In spite of over forty years of respectable growth in the economies of the less-developed world, problems of poverty are as severe as ever in many developing countries: the incomes flowing to the poorest 40% have climbed much more slowly than GNP, social welfare indicators such as longevity and infant mortality have shown little improvement in the lower quintiles, and processes of modernization and structural transformation have had little effect on the poorest strata. These generalizations are not true everywhere; Korea, Taiwan, and Sri Lanka represent exceptions (for different reasons). But this story is largely accurate for many more countries: for example, India, the Philippines, Brazil, Nigeria, and Mexico.
In this paper I present some of the work I have been doing on a project called "Putting the Poor First": an extended argument designed to show that it is both desirable and possible to give highest priority to improving the condition of the least-well-off strata of developing societies. My work has focused on rural poverty, both because most of the world's poor are rural and because rural poverty has in general proved more difficult to address than urban poverty. In this paper I first lay out the normative case for putting the poor first (Section I). I then turn to an analysis of some of the main features and causes of rural poverty (Section II). This analysis begins with a somewhat institutionally specific description of the forms that rural poverty takes. Analysis is provided of some of the main sources of income for the rural poor. This analysis emphasizes the centrality of the distribution of assets and sources of income in determining the distribution and character of rural poverty. I then turn to the more important part of the argument: analysis of the policy options available that would have the effect of privileging the poor (Section III). The point concerning the centrality of distributive institutions becomes crucial when we turn our attention to ideas about how to alleviate poverty, since it suggests that entitlement reform and, more broadly, institutional reform must play a central role in the process if poverty alleviation is to succeed. I then turn to a consideration of some of the strategies of development through which poverty-first development might occur. I close with a discussion of the political obstacles that stand in the way of poverty-first development (Section IV) and the role that democratization is likely to play within this process.
Important as absolute per capita growth rates are, we must also consider the distributive characteristics of various growth processes. And here again there is substantial variation. In many LDCs inequalities have grown sharply in the past three decades: Brazil, Central America, the Philippines, Thailand, and Nigeria, for example. In other LDCs, by contrast, inequalities have remained constant or fallen: Korea, Indonesia, China, and Nicaragua. Income inequalities may be measured in a variety of ways; but two common measures are the Gini coefficient (figure 1) and the share of income flowing to the poorest 40% of income earners. The Gini coefficients and income shares to the poor are represented in table 2 for a number of developing countries over the past three decades. Inequalities have generally worsened in most developing countries; the average ratio of income of the top quintile to the poorest two quintiles rose from 4.14 to 4.45 to 5.03 to 5.18 in the four periods between 1956-60 and 1971-75. (The average for 1976-80 in this data set is substantially lower, but this reflects a skewed sample for the final period.) This data demonstrates a downward trend in the share of national income flowing to the poorest 40% of population in developing countries.
Finally, we need to concern ourselves with the question of poverty. How has third-world economic growth affected the poor? Have the benefits of economic growth been broadly distributed over all income levels? Have incomes-and consequently welfare-risen for the poorest 20 to 40 percent of developing societies? This question is distinct from that of inequalities, since it is possible for inequalities to rise while per capita income to the poor rises as well. However, in a large number of developing countries the benefits of economic growth have not reached the poorest 20 to 40 percent: their share of income has fallen, and their absolute average income has remained approximately constant. Table 3 provides data on the welfare of the poor in selected developing countries; it shows quite dramatically that there are substantial differences in poverty performance across countries. Low income shares to the poorest income strata have direct welfare effects: malnutrition, disease, inadequate water, low education al levels, high infant and child mortality rates, and depressed longevity statistics. Some countries-e.g. Sri Lanka-have made impressive strides in raising the welfare of the poor, even in the absence of substantial economic growth. Other countries-e.g. Brazil and the Philippines-have witnessed a sharp decline in the welfare of the poor in the midst of respectable national economic growth. Table 4 presents regional aggregation of this data set for each of the variables considered. There is a general upward trend in the three chief welfare indicators represented here at the country and region level-life expectancy, infant mortality, and school enrollments, indicating a general improvement in welfare in developing countries during this decade. But these aggregate figures conceal substantial variation within each country, and it is reasonable to assume that much of the improvement indicated here is concentrated in the top three quintiles of income earners in each country. It should also be noted that there is substantial regional variation in each of these indicators; average infant mortality among countries in South Asia in 1986 was 138 per thousand, whereas the average figure for Southeast Asia was 48 per thousand.
