WP-S 2516 POLICY RESEARCH WORKING PAPER 2516 Does Decentralization Bolivia's 1994 decentralization led to changes in the Increase Responsiveness geographic allocation of to Local Needs? funds and to the to Loca Neec ws ? development of innovative institutions of local Evidence from Bolivia governance. The changes in the sectoral (and geographic) allocations of public funds jean-Paul Faguet show a strong relationship with objective indicators of social need, evidence that local priorities are being reflected. The World Bank Development Research Group Public Economics H January 2001 POLICY RESEARCH WORKING PAPER 2516 Summary findings Significant changes in public investment patterns-in As the smallest, poorest municipalities invested newly both the sectoral uses of funds and their geographic devolved public funds in their highest priority projects, distribution-emerged after Bolivia devolved substantial investment showed a strong, positive relationship with resources from central agencies to municipalities in need in agriculture and the social sectors. In sectors 1994. By far the most important determinant of these where decentralization did not bring about changes, the changes are objective indicators of social need (for central government had invested little before 1994 and example, education investment rises where illiteracy is the local government continued to invest little afterward. higher). Indicators of institutional capacity and social These findings are consistent with a model of public organization are less important. investment in which local government's superior Empirical tests using a unique database show that knowledge of local needs dominates the central investment changed significantly in education, government's technical and organizational advantage in agriculture, urban development, water management, the provision of public services. water and sanitation, and possibly health. These results are robust and insensitive to specification. This paper-a product of Public Economics, Development Research Group-is part of a larger effort in the group to analyze fiscal decentralization in developing countries. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Hedy Sladovich, room MC2-609, telephone 202-473-7698, fax 202-522- 1154, email address hsladovich@worldbank.org. Policy Research Working Papers are also posted on the Web at www.worldbank.org/research/workingpapers. The author may be contacted at j.p.faguet@lse.ac.uk. January 2001. (44 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center Does Decentralization Increase Responsiveness to Local Needs? Evidence from Bolivia Jean-Paul Faguet* * The author may be reached at J.P.Faguet@lse.ac.uk, Centre for Economic Performance and Development Studies Institute, London School of Economics, Houghton Street, London WC2A 2AE. This paper is taken from chapter 2 of my Ph.D. dissertation. This research was financed by a grant from the World Bank Research Committee. An ORS award and additional financial support were kindly provided by the ESRC. I am very grateful to my advisers Tim Besley and Teddy Brett for invaluable criticism, advice and encouragement through the numerous iterations of this work, and to Gunnar Eskeland for his many insights and constant support. I also thank Roli Asthana, Monica Baumgarten, Shantayanan Devarajan, Markus Haacker, James Putzel and seminar participants at the IDB, LSE, World Bank and at the LACEA99 conference for their thoughtful comments and suggestions. 1. Introduction The wisdom of decentralizing government has become popular currency in our time. At the end of a century that witnessed the sustained growth of the central state in both the developed and developing worlds, reformers and idealists have turned to decentralization as an antidote to ills as varied as governmental corruption, autocracy and repression, and public-sector inefficiency. But the public discussion of decentralization is often confusing, assuming the character of sweeping, cross-disciplinary claims about the effects of administrative measures on the quality and efficiency of both government and social interaction. Competing proposals, expressed in a lexicon that spans economics, political science, sociology and public administration are often hard to compare either as policy instruments or in terms of the effects they are designed to produce. Unfortunately, much of the empirical literature on decentralization is similarly messy and inconclusive, simultaneously examining issues as diverse and ill-defined as access to resources, participation, administrative capacity, employment, growth, and local and national development strategies. Having cast such a wide net, such studies subsequently fail to ground their research theoretically, and their empirical approach often descends into description and anecdote from selected cases of decentralization in very different countries. The radical and well-documented experience of Bolivia offers us the opportunity of conducting a methodologically rigorous study of decentralization, where we focus on a few questions which are among the most contentious in the field but have not been answered adequately in the literature. Restricting our scope to decentralization in one country allows us to control for external shocks, political regime, institutional and cultural effects, and other exogenous factors in a more systematic way than cross-country studies can. Furthermore, the Bolivian reform coincided with a huge upsurge in the generation of local-level and national data. These data are of surprising scope and quality (especially compared to Bolivia's national-income cohort) and include not only the usual infornation on fiscal flows and investment sums, but also numnerous variables covering political, institutional, administrative and even procedural (good-government type) indicators for all of Bolivia's 31 1 municipalities. Our use of such variables constitutes an innovation of this paper. The central question that we seek to answer is does decentralization increase the sensitivity of public investment decisions to local needs. Secondary questions include: (i) Under what conditions do the various effects we posit (local knowledge, central government's technical and organizational advantages, political weight) dominate? and (ii) What are the welfare implications of different levels of public goods provision under a variety of assumptions? In addition, this paper seeks to make a case by example of how to approach such questions empirically. We argue that locally specific economic and political decisions by local government and local civil society are important, and even defining, I characteristics of decentralization which must be studied if the phenomenon is to be properly understood. Before continuing, it is important to discuss precisely what we mean by "decentralization," a word used in the policy literature to refer to everything from the administrative deconcentration of executive agencies in autocratic regimes to privatization in democracies. For the sake of focus, this paper will concentrate on decentralization under democratic regimes. We shall see that the presence and nature of democratic controls play a large role in our ability to theorize about decentralization. We define decentralization as follows: Decentralization is the devolution by central (i.e., national) government of specific functions, with all of the administrative, political and economic attributes that these entail, to local (i.e., municipal) governments which are independent of the center within a legally delimited geographic and functional domain. The two reasons for choosing this usage are both powerful and fortuitous. First, the clarity of the proposition greatly sirnplifies analysis, allowing it to focus on discrete, well- defined decentralizing measures and exogenous variables in order to gauge the empirical effects of each on policy outputs. Second, the case of Bolivia involves precisely this form of decentralization (see section 2.1 below), implemented uniquely and vigorously. The remainder of the paper is organized as follows. Section 2 discusses Bolivia's decentralization program, and then examine in detail the changes in national resource flows which it brought about. Section 3 reviews the literature and then develops a model to analyze the tradeoff between local government's knowledge of local needs v. central government's technical and organizational advantage in the provision of public services in districts with heterogeneous preferences. We use a simple model of decentralization defined by two equations to examine the welfare implications of central v. local goods provision under different assumptions. Section 4 discusses our empirical methodology and then presents three sets of econometric results: two tests of whether decentralization changed public investment patterns across Bolivia's 311 municipalities, and a set of sectoral models of this change centered on objective variables of need. Conclusions and suggestions for further research along this path are in section 5. 