Addressing short-run food insecurity: Poverty traps and safety nets

Some principles of poverty targeting
Food-linked income transfers
Employment-linked income transfers

For the Asian developing countries, economy-wide growth is the main-and, as experience suggests, the only sustainable-solution to long-run food insecurity arising from chronic poverty, which itself is commonly caused by failure of entitlements on a day to day basis (regardless of the season or time of the year). Not all types of growth are, of course, equally effective in combating chronic poverty. As indicated earlier, growth benefits the poor, especially the large majority of them in rural areas, only if it increases the earnings of their only major asset - labour. Sustained increases in these earnings require inducing employment expansion in productive sectors of the economy, dismantling barriers to labour mobility, and improving access of the poor to social services such as education, health, and nutrition. But even then there are usually groups in society who are unable to participate-either partially or fully-during episodes of growth. The obvious case pertains to food-insecure individuals with physical disabilities or illnesses. While growth may eventually mediate their insecurity (in the sense that it allows greater capacity for a society to provide income transfers), avenues for coping mechanisms during the transition can be very limited. Another case pertains to those who are hurt-either absolutely or relatively-during the transition to sustained growth path simply because they do not have adequate assets, including skills, needed to take advantage of the opportunities offered by growth. Still another case, though closely related to the previous one, are those whose entitlements have been shrunk by public actions chosen to move the economy to the desired growth path, as well as those who fall into poverty traps owing to the reinforcing effects of adverse shocks and imperfect capital markets. This last case could be referred to as, borrowing from Sen (1981), "entitlement failure," meaning that the individuals’ entitlement sets no longer include enough food to stay adequately nourished or, in its worse form, to stay alive.[6]

Given the enormous fiscal requirements for growth and the consequently limited resources for direct poverty alleviation, how could public policy be designed so that these resources would yield maximum poverty reduction even in the short term? The developing countries in the region are so heterogeneous that no single solution is equally applicable for all. For one, there are significant differences not only in their economic structure but also in their institutional capacity, including quality of national and local governance. Nonetheless, each of these countries could harness general lessons from experience within the region and elsewhere, as well as new insights from the economics of mechanism design, that may prove useful in informing policy choices vis-à-vis poverty alleviation.

It is important to recognise that growth-promoting policy reforms, such as those associated with structural adjustment programmes, typically benefit some groups and hurt some other groups, at least during the transition to sustained growth. Households are affected by such reforms through any one of three avenues:

(1) changes in the price of goods and services consumed by households;

(2) changes in employment status as well as returns to production factors owned by households; and

(3) changes in transfers, both public and private, and in the provision of public services to households.

Some principles of poverty targeting

The success (or failure) of any poverty alleviation programme depends much on whether its design is incentive-compatible or not. A programme is incentive-compatible if (i) the intended or high-priority target groups have the incentives to seek the benefits of the program, and (ii) the unintended groups do not have the incentives to preempt the benefits of this programme. Though simple, this requirement has frequently not been given careful consideration in policy design and programme administration. As the cases noted below will demonstrate, this failure, together with the no-more-than lip service given to institutional and economic realities, is usually the main reason for the far-less-than-expected outcomes of poverty alleviation programmes.

Efficiency in the use of funds (or resources) for poverty alleviation underlies the principle of "targeting," in which benefits are channelled to the high priority group that a programme aims to serve. The more accurate a subsidy (cash or in-kind) is in reaching the poor, the less the leakage of benefits to unintended groups, and the less it costs to achieve the desired objective. But targeting requires the identification of the poor as distinct from the non-poor, as well as the monitoring of programme benefit flows to intended beneficiaries. This is a potentially costly exercise, both in terms of time and administrative outlay. The exercise may also involve making trade-offs between the need to focus programme benefits to those who need them most and the technical and political difficulties of doing so. The usual practice is to screen potential programme participants from the population on the basis of some means tests, such as reported household income, consumption-expenditures, or nutritional indicators. However, such tests, when used as indicator of eligibility, is often not as reliable as-or superior to-easily observable household or individual characteristics (Glawwe and van der Gaag 1990; Grosh and Baker 1995; Balisacan 1998b). But even if they are, the cost of periodically acquiring the information for hundreds of thousands-or tens of millions-of households should not be overlooked. The cost could easily either eat up a large proportion of the total budget for poverty alleviation, thereby diminishing the amount actually received by the intended beneficiaries, or "crowd out" spending in support of long-term growth and poverty alleviation.

