02)
Africa dummy ± ± ± ± ± ± ÿ0.16 ÿ0.12 ÿ0.13
(ÿ4.34) (ÿ2.49) (ÿ3.63)
South Asia dummy ± ± ± ± ± ± 0.003 0.003 0.02
(0.06) (0.50) (0.39)
Middle East dummy ± ± ± ± ± ± ÿ0:09 ÿ0:07 ÿ0:09
(ÿ1.01) (ÿ0.67) (ÿ0.97)
Intercept 0.15 0.28 0.06 0.23 0.02 1.07 0.16 0.25 0.30
(1.51) (6.09) (1.17) (3.36) (0.51) (0.68) (2.92) (2.08) (3.59)
# of observations 41 76 41 54 38 38 41 38 41

Adj. R-squared 0.38 0.27 0.38 0.26 0.41 0.40 0.59 0.51 0.55
T-Statistics are given in parentheses.
Signi®cant at 1%.
Signi®cant at 5%.
Signi®cant at 10%.

Table 2. Chain BÐordinary least squares regressions dependent variable: average real GDP/n growth 1970±92
Variable 10 11 12 13 14 15 16 17
Log GDP/n 1960 ÿ0.01 ± ÿ0.01 ÿ0.01 ÿ0.01 ÿ0.009 ÿ0.01 ÿ0.01
(ÿ3.01)a (ÿ3.04) (ÿ2.87) (ÿ1.87) (ÿ1.42) (ÿ2.38) (ÿ2.55)
Adult literacy rate 1970±72 ± ± ± 0.03 ± ± 0.03 0.03
(2.45) (2.19) (2.21)
Log life expectancy 1967 0.06 0.03 0.09 ± ± ± ± ±
(3.79) (2.09) (3.90)
Gross domestic investment as a % of 0.12 0.14 0.12 0.12 0.13 0.06 0.12 0.12
GDP 1960±92 (3.83) (2.72) (2.37) (1.99) (2.26) (1.06) (2.07) (2.09)
Income share of bottom 40% 1960±70 ± ± 0.001 ± 0.001 ± ± 0.001
(1.52) (1.70) (2.20)
Income share of bottom 20% 1960±70 ± ± ± ± ± ± 0.40 ±

Ratio of income share of top to bottom ± ÿ0:0007 ± ÿ0:0007 ± ÿ2:21 ± ±

quintile 1960±70 (ÿ2.17) (ÿ2.06) (ÿ0.05)

HDI* 1970 ± ± ± ± 0.06 0.07 ± ±
(2.38) (2.58)
Latin America dummy ± ± ± ± ± ÿ0:02 ± ±
Africa dummy ± ± ± ± ± ÿ0:01 ± ±
South Asia dummy ± ± ± ± ± ÿ0:003 ± ±
Middle East dummy ± ± ± ± ± 0.01 ± ±
Intercept ÿ0:17 ÿ0:14 ÿ0:28 0.12 0.03 0.04 0.07 0.07
(ÿ3.71) (ÿ2.30) (ÿ3.93) (2.74) (0.64) (1.08) (1.46) (1.54)
# of observations 73 38 36 24 35 35 24 24
Adj. R-squared 0.34 0.29 0.45 0.37 0.30 0.41 0.38 0.38
Continued next page
Table 2ÐContinued
Variable 18 19 20 21 22 23
Log life expectancy 1962 0.06 0.07 0.05 0.05 0.08 ±
(5.48) (3.98) (3.04) (2.65) (2.88)
Change in the log of life expectancy 0.13 0.17 0.14 0.12 0.14 ±
1962±82 (3.81) (2.48) (2.01) (1.65) (1.78)
Gross domestic investment as a % of ± ± 0.12 0.13 0.06 0.13
GDP (1970±92 average) (2.44) (2.39) (1.17) (2.42)
Income share of bottom 40% 1960±70 ± 0.002 0.002 ± ± 0.002
(2.83) (3.16) (2.96)
Ratio of income share of top to bottom ± ± ± ÿ0:0008 ÿ0:0002 ±
quintile 1960±70 (ÿ2.49) (ÿ0.62)
HDI* 1960 ± ± ± ± ± ÿ0:0007
HDI* shortfall reduction 1960±80 ± ± ± ± ± 0.09
Latin America dummy ± ± ± ± ÿ0:02 ±
Africa dummy ± ± ± ± ÿ0:005 ±
South Asia dummy ± ± ± ± 0.003 ±
Middle East dummy ± ± ± ± 0.004 ±
Intercept ÿ0:27 ÿ0:33 ÿ0:29 ÿ0:23 ÿ0:33 ÿ0:03
(ÿ5.37) (ÿ3.96) (ÿ3.56) (ÿ2.81) (ÿ2.81) (ÿ3.84)
# of observations 79 39 38 38 38 37
Adj. R-squared 0.27 0.30 0.38 0.32 0.44 0.33
T-Statistics are given in parentheses.
Signi®cant at 1%.
Signi®cant at 5%.
Signi®cant at 10%.