2. Decentralization in Bolivia 2.1 Popular Participation and the Decentralization Reform On the eve of the 1952 revolution, Bolivia was a poor, backward country with extreme levels of inequality, presided over by a "typical racist state in which the non- Spanish speaking indigenous peasantry was controlled by a small, Spanish speaking white 2 elite, [their power] based ultimately on violence more than consensus or any social pact."' The nationalist revolution which followed expropriated the "commanding heights" of the economy, and laid the foundations for the development of one of the most centralized state apparati in the region. The ruling Nationalist Revolutionary Movement embarked upon a state-led modernization strategy in which governing elites in La Paz directed a concerted drive to erase the social relations of the past and create a new, more egalitarian society.2 Political power was concentrated in the hands of the president, who directly appointed departmental governors and heads of the regional development corporations, among many others, and the legal and political instruments of local governance were by and large given little chance to develop. As a result, beyond the nine regional capitals (including La Paz) and an additional 25-30 cities, local government existed in Bolivia at best in name, as an honorary and ceremonial institution devoid of administrative capability and starved for funds. And in most of the country it did not exist at all (see point 4 below). This, very generally, is the background against which the Bolivian decentralization reform was announced in 1994. The genesis of the reform, along with the origins of the decentralization idea in Bolivia and the interest groups ranged for and against it, are treated in much detail in Faguet (2000b). The scale of the change in resource flows and political power that this law brought about make it a fascinating social experiment in decentralization, worthy of study. The core of the decentralization reform consists of four points:3 1. The share of all national tax revenues devolved from central government to the municipalities was raised from 10 percent to 20 percent. More importantly, whereas before these funds were apportioned according to ad hoc, highly political criteria, after decentralization they are allocated strictly on a per capita basis (see below). 2. Title to all local infrastructure related to health, education, culture, sports, local roads and irrigation was transferred to municipalities free of charge, along with the responsibility to administer, maintain and stock this with the necessary supplies, materials and equipment, as well as invest in new infrastructure. 3. Oversight Committees (Comites de Vigilancia) were established to oversee municipal spending of Popular Participation funds, and propose new projects. These are composed of representatives from local, grass-root groups within each municipality, and are legally distinct from municipal govermments. Their power lies in the ability to suspend all disbursements from the central government to their respective municipal governments if they judge that such funds are being misused or stolen, as well as the natural moral authority which they command. When suspension occurs, the center undertakes no arbitration, but simply waits for the two ' Klein, H., p.237. Author's translation. Klein is one of the classical authorities on Bolivian history. 2 Klein, H., pp.236-240. 3 Ley de Participaci6n Popular, Reglamento de las Organizaciones Territoriales de Base, Secretaria Nacional de Participaci6n Popular, Ministerio de Desarrollo Sostenible y Medio Amnbiente, 1994. 3 sides to resolve their dispute, relying on economic incentives to speed their agreement. Oversight Committees thus comprise a lean (their officials are unpaid), corporatist fonm of social representation which is parallel to elected municipal legislatures and serves somewhat like an upper house of parliament, as a check on the power of mayors and municipal councils.4 4. One-hundred ninety-eight new municipalities - 64 percent of the total - were created, and existing ones were expanded to include suburbs and surrounding rural areas, to the point where the 311 municipalities exhaustively comprise the entire national territory. The law heralded a new era of municipal government for the overwhelming majority of Bolivian towns and cities. In many parts of Bolivia where before the state was present, if at all, in the form of a local schoolhouse, health post and, perhaps, a military garrison or customs office, each reporting to its respective ministry, there was now for the first time elected local government accountable only to local voters. 2.2 Descriptive Statistics The extent of the change is perhaps best appreciated by examining the changes in resource flows that it catalyzed. Decentralization multiplied municipalities' share of public investment 17 times, from 0.7 to 12 percent of the total, and significantly altered its distribution. Consider figure 1, showing revenue-sharing between central and local governments for 1993, the last year prior to decentralization, and 1995, the first full year it was in effect, for the capital and second city of each of the country's nine departments. Total resources devolved from central to local governments increased by 72 percent. Though this is certainly significant, much more impressive is the change in the distribution of these funds. Before decentralization the nine departmental capitals shared 93 percent of all funds devolved from the center, leaving 7 percent for Bolivia's other 302 municipalities; the three leading cities, La Paz, Cochabamba and Santa Cruz, alone accounted for 86 percent of the total. After decentralization their shares fall to 38 percent and 27 percent respectively. The per capita criterion results in a massive shift of resources in favor of the smaller, poorer municipalities in Bolivia. Starting from a tiny or nonexistent base, these districts see enormous increases in their transfers, collectively exceeding 15,000 percent in Oruro, 43,000 percent in Chuquisaca, and 63,000 percent in distant Pando. The larger cities listed see more modest gains, 5 and only La Paz suffers a net reduction in transfers, itself a sign of how disproportionately it benefited under the old system. Within-department breakdowns similarly show movement from extreme skewing of resources in favor of the capitals to a more equitable distribution. 4I am indebted to Dr. Teddy Brett for this insight. This is possible only because of the large increase in total devolved funds. 4 Figure 1. Decentralization and the Regional Distribution of Public Funds Central-to-Local % of Departmental Revenue Sharing (Bs'000) Total City 1993 1995 % Change 1993 1995 La Paz 114,292 61,976 -46% 95% 34% El Alto 5,362 46,326 764% 4% 25% ROD 1,120 76,170 6704% 1% 41% total 120,774 184,472 53% Santa Cruz(*) 51,278 63,076 23% 95% 51% Montero 1,106 5,306 380% 2% 4% ROD 1,774 56,012 3058% 3% 45% total 54,157 124,394 130% Cochabamba(*) 25,856 38,442 49% 88% 34% Quillacoto 1,315 2,471 88% 4% 2% ROD 2,108 73,688 3396% 7% 64% total 29,279 114,601 291% Oruro 6,969 15,925 129% 99% 56% Challapata 29 1,090 3687% 0% 4% ROD 74 11,198 15022% 1% 40% total 7,072 28,213 299% Potosi 1,208 13,990 1058% 66% 24% Villazon 233 3,543 1420% 13% 6% ROD 394 39,813 10009% 21% 69% total 1,835 57,346 3026% Sucre 4,581 21,202 363% 94% 44% Camargo 244 2,214 809% 5% 5% ROD 56 24,374 43540% 1% 51% total 4,881 47,790 879% Tarija 3,219 10,063 213% 68% 35% Yacuiba 648 4,743 632% 14% 17% ROD 841 13,893 1552% 18% 48% total 4,708 28,699 510% Trinidad 480 4,892 920% 67% 22% Riberalta 87 6,599 7501% 12% 30% ROD 154 10,393 6645% 21% 47% total 721 21.884 2937% Cobija 99 502 408% 99% 57% ROD 1 379 63067% 1% 43% total 99 881 787% Total 223,525 608,280 172% ------ ------- Sources: Ministry of Finance, Ministry of Social Communication 1995 totals estimated due to incomplete reporting of budget data by both cities. ROD = Rest of Department S The most important change wrought by decentralization, however, is to the composition of investment. In our results in section 3 below, local government provides a level of public goods different from central government due to its more accurate detection of local preferences. Figure 2, which shows the investment priorities of central and local government before and after decentralization, provides initial evidence in support of these results. The front row corresponds to central government investment during 1991-93, and the rear row to local government investments during 1994-96. The differences are quite significant. In the years leading up to 1994 central government invested the largest sums in transport, followed by hydrocarbons, multisectoral (a hodgepodge of projects difficult to categorize), and energy. Together these four sectors account for 73 percent of total public investment during 1991-93. But after decentralization local governments invest most heavily in education, urban development, and water & sanitation, together accounting for 79 percent of municipal investment during this period. Of the sectors accounting for roughly three-quarters of total investment in both cases, central and local government do not have even one in common. Indeed, we have to descend to fourth place in the rear row to find