There usually are proxy indicators that correlate well with living standards, but which are quicker to measure, less costly to monitor, and less prone to manipulation than either incomes, expenditures, or nutritional indicators. References to certain spatial, demographic, and occupational characteristics of household members are, for instance, ample (e.g., World Bank 1990, Gaiha 1993, Lipton and Ravallion 1995, UNDP 1997). The typical household in agriculture is, for example, known to have a lower standard of living than its counterpart in industry and services, at least in developing Asian countries. Another observation is that, in many Asian countries, households headed by women tend to have lower standards of living than those headed by men, all other things remaining the same. The concentration of poverty has also found expressions along regional and ethnic lines.

Take the simple case where the only policy instrument is a set of lump-sum transfers differentiated by an easily observable characteristic (say, location of residence). As shown by Kanbur (1987), the budgetary rule for minimizing aggregate (national) poverty is quite straightforward: Maximum programme benefit should be directed to the population group with the highest poverty. Specifically, if the objective of anti-poverty programme is to minimize the aggregate poverty gap (i.e., the average depth of poverty for the entire population), then the poverty incidence (i.e., the proportion of the population deemed poor) is the appropriate indicator for budget allocation, with maximum benefit directed to the group with the highest poverty incidence. In the case where even the only information available is the regional profile of poverty, a regionally based targeting scheme proves to be far more efficient than, say, a universal income-transfer scheme in which everybody in the population, regardless of income, receives a transfer. Put differently, a regionally targeted income transfer costs much less than a universal transfer scheme in achieving the same reduction in aggregate poverty; its cost advantages are not likely to be wiped out by the administrative cost of targeting.[7] This is the case for Indonesia (Ravallion 1993), India (Datt and Ravallion 1993), and the Philippines (Balisacan 1994) where there is considerable regional variation in both the incidence and severity of poverty.

Combining several easily observable indicators has often worked well in improving the fineness of targeting, without resorting to highly costly means tests. In areas where institutional capacity for programme administration exists, it is desirable to combine geographical characteristics with information on dwelling attributes (e.g., type of roofing and/or walling materials) and household characteristics (family size, age and activity of children, work profile of parents, and so forth). Bangladesh’s well-respected Grameen Bank credit scheme, for example, screens participants on the basis of gender, landlessness, and rural residence.

In some cases, letting the intended beneficiaries select themselves both minimises targeting costs and results in well-targeted benefit incidence. Where applicable, self-targeting schemes satisfy the two requirements of an incentive-compatible design: the poor find it in their interest to avail of the benefits of the scheme, while the nonpoor do not have the incentives to pre-empt these benefits. The main feature of these schemes is that they impose a cost of participation that only the poor are willing to incur. An example is a food subsidy limited to commodities consumed mainly by the poor, or to lower-quality (perceived or otherwise) goods, which well-off groups would prefer to shun.[8] Another example is a public employment programme linked with wages lower than both legislated wages and wage levels prevailing in the formal labour market. Such programme is likely to attract only poor workers who need the work most.

Clearly, the targeting scheme that concentrates benefits to the poor is not necessarily the programme that has the greatest impact on poverty. Whether it is may depend critically on administrative cost, participation cost of the poor (as well as the nonpoor), and political economy considerations. It is possible, for example, that a targeted programme alters household behaviour (such as labour supply), institutions (including indigenous institutions functioning as social safety nets), and governance (both local and national) in ways that might actually worsen income distribution when compared to an untargeted programme. Political economy responses to targeted programmes may be also remarkably different from untargeted ones. Since targeted programmes help only a fraction of the population, and since those who are helped do not typically have the political clout, politicians maximising (perceived) gains in political support may not find targeted programmes to be as attractive as untargeted programmes. The relative importance of these factors in the Asian context becomes evident in the discussion below.

Food-linked income transfers

The common presumption for linking income transfers to food is that this scheme has a stronger effect on food consumption than the same real income transferred in cash. This is clearly the case for extra-marginal price subsidies (i.e., price subsidies for quantities that exceed those purchased prior to the subsidies) because they lower the price at the margin. However, the preference may also have to do with political feasibility: food-linked transfer (FLT) programmes may be perceived to be more effective than cash-transfer programmes in directly addressing a basic human need. In many other cases, FLT programmes are an outcome of a deliberate policy aimed at maintaining relatively low wages.