(18)±(22) were all highly signi®cant. Adult restricted the de®nition of HD in our Chain A
literacy 1970±72 was signi®cant in equations regressions to life expectancy shortfall reduc-
(13), (16) and (17); and the HDI* for 1970 was tion which is mainly a€ected by public expen-
signi®cant in equations (14) and (15), all with ditures on health and education, particularly
low coecients. The change in life expectancy, female, and less by household expenditures.
1962±82, was positive and signi®cant in all For Chain B, the relationship between HD and
cases but one (see equations (18)±(22)). The economic growth was stronger the higher the
change in HDI* shortfall reduction, 1960±80, investment rate and the more equally distrib-
was signi®cant in equation (23), but the initial uted the income. The regional dummies were
level of HDI* was not signi®cant. generally negative for both Africa and Latin
The domestic investment rate was always America in both chains but with small coe-
signi®cant except when the regional dummies cients. Since, on average, Africa experienced a
were included (equations (15) and (22)). The decline in GDP per capita over the period, this
lagged income distribution variables virtually indicates that the fall in per capita incomes was
all gave results with the expected sign (i.e. a not proportionately translated into a slowdown
more equal income distribution is associated or reversal in HD improvements. 59
with higher economic growth), and were almost
always signi®cantÐif with very low coe-
cientsÐexcept when the regional dummies were 4. VIRTUOUS AND VICIOUS CYCLES
included. Moreover, income distribution is AND LOP-SIDED DEVELOPMENT
apparently more strongly related to GDP
growth when changes as well as levels of human The existence of two chains linking HD and
development, however measured, are included. economic growth is thus strongly supported by
The ratio of income shares of the top to bottom both our framework, drawing on micro and
quintiles, 1960±70, showed signi®cance in macro studies in the literature, and our empir-
equations (11), (13) and (21); the share of the ical results. This means that an economy may
bottom 40% was signi®cant in equations (14), be on a mutually reinforcing upward spiral,
(17), (19), (20) and (23), but the lagged income with high levels of HD leading to high growth
distribution variable was not signi®cant in and high growth in turn further promoting HD.
equations (12), (15) and (22), although the signs Conversely, weak HD may result in low growth
were in the ``right'' direction. Regional and consequently poor progress toward HD
dummies for Latin America were signi®cantly improvement. The strength of the links in the
negative in both instances when deployed and, two chains in¯uences the extent of mutual
in one case, also for Africa. In all equations, reinforcement between HD and EG, in either
except when the regional dummies were intro- direction.
duced (equation (15)), the initial level of GDP Country performance can therefore be
per capita was signi®cant, with a negative sign, usefully classi®ed into four categories, virtuous,
indicating weak convergence, i.e. with a low vicious and two types of lop-sidedness, i.e.
coecient. lopsided with strong HD/weak growth (called
In summary, the two chains, taken as a HD-lopsided); and lopsided with weak HD/
whole, showed a signi®cantly positive e€ect of strong growth (EG-lopsided). In the virtuous
economic growth on HD and a signi®cantly cycle case, good HD enhances growth, which in
positive e€ect of HD on economic growth. turn promotes HD, and so on. In the vicious
With respect to speci®c links in each of the cycle case, poor performance on HD tends to
chains, our ®ndings broadly con®rmed the lead to poor growth performance which in turn
tested hypotheses, except for income distribu- depresses HD achievements, and so on. The
tion in Chain A. For Chain A, the higher social stronger the linkages in the two chains descri-
expenditure, the higher adult literacy, and the bed above the more pronounced the cycle of
higher the female education enrollment for a economic growth and HD, either in a positive
given level of GNP per capita, the larger the or dampening direction. Where linkages are
improvement in HD. The most surprising weak, cases of lop-sided development may
®nding, counter to our expectations, was that a occur. On the one hand, good economic growth
more equal distribution of income did not may not bring about good HD, if, for example,
improve HD performance; indeed, in some there are such weak linkages as a low social
equations the opposite result obtained. One expenditure ratio; on the other hand, good HD
explanation of this may well be that we performance may not generate good EG if

there is a dearth of complementary resources lopsided quadrant, i.e. 10 of the 13 are from
because of low investment rates. Such cases of that region. In the EG-lopsided category there
lop-sided development are unlikely to persist. are just four countriesÐEgypt, Pakistan,
Either the weak partner in the cycle eventually Mauritius and Lesotho.
acts as a brake on the other partner, leading to The important issue for policy purposes is
a vicious cycle case, or, if the linkages are how a country may move towards inclusion in
strengthened, possibly by policy change, a the virtuous cycle category. Much can be
virtuous cycle case results. learned about this by looking at the ways in
One way of classifying countries into the four which countries moved across categories over
categories is to compare their performance on time. Taking the movements of countries over
HD and EG with the average performance of the three decades 1960±70, 1970±80 and 1980±
all developing countries. Figure 2 presents this 92 (see Table 3 and Figure 3), we ®nd the
classi®cation for 1960±92 for all developing following:
countries for which data were available. The ÐOver half the countries in the vicious-cycle
vertical and horizontal grid lines represent the category in 1960±70, i.e. 18 out of 35, re-
average performance for the period, with mained in that category throughout. Most
countries weighted by their populations in of these countries were in sub-Saharan Afri-
1992. Most developing countries appear as ca, which started with very low levels of HD,
either virtuous (NE quadrant), or vicious (SW handicapping their growth potential; their
quadrant); a signi®cant number showed an low growth rates and, subsequently, the debt
HD-lopsided pattern, and very few an EG- crisis prevented them from generating the re-
lopsided one. A strong regional pattern emer- sources necessary for improvements in HD.
ges, with East Asia heavily represented in the ÐSix countries moved from vicious cycle to
virtuous cycle case (seven out of the eight cases EG-lopsided between the 1960s and the
are East or South East Asian, the eighth being 1970s, but four of them fell back to the vi-
Botswana). Of the 37 countries in the vicious cious cycle category in the 1980s. Three
cycle quadrant, 21 are from sub-Saharan moved from vicious to HD-lopsided, includ-
Africa, nine from Latin America. Latin ing Honduras, Algeria and Madagascar, and
America is also strongly represented in the HD- only Madagascar returned to the vicious

Figure 2. Classi®cation of country performance (1960±92). Note: The horizontal and vertical lines de®ning the four
quadrants represent developing country averages weighted by population.

