One Out of Ten: Social Cash Transfer Pilots in Malawi and Zambia

An explicit objective of the current social cash transfer pilots in Malawi and Zambia is to learn lessonsⁱ. Between them, these schemes, which are now operational in over ten districts, have unquestionably provided a wealth of valuable information on how to implement cash transfer interventions in southern Africa. But they could, and should, also be providing important lessons in how not to operate such schemes: we need to have the courage to recognize this, and to broadcast the weaknesses as readily as we proclaim the strengths.

Recent studies have highlighted two major flaws in the particular model that is being tested: one practical and one conceptual. Unless and until these are resolved, it is unlikely that the pilots will receive the necessary technical and political support to scale them up to national programmes.

The practical flaw is that community-based targeting of the poorest does not work. It doesn’t work now, even in geographically-constrained pilot areas, where additional technical support and resources can be mobilized to support weak government and community institutions, so it will never work at more extended, less rigorously scrutinized national levels. A recent studyⁱⁱ in Machinga district in Malawi demonstrates this powerfully and graphically. The study undertook a random sample survey of households in Mlomba, and gathered data on each household (production, revenue, assets and other variables) to estimate that household’s “income”. Income was calculated as the “disposable” money remaining per adult equivalent after the household had met its essential food energy needs, either through purchase or own production (the 48.9% of households with negative income do not even reach this minimum acceptable nutrition threshold). The resultant income distribution of the sampled households is shown in Figure 1.

The standard model of targeting in the majority of Zambian and Malawian SCT pilot districts is to use community structures to identify the most labour-constrained households from within the so-called “ultra- poor” (estimated to comprise the poorest 22% or so of the community in both Zambia and Malawiⁱⁱⁱ). By definition, therefore, beneficiary households should both (a) be labour-constrained and (b) fall in the lowest quintile (20%) of household income. In the Machinga study, however, only half met the criterion of being labour-constrained (using the pilot’s own definition), and only 24% fell into the lowest income quintile (corresponding broadly to the 22% figure for the “ultra-poor” in Malawi). The majority (29%) fell into the third (middle) income quintile, and – staggeringly – 32% of selected households fell in the two wealthiest quintiles. Selected households are shown in red on Figure 2. This means, in effect, that fewer than 12% of households selected by the community to receive the SCT met the Programme’s targeting criteria. As the study wryly notes, “the relationship between income and household selection to receive the SCT was found to be effectively random”!

This leads on to the serious conceptual flaw in the current SCT model: that it is impractical, and unethical, to target SCTs at only 10% of a population in which some 60% are poor, and a further 20% or so are highly vulnerable to poverty: the situation that prevails in most of sub-Saharan Africa. Another recent paper^iv, by RHVP’s Frank Ellis, argues the theoretical case convincingly. His paper examines the circumstances of small economic difference which gives rise to the oft-expressed sentiment that “… we are all poor here”. Using national budget survey data from Malawi, Zambia and Ethiopia, the paper demonstrates that there are only very minor differences in per capita consumption (as a rule of thumb no more than US$2 a month) between each of the lowest six income deciles. In other words, there is no more than US$9-10 a month separating an individual in the poorest decile from an individual in the sixth decile.

Current SCT models are therefore unable to meet their goals of reducing destitution, “without inevitably creating some proportion of ‘leapfrogging’ by recipients above the levels of per capita consumption of non-recipients in adjacent income deciles”. Put simply, let us imagine that it were possible to target accurately the poorest 10% of a community (which the preceding paragraphs have shown to be a pipe- dream!). If you were to provide an SCT of US$6 per month to an individual within that decile – an amount which is fully consistent with current transfer levels (e.g. to a household with school-going children in one of the Malawian or Zambian SCTs) – then that individual would thereby be catapulted four deciles to be among the middle-income members of the community. This raises complex practical issues (such as the need for frequent retargeting), and serious ethical concerns around inequity and social divisiveness – both of which may seriously erode political support for such SCT programmes.

A third study^v, also supported by RHVP, casts further light on the potential for social division created by flawed community targeting. Through a process of social mapping, targeting exercises and group discussions in six randomly-selected villages (two in each of Malawi’s three regions), this study concluded in every case that targeting was “inappropriate”. A variety of reasons was given for this: that targeting is against the sprit of umodzi (togetherness); that it creates tensions in the village and provokes reprisals (even witchcraft) from those excluded; and that non-beneficiaries withdraw from community development initiatives. Contrary to suggestions by its proponents that community-based targeting may enhance social capital, this study found that “targeting in a context of high poverty levels breeds suspicion, hatred, accusations and corruption”. Similarly, assertions that social empowerment is achieved through participation in targeting processes are flatly contradicted by the study’s findings that “in all the villages except one community based targeting does not qualify to be a democratic process, with community leaders dominating the decision-making process”. The study concludes that “asking people to select one poor family against another is tantamount to procedural injustice … In a context of high vulnerability, targeting for a precious resource … is a matter of death and life. It is not surprising that communities are unwilling to pass that judgment”.