It is important to separate out inequalities and the direction of change of inequalities, from the issue of poverty and the direction of change of poverty levels. For, as Gary Fields shows, it is entirely possible that poverty falls, the real welfare of the poorest rises, and relative inequalities increase. We may have social policy reasons for preferring less inequality to greater; but it is fallacious to assume that increasing inequalities necessarily entail increasing poverty. A simple numerical example shows that rapid growth with higher inequalities may improve the welfare of the least-well-off more than slow growth with low inequalities over a few years. If our ultimate concern is the absolute welfare of the poorest in a medium timeframe, then it may be preferable to favor growth over inequality. If, on the other hand, we are inherently concerned with equality (and not merely equality as an instrument for improving the welfare of the poorest), then we may choose the slow growth model. Whether growth with rising inequalities leads to immiseration or gradual improvement in the welfare of the poor depends on the rates of each; more basically, it depends on the form that growth takes. Consider the example of Brazil based on data in table 2. In 1971-75 Brazil is found to have an income ratio of 9.51, with the poorest 40% of the population receiving about 8% of the national income. Brazil's growth rate in 1986 was about 4.3%. If we assume that this rate of growth is uniformly distributed across all income earners (a highly unrealistic assumption), then the average income for the poorest 40% will rise from $91 to $95, while that of the richest 20% will rise from $855 to $892. If we take $125 as the poverty level, it will take about 40 years of growth at this rate to bring the average income of the poorest 40% up to the poverty level. On the more realistic assumption that the benefits of growth flow disproportionately to higher income groups, this disparity becomes even more pronounced.
These points make clear what was perhaps already well enough known to thoughtful observers: economic growth (improvement in per capita GNP) is not sufficient to produce improvement in the welfare of the poor. Instead, there are some growth strategies that have harmful effects on the poor and others that have poverty-reduction effects.
The problem before us, then, is this: how should the development policies adopted by LDC governments and advocated by international development agencies deal with the problems of inequality and poverty in the context of economic growth?
There are two general strategies through which we can identify the poor. First, we can identify poverty with low income. On this approach we are to imagine the population as ranked in order of income levels (individual or family). We are then to choose a poverty budget: a level of income that is just sufficient to satisfy the minimal subsistence needs of an individual or a family. The poor are then defined to be the portion of the income distribution of population that falls below this income. This identifies what we might refer to as the "absolutely poor."
As a variant, we might consider poverty as a relative notion: the poor are the least-well-off in society. But the poor in Sweden find themselves at a level of welfare vastly above the medium-well-off in Bangladesh. On this approach, we might arbitrarily identify the poorest 20% or 40% of income earners as the poor.
Figure 2.2. PPP-adjusted income distribution in India (World Development Report 1990)
Both of these approaches have one important analytical flaw: they identify radically heterogeneous groups. Some of the poor, on this account, are poor because they are aged; some are poor because they are physically handicapped or ill; some because they are unemployed; some because they have too little land to support subsistence needs; and so forth. An improvement from this point of view is a functional analysis of the income structure of a given economy-an analysis of the breakdown of income earners into groups defined by the source of their income. Thus we can distinguish landless workers, subsistence farmers, commercial farmers, itinerant merchants, landlords, owners of large businesses, owner-operators of small businesses, and so on. For whatever functional taxonomy of income-earners that we provide we can then consider the distribution of income within each group. Some of these groups will have low average income, low minimum income, and low dispersion around the mean; that is, most members of these groups are poor. We might then define the poor as members of functional income groups which are typically poor, along with a residual category including those who are poor because of their age, health status, or other non- economic circumstances.
This functional approach allows us greater analytical grasp. It permits us to identify the poor in terms of their economic relationships (their entitlement package, in A. K. Sen's term).
In the following, then, I will consider the poor in functional terms. Putting the poor first means choosing economic development strategies that are particularly effective in improving the incomes and welfare attaching to members of poor functional groups, or that have the effect of aiding the transition for members of poor groups to non-poor groups (e.g. landless workers in transit to urban industrial employment).