In Asia, food price subsidies, food stamp programmes, and food supplementation schemes (on-site feeding, take-home feeding, nutrition rehabilitation centres) have all been part of government interventions aimed at reducing malnutrition. The real income transfers embodied in these programmes have been frequently large. Estimates of available cases from 15 developing countries show that these transfers account for 15 to 25 percent of the incomes of the poorest quartile of the population (Pinstrup-Andersen 1993).

These estimates do not suggest, however, that the incomes of the poor could have been lower by 15 to 25 percent without these programmes. Since food is a wage good in underdeveloped societies, food price subsidies could depress real wages, thereby reducing the returns to the poor’s only major asset-labour. Evidence for Bangladesh (Ravallion 1990) points out this effect could be substantial. Food price subsidies may also affect household behaviour through such channels as work effort and interhousehold income transfers. Reduction in work effort is a possibility, especially for programmes with "benefit-reduction rate" features, i.e., the benefit is reduced when the income of the recipient increases. Evidence for Sri Lankan rural households show, for example, a significant effort reduction associated with targeted food stamp scheme: labour market participation fell by as much as 2.5 days per month for males and 2.9 days for females, or roughly 33 percent of the value of the subsidy benefits (Sahn and Alderman 1995).

Publicly provided food subsidies could also partially "crowd out" private interhousehold income transfers. In developing countries, interhousehold transfers, usually from richer to poorer households, could form a substantial proportion of the incomes of the poor. In the Philippines, for net-transfer recipient households, about 15 percent of total household income come from private transfers. The response of these transfers to public provision of social security, such as unemployment insurance, is strong: about 90 percent of the increase in household income from unemployment insurance is offset by reductions in private transfers (Cox and Jimenez 1995). This suggests that unemployment insurance would not do any good to targeted households; nearly all the benefits would accrue to the private donors, who are likely to come from high-income groups. Whether this observation also applies to food subsidies is not as yet known. Nonetheless, it appears that the income transfer embodied in food price subsidies tends to substantially overstate the net effect of the programmes on the real incomes of the poor.

In some cases, food subsidy programmes have been conceived primarily to stabilise the prices faced by the poor. This is the case of India’s and Sri Lanka’s food ration schemes, which were originally intended to provide households with access to rationed quantity at fixed prices. The schemes simply transferred price fluctuations from household recipients to the government or the rest of the food market. These reduced both chronic and transitory food insecurity at the household level (George 1988).

Box D Public Distribution of Foodgrains in South Asia The South Asian countries have a long tradition of public food distribution to poor households. The food, usually procured in the domestic market - although recourse to imports is not uncommon in some countries - is distributed through a public distribution system (PDS) or "fair price" shops at an "issue price" which is generally lower than prevailing market price. In the early years of the program, the gap between the procurement price and the issue price was usually low, resulting in large government subsidies to the PDS operations. The supply was not also well targeted to poor households, making the PDS arrangement not quite different from a "universal" one in which anyone, rich or poor, is entitled to receive the subsidy. In the end, it was the non-poor urban households and civil servants who benefited the most from the programme. The huge cost of running such a programme exerted severe fiscal pressure, thereby leading the governments in these countries to eventually curtail budgets for food subsidies.

To a large extent, discussions of hunger and starvation in South Asia have emphasized the role of foodgrain distribution and prices (see Annex A). Indeed evidence demonstrates that high and unstable foodgrain prices contributed substantially to excessive mortality during famines (Ravallion 1997). Episodes of famine in the region also show that mortality is inversely related with real wage rates for unskilled labour-largely agricultural workers-and that the absolute incremental impact of lower wages tends to rise as wage rates fall. The combined evidence thus suggest that either an increase in the price of staples or an income loss will put upward pressure on mortality, but the effects of these factors tend to be strong only when prices are initially high or incomes low. The evidence also shows that the impact of any sudden decrease in the entitlements of vulnerable groups (e.g., income loss or increase in food price) depends so much on the recent history of their consumption.