Table 3. Virtuous, vicious and lop-sided performance 1960±92

1960±70 1970±80 1980±92
Benin Vicious Vicious Vicious
Botswana Vicious Virtuous Virtuous
Burkina Faso Vicious Vicious Vicious
Burundi Vicious Vicious Vicious
Cameroon Vicious EG lop-sided Vicious
Central African Republic Vicious Vicious Vicious
Chad Vicious Vicious Vicious
Congo Vicious EG lop-sided Vicious
C^ote d'Ivoire EG lop-sided Vicious Vicious
Gabon EG lop-sided Vicious Vicious
Ghana Vicious Vicious Vicious
Kenya Vicious Virtuous Vicious
Lesotho Virtuous EG lop-sided Vicious
Madagascar Vicious HD lop-sided Vicious
Malawi Vicious EG lop-sided Vicious
Mali Vicious Vicious Vicious
Mauritius HD lop-sided EG lop-sided EG lop-sided
Niger Vicious Vicious Vicious
Nigeria Vicious Vicious Vicious
Rwanda Vicious Vicious Vicious
Senegal Vicious Vicious Vicious
Sierra Leone EG lop-sided Vicious Vicious
South Africa Virtuous Vicious Vicious
Sudan Vicious Vicious Vicious
Tanzania Vicious Vicious Vicious
Zaire Vicious Vicious Vicious
Zimbabwe Vicious Vicious Vicious
Latin America & Caribbean
Argentina Vicious Vicious HD lop-sided
Barbados Virtuous HD lop-sided HD lop-sided
Bolivia Vicious Vicious HD lop-sided
Brazil EG lop-sided EG lop-sided Vicious
Chile HD lop-sided HD lop-sided Virtuous
Colombia HD lop-sided Virtuous HD lop-sided
Costa Rica HD lop-sided HD lop-sided HD lop-sided
Dominican Republic HD lop-sided EG lop-sided Vicious
El Salvador HD lop-sided Vicious HD lop-sided
Guatemala HD lop-sided EG lop-sided Vicious
Haiti Vicious Vicious Vicious
Honduras Vicious HD lop-sided HD lop-sided
Jamaica Virtuous Vicious Vicious
Mexico Virtuous Virtuous HD lop-sided
Nicaragua Virtuous Vicious HD lop-sided
Panama Virtuous Virtuous HD lop-sided
Latin America & Caribbean
Paraguay Vicious EG lop-sided Vicious
Peru HD lop-sided Vicious HD lop-sided
Togo EG lop-sided Vicious Vicious
Trinidad & Tobago Vicious EG lop-sided HD lop-sided
Uruguay Vicious Vicious HD lop-sided
Venezuala HD lop-sided HD lop-sided Vicious
South Asia
India Vicious Vicious EG lop-sided
Nepal Vicious Vicious Vicious
Pakistan EG lop-sided Vicious EG lop-sided
Sri Lanka Vicious Virtuous Virtuous
Bangladesh Vicious Vicious Vicious
Continued next page

Table 3Ðcontinued
1960±70 1970±80 1980±92
East Asia
China HD lop-sided Virtuous Virtuous
Hong Kong Virtuous Virtuous Virtuous
Indonesia HD lop-sided Virtuous Virtuous
Korea Republic Virtuous Virtuous Virtuous
Malaysia Virtuous Virtuous Virtuous
Myanmar HD lop-sided Vicious Vicious
Philippines HD lop-sided EG lop-sided Vicious
Singapore Virtuous Virtuous Virtuous
Thailand Virtuous Virtuous Virtuous
Middle East
Algeria Vicious HD lop-sided HD lop-sided
Egypt EG lop-sided EG lop-sided Vicious
Morocco Vicious EG lop-sided HD lop-sided
Turkey Virtuous HD lop-sided HD lop-sided

Figure 3. Country HD-EG quadrant changes over three decades. Note: The country movements indicate the quadrants
in which countries are placed over the three decades but not their actual location relative to the axes.

cycle category. Kenya, which moved from ÐOf the eight countries which were EG-lop-
vicious to virtuous in the 1970s, also subse- sided in 1960±70, none stayed in that cate-
quently fell back to vicious. Only two coun- gory throughout, but all moved into the
tries managed to move from the vicious to vicious category. OneÐPakistanÐreverted
virtuous category on a sustained basisÐi.e. to EG-lopsided in the 1980s. Brazil and
Sri Lanka and Botswana. Egypt enjoyed relatively fast growth during

the 1960s and 1970s (over 3% in the 1960s Our analysis suggests that it is not possible to
and about 6% in the 1970s) but did not uti- move to virtuous via EG-lopsidedness, as this
lize this opportunity to improve the levels of proved a dead end.
HD substantially. In the case of Brazil the Hence the process we need to examine more
highly unequal income distribution (with a closely is that leading to a movement from
Gini of 0.634, one of the worst in the world) vicious to HD-lopsided, from HD-lopsided to
was one reason why growth did not translate virtuous, or the very unusual case, undoubtedly
into HD improvements. In both Pakistan dicult, of taking the direct route from vicious
and Egypt, public expenditure on health to virtuous.
and education was low, partly due to heavy Our earlier review of the links in the two
expenditure on the military, while Pakistan's chains suggests that to move from vicious to
HD performance su€ered especially from HD-lopsided one needs to strengthen the links
discrimination against females. in Chain A, which may be achieved by adopt-
ÐOf the 13 HD-lopsided countries in the ing some of the following policies:
1960s, only Costa Rica stayed in that cate- Ðthose leading to a shift in resource alloca-
gory throughout: four moved into a virtuous tion toward education and health services,
cycleÐChile, China, Colombia and Indone- especially those serving the majority of the
sia (Colombia later falling back to HD-lop- people, as apparently occurred in Argentina
sided). In these cases, early progress in with enhanced decentralization.
human development meant that they were Ðthose generating a more equitable income
able to take advantage of policy reforms distribution (for example, through land and
for generating growth. Egalitarian income tax reform or a move toward a more
distribution also assisted the movement to- employment-intensive pattern of output).
ward a virtuous cycle. Four moved initially Algeria is an example, with land reform in
from HD-lopsided into the vicious category the 1960s and some large-scale employment
(Venezuela, Myanmar, Peru, and El Salva- programs in the 1970s. 64
dor 60Ðthe latter two moving back into Ðthose providing extensive opportunities
HD-lopsidedness in the 1980s). 61 ThreeÐ for the unemployed, for example Bolivia's
Dominican Republic, Guatemala and the Social Fund in the 1980s, which received
PhilippinesÐinitially moved to EG-lopsided substantial support from donors.
and subsequently fell back into the vicious Movement from the HD-lopsided to the
category. Among the reasons for the failure virtuous category requires strengthening the
to move into high economic growth were links in Chain B by, for example:
the debt situation, poor economic policies Ðtaking advantage of an improved HD to
and internal disturbances. Consequently, promote economic growth through policy
they were unable to maintain the rate of pro- reform;
gress in HD because of slow economic Ðincreasing the investment rate, possibly
growth. assisted from the outside;
ÐThirteen countries were in the virtuous cy- Ðimproving the distribution of income.
cle category in the 1960s; ®ve retained this Chile is an example where strong economic
position throughout; and ®ve fell back to reforms, combined with a high level of HD,
the HD-lopsided and three to the vicious accounted for improved economic growth.
category. 62 Mostly, the countries that fell Relatively good income distribution in China,
back were subject to the depressing e€ects along with policy reform and very high invest-
of the 1980s debt crisis on economic growth. ment rates, led to accelerated economic growth.
It is important to note that lopsidedness was In Indonesia, a similar process was heavily
a temporary condition in all cases except Costa supported by external capital ¯ows.
Rica. 63 Our most signi®cant ®nding is that, We do not wish to argue here that any
while HD-lopsidedness permitted movement particular set of policies would achieve a
towards a virtuous cycle (occurring in about a particular movement across categories; rather
third of the cases), in the case of EG-lopsided- we want to emphasize an important conclusion
ness, all the cases reverted to a vicious cycle. about the sequencing of policy change, i.e. that
Very few countries managed to go directly from HD must be strengthened before a virtuous
vicious to virtuous; but some succeeded in cycle can be attained. Policy reforms which
moving to HD-lopsided, from where it was focus only on economic growth are unlikely to
possible to move into the virtuous category. succeed. Countries in a virtuous cycle category