It is possible for rural development plans to successfully increase agricultural productivity and per capita rural incomes, and yet simultaneously increase stratification and poverty at the bottom end. These effects raise serious problems of distributive justice and social policy. This paper has three main parts. First, it presents a schematic account of some of the considerations of distributive justice relevant to rural development. Second, it considers the tendencies concerning distributive justice that are contained in development schemes that work primarily through in vestments in private farming systems (capitalist development schemes). These schemes have a tendency to increase farm productivity and per capita income while at the same time increasing inequalities and creating a "surplus population" of rural poor. Finally, the paper considers socialist rural development plans that are designed to avoid these tendencies towards inequality and rural poverty. The paper considers whether such devices as fundamental land reform, producers' co operatives, and collective farming can potentially lead to a process of rural development that successfully increases agricultural productivity while at the same time reduces inequality and poverty at the lower end.
Let us consider briefly the main tasks of rural development in any developing society. Central among these are raising farm output, enhancing food security, and increasing rural in comes. A second set of goals involves improving equity in the distribution of wealth and income. Finally, developing economies are concerned with various aspects of economic modernization, including particularly the introduction of more efficient production technologies and the facilitation of structural transformation from traditional production sectors to modern production sectors.
There is another aspect of development policy formation that is often overlooked by development economists; this is the role of political goals within the development process. Regime stability, security interests, and the domestic political interests of the ruling party all play an important role in development policy formation in the developing world. And in many states-China, for example,-we may add to this list the set of ideological goals that have driven policy at various points: creation of a new man, reducing the social importance of material incentives, and enhancing the prestige and leadership role of the regime in power.
What is required in order for these development goals to be achieved? First, it is evident that most of these goals require the introduction of innovations increasing productivity in agriculture, particularly of land and labor. This is the kernel of truth in Schultz's arguments about traditional agriculture; through long adaptation, traditional agriculture had adjusted in such a way as to extract the highest possible yields from traditional technologies and inputs. In order to enhance food security it was necessary that grain outputs should increase at faster than the rate of population increase, and this required the introduction of modern technologies and inputs into cultivation. These include particularly adoption of modern seed varieties, chemical fertilizers and pesticides, power machinery, electrification, and the extension of irrigation. [4]
A second means of development has to do with the organization of the institutions of production: the size of the unit of production, the investment funds available to the unit, the incentives defining the environment of choice of the participants, and the role of market processes in directing production decisions.
A third means of development focuses on the infrastructure of the rural economy: the efficiency and cost of transportation, the marketing system, and the system of grain storage. Here the role of the state is generally reckoned to be large in any developing country, since these features of the economy have many of the properties of public goods. But in an economy in which a fifth of the harvest may spoil during storage or in which the cost of transport from rural market to urban consumer is equal to the cost of growing the grain, development in these areas can have a major effect on output.
Through what policy tools might a state within a developing society attempt to reform technology, organization, and infrastructure? There are various dichotomies available: for example, plan versus market, compulsion versus voluntary adoption, or national policy versus regional variation.
There is an extensive literature within development studies that is organized around the problems of inequalities and poverty in development. [5] In order to design a strategy of economic development that puts the poor first, we need to have an analysis of the causes and circumstances of poverty in the developing world. The poor have few assets to sell within a market economy. They are land-poor or landless, and are dependent on the sale of unskilled labor for income. And the institutional arrangements of LDCs-the property system, national political arrangements, and local power relationships-commonly leave the poor with little access to land and little political power through which to influence state policy. This analysis suggests that there are three broad avenues for improving the income of the poor: by improving their access to productive assets (chiefly land and education), by increasing the demand for labor, and by increasing the flow of state resources into amenities for the poor. This in turn suggests several strategies for poverty-reduction: asset redistribution programs (land reform, for example), economic programs that have the effect of increasing unskilled employment, [6] and what Dreze and Sen refer to as "public policy" spending-provision of health and education services to the rural poor [Dreze, 1989].
On this approach, then, we have a simple definition of the poor; they are those persons or households whose income falls below the poverty line. This definition is crude, however. For one thing, it cannot distinguish between the case in which large numbers of individuals fall just below the poverty line and the case in which there is a wide dispersion of income below the poverty line. The poverty-count approach cannot measure the depth of poverty-a failing that can be to some extent lessened by computing the poverty shortfall-the amount of additional income that would be needed to bring all the poor up to the poverty line.