It appears from the above discussion that a strong case can be made for food subsidy programmes to vulnerable groups, especially those experiencing a recent history of consumption declines. However, the fiscal cost of such programmes, apart from the behavioral effects noted above, should be a key consideration. This in large part determines the feasibility, sustainability, and equity and nutritional impacts of such programmes (see Box D). Universal price subsidies, in which everyone (poor or nonpoor) gets a subsidy, are unsustainable because they tend to not only have high fiscal costs but also "crowd out" public investments critical to the achievement of longer-term food security. They are inequitable because they tend to concentrate benefits to the relatively well-off groups.[9] In the Philippines, about 60 percent of the benefits from the universal rice price subsidy accrued to the non-poor (Balisacan 1994). Large cuts in fiscal costs could be made without reducing nutritional benefits, provided this is accompanied by better targeting and improved administration. Geographical targeting of food-linked programmes, for example, is a promising form of intervention in countries in which the poor tend to be concentrated and where administrative capacity is quite weak. Such programmes have been successfully employed in Colombia and the Philippines. Food stamp or ration card programmes, on the other hand, have high potential in areas where there is sufficient administrative capacity. Combined with appropriate targeting of household members (e.g., children between six and twenty-four months of age, and pregnant and lactating women are more likely to be more malnourished than other household members), these programmes could have low fiscal costs relative to the achieved reduction in malnutrition.

Another consideration in designing FLT programmes must be that these treat the causes-not the symptoms-of the food-insecurity or malnutrition problem. As already pointed out, the usual root of the malnutrition problem is long-run poverty, which in turn arises from either incapability of certain household groups to raise their entitlements, or the diminution (or even collapse) of their entitlements owing to unforeseen shocks, such as drought or flood. Without enhancing these groups’ income-generating capacity, there will not be a situation in which income transfers are no longer needed. The challenge is to use food-linked transfer programmes that will lead to a greater access to food in the short run while creating a capacity within the household to secure food independently in the longer run. Opportunities exist for this mode of intervention, though not quite exploited in many of the countries in the region.

One such opportunity is the formation of human capital through nutrition, health, and education improvement, coupled with the use of food, technical assistance, and credit to facilitate the development of small-scale enterprises and other self-help activities for the poor. Another is the linking of food subsidies to the development of infrastructure, such as food-for-work programmes. This will not only raise the incomes of household recipients during programme implementation but, more importantly, also enhance employment generation for the poor after the programme ends.

Indeed evidence shows that poverty alleviation programmes exploiting synergies that exist among food, family care, and access to health and infrastructure services, have a far better chance of stemming the root causes of malnutrition and long-run food insecurity than what independently running programmes-nutrition programmes, food subsidy programmes, credit programmes, and so forth-could achieve.

Employment-linked income transfers

In Asia, employment-generating programmes have received much attention partly because of the rapid growth of labour force, partly by the increasing asset-scarcity of the poor (e.g., declining land-labour ratio), and partly by the inadequacy of current patterns of economic growth to absorb labour, especially unskilled labour. Even in East Asia where growth during the last two decades was rapid vis-à-vis the rest of the world, poverty remains a central concern, with about a third of the world’s poor living in the region. Moreover, the region’s rising inequality-not to mention the current financial turmoil-appears to be standing in the way of sustained economic growth and, hence, continued expansion of employment opportunities for the poor.

Labour-intensive employment programmes, if properly designed and implemented, hold high promise as instrument for addressing both short-term relief and longer-term asset creation. Since they strike at the root causes of long-term household food insecurity, they are superior to "workfare" or "make-work" projects, in which employment is provided but the asset-generation component is disregarded. This is not to say, however, that workfare is never a socially justifiable programme. In the midst of extensive natural or man-made disasters, workfare might be preferable to large-scale migration, the burgeoning of slums around urban centres (often following or during a crisis), or relief camps.

In such times, given the need for quick, large-scale relief, maximizing employment to smooth consumption might be the top priority; the quality of assets created might be just a secondary goal. The intervention can mean the difference between life and death; it can also prevent damaging responses that lead to poverty traps, such as when productive assets are sold, thereby inhibiting long-term poverty reduction.

However, even in such times, experience suggests that consumption-smoothing benefits during a crisis need not compromise asset creation; the state of the government’s readiness in dealing with disasters greatly influences its capacity to provide income transfer in the short run while building in place the foundation for long-term food security. Moreover, as noted above, the processes at work governing transient poverty may be essentially the same ones influencing chronic poverty. Addressing the latter will go a long way in addressing the former, though not necessarily the other way around. In general, workfare is not a viable and sustainable option for dealing with long-term food security goal.