may well slip back into HD-lopsidedness, if, for In summary, we have demonstrated the
some reason, growth slows down, but as long existence of an iterative process between the
as HD stays high such cases have a good ultimate objectiveÐimprovements in HDÐ
chance of resuming their virtuous cycle pattern. and economic growth as a necessary but not
Whenever either or both chains appear to be sucient condition for achieving such
weak, leading to lop-sided or vicious cycles, it is improvements. Moreover, by investigating the
important to identify where the weak links are relative importance of various links connect-
and what the appropriate policies might be to ing HD and EG we have identi®ed the
strengthen such links. These policies must, direction policy might take to strengthen such
moreover, be viewed in an evolutionary links. One important conclusion concerns the
context. Even countries initially successful in desirable phasing of policy change. Economic
both HD and EG will need to change their and social policy has tended to focus priority
policies as development proceeds in order to on getting the economic fundamentals ``right''
sustain their success. In an early phase, for as a necessary precondition for economic
example, priority should be given to primary growth, while arguing that HD improvement
education and some comprehensive health must await such economic growth. Our
interventions, both from the perspective of ®ndings do not deny the importance of
improving HD and that of increasing economic economic reforms, but emphasize that a
growth. At a later stage, the roles of science and focus on HD must be included from the
technology institutions and higher education beginning of any reform program. Economic
become more important for strengthening growth itself will not be sustained unless
Chain B, while, along Chain A, hospitals preceded or accompanied by improvements in
assume a greater role than before. HD.


1. See especially the UNDP's Human Development 8. Strauss and Thomas (1995). Of 38 studies in
Reports, starting in 1990. di€erent countries one-third indicate that at least one-
half of additional income is spent in this way. See also
2. See, e.g., Sen (1984); Streeten, Burki, ul Haq, Hicks Bouis and Haddad (1992).
and Stewart (1981); Fei, Ranis and Stewart (1985).
9. E.g., Alderman, Behrman, Khan, Ross and Sabot
(1995, 1996a); Behrman and Wolfe (1987a,b); Birdsall
3. The ®rst UNDP Human Development Report stated
(1985); Deolalikar (1993); King and Lillard (1987).
that: ``The basic objective of development is to create an
enabling environment for people to enjoy long, healthy 10. Williamson (1993).
and creative lives'' (UNDP, 1990, p. 9), and de®ned
human development as ``a process of enlarging people's 11. Birdsall, Ross and Sabot (1995).
choices'' (p. 10).
12. Blau (1986); Harbert and Scandizzo (1982);
4. One notable exception is Adelman and Morris Thomas, Strauss and Hanriques (1990, 1991); Thomas,
(1973) who argued that educational investment should Lavy and Strauss (1992).
be given sequencing priority.
13. Gertler, Locay and Sanderson (1987).
5. E.g. Fields (1989); Deininger and Squire (1996).
14. von Braun (1988).
6. Bruno, Ravallion and Squire (1995), for example, in
a study of 20 developing countries, show that a 10% 15. Garcia (1990).
increase in average income per capita, 1984±93, was
associated with a 20% fall in the proportion of people 16. Hoddinott and Haddad (1991).
living on less than $1 a day. See also Ravallion and Datt
(1991). 17. See UNDP (1991).

7. Lipton (1977); Ranis (1979); Stewart (1977); Lipton 18. See UNDP (1991) Chapter 3, and UNDP (1996)
and Ravallion (1995). Chapter 3.

