Decentralised targeting of transfer programmes: A reassessment

Ahead of the release of ‘Decentralised Governance: Crafting Effective Democracies Around the World’, Dilip Mookherjee brings together some of the arguments against decentralisation of welfare programmes, including the incidence of political clientelism and elite capture, and summarises attempts made by developing countries to undertake hybrid ‘recentralisation’ initiatives. While acknowledging that direct benefit transfer (DBT) programmes can limit the scope for misallocation and corruption, he probes the ability of DBTs to be responsive to localised shocks and recentralisation’s implications for fiscal federalism.

The first phase of modern economic development between 1950-1990 was characterised mostly by centralised, top-down development programmes. Planning, execution and service delivery were delegated to a bureaucracy appointed by and accountable to a central government at either the federal or state level. However, there was growing disenchantment with this system owing to targeting failures, leakages, losses, corruption, and the lack of responsiveness to local needs. Consequently, the last three decades witnessed a shift towards decentralised, bottom-up implementation where authority was shifted to local government officials elected by local citizens. The primary motivation was to bring information about local needs and capacities into decision-making and align the incentives of officials more closely with the interests of local citizens.

The argument that decentralisation enhances state accountability is, however, highly controversial. Counterarguments have been made by designers of both the US and Indian constitutions, on the ground that local governments are more prone to ‘capture’ by local elites compared with central governments, especially in areas of high inequality and poverty. These arguments are complemented by additional concerns that local democracy tends to be characterised by higher levels of political clientelism, wherein local incumbents manipulate allocation of benefits to loyal supporters or swing voters to increase chances of being re-elected. A large literature (reviewed in Mansuri and Rao (2013) and Mookherjee (2015)) has documented systematic evidence of such problems and how they have distorted allocation of benefits both within and across local communities. This motivates a reappraisal of the choice between centralised and decentralised governance mechanisms, and the exploration of possible hybrid forms which might perform better.

Efforts at ‘recentralisation’

Some countries have recently experimented with ‘recentralisation’ initiatives, which have enhanced monitoring or reduced discretion provided to local government officials, often utilising emerging new technology and ‘big data’ capabilities of central governments. Examples of this include field experiments with ‘e-governance’ by MGNREGA, a large employment and social insurance programme operating in different Indian states (Banerjee et al. 2019, Muralidharan et al. 2016, 2020). These experiments achieved varying degrees of success in reducing leakages and corruption. Moreover, they were not aimed at reducing partisan or personal biases of local government officials in the allocation of benefits, either within or across communities.

Other research reports the results of experiments or simulations of reforms in Kenya (Hoffman et al. 2017), Brazil (Finan and Mazzocco 2021) and in Indian states such as West Bengal (Mookherjee and Nath 2021) which used formula-based grants instead of relying on the discretion of officials in intermediate layers of government to allocate funds across different areas under their jurisdiction. However, last-mile delivery within local communities continued to be delegated to elected officials. The results of these reforms also appear to have been mixed. In Kenya and Brazil, the inter-community allocation biases associated with elite capture were moderated, but at the cost of increasing local corruption. In India, formula-based grants failed to improve pro-poor targeting owing to limited information about the scale of benefits administered by local governments, which form the base of the formula used. Intra-community mistargeting continued to arise from delegation of last-mile delivery within communities.

More far-reaching recentralisation initiatives were undertaken in Pakistan (Habeeb and Vyborny 2021) and Indonesia (Banerjee et al. 2021). These eliminated the scope for discretion of local officials at all levels by directly delivering benefits to citizens predicted to fall below a poverty line based on an asset-based proxy-means test (PMT) formula¹. The Indian government has also initiated a switch from traditional welfare programmes to direct benefit transfers. These reforms necessitate investment in a country-wide administrative database of individuals, households and assets, combined with registration of beneficiaries and direct delivery mechanisms through bank accounts or mobile phones. The improvements in targeting of these programs appear to be impressive, and therefore raise interesting new questions about the feasibility or desirability of other developing countries adopting them.

Issues with direct benefit transfer programmes

Direct benefit transfer programmes indicate a transition to a Western-style social security system where the central government administers most private transfers programmes, with the role of local government restricted to the provision of local infrastructure and collective goods such as education and health. Besides the advantage of limiting scope for elite capture, corruption, clientelism and misallocation of welfare benefits, the system would become more transparent and make citizens less dependent on local intermediaries. Moreover, benefits would be portable across locations: entitlement to welfare benefits would not be jeopardised because of migration between districts or states.

One concern with direct benefit transfer programmes could be the adequacy of asset-based PMT measures in accurately measuring poverty, since they ignore temporary shocks experienced by households. Recent assessments of the extent to which local government officials are well-informed about household level shocks and respond with benefit allocations are mixed (Basurto et al. 2019, Trachtman et al. 2021). However, there is scope for augmenting PMT-based allocations to incorporate locality-based environmental shocks.

A more substantive concern deals with the feasibility and accuracy of centralised administrative databases and e-delivery mechanisms. To what extent can poor countries with low state capacity develop and operate such systems effectively? There is a need to understand better how Pakistan and Indonesia managed to develop and operate these systems.

A final concern is the potential implications of such far-reaching recentralisation programmes for fiscal federalism and the balance of political power between central and regional governments. Cheema et al. (2006) have described how successive waves of decentralisation and recentralisation in Pakistan were related to the conflict between authoritarian rulers at the central level and populist democrats at the regional level. Should direct benefit transfers be administered by state or central governments? To what extent can the central government be trusted to abide by the PMT-based formula for benefit transfers, instead of manipulating actual transfers to undermine the power of regional governments governed by different parties? These questions await further discussion and research.

This article is a summary of the forthcoming chapter ‘Decentralized targeting of transfer programmes: A reassessment’, in Jean-Paul Faguet and Sarmistha Pal (eds), Decentralised Governance: Crafting Effective Democracies Around the World, London, LSE Press.

Note:

Proxy-means testing uses multivariate regression to correlate certain proxies, such as assets and household characteristics, with poverty and income (Commonwealth of Australia, 2011).