Second, the low-income criterion of poverty fails to take into account non-income factors that influence the well-being of the poor. Particularly important are public amenities provided for the population at large, or targeted sub-groups; but various forms of informal assistance, subsistence farming, and other kinds of in-kind receipts affect the condition of the poor in a given society.
Sen's basic insight is that well-being is best defined in terms of the individual's capability to become a fully functioning human being. "In assessing the standard of living of a person, the objects of value can sensibly be taken to be aspects of the life that he or she succeeds in living. The various `doings' and `beings' a person achieves are thus potentially all relevant to the evaluation of that person's living standard" (Sen 1985:29). If we were fortunate enough to live in a world in which all persons, rich and poor, were fully capable of realizing their human capacities, then the issue of poverty and wealth would be of secondary concern. In our world, however, the limitations on personal development imposed by poverty are all too obvious: malnutrition, illiteracy, poor health, boring and dangerous conditions of work, and early mortality are plainly serious obstacles in the way of full human development for the poor.
Second, it must be noted that increasing the realization of capacities can be achieved through other means besides simply raising incomes; putting the point in another way, it is possible for the poor in one society to have higher income and lower capacity realization than the poor of another society, due to differences in the public provision of capacity-enhancing amenities. Societies in which there is extensive provisioning of education or health services, for example, will have a higher level of well-being in its poor population-even though the absolute income flowing to this stratum may be as low or lower than that of other societies.
This point brings to the fore a third qualification of the income-based criterion of poverty. Given the possible divergence between income and capacity-realization, we need to have other ways of measuring the extent and depth of poverty in different countries. It is here that a variety of "quality of life" indicators prove their merits. For it is possible to measure other variables besides income that have a more immediate relation to capacity-realization. Malnutrition is directly and patently incompatible with full realization of human capabilities; so, other things being equal, one society with a higher level of malnutrition than another is worse-off from the point of view of the condition of the poor. Longevity is a general indicator of the quality of health services available to the population (and the poor, since poor health care for a large share of the population will translate into reduced life expectancy on average). Infant mortality statistics are generally taken to be another sensitive indicator of the health and nutrition status of the poor; downward fluctuations in the latter lead to significant increases in the former. Likewise, data about school enrollments at various levels-primary, secondary, post-secondary-provide important information about the extent to which a given society is succeeding in providing education to its poor; and lack of education is plainly intimately related to obstacles in the way of capability-realization. (Literacy statistics can serve the same purpose.)
Measures of these non-income variables provide a fairly sensitive indicator of the condition of the poor in a way that permits informative cross-cultural comparisons. Several important indices of well-being have been constructed using such information. Central among these are the Physical Quality of Life index and the Human Development index.
Much of the harm of poverty is tangible and material: high rates of infant mortality, poor nutritional and health status, and so forth. However, the emphasis on capabilities and functionings above should alert us to the fact that poverty has an intangible side as well. For the underlying value, recall, is that of the fully developed human being-the person in realization of his or her capabilities and functionings qua human being. And among the diminishments imposed by poverty are enduring assaults to human dignity over the whole of a human life. The person who cannot afford minimally decent clothing will often be ashamed to present himself in public. The underemployed housemaid may be compelled to accept indignity and disrespect from her employers rather than risk losing her job. The tenant farmer with low income and little power will be obliged to kowtow to his landlord rather than face eviction. Each of these situations is one in which we find a human being in circumstances of indignity; and it would be hard to imagine the person being able to sustain a robust sense of self-worth and self-respect in these circumstances. [7]
These forms of fluctuation of income capacity point up the very great importance of income stability as a factor underlying the well-being of the poor. A somewhat higher average annual income may involve long periods of unemployment, and subsequent deprivation throughout significant parts of the year.
The other important kind of temporal fluctuation in poverty is life-cycle variation. A poor family is in its best circumstances when both parents are present and healthy and when children are old enough to contribute their labor to the family's well-being as well. At the beginning and end of this process, by contrast, the earning capacity of the family is reduced. During pregnancy and infancy the mother's capacity to labor is often diminished to some degree, during the early years of childhood the children are hungry mouths rather than sources of labor. At the other end of the cycle, aging, illness, and death once again reduce the income-earning capacity of the family. (Jean Dreze has written very movingly of the terrible situation of widows in rural India; Dreze 19??:??.) So when we think of the situation of the rural poor, it is important not to imagine a sort of homogeneous level of deprivation. Instead, there will inevitably be a range of experiences, from the disadvantaged but viable to the horrendously deprived at the bottom. And the various measures of well-being discussed above-infant mortality and health and nutrition status-are certain to be correlated with these variations.