a sector - transport - that ranks highly in the front row as well, and even so it's share of the total has fallen by five-sixths. Thus, we find evidence that local and central government have very different investment patterns. Figure 2: Local v. Central Government Investment 25% Sectoral il20% Investment 15% Total 10% l*Ia II%0 1 I0 Sector 6 Lastly, it is instructive to examine how investment was distributed geographically among Bolivia's municipalities before decentralization, and compare that to the current regime. Although detailed maps of project locations and types are not currently available, we can get a very rough sense of the distribution behind the sums by examining figures 3-5 below. These place all of Bolivia's municipalities in a row on the horizontal axis and measure investment per capita as vertical displacement. If the allocation of investment were extremely skewed in favor of a few municipalities, we would expect to see most values lying near the bottom of the graph and a few points strewn high above them. If the distribution of investment were reasonably equitable across space, we would expect to see most points in a broad band at some intermediate level. Figure 3, per capita investment before decentralization, seems to conform to the first pattern. It is certainly skewed, with investments in one district6 of over Bs.50,000 per head, and two more7 in the neighborhood of Bs.20,000 per head, while the vast majority seem to sit on or near zero. Compare this to the national average for this period of Bs. 1,400 per head and we see the extent of the imbalance. But the degree of skewing itself distorts the vertical axis and compresses the lower range, where most of the values are. We turn to Figure 4, which excludes the upper twelve observations and shows only those below Bs.2,000 per capita, in order to examine these more carefully. Though the distribution now appears less unequal, there is still monotonically increasing density as we move downwards, and a preponderance of observations on or near the horizontal axis - 146 in fact, or half of the 298 in the plot. Our initial impression is confirmed. Investment under centralized government was terrifically skewed in favor of a few municipalities that received enormous sums, a second group where investment was significant and the bottom half of districts that received nothing. Compare this with figure 5, which shows municipal investment after decentralization. This chart shows no district over Bs.700 per capita, a broad band with greatest density between Bs.100-200, and only a few points touching the axis. Average municipal investment for this period is Bs.208 per capita, and thus our band contains the mean. (The investment sums here are much lower because they exclude central government funds.) The overall distribution is thus much smoother and more equitable than figure 4. Although these are crude indicators, it would seem that central government, with a much larger budget and free rein over all of Bolivia's municipalities, chose an unequal distribution of investment across space, while decentralized government distributes public investment more evenly throughout the country. 6 Sabaya, Oruro, population 2,074. 7 Chimore, Cochabamba, site of major highway works, and Ascenci6n de Guarayos, Santa Cruz. 7 Figure 3: Investment Per Capita, 1991-93 60,000 50,000 40,000 (U 20,000 - * 10,000 0 0 50 100 150 200 250 300 Municipal Identity No. Figure 4: Investment Per Capita, 1991-93 2,000 1,500 4 L01,000 4 g~~~ _. C) 500 o 50 100 150 200 250 300 Municipal Identity No. Figure 5: Local Investment Per Capita, 1994-96 700- 600- j% 500 - 400 a300 ,**4* 4 0200* .* z* 100 0 50 100 150 200 250 300 Municipal Identity No. 8 3. Theory 3.1 The Literature Economists and political scientists have often disagreed on the question of the needs-responsiveness of central v. local government. This is largely due to the focus that each discipline gives to the problem. Economists such as Oates and Besley and Coate (see below) tend to assume a better match between local government outputs and local preferences, and accordingly find local government preferable when this advantage is not outweighed by spillovers or inefficiencies in central government provision of public services arising from distortions in their financing or production and allocation. Economists do not agree on how this better matching come about, however, with some ascribing it primarily to the character of the information involved, and others to local elections or institutions. Political scientists, on the other hand, (see for example Crook and Sverrisson (1999) and Smith (1985)) tend to concentrate more on interest group capture of the local political process, and the distortions of political representation in small electoral environments. When these phenomena exist, interest groups will gain a decisive influence over local government, and decentralization will tend to favor these small local groups disproportionately over everyone else. In this context, centralization can be preferable, as interest groups which are sufficiently big locally to distort the local political process will tend to be small in comparison to national government, which can then match policy to (general) local needs in a disinterested fashion. We incorporate specific forms of these insights into our model and then test them below. We first examine the empirical literature on decentralization, and then turn to theory. A large part of the empirical work on decentralized provision of public services reports mixed results which, taken together, are inconclusive. Much of this literature approaches the subject from a very broad perspective, examining such issues as fiscal flows, taxation, expenditure and investment alongside very different questions such as managerial efficiency, government responsiveness and political representativeness. The breadth of these studies' scope combined with their and small sample size make controlling empirically for all the exogenous economic, social and institutional factors involved in decentralization impossible. They also generally fail to specify a coherent theoretical framework which credibly links all of the phenomena in question to specific decentralization measures in very different national and cultural contexts. Attempting to summarize such work can be a frustrating task as its findings are both numerous and diverse, and isolating cause-and-effect relationships is difficult. Examples of the results in Andersson, Harsman and Quigley, (1997), Bennet (1993), Cheema and Rondinelli (1983), Rondinelli et al. (1984), Rondinelli (1981), and Veira (1967) include: 9 1. The performance of decentralized administrative units in Algeria, Libya and Tunisia has been positive in some cases, but has not always met the original goals of policy reformers. 2. Decentralization and privatization of state activities have a tendency to create greater inequities among communities and regions with different levels of organizational capacity, opening the door for local elites to play a disproportionate role in the planning and management of projects. 3. Devolution in Papua New Guinea increased popular participation in government, and has improved the planning, management and coordination capacity of provincial administrators, but has added to government bureaucracy and so weakened it's ability to attract foreign investment and stimulate long-term economic growth. 4. Decentralization has increased the access of people in previously neglected rural regions and local communities to central government resources, if only incrementally, in most of the developing countries where it has been tried. 5. The administrative and technical capacity of local organizations is said to be slowly improving, and new organizations have been established at the local level to plan and manage development. 6. National development strategy now increasingly takes account of regional and local level planning. 