19. Decentralization of publicly provided services which do not allow for this may understate the total
recently has been introduced in a wide range of returns.
countries. Tentative conclusions about their e€ectiveness
are mixed, with apparent relative success in promoting 30. Boissiere, Knight and Sabot (1985); Alderman,
eciency and contributing to HD in Indonesia, Malay- Behrman, Khan, Ross and Sabot (1996b).
sia, Chile and Karnataka in India, but less so in
Argentina, Bangladesh and Brazil. Mostly, local govern- 31. Schultz (1975); Welch (1970); Rosenzweig (1995);
ments have been severely constrained in their ability to Foster and Rosenzweig (1994); Behrman, Rosenzweig
raise taxes as well as in the freedom of allocative and Vashishtha (1995).
decision-making, and full democratic devolution has
been rare. See Behrman (1995b); Prud'homme (1995);
32. Birdsall (1993).
Klugman (1995); Ranis and Stewart (1994); Tanzi
33. Jamison and Moock (1984). A similar level of
education was estimated to increase farm productivity
20. UNDP (1996, p. 71). These calculations adopt a
by 10% or more, according to studies in India and
narrow de®nition of social priority expenditure, includ-
Pakistan (Azher, 1991; Butt, 1984; Duraisamy, 1992).
ing pre-primary and ®rst-level education plus primary
Studies in Malaysia, Ghana and Peru show that, on
health care only.
average, an extra year of schooling of a farmer is
associated with an annual increase in output of 2±5%
21. Riddell, Robinson deConinck, Muir and White (Birdsall, 1993). Examination of the early stages of
(1995). India's Green Revolution indicates that the bene®ts
from schooling arose more from conferring on farmers
22. See, e.g., the review by Strauss and Thomas (1995). the capacity for learning from their own experience or
from those of others, rather than from providing them
23. See e.g., Rosenzweig and Schultz (1982); Wolfe and with an initial information advantage (Foster & Rose-
Behrman (1987); Barrera (1990). nzweig, 1995).

24. Thomas (1990). 34. Deraniyagala (1995).

25. Lavy, Strauss, Thomas, deVreyer (1995). 35. See Behrman and Deolalikar (1987) and surveys in
Behrman (1993, 1996).
26. A term used broadly to describe civil society
networksÐsee Coleman (1988); North (1990) and Helli- 36. See Cornia and Stewart (1995).
well and Putnam (1995) for the development of the
concept and some empirical evidence. Tendler and
37. Strauss (1986); Immink and Viteri (1981); Wolge-
Freedheim (1994) provide evidence.
muth, Latham, Hall and Crompton (1982).

27. This clearly does not detract from the intrinsic

value of improving the lives of the unemployable, who 38. Behrman and Deolalikar (1989); Behrman and
may account for a signi®cant proportion of the popu- Lavy (1995); Deolalikar (1988); Foster and Rosenzweig
lation. (1993); Haddad and Bouis (1991); Pitt, Rosenzweig and
Hassan (1990); Sahn and Alderman (1988); Strauss and
Thomas (1996).
28. See surveys in Behrman (1990a,b,c, 1995a); Behr-
man and Deolalikar (1988), King and Hill (1993); King
and Bellew (1988); Psacharopolous (1994); Schultz 39. Selowsky and Taylor (1973).
(1988, 1993a,b); Strauss and Thomas (1995).
40. Selowsky (1981).
29. Following the work of Psacharopolous, the conclu-
sion that the returns to education are highest for primary 41. Schultz and Tansel (1993).
education has become part of the accepted wisdom; but
in fact this is by no means always the caseÐsee, e.g., 42. Perotti (1993).
Schultz (1993a,b). However, since secondary education
necessarily requires prior attendance at primary school, 43. See Romer (1990); Grossman and Helpman (1991);
the usual measures of returns to primary education Gemmell (1995).

44. E.g., Barro (1991); Barro and Lee (1993, 1994). 57. HDI* is a modi®ed version of the Human Devel-
opment Index of UNDP, including only the nonincome
45. Wood (1994); Owens and Wood (1995). components, i.e. educational attainment as measured by
a combination of adult literacy (2/3 weight), mean years
46. See Kenen (1965); Keesing (1966); Pack and of schooling (1/3 weight), and longevity as measured by
Westphal (1986); Leamer (1993). life expectancy at birth.

47. E.g. Michaely (1977); Krueger (1978); Ram (1985); 58. When changes over time were introduced, the dates
Rana (1988); Edwards (1993). of the initial level of HD were also changed, so that log
life expectancy was for 1962 and HDI* for 1960.
48. Psacharopolous, Morley, Fiszbein, Lee and Wood
(1992, p. 48). 59. One explanation is that there were important
speci®c interventions (notably immunization) that
a€ected HD positively throughout the world in the
49. Bourgignon and Morrisson (1990).
1970s and 1980s. In a sense these weakened the links in
Chain A by rendering HD improvements more indepen-
50. Bourgignon (1995). dent of incomes. In Africa, this would translate into a
reduced negative impact on HD of falling incomes.
51. See Birdsall, Ross and Sabot (1995); Jayaraman
(1995); Strauss and Thomas (1995); Thomas, Strauss 60. Iraq made the same move between the 1960s and
and Henriques (1991); Behrman and Wolfe (1987a). 1970s, but data are not available for the later period, when
con¯ict is likely to have damaged both HD and growth.
52. Ainsworth, Beegle and Nyamete (1995).
61. The Philippines moved from HD-lopsided to EG-
53. OECD/DAC (1995); World Bank (1995a). lopsided and then returned to HD-lopsided in the 1980s.

54. Alesina and Rodrik (1994); Alesina and Perotti 62. Lesotho moved from virtuous to vicious by way of
(1994); Persson and Tabellini (1994); Birdsall et al. EG-lopsided.
63. One of the explanations of why Costa Rica was
55. Alesina and Rodrik (1994). able to sustain HD achievements despite low economic
growth resides in its early, strong and sustained
56. As incorporated in the Human Development commitment to HD, exempli®ed by abolition of its
IndexÐsee UNDP (1995). We also tried regressions army in 1948 and its heavy investment (at 10% of GDP)
with changes in adult literacy rates and a combined, on health and education during 1970±92.
equally weighted, measure of the two. The results were
similar but the number of observations more limited. 64. El-Ghonemy (1990).