Figure 1. Lorenz curve of income
Let us dwell for a moment on the mathematics of income and distribution. One measure of the affluence of an economy is its gross national product (GNP) per capita. But there is substantial variation in the pattern of distribution of income across economies; some economies have a very pronounced skew toward higher income groups, whereas others have a more substantial degree of income equality. Income inequalities can be measured in a variety of ways; the goal is to arrive at a way of characterizing the degree of dispersion of income across groups. A common tool for representing the dispersion of income is a graph of income representing cumulative shares of income across cumulative shares of population (referred to as a Lorenz curve; figure 3.0). A society in which income is equally distributed across all persons will have a straight-line Lorenz curve at 45 degrees to the origin. The Lorenz curve for a particular income distribution permits us to read off how much of the national income is flowing to the ith percentile of income earners.
Corresponding to each Lorenz curve is a simple measure of inequality-the Gini coefficient. This construct measures the degree of inequality represented by a given Lorenz curve as the ratio of the area enclosed by the Lorenz curve and the 45 degree line to the area below the 45 degree line; thus perfect equality corresponds to a Gini coefficient of 0 and perfect inequality corresponds to a coefficient of 1. It is important to note, however, that the Gini coefficient represents less information than the full Lorenz curve; different Lorenz curves may possess the same Gini coefficient.
The fact of differences in the distribution of income across economies means that two countries with the same per capita GNP may have substantially different amounts of income flowing to the poorest income groups. So if we are concerned with poverty we need to pay particular attention to the pattern of distribution of income, and the amount and dispersion of income flowing to the poorest 20 to 40 percent of income earners. This suggests that we need to associate GNP data with a disaggregation of income across the population.
In practice we rarely have income distribution data as detailed as that represented by a Lorenz curve distribution for any country. Instead, available data generally represent an aggregation of data representing distribution of income across quintiles. World Bank tables provide estimates of percentage of income flowing to quintiles and upper deciles; however, this data is only available for a minority of reporting countries (21 out of 89 low- and middle-income countries; WDR 1990). Other sources may provide only an estimate of the ratio of income shares of the top and bottom quintile, or an estimate of the Gini coefficient of income; Human Development Report 1991, for example, reports the income share of the lowest 40% of households (24 of 160 countries), the ratio of the top quintile to bottom quintile (20 of 160 countries), and gini coefficients (28 of 160 countries). It is desirable to be able to convert information provided in these various forms into an approximation of the Lorenz distribution of income that underlies the data. This can be done as a relatively simple spreadsheet exercise by constructing a linear Lorenz distribution consistent with the constraints imposed by the data source (e.g. quintile share data, quintile ratio data, or Gini data). Figure 3 presents the results of this exercise for Brazil, Egypt, and India. Once we have constructed the linear Lorenz distribution corresponding to a given data estimate, we can also calculate the percentage of the population falling below a given poverty budget. (This is convenient because poverty data are even more difficult to get than distribution data.)
Figure 2. Quintile incomes for Egypt
Quintile share data can be converted into an estimate of the average income flowing to each quintile. This disaggregation of income is useful because it permits us to focus on the incomes flowing to the poorest 40%. Egypt's purchasing power parity-adjusted GNP per capita in 1987, for example, was $1357. United Nations sources [8] provide an estimate of the income share of the lowest 40% and an estimate of the ratio of the top quintile to the bottom quintile. These estimates can be interpolated to provide a quintile distribution: the bottom quintile of income earners received 6% of income; the second lowest quintile received 10%; the third received 13%; the fourth received 20%; the ninth decile received 16%; and the tenth decile received 35%. These data may be broken out into an estimate of the average income of each quintile (figure 2). If we adopt a PPP-adjusted poverty budget of $850, this disaggregated data suggests that the bottom two quintiles of income earners fall below the poverty line. [9]
The data underlying Figure 2 can be converted into a Lorenz distribution, making use of the quintile inflection points. This Lorenz distribution corresponds to a Gini coefficient of 0.419.
