Decentralization: China and Russia

Transcription

1 March 2008 Delhi School of Economics MA in Economics Course 902: Issues in Economic Systems and Institutions Teaching Notes 1 Decentralization: China and Russia Decentralization demands more centralization and more sophisticated political skills at the national level. Ex-Colombian minister of State Rudolf Hommes (cited in Tendler, 1997 p.143). Decentralization has been much in vogue in the development community in the past decade or so. This has flowed from more than one source: reaction to the perceived failures of centrally managed design and implementation; hopes that more local government will be more responsive to local preferences, needs and opportunities, including of the poor; or simply following actual decentralizations, that have often been driven by political rather than economic forces, notably in Latin America and Russia in the late 1980s and 1990s. The hope that decentralization would bring more effective, and more poverty-focused, development has often been linked to participation, both through formal democratic processes, and perhaps especially in development circles in the heightened emphasis on the supposedly positive role of civil society in aggregating local interests and pressing governments for action. Decentralization has taken a variety forms. It has involved mere devolution of service delivery (backed by central grants); genuine fiscal devolution, with revenue-raising capacities of local government; and political decentralization with local elections. The level of decentralization has also varied considerably, from center to province or state (that in China and India have populations larger than most countries), to the village or community, as in the related vogue for community-driven development amongst many donors (notably the World Bank) in the last few years. The optimism over decentralization-plus-participation contrasts with an older tradition of skepticism, for which the classic example is the concern of James Madison (one of the designers of the US constitution) with local elite capture, aptly caught in the following quote: 1 These notes were prepared by Michael Walton and are solely for teaching purposes.

2 2 the smaller the society, the fewer probably will be the distinct parties and interests composing it; the fewer the distinct parties and interests, the more frequently will a majority be found of the same party; and the smaller the compass within which they are placed, the more easily will they concert and execute their plans of oppression (cited in Bardhan and Mookherjee, 2005, from an original reference in the Federalist, 1787) This is illustrative of theoretical ambiguity. There are also striking empirical contrasts. China undertook a major decentralization after High levels of provincial investment and growth ensued. Local governments have supported the expansion of new firms, and China scholars see this as an instance of market-preserving federalism (Qian and Weingast, 1997). Russia underwent a major decentralization after its political and economic transition of the 1990s. Growth was dismal. Local governments often became a block to growth, and were often captured by large firms (Blanchard and Shleifer, 2000, Zhuravskaya, 2000). How do we put some conceptual order on these apparent contrasts? And what does this imply for approaches to policy? In this session, we look at concepts in relation to the contrasting experiences of China and Russia. The overarching message is a meta-theme of the course: both design and context matter, but structured analysis is a tool for working through policy choice in real settings. Concepts and application to the China-Russia comparison It is useful to distinguish two generations of theory on decentralization (Qian and Weingast, 1997) 2. The first generation treated government decision-makers as benevolent social planners, albeit often with limited information. Two mechanisms whereby decentralization can bring benefits were emphasized. First, local governments may have better information on the preferences of their local populations. Second, with inter-jurisdictional mobility, people and firms can move to match their preferences to the array of public goods offered by local governments. 3 Whether a service should be delivered under central or local auspices then depends on the degree of externalities across jurisdictions: defence, national; school delivery, local. The second generation treats governments, politicians and other actors as agents pursuing objectives that may be imperfectly aligned with the principals they supposedly represent. This involves a sharp focus on how incentives for agents are shaped by institutional structures. For example, in the theoretical analysis of poverty-targeted programs in Bardhan and Mookherjee (2005) the central contrast is between the behavior of a corrupt central 2 For the purposes of this note, we are concerned with issues that are common to decentralization and the narrower frame of federalism. 3 The first view is originally associated with Friedrich Hayek and the second with Charles Tiebout.