7. The absence of or weakness in supporting institutions needed to complement the managerial capacity of local governments, as well as weaknesses in the linkages and interaction between local and central administrations, have led to disappointing results from decentralization in Africa and Asia. Such studies tend to show that decentralization has achieved moderate success in some countries, moderate failure in others, and both in many, with the underlying reasons poorly identified. It is, as a result, difficult to judge whether specific decentralization "failures" were due to inappropriateness of the policies implemented or weaknesses in their implementation, and more difficult still to recommend improvements. The theoretical debate on the effects of decentralization on social welfare and efficiency is of higher quality. In terms of productive efficiency, central government should be naturally superior so long as returns are at least slightly increasing. Any economic case for decentralization must therefore invoke a counterbalancing source of efficiency in which local government has an advantage. Different authors have approached the problem in different ways. Tibet's (1956) seminal work, reviewed in Rubinfeld (1987), posits a world where individuals move costlessly among localities that offer different levels of provision of a public good, and finds that the competitive equilibrium in locational choices which results provides an efficient allocation of local public goods. Though the starting-point for many Io analyses of decentralization, this work ignores central-government provision of public goods, and is thus an inappropriate foundation for the present empirical study. More importantly, it assumes a highly mobile population and fixed governments, which, more than unrealistic, we consider exactly backwards. It seems self-evident that government is the relatively mobile element in most local democratic systems, changing every electoral period or two, whereas the population is essentially fixed over the 4-5 years that electoral periods typically comprise. By invoking infinitely transportable individuals as the mechanism which joins the supply of public goods to demand, Tiebout fundamentally misses the point. "Voting with one's feet" in this way is undoubtedly a valid mechanism for preference revelation at the margins, and may be more important for particular public goods, such as education. But the principal mechanism for joining demand and supply must involve the political process. Indeed this is arguably why local government exists at all. Oates (1972) examines heterogeneity in tastes and spillovers from public goods through a model in which local govemment can tailor public goods output to local tastes, whereas central government produces a common level of public goods for all localities. He finds that decentralization is preferred in systems with heterogeneous tastes and no spillovers; with spillovers and no heterogeneity, centralization is superior on efficiency grounds. But Oates' results rest largely on his assumption of uniform central provision of public goods which, though an empirical regularity, is theoretically ungrounded and problematic when viewed in the Bolivian context. Close scrutiny of the data (see section 2.2 above) shows that central government investment patterns were non-uniform during the period we examine. Investment flows were concentrated in a few municipalities to such an extent that public investment actually became uniform after decentralization. We thus require a theory, which does not restrict central government choice so strongly. Besley and Coate (1998) provide a model in which this restriction is lifted. Like Oates, they invoke uniform taxation to finance public goods provision. But they then devise a model of central policymaking in which elected representatives bargain over public goods provision in multiple districts. For heterogeneous districts, they find that decentralization continues to be welfare superior in the absence of spillovers, but centralization is no longer superior when spillovers are present. They also find that higher heterogeneity reduces the relative performance of centralization for any level of spillovers. This model is both more representative of how real central governments operate, and more in keeping with the facts of the Bolivian transition from centralized to decentralized provision. Our results below can be interpreted as an indirect test of their findings, given reasonable assumptions about representative local utility functions. Thus construed our results weakly support their findings. Bardhan and Mookherjee (1998) develop a model of public service provision which examines the implications of decentralization for the targeting and cost-effectiveness of public expenditure. They find that for provision of a merit good available on competitive 11 markets to the poor, decentralization dominates with respect to intercommunity targeting and cost-effectiveness, though not necessarily for intracommunity targeting. For the provision of infrastructure, decentralization dominates only if local governments are not vulnerable to capture, local government has adequate financing, interjurisdictional externalities do not exist, and local governments have all the bargaining power vs. public enterprise managers. Somewhat more tangentially, Persson, Roland and Tabellini (1997) examine how the separation of powers can lead to political accountability. They examine how voters can combine incentives produced by elections and the separation of powers to control moral hazard and reduce politicians' rents under a variety of constitutional regimes (presidential, parliamentary, etc). Under appropriate checks and balances, they find that separation of powers helps voters elicit information about both politicians and the state of nature. Though it examines a different question, this paper is highly relevant to our empirical work, as the separation of powers is central to the design of the Bolivian system of decentralization. 3.2 The Model A country is made up of T districts, each with population nj where the subscriptj denotes district. Individuals, subscripted i, have linear utility Ui = xi + 6b(gj) where xi is the amount of private good consumed by individual i, gj is the amount of public good available in districtj, and 0i is individual i's preference for public good gj. We use 0,,, to denote the local median preference for the public good in districtj. We define local welfare as median utility, Umj = Xmj + 03mjb(gj). The function of government is to provide public goods, which it finances with a local head tax. We allow central government to have a cost advantage in the provision of public goods, such that the head tax needed to finance a given level of provision under central government is ag/nj with OUi. 13 Athough the symmetric misestimation of local preferences is a desirable feature of the model on grounds of generality, it is not clear that it is relevant to the experience of Bolivia Section 2 shows that central government ignored one-half of Bolivia's municipalities in the period before decentralization, and qualitative evidence presented in Faguet (2000b) indicates that central underinvestment, not overinvestment, was the persistent complaint from the grass-roots level. For the sake of simplicity, we assume from this point on that 0-.m = 0 and analyze central government's assessment of local preferences via the pOrn term. The central government equilibrium is now defined by b '(gd) = od(np0,,d. Where pp, the cost advantage is dominated by the center's inaccuracy in measuring local preferences, and gcg,. Citizens prefer central government. These results are summarized in figure 7. Figure 7. Indifference condition: b (g)= b (g,)> a = I Assuming Condition Result Preference O- =O Ia>p gcgl Central For simplicity, the analysis above depicts the function of the public sector as the provision of a single public good g, and examines the effects of competing political and institutional factors on that provision. In reality, of course, local and central governments provide many public and private goods and services, and perform a large variety of functions which this approach is too simple to capture. Cost advantage and assessment inaccuracies are likely to affect these different activities in different ways. Section 4 examines this question empirically by comparing central and local investment patterns across ten different sectors for Bolivia before and after a radical decentralization reform. We investigate whether public investment patterns were different under local government than under central government, and if so what economic and social factors explain this difference. 4. Empirical Tests: Decentralization and Investment 4.1 Methodology Our objective is to test whether decentralization changed the pattern of public sector investment in Bolivia, and if so to find the determinants of that change. It is possible that public investment did not change with decentralization. In this case decentralization may be 15 desirable for political reasons of representation, for example, or undesirable for reasons of administrative effectiveness. But from an economic perspective decentralization and centralization would be largely equivalent. On the other hand, if decentralization did change investment patterns it becomes important to try to characterize this change in terms of welfare and distribution, and determiine which social and institutional factors were most important in defining it. Ideally we would measure public goods in quality-adjusted units of output, separated by type. But such information is unavailable for Bolivia, and instead we measure investment inputs in the form of resources expended on public investment projects. This approach has the advantage of using natural, noncontroversial units, and of facilitating comparisons across different sectors. We separate these flows into 13 distinct sectors, Education, Urban Development, Water & Sanitation, Transport, Health, Energy, Agriculture, Water Management, Communications, Industry & Tourism, Multisectoral, Hydrocarbons, and Mining & Metallurgy, and analyze the first ten. We drop Multisectoral because it includes a sufficient diversity of projects as to be functionally meaningless as a category, and thus difficult to interpret. We ignore Hydrocarbons and Mining because almost no