Figure 2a. Lorenz distribution for Egypt (based on Figure 2)
The fact of differences in the distribution of income across economies means that two countries with the same per capita GNP may have substantially different amounts of income flowing to the poorest income groups. So if we are concerned with poverty we need to pay particular attention to the pattern of distribution of income, and the amount and dispersion of income flowing to the poorest 20 to 40 percent of income earners. This suggests that we need to associate GNP data with a disaggregation of income across the population.
Figure 3 illustrates this set of facts. In this chart the national incomes of India, Egypt, and Brazil are disaggregated over their populations. (These income data reflect PPP-adjusted dollars, based on 1987 data. The graph should be interpreted as representing a value for the average income flowing to the nth percentile of income earners.) There are substantial differences in the national income of these three economies; but as the chart demonstrates, the condition of the poor is strikingly similar in the three cases. Brazil's per capita income (PPP-adjusted) is $4307, Egypt's $1357, and India's $1053. Brazil, then, is substantially better off than Egypt or India. However, Brazil's income distribution is much more skewed than that of either Egypt or India; these income distributions correspond to Gini coefficients of .358 for India, .419 for Egypt, and .591 for Brazil. The poorest quintiles of each of these countries receive approximately the same income. India and Egypt have about the same levels of income through the 80th percentile, after which Egyptian income rises more rapidly than Indian. And the level of income of the Brazilian population begins to rise above those of Egypt and India after the poorest quintile, slowly at first and then very rapidly above the 80th percentile. It is not unreasonable to interpret these data as showing that Brazil's relative affluence is chiefly concentrated on the upper quintiles of income earners, whereas the poor of Brazil are about as badly off as those of Egypt or India.
Figure 3. Income distribution data
Source: Income data reflect Human Development Report 1990. [10]
LF laissez-faire growth: choose those policies and institutional reforms that lead to the most rapid growth: unfettered markets, profit-maximizing firms, minimal redistribution of in come and wealth.
PF poverty-first growth: choose those policies and institutional reforms that lead to economic growth favorable to the most rapid growth in the incomes flowing to the poorest 2 quintiles
WF immediate welfare improvement: direct as much social wealth as possible into programs that immediately improve the welfare of the poor (education, health, food subsidies, housing subsidies)
These strategies are stylized and reflect a set of assumptions about the dynamics of growth and distribution. (Distribution is measured by the percentage of GNP flowing to the poorest 40%.) I have assumed that the LF strategy has a consistently higher rate of growth, that PF be gins with a growth rate 1% lower than LF, and that WF begins 2% lower (5%, 4%, 3%). I have assumed that each growth rate begins to fall in later decades (reflecting the notion that exponential growth is not permanently sustainable). Second, I have assumed that both LF and PF show a Kuznets-U pattern of distribution over time, with inequalities increasing and then declining, but that PF declines less and recovers sooner. Figure 4 illustrates these assumptions. On the basis these assumptions I have computed GNP and income to the poor for each strategy over a fifty-year period; the results are provided in panels C and D of figure 4.
Figure 4. Three development strategies
Several points emerge clearly from inspection of these graphs. First, the laissez-faire strategy succeeds in accomplishing its central claim: it produces a substantially higher GNP at the end of the 50-year period-a 33% advantage over the poverty-first strategy and a 110% advantage over the welfare-first strategy. Over the long term (75-100 years) all income groups are better off with the highest growth rate strategy: even though this strategy gives the lowest relative share to the poor, the more rapid increase in total GNP more than compensates. [11] From the point of view of the welfare of the poor over the bulk of the period of development, however, the LF strategy does less well. For the first twenty-five years the incomes to the poor are higher on both the PF and the WF strategies-even though LF's GNP is substantially ahead of both alternatives. Second, in the near short term (0-25 years) the poorest groups are most advantaged by the immediate-welfare strategy. During this period incomes to the poor improve very slowly on the WF strategy, and improve significantly on the PF strategy. The critical period, though, is the medium term. In this period the poverty-first strategy passes the immediate-welfare strategy, and it retains the advantage over the maximize-growth strategy as well. The immediate-welfare strategy loses ground in the medium term because it has made too little productive investment in the national economy and has directed too much of the available surplus toward immediate welfare improvement. Incomes to the poor then stagnate (along with GNP as well) and improve only slowly from this point on.