For a practical illustration of these conceptual problems, demonstrating that inaccurate targeting merely compounds much more invidious inter-household equity issues, we need look no further than the Machinga study cited above. If we add the value of the SCT received over the course of a year by each household to its pre-transfer disposable income, we can represent the impact graphically. Figure 3 shows the effect on recipient households (in red).

If we then re-order the same graph in ascending order of disposable income per adult equivalent (Figure 4), we can see that the beneficiary households are all now (after just one year of receiving the SCT) grouped in the top half of the income distribution chart. That is inequitable.

With the justification that a stated objective of the current raft of SCT pilots in Malawi and Zambia is to learn lessons, we need to be honest enough to recognise the fatal flaw in the prevailing model: that community-based targeting of an inadequate 10% quota is an unacceptable model for national replicability in sub-Saharan Africa. New targeting approaches – such as categorical schemes (social pensions, child benefits and disability grants), or targeting for exclusion rather than inclusion – need to be tested; beneficiary numbers need to be significantly raised to reflect national levels of poverty and vulnerability; and transfer amounts need to be adjusted to levels where they do not cause some lucky beneficiaries to leapfrog the standard of living of non-beneficiaries in the same communities. Recognising this would be an important step in the key process of gaining political support for the national implementation of comprehensive social protection schemes.

ⁱ Both countries’ schemes have as their third objective: “To generate information on the feasibility, costs and benefits, and on the positive and negative impacts of a Social Cash Transfer scheme”.

ⁱⁱ John Seaman, Celia Petty and Patrick Kambewa, “The Impact on Household Income and Welfare of the pilot Social Cash Transfer and Agricultural Input Subsidy Programmes in Mlomba TA, Machinga District, Malawi” (June 2008).

ⁱⁱⁱ The use of such estimates, derived from national data, itself worsens the targeting problem: a single proportion (such as 22% in Malawi) clearly cannot be expected to apply evenly across geographical and social space, even if it can be delineated satisfactorily at a national aggregate level. It follows that it will over-capture the kind of households it seeks to target in some places (wrong inclusion) while under-capturing such households in other places (wrong exclusion).

^iv Frank Ellis, “‘We Are All Poor Here’: Economic Difference, Social Divisiveness, and Targeting Cash Transfers in Sub- Saharan Africa” (Sept 2008).

^v Overtoun Mgemezulu, “The Social Impact of Community Based Targeting Mechanisms for Safety Nets” (August 2008).

Missing the Target

This blog was published jointly with RHVP colleagues as a Wahenga Comment

The boundaries of social protection are hard to define. Especially recently, as it has become “flavour of the month” in development circles, we have seen governments, donors and NGOs falling over themselves to repackage a whole variety of traditional development interventions as “social protection”. But, however wide you cast the social protection safety net, everybody should be agreed on one thing: that there is a core constituency who should be the priority focus of social protection. These are the poorest and weakest in society.

 Conventional thinking, especially in a neo-liberal ideology, goes on to suggest that the best way to support this group is to target them for specific interventions. The argument is that limited resources are more efficiently used when allocated to this particular sub-group. And so we have seen a proliferation of programmes, especially in Africa, that attempt to target the poorest and weakest in society, in order to provide them with some kind of social transfer. But does such targeting work? The answer is rarely, or never.

First, targeting the poorest is notoriously difficult to do, especially in an African context where a substantial proportion of the population is poor (50%-80% in many countries), and where the differentials between poor households are minimal (viz. the often-heard refrain that “we are all poor”). A variety of targeting mechanisms has been tried in order to identify the most needy; but all have significant weaknesses. Indeed many of them have inherent flaws that mean they will systematically miss the target group:

• Self-targeting through public works programmes by definition excludes those with no labour.
• Community-based targeting (especially at large scale) tends to exclude those with no voice and no status, by perpetuating local power structures (for example, a study of Malawi’s Targeted Inputs Programme found that random selection would have resulted in better targeting of the poorest than relying on the local community).
• Means-testing excludes those who lack official documents and bureaucratic savvy (for instance, only 10% of eligible recipients accessed South Africa’s Child Support Grant, until the means-testing bar was lowered).
• Targeting through the markets, for example using input subsidies, excludes those with no cash (because they cannot afford even the subsidised price), and with no land (because they cannot use the subsidised product).

