Discussion of ‘Fiscal Decentralisation as an Example of Institutional Ineffectiveness’

Summary

It is widely accepted that countries’ institutions play a major role in their economic development. Yet, the way they affect, and are affected by, development, and how to reform them are still poorly understood. In this companion volume, State and Business in Tanzania diagnoses the main weaknesses, root causes, and developmental consequences of Tanzania’s institutions, and shows that the uncertainty surrounding its development paths and its difficulty in truly ‘taking off’ are related to institutional challenges. Based on a thorough account of the economic, social, and political development of the country, this diagnostic offers evidence on the quality of its institutions and a detailed analysis of critical institution- and development-sensitive areas among which state-business relations rank high, even though the institutional features of land management, civil service and the power sector are shown to be also of prime importance. This title is also available as Open Access.

I The Concept of Decentralisation

Decentralisation is a rather vague term, used to denote quite different concepts: political, administrative, fiscal, and market decentralisation.⁶ While these concepts overlap, they also differ in major ways. To start with the most radical of these four concepts, market decentralisation shifts authority and responsibilities out of the government to the private sector. Typically, under such privatisation the government retains considerable power as a shareholder (sometimes with special powers as holder of a ‘golden share’), or as regulator of the private firms providing public services. The central issue with this type of decentralisation is that the two agents, the government and the private provider, do not have the same objectives or information. This gives rise to a principal–agent problem.⁷

Fiscal decentralisation is in my view best thought of not as a form of decentralisation in its own right, but rather as a necessary complement of the three other types of decentralisation. After all, whenever central government functions are shifted to regional or local governments or to private providers, these must be enabled to fulfil their function. At a minimum, they must have an adequate source of revenue: local taxes, user charges, or a transfer of central government revenue.

Administrative decentralisation transfers authority and responsibility for the planning, management, and provision of public services to a lower level of government or to a special agency. In its weakest form, deconcentration, authority and responsibility remain with the central government but are shifted from civil servants based in, for example, the capital to those based at a regional centre. In a stronger form, delegation, the central government surrenders some of its control over planning and execution to an agent that has considerable autonomy while remaining accountable to the central government.⁸

Political decentralisation shifts political control over public decision making to representatives of those directly affected. Typically, control is transferred to an elected body at a lower level, such as a city council or regional legislature. (When political decentralisation is far reaching, it is sometimes called devolution. But the difference is only one of degree: there is no clear boundary.)

II The Rationale for Decentralisation

What are the pros and cons of decentralisation? There are at least four different rationales, related to scale, externalities, preferences, and rent-seeking.⁹ First, economies of scale in some aspects of public services provision obviously favour centralisation. Secondly, decentralisation is problematic if decisions taken locally have effects outside the jurisdiction. Such externalities will lead to underprovision if the spillover effects are positive or overprovision in the case of negative effects. In this case, centralisation to a sufficiently higher level has the advantage that the externalities are internalised: the effects in other jurisdictions that would be ignored at the local level are taken into account when the decision is transferred to a higher level. (The subsidiarity principle implies that decisions should be taken at the lowest possible level, but not lower. This in itself is too vague to be helpful, but defining the appropriate level in terms of externalities gives the concept teeth.)

Thirdly, if preferences are heterogeneous across space, then public services must be provided in different forms in different locations, and this calls for at least administrative decentralisation. This is likely to work only if complemented with some form of political decentralisation so that local preferences can be articulated at the political level and policymakers are held accountable for the way they respond to those preferences. Decentralisation then has the advantage that it generates better information on the demand for public services and (through political accountability) strong incentives for the public sector to tailor provision to that demand.

Finally, decentralisation may increase the scope for rent-seeking. Hence, public sector functions are sometimes centralised in order to reduce this.¹⁰ However, it is important to note that the direction of the effect of decentralisation on rent-seeking is not clear. Rent-seekers who try to influence decisions at higher government levels may face more countervailing power there (either from competing rent-seekers or from public sector agents who oppose them) than those who operate at lower levels. On the other hand, at the local level their actions may be easier to observe and therefore to resist. The net effect of these two forces, countervailing power and asymmetric information, is not clear.¹¹

III The Reasons for Decentralisation in Tanzania

This brief review of the nature and rationale for decentralisation will help to structure my discussion of this chapter by Likwelile and Assey.

The chapter clearly shows how special the Tanzania case is. Local government was weakened dramatically after the 1967 Arusha Declaration. The LGAs were revived in the early 1980s but were given remarkably little power: the central government and the party remained firmly in control. The law even entitles the central government to do anything that ‘shall facilitate or secure the effective, efficient, and lawful execution by the District Authorities of their […] duties’.

Likwelile and Assey give a clear and very useful account of the evolution of local government in Tanzania. (This is neatly summarised in Table 6.1.) But they say very little about the reasons for the various changes they document. The centralisation in the late 1960s clearly had very little to do with the rationales I have listed: scale economies, externalities, heterogeneous preferences, or rent-seeking.¹² Instead, it was mainly driven by ideology. The subsequent decentralisation is more puzzling. Why did it happen? Because donors pressed for it? Because rent-seekers felt it would give them new opportunities? Because the leadership believed in it? What then did they think it would achieve? Why did reforms oscillate between centralisation and decentralisation?