Adelman, I., & Morris, C. T. (1973). Economic growth Alderman, H., Behrman, J. R., Ross, D., & Sabot, R.
and social equity in developing countries. Stanford, (1995). Decomposing the gender gap in cognitive
CA: Stanford University Press. skills in a poor rural economy. Journal of Human
Ainsworth, M., Beegle, K., & Nyamete, A. (1995). The Resources, 31.
impact of female schooling on fertility and contracep- Alesina, A., & Perotti, R. (1994). The political economy
tive use: a study of fourteen sub-Saharan African of growth: a critical survey of the recent literature.
countries. SMS Working Papers, 110, World Bank, The World Bank Economic Review, 8.
Washington, DC. Alesina, A., & Rodrik, D. (1994). Distributive politics
Alderman, H., Behrman, J., Khan, S., Ross, D., & and economic growth. Quarterly Journal of Econom-
Sabot, R. (1996b). The returns to endogenous human ics, 109.
capital in Pakistan's rural wage market. Oxford Azher, R. A. (1991). Education and technical eciency
Bulletin of Economics and Statistics. during the green revolution in Pakistan. Economic
Alderman, H., Behrman, J., Khan, S., Ross, D., & Development and Cultural Change, 39.
Sabot, R. (1996a). The income gap in cognitive skills Barro, R. J., & Lee, J. -W. (1993). International
in rural Pakistan. Economic Development and Cultur- comparison of educational attainment. Journal of
al Change, 44. Monetary Economics, 32.

Barro, R. J., & Lee, J. -W. (1994). Losers and winners in Green Revolution experience. In H. Siebert, Locat-
economic growth. In Proceedings of the World Bank ional competition in the world economy. Kiel: Kiel
annual conference on development economics, 1993. Institute of World Economics.
Washington, DC: IBRD. Birdsall, N. (1985). Public inputs and child schooling in
Barro, R. (1991). Economic growth in a cross-section of Brazil. Journal of development Economics, 18.
countries. Quarterly Journal of Economics, 106. Birdsall, N. (1993). Social development is economic
Barrera, A. (1990). The role of maternal schooling and development. World Bank policy research working
its interaction with public health programs in child papers, WPS, 1123, World Bank, Washington, DC.
health production. Journal of Development Econom- Birdsall, N., Ross, D., & Sabot, R. (1995). Inequality
ics, 32 (1), 69±91. and growth reconsidered: lessons from East Asia.
Behrman, J. R. (1990b). The action of human resources World Bank Economic Review, 9.
and poverty on one another: what we have yet to learn. Blau, D. (1986). Fertility child nutrition and child
LSMS Working Paper 74. Washington, DC: World mortality in Nicaragua: an economic analysis of
Bank. interrelationships. Journal of Developing Areas, 20.
Behrman, J. R. (1990c). Women's schooling and nonmar- Boissiere, M., Knight, J. B., & Sabot, R. H. (1985).
ket productivity: a survey and a reappraisal. Mimeo, Earnings schooling ability and cognitive skills.
University of of Pennsylvania, Philadelphia. American Economic Review, 75.
Behrman, J. R. (1993a). The economic rationale for Bouis, H. E., & Haddad, L. J. (1992). Are estimates of
investing in nutrition in developing countries. World calorie-income elasticities too high? A recalibration
Development, 21. of the plausible range. Journal of Development
Behrman, J. R. (1995a). The impact of distributive Economics, 39.
policies, governmental expenditure patterns and decen- Bourgignon, F. (1995). Equity and economic growth:
tralization on human resources. Philadelphia: Mimeo, Permanent questions and changing answers? Back-
University of Pennsylvania. ground paper prepared for the 1996 Human Devel-
Behrman, J. R. (1995b). Intra-household allocation of opment Report, UNDP, New York.
resources: is there a gender bias? In Sex di€erentials Bourguignon, F., & Morrisson, C. (1990). Income
in infant and child mortality. Population Division, distribution development and foreign trade: a cross-
United Nations, New York. sectional analysis. European Economic Review, 34.
Behrman, J. R. (1996). Impact of health and nutrition on Bruno, M., Ravallion, M., & Squire, L. (1995). Equity
education. World Bank Research Observer, 11. and growth in developing countries: old and new
Behrman, J. R., & Deolalikar, A. B. (1987). Will perspectives on the policy issues. Paper presented at
developing country nutrition improve with income? International Monetary Fund Conference on Income
a case study for rural south India. Journal of Political Distribution and Sustainable Growth, Washington,
Economy, 95. DC, June 1±2.
Behrman, J. R., & Deolalikar, A. B. (1988). Health and Butt, M. S. (1984). Education and farm productivity in
nutrition. In H. B. Chenery & T. N. Srinivasan. Pakistan. Pakistan Journal of Applied Economics, 3.
Handbook of development economics, vol. 1. Amster- Coleman, J. (1988). Social capital in the creation of
dam: North Holland. human capital. American Journal of Sociology, 94
Behrman, J. R., & Deolalikar, A. B. (1989). Wages and (suppl. S95-S120).
labor supply in rural India: the role of health, Cornia, G.A., & Stewart, F. (1995). Two errors of
nutrition and seasonality. In D.E. Sahn, Causes and targeting. In F. Stewart, Adjustment and poverty:
implications of seasonal variability in household food options and choices. London: Routledge.
security. Baltimore, MD: The Johns Hopkins Deininger, K., & Squire, L. (1996). A new data set
University Press. measuring income inequality. World Bank Economic
Behrman, J.R., & Lavy, V. (1995). Production functions, Review, 10.
input allocations and unobservables: the case of child Deolalikar, A. B. (1988). Nutrition and labor produc-
health and schooling success. Mimeo, University of tivity in agriculture: Estimates for rural south India.
Pennsylvania, Philadelphia. Review of Economics and Statistics, 70.
Behrman, J. R., & Wolfe, B. L. (1987a). How does Deolalikar, A. B. (1993). Gender di€erences in the
motherÕs schooling a€ect the familyÕs health nutrition returns to schooling and school enrollment rates in
medical care usage and household? Journal of Indonesia. Journal of Human Resources, 28.
Econometrics, 36. Deraniyagala, S. (1995). Technical change and eciency
Behrman, J. R., & Wolfe, B. L. (1987b). Investments in in Sri Lanka's manufacturing industry. Ph.D. disser-
schooling in two generations in pre-revolutionary tation, University of Oxford, UK.
Nicaragua: the roles of family background and Duraisamy, P. (1992). E€ects of education and extension
school supply. Journal of Development Economics, contacts on agricultural production. Indian Journal
27. of Agricultural Economics, 47.
Behrman, J. R. (1990a). Human resource led develop- Edwards, S. (1993). Openness trade liberalization and
ment: review of issues and development. ARTEP/ILO, growth in developing countries. Journal of Economic
New Delhi. Literature, 31.
Behrman, J. R., Rosenzweig, M. R., & Vashishtha, P. El-Ghonemy, M. (1990). The political economy of rural
(1995). Location-speci®c technical change, human poverty: the case for land reform. New York: Rout-
capital and local economic development: The Indian ledge.