municipalities invest in either, rendering comparisons across regimes impossible. For each of the remaining ten sectors we estimate the model, Gmt = I3Cam + 132CLm + 0338t + £mt (5) where am and 6t are vectors of state and year dummy variables as per above, and a*m is the product of am and a decentralization dummy variable which takes the values 0 before 1994 and 1 after (i.e., postdecentralization).11 We thus decompose investment patterns into three terms: a state effect, am, which captures all of the characteristics of a state fixed in time, a year effect, °t, which captures year shocks and time-specific characteristics, and a decentralization-interacted state effect, U*m, which captures state-specific characteristics commencing in 1994 which were previously absent. As decentralized public goods provision began in 1994, this term will capture the effects of local government, local civic associations and other local institutions that sprang up with the reform, and social and political dynamics more generally that impact upon local government but lay dormant under central rule. Our data cover the period 1987-96. We then perform three tests: 1. P1=,2 Means test. This is a simple t-test to determine whether the means of the am and a*m coefficients are significantly different for each sector. Significantly different coefficients indicate that decentralization caused a change in national investment patterns in a given sector through the effects and actions of local governments. Thus a*m takes the value 0 for all municipalities and all years before 1994, and is identical to ac for all years firom 1994 onwards. 16 2. PIrn = P2m Individual tests. This F-test checks municipality-by-municipality whether the decentralization-interacted state coefficients are different from the simple state coefficients for investment in a given sector. A significant F-test constitutes evidence that decentralization caused a change in local investment patterns in a particular municipality. Significance in many municipalities constitutes strong evidence (stronger than above) that decentralization changed national investment patterns. 3. Lastly, we place the differences in state dummy coefficients on the left-hand side (LHS), and estimate the model, P2rNI3ml= gSm + IlZm + Cm (6) for each of ten sectors, where S is a scalar or vector of the existing stock of public services (variously defined, as we will see below) at an initial period, and Z is a vector of institutional and civic variables, both indexed by municipality m. This approach allows us to isolate those changes in investment patterns resulting from a move to a decentralized regime, and then find its determinants. Notice that equation (6) is a general-form and not structural model, and hence our results will not be sensitive to specific theoretical assumptions. Our LHS variable should by construction be unrelated to all factors which remain constant between the two periods, and thus we omit socioeconomic, regional and other variables (used in Faguet 2000a) which do not vary between the centralized and decentralized regimes. We will employ these variables elsewhere to investigate the determinants of public sector investment under each regime separately, where a richer menu of explanatory variables is called for. We assume that the variables in Z, as well as the stock of public services in the ten sectors of interest to us, S, are constant over the period in question. For most of the demographic and socioeconomic variables in question, which tend to show change that is statistically significant only over longer periods of time, this is reasonable. It is less reasonable in the case of the S variable. Unfortunately the data leave us no choice. The huge number of variables that might enter Z permit literally hundreds of specifications of equation (6) above. To facilitate analysis, and in order to combine very specific variables into more meaningful and conceptually defensible indicators, we characterize these variables according to the following groups: 1. Civil Institutions 3. Training & Capacity-Building 2. Private Sector 4. Information Technology 5. Project Planning and construct principal component variables (PCVs) for each. Principal component analysis is a data reduction technique in which variables are added together linearly in order to find the unit-length combination which maximizes variance. This is explained in detail in Annex 1 below. Our interpretations of the PCVs is summarized in Figure 8. The PCVs and their 17 constituent variables, as well as variables of need, are summarized in figure A1.2. Equation (6) can thus be written as P2m7PI = CSm + IliZim + ... + 'l5Z5m + Cm, (7) where subscripts 1 to 5 denote the groups above. Figure 8: Interpretation of PCVs PCV Interpretation - Variable increases in... listed in order of PCV Group No. importance, where applicable (see Annex 1 for details) Civil Institutions I Strength of local civil institutions and organizations Training & Capacity-Building 1 Intensity of the local capacity-building efforts undertaken by/for local government Project Planning 1 Inforned project planning which follows consensual and open procedures In theoretical terms, the main coefficient of interest is 4, which we interpret as an indicator of the degree to which investment is based on need. We define "need" as the marginal utility arising from a particular type of public service, N = U'(g), where N is need and utility is defined as in the model in section 3.2. In the language of the model, we can let 0,,=U'(g). Hence need falls as the stock of g rises, and vice versa. We use two types of information as indicators of the stock of public services: (1) the penetration rates12 of public services or benefits in the local population, r, or the population without access to the same, 1 -r,13 and (2) the initial per-capita stock of infrastructure (at the outset of decentralization). Examples of these are: (1) the literacy and illiteracy rates, the share of population without water or sewerage; and (2) the number of sports facilities and markets per capita in 1994. Of these we consider type 1 variables to be truer indicators of need, as they better capture the criterion of public service use by the population, and are likely to be better measures of the flow of benefits produced by public investments. Type 2 variables indicate existence more than exploitation by the local population, and hence should be less accurate indicators of need. We use type 2 variables in our regressions when type 1 variables are unavailable. It is also important to note that need for us is a relative concept, rising and falling with U'(g). This is an important distinction, as the semantics of its common usage imply that need is an absolute, and even discrete, concept, existing in some places (at some times) but not in others. By contrast, "need" for us is a continuous function, present in different degrees in all places always. 12 Note that "rate" here denotes a stock and not flow concept. 3 We use both for education, and obtain the expected variation in sign in our results (see below). 18 Following the argument in section 3.2, we expect 4 to be negative and significant when Sm is measured by the penetration rate r, and positive and significant when Sm is measured by (1-r). In the analysis that follows we assume Sm is measured by r. A negative coefficient suggests that decentralized government invests more heavily in a type of public good where it is scarce, and hence presumably where it is more strongly preferred. Decentralization would thus lead to a more progressive investment pattern in terms of objective need than obtained under centralized government. A positive coefficient implies that decentralized government behaves regressively, accentuating the preexisting differences in public goods endowments. We interpret this as evidence that the relationship we posit in 3.2 is exactly backwards, and central government allocates public investment with more sensitivity to need than local government. A coefficient equal to zero suggests that local government does not take the existing stock of public goods into account at all in making its investment decisions, implying that our theory is misguided and local preferences should not appear in the model. The variables in Z are not included as mere controls, however. We are interested in their coefficients, i, insofar as they help explain the institutional, civic and procedural determinants of decentralized investment decisions, and so constitute indirect tests of our theoretical argument above. The arguments put forward by political scientists14 for local government's superior assessment of local preferences and needs include greater sensitivity to grass-roots demand, greater accessibility of local lobby groups to local government, and greater political accountability to the local populace. Some of the ways in which this can happen include the use of participative planning techniques, and the existence of private sector and civic organizations that are strong and dynamic. Remember that these factors were not relevant to central decisionmaking, which occurred in the center. Hence we interpret positive coefficients on these PCVs as weak evidence that local government assesses preferences more accurately than central government, implying that the value of p is less than 1 and the difference between real preferences and those perceived by the center (0(m-0-m) is high. 4.2 Results Figure 9 shows our results from the means test PI = 02. Mean values are significantly different at the 0.1 percent level for education, water & sanitation, agriculture, transport, urban development and communication, and at the 1 percent level for industry & tourism and water management. In health, values are significantly different at only the 13 percent level, and even worse for energy. The evidence is that decentralization changed national investment patterns in each of the first eight sectors. Examination of the P2 values indicates that the effect of local government on average investment under decentralization was to increase investment in education, urban development, water management and 14 See for example Wolman in Bennet (1990). 