3 3 government bureaucrat working at the local level with informational advantage vis-à-vis his superiors, and the political capture of local governments. As they put it: It is worthwhile to clarify the essence of the contrast between centralization and decentralization in our model. The nature of the incentive problem is qualitatively different across the two systems: Authority is shifted away from unsupervised central bureaucrats that seek to maximize bribe incomes, toward elected officials of local governments whose objective is to maximize the chances of staying in power. Economic corruption tends to be replaced by political corruption (in the form of diversion of services to local elites.) The difference between the two systems [stems]..from differing patterns of accountability arising from different information and incentive systems. (p. 679) In similar vein, the World Bank s World Development Report 2004 on service delivery is essentially based on a principal-agent framework, with imperfect alignment in objectives between principals and the variety of agents. This emphasizes the array of actors, with citizens (the ultimate principals) interacting with central government, local government and service providers (Figure 1). It argues that attention to how accountability relationships fail or work, in affecting the resources and incentives of providers, is fundamental to understanding the nature and quality of service delivery. The same principle applies to a broader range of services, including public goods and services for the business sector. Policy should focus on shaping accountability relationships, or on action that improves behavior conditional on those parts of the accountability structure that cannot be changed. Figure 1. The World Bank s accountability framework applied to decentralization Decentralization National policymakers Local policymakers Poor people Providers Source: World Bank, This is only a simplified listing of the actors. The question is how to put some structure on the incentive properties of different institutional designs and how these interact with

4 4 different initial conditions. We highlight a few aspects, drawing on a core statement of the problem that parallels approaches to dealing with moral hazard. The principal-agent problem, and how decentralization of information and authority can resolve it. We can structure the initial question in terms of principal-agent problem, now with the central government as principal and local government as agent. The central government wants to generate local economic activity. Action of local government influences this: say growth, dy i, in region i, is a function of provision of a variety of local public goods and reforms (pg i ). dy i = f(pg i ) (1) However, effort of the local government is not monitorable. How can the government structure incentives to induce high effort? Under a market-preserving federalism design, the activity is, in essence, devolved to local governments, with a high level of incentives and little risk-sharing in the contract. The mechanism for this is fiscal decentralization: local governments not only get to raise revenues but also to keep what they raise. They thus have an interest in an expanding the tax base dy, and in choosing policies to foster this (some of which are financed by the taxes raised.) This is based on the view that, growth-promoting public good provision is a function of marginal tax retention, tr i : pg i = g(tr i ) (2) A stylized version of the contract is depicted in Figure 1. Figure 1. Fiscal federalism implies high returns to local (tax) effort Reward s High Low Effort

5 5 How was this done? In China s case by reforming the fiscal inter-relations from the pre system of eating from one big pot to the fiscal contracting system known as eating from separate kitchens, that applied from Without going into details, for center-province relations this had three essential elements: (i) provincial revenue accounted for a high fraction, some 66% of total (provincial-plus-central) revenues; (ii) long-term (typically 5-year) fiscal contracts on any inter-governmental transfers were agreed, that effectively allowed provinces to keep marginal taxes; and (iii) extrabudgetary revenues was made legal (Jin, Qian and Weingast, 2005). While less fully documented, similar arrangements were made between provinces and lower levels of government, down to local level of townships. In other words, local governments became residual claimants on the surplus created in their jurisdiction. From 1994 there was a further fiscal reform, the separating tax system that formally assigned taxes between province and center, with income taxes on all non-central government businesses, personal income taxes, and the VAT shared. Provinces thus still got to keep a large part of taxes on provincial tax bases. This is in principle a high incentive contract. But is it credible? There are two threats to its incentive properties: in the high effort case, through the central government taking part of the provincial government s taxes the center predation problem; and in the low effort case, through the central government helping out the soft budget constraint problem (Qian and Weingast, 1997). The credibility problem is reduced through the treatment of information. Whereas insufficient information was part of the initial monitoring problem, too much information can make it harder for the central government to credibly commit not to predate or bail out. Qian and Weingast argue that allowing local governments extra-budget and off-budget accounts, about which they had limited knowledge, was a form of making the commitment more credible. Evidence from China and Russia There is evidence for the stylized equations (1) and (2) and the role of marginal tax retention for both China and Russia. For China, we highlight two results from Jin, Qian and Weingast (2005). First, pre-1979 there was little correlation between revenues and expenditures (indicating very low marginal tax retention); after the reforms this rose substantially, and was 100% for many provinces (Table 1).