Fei, J. C. H., Ranis, G., & Stewart, F. (1985). Towards Kenen, P. B. (1965). Nature capital and trade. Journal of
viable balanced growth strategies: A locational Political Economy, 73.
perspective. Paper prepared for UNIDO. King, E. M., & Bellew, R. (1988). Education policy and
Fields, G. S. (1989). Changes in poverty and inequality schooling levels in Peru. Mimeo the World Bank,
in developing countries. World Bank Research Washington DC.
Observer, 4. King, E.M., & Hill, M.A. (1993). Women's education in
Foster, A. D., & Rosenzweig, M. R. (1993). Informa- developing countries: barriers, bene®ts, and policies.
tion, learning, and wage rates in low-income rural The Johns Hopkins University Press, published for
areas. Journal of Human Resources, 28. the World Bank, Baltimore.
Foster, A. D., & Rosenzweig, M. R. (1994). Technical King, E. M., & Lillard, L. A. (1987). Education policy
change and human resources and investments: Conse- and schooling attainment in Malaysia and the
quences of the Green Revolution. Mimeo, Philadel- Philippines. Economics of Education Review, 6.
phia: University of Pennsylvania. Klugman, J. (1995). Decentralization: A survey of the
Foster, A. D., & Rosenzweig, M. R. (1995). Learning by literature from a human development perspective.
doing and learning from others: Human capital and Human Development Report Oce Occasional Papers
technical change in agriculture. Journal of Political No. 12.
Economy, 103. Krueger, A. (1978). Foreign trade regimes and economic
Garcia, M. (1990). Resource allocation and household development: liberalization attempts and conse-
welfare: A study of personal sources of income on food quences. Cambridge, MA: Ballinger.
consumption, nutrition and health in the Philippines. Lavy, V., Strauss, J., Thomas, D., & deVreyer, P. (1995).
Ph.D. dissertation, Institute of Social Studies, The The impact of the quality of health care on children's
Hague. nutrition and survival in Ghana. LSMS Working
Gemmell, N. (1995). Evaluating the impacts of human Papers, 106. Washington, DC: World Bank.
capital stocks and accumulation on economic growth: Leamer, E. (1993). Factor-supply di€erences as a source
Some new evidence. Center for Research in Economic of comparative advantage. American Economic Re-
Development and International Trade, UK: Univer- view, 83.
sity of Nottingham. Lipton, M., & Ravallion, M. (1995). Poverty and policy.
Gertler, P., Locay, L., & Sanderson, W. (1987). Are user In J. R. Behrman & T. N. Srinivasan. Handbook of
fees regressive? The welfare implications of health development economics, Vol. 3. Amsterdam: North
care ®nancing proposals in Peru. Journal of Econo- Holland.
metrics, 36. Lipton, M. (1977). Why poor people stay poor: urban bias
Grossman, G. M., & Helpman, E. (1991). Innovation and in world development. London: Temple Smith.
growth in the global economy. Cambridge, MA: MIT Lucas, R. E. (1988). On the mechanics of economic
Press. development. Journal of Monetary Economics, 22.
Haddad, L., & Bouis, H. (1991). The impact of Michaely, M. (1977). Exports and growth: an empirical
nutritional status on agricultural productivity: Wage investigation. Journal of Developing Economies, 4.
evidence from the Philippines. Oxford Bulletin of North, D. (1990). Institutions and their consequences for
Economics and Statistics, 53. economic performance. In K. Cook & M. Levi, The
Harbert, L., & Scandizzo, P. L. (1982). Food distribution limits of rationality (pp. 383±401). Chicago: Univer-
and nutrition intervention: the case of Chile. Sta€ sity of Chicago Press.
Working Paper 512. Washington, DC: World Bank. OECD/DAC (1995). Development Cooperation: E€orts
Helliwell, J. F., & Putnam, R. (1995). Economic growth and Policies of the Members of the Development
and social capital in Italy. Eastern Economic Journal, Assistance Committee, OECD, Paris.
21. Owens, T., & Wood, A. (1995). Export-oriented indus-
Hoddinott, J., & Haddad, L. (1991). Household trialization through primary processing? IDS Working
expenditures, child anthropometric status and the Paper 19, Institute of Development Studies. Univer-
intrahousehold division of income: evidence from the sity of Sussex, Brighton.
C^ ote d'Ivoire. Research Program in Development Pack, H., & Westphal, L. (1986). Industrial strategy and
Studies, Discussion Paper 155, Woodrow Wilson technological change: theory versus reality. Journal
School. of Development Economics, 22.
Immink, M., & Viterri, F. (1981). Energy intake and Perotti, R. (1993). Political equilibrium income distri-
productivity of Guatemalan sugarcane cutters: An bution and growth. Review of Economic Studies,
empirical test of the eciency wage hypothesis. 60.
Journal of Development Economics, 9. Persson, T., & Tabellini, G. (1994). Is inequality harmful
Jamison, D., & Moock, P. (1984). Farmer education and for growth?. American Economic Review, 84.
farm eciency in Nepal: The role of schooling, Pitt, M. M., Rosenzweig, M. R., & Hassan, M. N.
extension services and cognitive skills. World Devel- (1990). Productivity health and inequality in the
opment, 12. intrahousehold distribution of food in low-income
Jayaraman, R. (1995). On the meta-production front: countries. American Economic Review, 80.
An evidence-gathering exercise. Processed for Prud'homme, R. (1995). The dangers of decentraliza-
UNDP, New York. tion. The World Bank Research Observer, 10.
Keesing, D. B. (1966). Labor skills and comparative Psacharopolous, G., Morley, S., Fiszbein, A., Lee, H., &
advantage. American Economic Review, 56. Wood, B. (1992). Poverty and income distribution in