19 perhaps health, no change in energy, and decrease investrnent in agriculture, transport, communication, industry and tourism, and (puzzlingly given the increase in water management) water & sanitation. But figure 10 shows that the number of municipalities investing in these sectors increased for all except agriculture. This implies that the concentration of investment fell, as more municipalities invested in a large number of (often-smaller) projects in nine sectors. Figure 9. Test 1: Coefficients Equal? Test 1-f2 - 0 Test Test Sector Variable Mean Std Error t-statistic P Value Education 3 0.00128 0.00032 -22.798 0.0000 P 2 0.01685 0.00042 Water & Sanitation D 0.00374 0.00043 17.343 0.0000 P2 -0.01174 0.00049 Agriculture D 0.00867 0.00080 8.667 0.0000 P 2 -0.00535 0.00086 |Transport _ 0.05464 0.00890 5.967 0.0000 D 2 -0.05152 0.00890 Urban Development I 0.00307 0.00049 -5.324 0.0000 P 2 0.00791 0.00053 |Communication D p 0.00191 0.00032 4.011 0.0001| m 2 -0.00055 0.00031 Industry& Tourism P1 0.00101 0.00023 3.768 0.0002 P 2 -0.00071 0.00023 Water Management I1 0.00075 0.00018 -2.932 0.0034 2 0.00182 0.00020 Health 0, 000258 0.00038 1.540 0.1238 1520.00141 0.00041 20 Figure 10: Number of Municipalities Receiving Investment, by Sector (in municipality-years) % Sector Before After Change Urban Development 66 675 923% Education 75 685 813% Health 95 484 409% Water Management 46 175 280% Communications 38 97 155% Water & Sanitation 202 506 150% Energy 180 259 44% Industry & Tourism 44 60 36% Transport 357 444 24% Agriculture 343 309 -10% Figure 11 shows the number of municipalities where we can reject the hypothesis 13Im = B2m,that is, the number of municipalities where decentralization changed investment patterns significantly during the first three years. As we might expect, decentralization did not change investment equally in all sectors. The test is significant in about Y4 of municipalities for water & sanitation and education, and in '/ of municipalities for urban development and water management, but in only '/5 of municipalities for agriculture and health and fewer in other sectors. This test suggests that investment pattems changed significantly for water & sanitation, education, urban development and water management, did not change for industry & tourism, energy, communication and transport, with agriculture and health on the border between significantly different and not. It is notable that the only sector which fails both tests is energy. Taking into account our results from test 1, we conclude that agriculture spending did change significantly between the two periods, while for health it may have but the evidence is inconclusive. Thus we add two sectors to the two above for which decentralization did not significantly change investment patterns across Bolivia's 311 municipalities. From this point we focus our analysis on water & sanitation, education, urban development, water management, agriculture and (marginally) health. 21 Figure ll:Test 2: Coefficients Equal? Test PIm-P2m = 0 No. % Sector Significant Significant Water & Sanitation 224 76% Education 209 71% Urban Development 107 36% Water Management 105 36% Agriculture 65 22% nea .. :::::.::: 4::: 17N: We can best understand this result by considering the following: 1. One-half of all municipalities in Bolivia received no public investment at all during the three years before decentralization, and these are for the most part the poorest municipalities. As all municipalities have funds to invest postdecentralization, the most pronounced changes in investment patterns are accounted for by the poorest municipalities. 2. Given high levels of poverty and low levels of public investment before decentralization, poor municipalities have a need for investment in more than one sector. 3. Rather than spread resources around thinly, most reasonably choose to concentrate investment in a few, high-priority sectors during the initial years of decentralization. Hence our results are driven by investment by the poorest districts responding to their greatest needs. By revealed preference we can infer that local administrations in these areas prioritize basic social services projects above productive projects, and productive (i.e., income-enhancing) projects in turn above economic infrastructure. Hence they will tend to invest in education and water before agriculture, and agriculture before transport or communication. Because only a few years of post-decentralization data are available, we expect the F-test to fail in low-priority sectors, as poor municipalities received little or no investment under central government and continue to invest little under decentralization. In high-priority sectors, however, investment will leap upwards from a very low base if decentralization matters. This is indeed what happens. Decentralization leads to an increase in investment in water & sanitation and education in 3/4 of all municipalities, and urban 22 development and water management in 1/3. There are moderate changes in investment patterns in agriculture and health, and very little change in transport, communication, energy and industry & tourism. We conclude that decentralization did change the pattern of Bolivian public investment, and this difference was strongest in the social services and urban development. Test 3 investigates the determinants of the difference in dummy state variables, P2 - [I, equivalent to tfie increase in investment due to decentralization. We examine our results sector-by-sector, beginning with education. Education Figure 12: Test 3: P2m Im = IISm + fliZm + *-- + 5Z5m + Sm Model* Independent Variable I II III IV V Private Sector PCV1 -0.000983 -0.00121 -0.00106 -0.0003 -0.00056 (-2.466) (-3.004) (-2.689) (-1.004) (-1.619) Civil Institutions PCVl 0.000973 0.00101 0.00103 (1.752) (1.774) (1.839) Information Technology PCV I 0.00118 (1.010) Illiteracy Rate (Over-6's) 0.00018 (2.505) Local Education Authority 0.005603 0.00534 0.00543 0.0053 0.00479 (1.421) (1.356) (1.378) (1.354) (1.379) R-square 0.0176 0.0136 0.0162 0.0155 0.0172 lProb>F 0.001 0.0025 0.0016 0.0128 0.0104 * OLS regressions reported with robust standard errors t-stats in parentheses; PCV I1st principal component variable. 23 All of our models for education are jointly significant at the 2 percent level or higher. We see that investment nrses under decentralization where the illiteracy rate is higher, and investment is thus progressive in termns of need. This implies that local government is more sensitive to local need than central government. This finding is not sensitive to specification or to the measure of illiteracy used, as we see in Figure 12, where the literacy rate is significant and negative. In terns of the model of section 3.2, our results imply that pF 0.Q000 0.0743 0.0000 0.0000 *OLS regressions reported with robust standard errors t-stats in parentheses; PCVI =Ilst pricipal component variable. 24 All multivariable models for water & sanitation are jointly significant beyond the 0.1 percent level, and even the univariable model is significant at the 10 percent level. Investnent rises under decentralization where more people have no sewerage. It also rises where the percent of the population without access to drinking water increases, though this finding is sensitive to specification and drops out when other variables are included in the model. Thus local govermments invest more where need is greatest, and investment is progressive in terms of need. This implies that pF 0.0000 0.0000 0.0000 l * OLS regressions reported with robust standard errors t-stats in parentheses PCV1I = I1st principal component variable. ** Defined as other than football fields, multi-use courts and coliseums. 