Since those in the core constituency for social welfare typically have no labour, no voice, no documentation, no cash and no land, they risk being missed by each of the above targeting mechanisms. Perhaps in African countries it would be better to accept this, and – if you have to target at all – to target the better-off, who are a much smaller and more easily characterised group, for exclusion from social benefits (although this might be dangerous from the perspective of political support – see below).

Second, targeting is highly demanding in time, resources and institutional capacity. Studies, even from staunch advocates of targeting such as the World Bank, show that the administrative costs of a targeted programme are in the region of 15%-30%, compared with 5%-10% for universal programmes. And, since good targeting requires good information and strong administrative capacity, we cannot expect it to work well in the majority of countries in sub-Saharan Africa. Again the World Bank[i] admits: “institutional capacity in very poor countries tends to be very limited, making targeting mechanisms even more difficult to administer”. Thus a much-quoted World Bank review[ii] of 122 antipoverty targeting interventions in 48 countries concluded that: “the median targeting programme in Africa transfers 8 per cent less resources to poor individuals than universal programmes”[iii].

Third, the efficiency of targeting raises serious questions[iv]. Most targeted programmes attempt to gauge the accuracy of their targeting by measuring the percentage of the transfers that reaches the poorest. They boast, for example, that 70% of transfers are made to the poorest quintile. This reflects a preoccupation with minimising “leakage” to the non-poorest.  But if one’s interest is in poverty reduction, then the issue of under-coverage is of equal – or arguably even greater – importance, than that of leakage. Traditional assessment of targeting accuracy takes no account of this. As an example, let us look at the much-vaunted Kalomo social cash transfer scheme in Zambia. Even if one accepts a quota system as being an ethically-justified approach to social protection (where only a fixed 10% of any community receives the benefit); and even if one overlooks the fact that such an approach completely ignores the geographical distribution of poverty (ie that the poorest 10% in one area is likely to be significantly more, or less, poor than the poorest 10% in another area), one is still dealing with a highly inefficient targeting mechanism. In terms of inclusion error, Kalomo’s own Final Evaluation Report[v] suggests that “42% of all beneficiary households should not have benefited” (even based on the scheme’s already very restricted criteria) – this is a high, though by no means exceptional, level of leakage. But of much more concern is the likely level of exclusion error: in a country where 70% of the population is poor, the arbitrary 10% cut-off necessarily involves very significant under-coverage. Taking this too into account, it is likely that the actual efficiency of the targeting is lower than if it had been a universal programme.

Finally, there is the issue of the political economy of targeting, where such under-coverage as described above becomes particularly significant. By targeting, especially in the African context of pervasive poverty, you are essentially creating a privileged group among the poor, an inequitable practice in the first place, but one which will inevitably result in perverse incentives, arbitrary treatment, corruption and patronage. As a recent UNRISD report[vi] comments laconically: “stated simply, combating social exclusion with programmes that exclude a high number of the socially marginalized does not look like a successful formula for reaching the poor”. And this is supported by considerable evidence, from developed and developing countries alike, showing that targeted programmes are less popular than universal ones. As the number of beneficiaries decreases, the balance between those receiving and those paying for a benefit changes, and political support ebbs away: the drive to reduce a programme’s leakage in reality undermines the popular appeal of that programme. As Amyrta Sen[vii] observed: “benefits meant exclusively for the poor often end up being poor benefits”.

The sooner we recognise this and move to universal rather than targeted social transfers, the sooner social protection – in whatever guise – will have a genuine impact on poverty.

---

[i] Grosh M E (1994) “Administering Targeted Social Programs in Latin America: From Platitudes to Practice”, World Bank.

[ii] Coady, D, M E Grosh and J Hoddinott (2004) “Targeting of Transfers in Developing Countries: Review of Lessons and Experience” World Bank.

[iii] As quoted in wahenga.comment (2007) “Should we forget about targeting?”, RHVP.

[iv] This section draws on the excellent analysis in Dutrey, A (2007) “Successful Targeting? Reporting Efficiency and Costs in Targeted Poverty Alleviation Programmes”, UNRISD, to which the reader is strongly recommended.

[v] GTZ (2007) “Final Evaluation Report: Kalomo Social Cash Transfer Scheme”, GRZ/GTZ.

[vi] Dutrey, A (2007) “Successful Targeting? Reporting Efficiency and Costs in Targeted Poverty Alleviation Programmes”, UNRISD.

[vii] As quoted in “RHVP Policy Brief No 6: Targeting Social Transfers”, RHVP.