Latin America: the story of the 1980s. Washington, in human capital. Economic Development and Cultur-
DC: World Bank. al Change, 22.
Psacharopolous, G. (1994). Returns to investment in Sen, A. K. (1984). Rights and capabilities. In A. Sen,
education: a global update. World Development, 22 Resources, values, and development. Cambridge, MA:
(9), 1325±1343. Harvard University Press.
Ram, R. (1985). Exports and economic growth: some Stewart, F. (1977). Technology and underdevelopment.
additional evidence. Economic Development and London: Macmillan.
Cultural Change, 33, 415±425. Strauss, J. (1986). Does better nutrition raise farm
Rana, P. (1988). Exports policy changes and economic productivity?. Journal of Political Economy, 94.
growth in developing countries after the 1973 Strauss, J., & Thomas, D. (1995). Human resources:
oil shock. Journal of Developing Economies, 8, Empirical modeling of household and family deci-
261±264. sions. In J. R. Behrman & T. N. Srinivasan.
Ranis, G. (1979). Appropriate technology in the dual Handbook of development economics, Vol. 3. Amster-
economy: re¯ections on Philippine and Taiwanese dam: North Holland.
experience. In A. Robinson, Appropriate technologies Strauss, J., & Thomas, D. (1996). Health, wealth and
for Third World development. New York: Macmillan. wages of men and women in urban Brazil. Journal of
Ranis, G., & Stewart, F. (1994). Decentralization in Econometrics.
Indonesia. Bulletin of Indonesian Economic Studies, Streeten, P., Burki, S. J., ul Haq, M., Hicks, N., &
30. Stewart, F. (1981). First things ®rst: meeting basic
Ravallion, M., & Datt, G. (1991). Growth and redistri- needs in the developing countries. New York: Oxford
bution components of changes in poverty measure: A University Press.
decomposition with applications to Brazil and India in Tanzi, V. (1995). Fiscal federalism and decentralization:
the 1980s. LSMS Working Papers #83. Washington, A review of some eciency and macroeconomic
DC: World Bank. aspects. Mimeo, International Monetary Fund,
Riddell, R., Robinson, M., deConinck, J., Muir, A., & Washington, DC.
White, S. (1995). Nongovernmental organizations and Tendler, J., & Freedheim, S. (1994). Trust in a rent-
rural poverty alleviation. New York: Oxford Univer- seeking world: health and government transformed
sity Press. in northeast Brazil. World Development, 22 (12),
Romer, P. M. (1990). Endogenous technological change. 1771±1791.
Journal of Political Economy, 98. Thomas, D., Strauss, J., & Henriques, M. H. (1990).
Rosenzweig, M. R., & Schultz, T. P. (1982). Child Child survival height-for-age and household charac-
mortality and fertility in Colombia: individual and teristics in Brazil. Journal of Development Economics,
community e€ects. Health Policy and Education, 2, 33.
305±318. Thomas, D., Strauss, J., & Henriques, M. H. (1991).
Rosenzweig, M. R. (1995). Why are there returns in How does mother's education a€ect child height.
schooling? American Economic Review, 85 (2). Journal of Human Resources, 26.
Sahn, D. E., & Alderman, H. (1988). The e€ect of Thomas, D. (1990). Intrahousehold resource allocation:
human capital on wages, and the determinants of An inferential approach. Journal of Human Re-
labor supply in a developing country. Journal of sources, 25, 4.
Development Economics, 29 (2). Thomas, D., Lavy, V., & Strauss, J. (1992). Public policy
Schultz, T. P. (1988). Education investments and and anthropometric outcomes in the C^ ote d'Ivoire.
returns. In H. Chenery & T.N. Srinivasan, Handbook LSMS Working Paper 89. Washington, DC: World
of Development Economics. Amsterdam: North Hol- Bank.
land. UNDP (various). Human Development Report. New
Schultz, T. P. (1993a). Returns to women's education. In York: Oxford University Press.
E. M. King & M. A. Hill, Women's education in von Braun, J. (1988). E€ects of technological change in
developing countries: barriers, bene®ts, and policies. agriculture on food consumption and nutrition: rice
The Johns Hopkins University Press, published for in a West African setting. Economic Development and
the World Bank, Baltimore, MD. Cultural Change, 37.
Schultz, T. P. (1993b). Investments in the schooling and Welch, F. (1970). Education in production. Journal of
health of women and men: quantities and returns. Political Economy, 78.
Journal of Human Resources, 28. Williamson, J. (1993). Human capital deepening, income
Schultz, T. P., & Tansel, A. (1993). Measurement of inequality and demographic events along the Asia-
returns to adult health. LSMS 95, World Bank, Paci®c Rim. In N. Ogawa et al. Human resources in
Washington, DC. development along the Asia-Paci®c rim. Singapore:
Schultz, T. W. (1975). The value of the ability to deal Oxford University Press.
with disequilibria. Journal of Economic Literature, Wolfe, B. L., & Behrman, J. R. (1987). Determi-
13. nants of women's health status and health-care
Selowsky, M. (1981). Nutrition health and education: utilization in a developing country: a latent
The economic signi®cance of complementarities at variable approach. Review of Economics and
an early age. Journal of Development Economics, 9. Statistics, 56.
Selowsky, M., & Taylor, L. (1973). The economics of Wolgemuth, J. C., Latham, M. C., Hall, A., &
malnourished children: an example of disinvestment Crompton, D. (1982). Worker productivity and

nutritional status of Kenyan road construction IDS Development Studies Series. Oxford: Oxford
laborers. American Journal of Clinical Nutrition, 36. University Press.
Wood, A. (1994). North-South trade, employment and World Bank (1995a). Global economic prospects and the
inequality: Changing fortunes in a skill-driven world. developing world. Washington, DC: World Bank.