The chapter is almost silent on these questions. There are very brief references to the economic crisis of the early 1980s,¹³ the collapse of public services provision at that time, and ‘pressure’ from donors. We are not told what the objectives of the reforms were. Hence, the chapter offers a fascinating factual account of the reforms, but no political economy analysis of the changing incentives for the various agents involved, nor an empirical analysis of the results of the reforms.

Reading between the lines, the real story appears to be that the central government, while often describing decentralisation as a way to promote good governance (perhaps even a panacea), has in fact never fully accepted decentralisation. Its ability and desire to resist in practice what it preached waxed and waned over time, for reasons that are unfortunately not discussed. As a result, Tanzania has moved back and forth between centralisation and decentralisation, responsibilities have often been unclear, and fiscal decentralisation has seldom been in line with political and administrative decentralisation. If this interpretation is correct, then the incentives for the central government to push for or to resist decentralisation should be at the centre of the story. Focusing instead on fiscal decentralisation, as the chapter does, amounts to studying the symptoms rather than the disease.

IV Fiscal Decentralisation and Local Revenue Collection

At the outset, in Section I, the authors indicate that their focus is ‘on the administration of local revenue mobilisation given the centrality of financial resources in empowering local authorities to deliver on their mandate and improve their performance’. This suggests that to understand decentralisation and development in Tanzania the key process to study is fiscal decentralisation.

This seems hard to justify, for two reasons. First, as the authors themselves stress, local authorities do not have the capacity and integrity required for a successful decentralisation. Therefore, fiscal autonomy may lead to mismanagement and corruption. Secondly, fiscal decentralisation is a process that makes political or administrative decentralisation possible. That may lead to better governance and thereby to better service provision. Whether it actually did so in Tanzania is an important empirical question, but that question is not addressed in the chapter.

In Tanzania LGAs are almost exclusively financed through transfers from the central government: these account for more than 90 per cent of their revenue.¹⁴ The authorities are allowed to raise their own revenue but usually lack the capacity to do so.¹⁵ Possibly, this heavy reliance on transfers undermines horizontal accountability.¹⁶

It is worth stressing that local taxation is neither necessary nor sufficient for horizontal accountability. It is not necessary since, contrary to what the chapter suggests, effective accountability can be achieved without fiscal decentralisation. Reinikka and Svensson (2004) show in a famous study of the use of government transfers for education in Uganda that posting news in the village about the transfer was sufficient to trigger a powerful process whereby villagers held teachers accountable. In this case, accountability was achieved without any institutional change, simply by empowering people with information. Similarly, in Tanzania, villagers in Mkenge (Bagamoyo District) learned in a non-governmental organisation training programme about their rights. They then proceeded in 2010 to throw out the village council with a vote of no confidence for failing to account adequately for public expenditure. Tanzania has also practised an innovative system of village meetings in which spending priorities are discussed.¹⁷ While local leaders predictably try to capture these meetings, they are often overruled so that government transfers are spent on, say, education rather than roads. These examples suggest that the absence of local revenue mobilisation does not preclude effective accountability.

There is a third point to consider, made by Besley and Persson (2013): local taxation requires highly qualified staff for assessment, monitoring, and enforcement.¹⁸ This may make fiscal autonomy very costly.

In summary, there is a trade-off to consider, which involves three effects. The traditional argument is that local taxation will lead to better accountability and hence to better public services delivery, in line with local preferences. These benefits have to be weighed against the cost of local taxation, which may be relatively high if economies of scale are important: the Besley and Persson point. But it may be possible to avoid this trade-off, namely by achieving accountability without relying on the scrutiny of expenditure by taxpayers. The village meetings are one way of doing this.¹⁹ An optimal policy that takes these three effects into account clearly cannot be identified without empirical analysis. But it is important to point out that the desirability of fiscal autonomy – taken for granted in much of the chapter – is by no means obvious.²⁰

V Institutional Weaknesses

Likwelile and Assey argue that raising local revenue faces several institutional weaknesses, including discretion, arbitrariness, unpredictability, and inconsistencies in decisions by the central government. One of the examples they discuss is the oscillation between centralisation and decentralisation in the collection of property taxes. In the pre-2008 phase of decentralised collection, corruption was rampant. This suggests that centralisation may be needed to reduce rent-seeking. But the chapter also mentions that in the subsequent centralised phase, ‘Imperfect information flowing to central operators created opportunities for corruption’. This leaves the reader with a puzzle: if corruption, arguably due to different causes, could flourish in both regimes then what, if any, is the effect of decentralisation on corruption? Apparently, there is no empirical work on this. Clearly, there is a need for an empirically based study that compares corruption under decentralisation and centralisation.

VI Reforms

The chapter’s section on reform (Section VI) is quite brief but lists a whole series of possible reforms.