The effectiveness of an employment-linked transfer (ELT) programme as a short-run food security measure depends on the benefits conferred to intended (as well as unintended) beneficiaries, administrative and participation costs, behavioral responses to the program, and the way resources are raised to finance the programme. Conceptually, the income transfer received by a programme participant is the wage offer times the duration of employment, net of participation cost and the foregone earning from an alternative employment.

For the programme to be incentive-compatible (i.e., for the programme to attract only the poor), the wage rate has to be kept low. While this results in lower transfer per (poor) beneficiary, it makes participation disproportionately biased in favour of the poor and helps keep overall costs down. Given a budget, a wage rate set high (say, at market rate) will not only attract the nonpoor to the programme but also necessitate job rationing. As noted above, rationing jobs to the intended poor beneficiaries when information about living standards of the population is imperfect is a potentially costly activity. Information acquisition raises the program’s administrative cost and, given a budget, effectively reduces the amount transferred to poor beneficiaries.

It follows from the above that the cost-effectiveness of an ELT programme can be gleaned from:

(a) the wage offer relative to prevailing market wage (or the statutory minimum wage) for unskilled labour; and

(b) the cost per one dollar of transfer to the poor.

Data on cost per dollar of income transferred through ELT programmes are even more scarce. It is thus not possible to say much about the cost effectiveness of such programmes. But if India’s experience is an indication, such programmes could be rather costly. In India’s nationwide programme of public works, the transfer of a dollar to the poor required an expenditure of about four dollars (including the amount of the transfer). The high cost resulted partly from the substantial leakage of the benefits to the nonpoor, largely because the programme wage was higher than the market wage rate (Subbarao 1997).

Often overlooked in the design of an ELT programme as an anti-poverty measure is the manner by which resources are raised to finance the programme. If the programme is entirely aid-financed, and unless the aid is large enough to alter domestic resource allocation through its effect on factor and product markets, the benefits to worker participants are straightforward. However, if the programme is financed out of general tax revenues, these benefits would have to be compared with those from alternative ways of spending the same amount of budgetary outlay, including the possible crowding out of activities conferring nonlabour income to the poor (e.g., educational or health services). This requires a simulation of counterfactual situation, an exercise seldom seen in operational work. If tax revenues would have to be raised to finance the program, the type of tax, its incremental cost to the society, and its incidence on the poor would have to be determined and compared with the expected benefits of the ELT programme. Again, this is seldom done in evaluation of ELT programmes. Yet, it is precisely this sort of exercise that is critical to informing policy choices vis-à-vis poverty alleviation.

Another area frequently overlooked in ELT programme design and implementation is the implication of economic, technical, and managerial constraints on programme scale. Success in varying degrees often characterizes most pilot ELT projects, but success proves to be difficult to maintain as these are scaled up-oftentimes to cover the entire country, as in India’s Jawahar Rojgar Yojna. As pilot projects become large-scale projects, they oftentimes run up against constraints on the demand side for the products and on the supply side for the inputs. For example, if the output is nontradeable and the demand for it is relatively income-inelastic (as is the case for certain staples), then the ELT project will cause the output price to fall, possibly hurting other producers-who may be poor-not involved with the project. Moreover, the highly management-intensive organization in pilot projects is often not transferable to larger projects. Difficulties in planning, coordination, collaboration, and performance evaluation tend to mount as pilot projects become large projects.

The principal-agent problems that beset even small organizations-for example, the difficulty of the principal (here, the society) in acquiring the information about the behavior of the agent (here, the project executors) who is entrusted to carry on the task of the principal-are also more pronounced in large projects. As observed by Lamptey and Sai (1985) for the case of health and nutrition projects, and by Siamwalla (1993) for the case of credit-based asset creation and employment programmes, informational difficulties are particularly acute for large projects if:

• administration does not ensure decentralization,
• local administrative control is weak, and
• beneficiary/community participation in project design and implementation is absent. Given the budgetary implication of such projects, high-level political commitment is also necessary. In the design of most large ELT projects in Asia (as elsewhere), these concerns have hardly gone beyond lip service.

The reasons vary considerably across countries and, within a country, in various times, but a common element is the quality of governance, to which we return later in this paper.