This exercise has several important lessons. First, these idealized strategies show that development policy forces us to choose among various things: average income versus income to the poor; rate of growth versus rate of improvement in welfare of the poor; improvement in the present versus well-being in the future. Equally importantly, however, the exercise shows that if we pay attention only to efficiency and growth, the interests of the poor in the short and medium term will not be well-served. Third, this example makes it clear that privileging the interests of the poor does not entail neglecting economic growth. It is plain that sustained economic growth is the only longterm solution to the problem of poverty. National economic plans that work primarily toward channeling existing income into welfare assurance plans have shortterm benefits but promise longterm stagnation. (This may be the case of Sri Lanka in the 1970s and 1980s.) Only if a national economy is able to produce substantially greater per capita income and wealth will it be possible to create and sustain a process of improvement in the condition of the poor. This example makes it clear that other things being equal, economic growth is desirable from a poverty-alleviation point of view. Growth makes possible a sustained improvement in the income and welfare of the poorest strata of the developing economy. But whether that improvement occurs or not depends on the particular characteristics of the growth process and the institutions and institutional innovations through which incomes are distributed. If growth is stimulated by capital-intensive investment for world markets, there will be only sluggish increase in the demand for labor, which means that the pool of modern sector labor will expand only slowly and modern-sector wages will rise only slowly. To give lexical priority to poverty alleviation, then, entails that we should rank strategies by their impact on the income and welfare of the poor, irrespective of overall growth rates.
Equally important, however, the exercise shows that if we pay attention only to efficiency and growth, the interests of the poor in the short and medium term will not be well-served.
These lessons also suggest that economic development planning should
be time-sensitive. In a distributive environment in which there is extensive
poverty and deprivation, policy should sacrifice some economic growth in
exchange for more rapid improvement in the welfare of the poor. As the
situation of the poor begins to improve substantially the mix of policy
tools should then adjust toward a higher-growth strategy. And in fact there
is a mixed strategy that suggests itself upon reflection. It would certainly
be possible to shift the balance of strategy priorities over time, favoring
the poor in the early stages and favoring growth in later stages (as the
absolute welfare of the poor improves substantially). This would be a time-sensitive
strategy: choose poverty-first strategy while there is widespread abject
debilitating poverty; begin to shift to growth-first strategy as the poor
pass the level of abject poverty in order to maximize the well-being of
the least-well-off in future generations.
[2] World Development Report 1989.
[3] We need also to ask, whose goals: local political authorities, international lenders, U.S. foreign policy makers, or local people?
[4] See Mellor (1976) and Hayami and Ruttan (1971) for discussion of the problems of implementing new technologies in agriculture in the developing world.
[5] Particularly important are writings by Irma Adelman, Gary Fields, Atul Kohli, Keith Griffin, Hollis Chenery, and Ronald Herring.
[6] Keith Griffin describes the requirements of a poverty-first strategy of development as involving the following elements: "(i) an initial redistribution of assets; (ii) creation of local institutions which permit people to participate in grass roots development; (iii) heavy investment in human capital; (iv) an employment intensive pattern of development, and (v) sustained rapid growth of per capita income" (Griffin 1988:31).
[7] See James Scott's eloquent discussion of the social psychology of domination and subordination (1990).
[8] Human Development Report 1990, table 16.
[9] Purchasing-power parity is a measure of income that takes into account differences in prices in different economies.
[10] Income distributions have been estimated on the basis of limited quintile-distribution data. This income data is reported in HDR 1990. The income curves represented here are my approximation of income distribution given HDR data on shares flowing to each quintile.
[11] Empirical data suggests, however, that this process may be extremely slow. In 1971-75 Brazil had an income ratio of 9.51, with the poorest 40% of the population receiving about 8% of the national income. Brazil's growth rate in 1986 was about 4.3%. If we assume that this rate of growth is uniformly distributed across all income earners (a highly unrealistic assumption), then the average income for the poorest 40% will rise from $91 to $95, while that of the richest 20% will rise from $855 to $892. If we take $125 as the poverty level, it will take about 40 years of growth at this rate to bring the average income of the poorest 40% up to the poverty level. On the more realistic assumption that the benefits of growth flow disproportionately to higher income groups, this disparity becomes even more pronounced.