6 6 Table 1 Chinese provinces experience high marginal revenue retention rates after 1980 Source: Jin, Qian and Weingast, 2005 Second, through exploiting the variation in the provincial marginal revenue retention rate tr i, the authors could look at impacts on the growth of the non-state sector, and find a significant positive impact, illustrated in Table 2. They also found a significant positive impact on measures of reform of the state sector (proxied by the change in the share of contract workers in state employment.) 4 4 The authors could not directly test for endogeneity, but do explore possible bias: in particular they find tr i is negatively related to lagged non-state sector growth, that would bias the core results against their hypothesis.

7 7 Table 2. The provincial marginal revenue retention rate is associated with higher growth of the non-state sector Source: Jin, Qian and Weingast, 2005 Was Russia different? Yes, both in design and outcome, in ways that are consistent with the theory. In a similar study based on the budgets of 35 Russian cities for , Zhuravskaya (2000) finds both a much lower share of own revenues (at 20% at the end of the period) and that virtually all marginal own-revenues were effectively extracted by the central government (that is tr i was close to zero). This is attributable to two factors: first, fiscal contracts lacked the incentive properties of Chinese contracts they were reckoned by observers to be based primarily on political negotiations; and second, the central Russian government was desperate for revenues in the 1990s, rendering non-credible any commitment not to predate on local government revenues. Furthermore, Zhuravskaya finds evidence that greater city-level fiscal incentives (proxied by revenues shared with higher levels of government moving in the same direction as own revenues) is associated with a greater number of newly privatized and new businesses in the region. Finally, she finds indirect evidence of more efficient public goods provision with a significant (beneficial) influence of the incentives proxy on infant mortality and a measure of availability of schools.

8 8 Is this the full story? Almost certainly not. There are a number of other factors at work, that make the story more nuanced and context-specific. We sketch three here: on the political power of the center, on the influence of local power structures, and on interjurisdictional competition. The role of the political power of the center. High marginal tax retention ratios is not the only way to incentivize local governments in pro-growth directions. Blanchard and Shleifer (2000) suggest an extension of the incentive problem. A local government chooses between whether to pursue growthfostering or growth-killing policies, the latter through protection of old firms and restricting or predating on new firms. The central government can offer the carrot of high levels of tax retention from new growth, as in the above account. If it has the power it can also influence the probability that local government will stay in power. Let p f be the probability of staying in power if a local government fosters growth, p k the probability of staying in power if it kills growth, b the benefit received from killing growth (e.g. in bribes from old firms), then, with appropriate normalization, a simple formulation of the local government s decision is that it will choose to foster growth if p f.tr.dy > p k.b, or equivalently p.tr.dy > b (3) with p = p f / p k Here p reflects, in part, the central government s power to influence the survival of local governments (see next point on links with local support). Blanchard and Shleifer then argue that a key difference between China and Russia was the strength of the center: in China a strong center, reinforced by the communist power, they suggest the central government had the power to act against growth-killing provincial governors. In Russia, the central government was (circa 2000) neither strong enough to impose its views, nor strong enough to set clear rules about the sharing of the growth. They argue that power at the center is necessary to make decentralization work. The role of local power structures and elite capture The account so far has a highly simplified account of the relationship between local government and local conditions. This can be given more structure. In an approach that is the mirror opposite of the omnipotent central government appointing local government (that could lead to very high p in equation (3)), the latter depend on their capacity to sustain the support of their constituents. Bardhan and Mookherjee (2000, 2005a) develop an approach based on different parties and elections, in which support for one party s program depends on the preferences of different groups in a population, the extent to which different groups are politically aware, the extent of electoral competition, interest group cohesiveness, and the capacity of richer groups to use campaign finance (or other forms of influence) to shift the opinions of the uninformed. There can be disparities in