27 All of our models for urban development are significant at the 0.1 percent level. In this sector we use the initial (i.e., predecentralization) stock of infrastructure directly as our measure of need. Investment under decentralization increases as the initial number of markets per capita increases, and as the number of general sports facilities per capita increases as well. Investment is thus regressive in terms of need in this sector, as opposed to the others considered above, and this finding is not sensitive to specification. Thus it would seem to be central government that more accurately assesses local need in this sector, and local government that misestimates it. Investment increases with the number of private sector firms, which is as we would expect given that urban development projects often result in lucrative contacts for these firns. Investment is unaffected by participative planning techniques and civil institutions, implying that it is not a high priority at the grass-roots level. Lastly, neither training programs nor IT affect investment in urban projects. These results suggest that we should modify our model to includepi and pc, the probabilities that local preferences will be accurately assessed by local and central governments respectively. Indeed, there are sound political economy reasons related to interest group formnation and collective action for believing that local government will be more sensitive to demand in some sectors than in others. The important question would then become, which form of government is better at assessing local preferences, pi >< pc? On the other hand, of the six sectors we analyze this is the only one where our indicators of need are unsatisfying (type 2 variables in the characterization of section 4.1), and the only one where we find a broadly regressive pattern of investrnent in terms of need. It would thus seem wise to reestimate these equations with better indicators before concluding that the model is inadequate. We leave this for future work. 28 Health Figure 17: Test3: 2m-ml= Sm + 11ltZIm+***+Tl5Z5m+ 6m Model* Independent Variable I II III Private Sector PCV1I 0.000348 0.000234 0.000555 (0.527) (0.526) (0.992) Civil Instituti ns PCVI si oa00165 (-0.304) Information Technology PCge -0.001108 (-1.245) ih ealth Care, Other variables7 0.000400 0.000400 (1.529) (1.719) (1.628) Local Health Authorit n -0.000754 0.000393 0.000815 (-0.394) (0.169) (0.367) R-square 0.0187 0.0207 0.0202 lProb>F 0.8545 0.6514 0.5682 *OLS regressions reported with robust standard errors t-stats in parentheses; PCV I = I1st pTicipal component variable All of our health models are collectively insigniuficant, and hence we cannot make any clainis based on them. This is no great blow, however, as none of the variables in our main model are significant either. In the first alternative model "Other" healtb care is poiive and significant, and in the second alternative it is nearly so. This variable measures the perceDtage of households that have recourse to health care outside the fonrmal public and private health networks. We might interpret it as an indicator of pent-up demand for health services, and hence its positive coefficient as weak evidence that local governments invested in health care where demand was greatest, and p<1l. No other variables - the private sector, civil institutions, participative planning, IT or training schemes - appear to have any effect on investment, nor do local sectoral institutions in the form of Local Health Authorities. But 29 as we observed above, none of these models is collectively significant, the small trend we do find is sensitive to specification, and the difference between state variables examined above is marginally significant for this sector in both tests. We thus conclude that we can make no claims about investment in health. Results for Transport, Communication, Energy and Industry & Tourism, sectors for which differences in state variables are not statistically significant, appear for the sake of completeness in Annex 3. 4.3 Summary Our results show that decentralization significantly changed national public investment patterns. Investment changed unambiguously in education, water & sanitation, water management, agriculture and urban development after the 1994 reform, and there is some evidence that it may have changed in health, transport, communication and industry & tourism as well. Furthermore, these changes are strongly and positively related to real local needs. In education, water & sanitation, water management, and agriculture, postdecentralization investments are higher where illiteracy rates are higher, water and sewerage connection rates lower, and malnutrition a greater risk respectively. In a decentralized context dominated by the actions of some 250 small, poor municipalities that make up 80 percent of the Bolivian total, public investment is strongest in human capital and social services. And within these sectors investment is progressive in terms of need. Investment rose by number of municipalities in all of the sectors we examine except agriculture, and the effect of local government on average investment was positive in the social sectors and urban development, and negative in economic infrastructure and agriculture. We can combine our various results to distinguish between the cost advantage and needs-assessment effects that we posit in section 3. We interpret the average rise in investment (i.e., across all municipalities) in education, health, water management and urban development after decentralization as due entirely to the need-orientation of local government, and evidence that the center cannot produce these services at lower cost than the periphery. The fall in average investment in agriculture, by both volume and number of municipalities, combined with the significance of need, is evidence that the center was overinvesting in this sector, and that given the choice municipalities prefer to redirect resources elsewhere. The fall in average investment by value in water & sanitation, combined with an increase in the number of districts investing and the significance of need, implies that the central government concentrated investment in too few projects and districts; local government thus reallocates resources in a larger number of smaller projects where need is greatest. And lastly, the systematic fall in investment by value throughout Bolivia in transport, communication and industry & tourism, combined with modest increases in numbers of municipalities investing and the irrelevance of need, implies weakly 30 that the center may have had a cost advantage in these sectors, leading volumes to fall after decentralization. After needs, the next most important indicator is participative planning techniques. We expect such planning techniques to contribute to needs-based investment insofar as they help local governments to sense Om accurately. But where they enter significantly their sign is negative. This may be because such activities are expensive and divert resources and attention from implementing investment projects, irnplying that they may not be worthwhile. Alternatively, it could be due to the avoidance of projects that are not desired by the grass-roots, implying that these activities are valuable. The latter is only really a possibility if there is a general level of overinvestment in the Bolivian public sector, as otherwise we would expect good participative planning to increase investment in at least some sectors. We consider this possibility highly unlikely, and conclude that participative planning does not seem to improve local government's ability to sense local preferences. In econometric terms, the most interesting single feature of our results is that the only terms that are consistently significant across the five principal sectors we analyze are indicators of need. These relationships are robust and insensitive to specification. By contrast social, institutional and procedural variables are infrequently significant across sectors, and seem to account for little total variation. Indeed, the only effect we find for private sector firms is to transfer resources from education to urban development. Civil institutions are significant only for education, where they increase investment, and insignificant everywhere else. Training, capacity-building and IT are insignificant for all sectors. This implies that the differences in investment patterns chronicled above are not related to the number of private enterprises or civil institutions, or driven exogenously by training programs or informnation technology, but are instead determined by local needs. We conclude that decentralization led to an increase in investment in those municipalities least well endowed with infrastructure, and with the worst demographic indicators in the respective sectors which we examine. This is exactly the opposite of what many academics and policymakers predict, and what other researchers have found in the past. Given this finding, it is imnportant that we investigate the social and institutional mechanisms that cause these changes. We turn to these questions in Faguet (2000a) and (2000b). 