First, the authors argue that the fiscal decentralisation agenda should be revisited so as to develop a set of rules for local and central governments and ensure predictability. While this is vague, they also take a clear position: given the poor fiscal capacity of the lower-level governments, ‘it makes sense […] to have the central government collect revenues and establish clear legal mechanisms to transfer part of those revenues to LGAs based on recognised resource endowment, the need in terms of public services to be provided, and a fiscal capacity-building component’. This is probably the key statement in the chapter, and I will return to it.

Secondly, predictability should be ensured by ‘rationalis[ing] the conduct of discretionary decisions and actions by the central government’. This sounds like motherhood and apple pie: it is hard to see how one could disagree. Predictability is, of course, a good thing, but how the central government’s scope for discretionary actions should be rationalised is not at all clear. This would require a thorough analysis of the pros and cons of centralisation, culminating in a convincing diagnosis of the problem.

Returning to the key statement about the desirability of the reliance on transfers from the central governments, a few comments are in order. First, this statement seems somewhat contradictory with the authors’ advocacy elsewhere in the chapter of the desirability of fiscal autonomy, unless it is understood as referring to a temporary measure motivated by lacking local capacity. Fiscal autonomy, of course, requires a drastic reduction in the reliance of the LGAs on transfers from the central government. Secondly, one reason for their favouring transfers appears to be the evidence they cite from Masaki (2018) that transfers crowd in local revenue. There is a technical reason to be somewhat sceptical of this result.²¹ More importantly, while Masaki finds a large elasticity of crowding in (0.6), given the very low share of local revenue in total revenue this translates in fact into a very weak effect. Since transfers finance 91.5 per cent of domestic expenditure, an extra government transfer of TZS 1.00 crowds in only about TZS 0.06 of domestic revenue. Therefore, while the effect is statistically significant, it is in economic terms almost trivial. Thirdly, the acute need for ‘more tax evaluators and collectors, and greater capacity to monitor and penalise non-compliance’ are good arguments (following Besley and Persson, 2013) for relying on central tax collection and transfers to LGAs. However, and this is my main objection, this does not clinch the case: these disadvantages of local fiscal autonomy must, as noted before, be weighed against the advantage of fiscal autonomy leading to improved accountability, taking into account the scope for achieving accountability in other ways. Finally, clearly the status quo cannot continue indefinitely. The authors see, if I understand them correctly, continued heavy reliance on transfers as a temporary measure.²² Local capacity must be built up (largely financed by these transfers) so that eventually local revenue can be substituted for transfers. This is sensible but, as the authors recognise themselves, raises the question of how long this will take and, crucially, what incentives LGAs have to use the transfers for such capacity building.

VII Conclusion

The experience of decentralisation in Tanzania is of special interest for two reasons. First, the starting position was an extreme one: the central government and the party exercised a degree of control that was unique in Africa. Secondly, the subsequent decentralisation quickly turned into a bewildering oscillation between more and less central control. That this back-and-forth process was possible, was in part due, as the chapter rightly stresses, to ambiguities in the law. In this respect too, Tanzania seems to be a special and instructive case. The chapter gives an excellent description of this strange Echternach procession (three steps forward, two steps back). However, we learn what happened, but not the reason why.

The effects of decentralisation (political, administrative, or fiscal) on welfare, notably through the provision of public services, are the outcomes of a very complicated process. There are changes in formal power, in the opportunities and incentives for rent-seeking, in corruption, in the scope for holding officials accountable, and in the information available at different levels of government. Since there are many opposing effects in this tangled process, theory is of little help to decide on the desirability of centralisation or decentralisation. The chapter makes almost no use of evidence. There is some evidence for Tanzania, but empirical work is an urgent priority. Ideally experimental methods should be used to estimate the effects of decentralisation. When randomisation is not feasible, comparing outcomes across locations with different levels of decentralisation can still generate very useful information. Even less formal evidence can be helpful. Anecdotal evidence, for example on the effects of decentralisation on corruption or on the circumstances that enable village meetings to hold officials accountable, can begin to build a body of evidence.

In the absence of firm evidence there is little that can be concluded. Is fiscal autonomy desirable? Early in the chapter, the authors appear to take this for granted. However, they later argue that it would lead to corruption and to poor management because of the weak capacity of LGAs. If this is correct, should we then consider Tanzania’s incomplete and half-hearted fiscal decentralisation as a blessing in disguise? This seems to be a corollary of the authors’ conclusion that reliance on central government transfers should continue until the LGAs have built sufficient capacity for raising their own taxes.

Most importantly, the end of the chapter narrowly focuses on the need for capacity building. The political economy issues, including the incentives for corruption, are discussed in the earlier parts but slowly disappear over the horizon.

I have argued that the chapter’s focus on fiscal decentralisation amounts to studying the symptoms rather than the disease. The central issue is what incentives the agents involved – the central government, local authorities, and various groups of rent-seekers – face. These incentives will in part determine what groups will push for decentralisation or will resist it. Theoretical analysis cannot take us very far here. An empirically based study that compares the effects of, and the incentives for, rent-seeking under decentralisation and centralisation is an urgent priority.