9 9 political awareness across groups, with evidence suggesting that poorer groups are typically less politically aware than richer groups. We do not present the formal results here, but the intuition is clear: even in a democratic system, there is scope for elite capture, and this will tend to rise with lower political competition, greater interest group cohesiveness, lower political awareness and greater disparities of awareness across groups. Moreover, it is theoretically ambiguous whether capture is greater at local or national levels. These considerations apply with equal or greater force in non-democratic or imperfectly democratic systems. Some evidence on local capture comes from Russia, from a study that exploited the publicly accessible character of local Russian laws, and could document instances in which specific large companies were favored, for example through legislated tax breaks or government subsidies (Slinko, Yakolev and Zhuravskaya, 2005). They looked at firms that had been amongst the largest five by sales in each region in one year of They found over 40% had received a legislated preference, and none from outside their list. Such politically favored firms performed better than average, while growth of firms without political connections, small-business dynamism and tax capacity were all lower where there was greater capture. In this period there was no significant effect on overall regional growth, however. This is just one example of evidence of local capture. When such capture is significant, there is a further role for central government action in making decentralization work. This is one of the themes of Tendler s (1997) rich analysis of the very different context of delivery of a variety of services, including go small businesses, in the poor and unequal state of Ceará in Brazil. The role of inter-jurisdictional competition A further story emphasized in the market-supporting federalism account is the influential of competition for mobile factors on incentives (Qian and Weingast, 1997). Local governments have incentives to offer better public goods to keep and attract mobile capital. This also increases the opportunity costs of bailouts to poorly performing firms. This is argued to have played a role in China, for example with respect to foreign capital. However, this beneficial story is not guaranteed. Where regions are heterogeneous, some low-performing regions may stand little chance of attracting mobile capital, and it may be more advantageous to pursue predating, patronage-based and growth-killing policies (see review of papers in Bardhan and Mookherjee, 2005b). Think of the electoral strategy of the government of Bihar in India, in which a corrupt government is sustained by delivering private benefits to its political base (in recent years a lower-caste base) rather than growth-fostering public goods, keeping it in a low-level equilibrium. And even amongst regions who can compete for capital, this can take the form of competitive taxbreaks and subsidies as in the fiscal wars amongst Brazilian states with an equilibrium of lower resources for public goods.

10 10 A context-specific overall message An important lesson from both the theoretical and empirical literature is that general statements on the positive or negative impact of decentralization are likely to be wrong. But this does not imply that anything goes, or that everything depends. It is rather that: (a) context matters, but in structured ways; (b) design clearly matters, but the effect is a product of interactions with context; (c) theory is a powerful tool for structuring analysis of both (a) and (b); (d) while one category of policy lesson is to shape design taking context into account, another is to work on changing the context in ways that are more likely to be beneficial for decentralization, for example through social mobilization and raising political awareness. Questions and themes It is clear that both design and context help explain the contrasting experiences of China and Russia with decentralization. This raises some interesting policy questions Given Russia s initial conditions, could decentralization in the 1990s have been designed differently in technical terms to achieve a more growth-promoting result? Was there a technical solution that would have been politically supportable? And how do we analyse the restoration of central power under the Putin administration, in light of the incentive features of decentralization? How do we interpret the reported evidence of growing regional protectionism and local corruption in China? Even if decentralization had elements of beneficial marketpreserving federalism, does it need to be adapted? How does it need to be modified to deal with rising regional inequalities? What might these experiences imply for countries in the middle of long processes of decentralization, such as Mexico? References Bardhan, Pranab and Dilip Mookherjee Capture and Governance at Local and National Levels. American Economic Review 90(2), pps a. Decentralizing Antipoverty Program Delivery in Developing Countries. Journal of Public Economics 89, pps b. Decentralization, Corruption and Accountability: An Overview. Forthcoming in Rose-Ackerman ed. Handbook of Economic Corruption. Edward Elgar.

11 11 Jin, Hehui, Yingyi Qian and Barry R. Weingast Regional Decentralization and Fiscal Incentives: Federalism, Chinese style. Journal of Public Economics 89, pps Qian, Yingyi and Barry R. Weingast Federalism as a Commitment to Preserving Market Incentives. Journal of Economic Perspectives 11(4) pps Slinko, Irina, Evgeny Zakovlev and Ekatarina Zhuravskaya Laws for Sale: Evidence for Russia. American Law and Economics Review 7(1), pps Tendler, Judith Good Government in the Tropics. Baltimore and London: The Johns Hopkins University Press. World Bank World Development Report 2004: Making Services Work for Poor People. New York: Oxford University Press. Zhuravskaya, Ekaterina Incentives to Provide Local Public Goods: Fiscal Federalism, Russian Style. Journal of Public Economics 76, pps