5. Conclusions Our results confirm that decentralization did change local and national investment patterns in Bolivia, and that local preferences and needs are key to understanding these changes. Taken together, the pattern of centralized public investment and the structure of the decentralization program imply that these results are largely driven by the smallest, poorest municipalities investing newly devolved public funds in their highest-priority projects. We find that investment in education, water & sanitation, water management and 31 agriculture are progressive in terms of need, implying, in the language of our model, that central government's p<1. Even in agriculture, where total investment fell between the pre- and post-decentralization periods, our evidence indicates that the remaining investment was reallocated amongst districts according to need. The results also point to the existence of a poverty trap in the water management sector, where decentralized investment falls in the neediest districts as need increases. Within this range of the stock of public services, local government fails to respond to need and central government provision is superior. Our model can explain this indirectly, if in these neediest districts the costs and complexity of making initial investments in water are so great (from developing water sources, laying water mains and building treatment plants, for example) that local governments cannot undertake them alone, but once these initial investments are made the marginal costs of extending the system are manageable. In the language of the model, central government has a cost advantage over local government for initial investments, ax<1, an advantage that disappears at intermediate and higher levels of provision. By demonstration, this paper seeks to make a case for conducting empirical research on decentralization and fiscal federalism in the marmer in which we have done. Much of the empirical work on decentralization to date focuses on the share of national expenditures conducted by different levels of government, and ignores the many insights waiting to be uncovered by moving down to the level of the local political economy and conducting a careful comparison of spending and investment patterns with economic, institutional, social and demographic indicators. The data presented here is from one of the poorest countries in the Western hemisphere, and took years to collect, clean and organize. But as this paper demonstrates, its quality is sufficient to permit significant and (for many) counter-intuitive results. Applying a similar methodology to more sophisticated countries in the region, not to mention Europe and North America, might prove very fruitful. Lastly, the above analysis leaves open the question of how political power is distributed in a central government, the institutional mechanisms by which governments sense and take up local demand for public services, and the precise nature of the organizational or technical advantages or scale economies which might benefit one level of government over another. That is, p, e3m and a are all exogenous here. Research is needed to understand these processes and endogenize them in our models of public goods provision. Several authors have made progress in this direction but more work is needed. 32 Annex 1. Methodology, Including Principal Component Analysis and Interpretation N.B. This annex is general to all of the papers originatingfrom the study "Participatory Planning and Decentralization in Bolivia. " It describes the strategy used to arrive at the principal component variables used in this paper as well as Faguet (2000a). Hence some of the variables and categories referred to below do not appear in this paper but are exploited elsewhere. Methodology Our empirical strategy is iterative, and begins by finding the best idiosyncratic model of public investment for each of the ten sectors of interest. Hence we fit the equation Gm = 4Sm + rlZ + cm, (Al) separately for central public investment (1991-3) and local public investment (1994-7) where Gm is aggregate investment per capita in the public good subscripted by municipality, Sm is a scalar or vector of the existing stock of public goods of that type (variously defined) at an initial period, and Z is a vector of socio-economic, demographic, regional, political, institutional, administrative and procedural variables which might affect investment decisions. Our use of the Z term follows the literature on the demand for public goods exemplified by Bergstrom & Goodman (1973) and Rubinfeld, Shapiro and Roberts (1987) within the context of the available data. In particular, no income data is available at the municipal level in Bolivia, and so we substitute several alternative indicators of income and wealth, including for example type of cooking fuel, and housing size, quality and related characteristics. But we expand the scope of the Z indicators considerably from that of previous authors by including measures of the strength of local political forces as well as municipal institutional capacity. This innovation allows us to investigate the micropolitical basis of local government decision-making, which we explore in detail in Faguet (2000a). We allow no constraints across sectors on the particular variables admissible in Z. We use the Huber/White estimator of variance to produce consistent standard errors in the presence of non-identically distributed residuals. This produces ten different models of public sector investnent, one for each sector. Individually these models are quite satisfactory, with high R2 and few variables insignificant. But because of large variation in the specification of the Z vector, comparison across sectors is problematic. Additionally, on a theoretical level these models would seem to assert that public investment in different sectors happens according to different processes, in which different variables intervene. This is evidently unacceptable. 33 In our second iteration we return to equation (Al) and estimate it, holding the Z vector constant across all sectors. But we take advantage of the previous stage by using only those variables found significant there; in this sense the previous stage constitutes a method for reducing the 1200+ indicators to a subset of 197. But even so we still suffer from a dimensionality problem. We then employ a method of forward and backward substitution and elimination in order to reduce this subset to 22 variables encompassing the 13 categories of Z, in specifications of 23-30 variables overall (see Faguet 2000c, Annex 3). These models benefit from being readily comparable across sectors. The ratio of significant to insignificant variables drops sharply compared to the first stage, however, and R2 values are somewhat lower. The insignificance of the variables chosen is not entirely separable from the issue of comparability, however. It is evident from these results that none of the variables is significant in most of the sectors, and many are significant in only 2 or 3. How do we interpret a given variable across sectors, knowing as we do that an alternative one from the same group would produce a different pattern of significance and insignificance? The training & capacity-building variables in Faguet 2000c (Annex 3), for example, are insignificant in most of the models. What importance do we attach to this when we know from stage 1 that there is at least one alternative training variable which would be significant for each sector where the current ones are not? We evidently cannot assert for any sector that politics does not matter; we must conclude that the comparability constraint forces us to omit from our models information that is important in explaining investment behavior. Indeed, given that there are 197 variables, many of them quite specific, which have explanatory power over our dependent variable, any subset of 20, 30, or even 100 will omit valuable information. We require a solution which allows us to retain the full breadth of information, and yet produce a specification which is both parsimonious and comparable. We turn to principal component analysis, a data reduction technique in which the objective is to find the unit-length combinations of explanatory variables with the highest variance. We follow Maddala (1977) in calculating variables zi to zk where z is a linear combination of the x variables, z, = a,x, + a2x2 + ...+ aLxL z2 = bix, + b2x2 +... + bLxL etc.l5 ranked in order of variance, with highest first. Principal component analysis regresses y on zi, Z2, ..., Zk, where k < L and z's are constructed so as to be orthogonal. So long as the z's chosen represent combinations of variables that have economic meaning and can be interpreted, this affords us a method for estimating parsimonious models with limited loss of information. 15 For further treatment of this topic, see also Greene (1997), and Jackson (1991). 34 We calculate a set of principal component variables (PCVs) based on the raw variables retained in stage 1. We discard all those with low eigenvalues, as per normal procedure, and then find the subset of the remaining ones which optimally estimate equation (Al), where Z is a vector of PCVs. Figure Al.l contains the eigenvectors associated with each of the PCVs used in this paper. The factor loadings on the raw variables can be read vertically down each column. The numnbered column headings denote which PCV is referred to. Our interpretation of each PCV is explained below. Interpretation of PCVs Civil Institutions: This is an indicator of the number organizations and institutions of local civil society. It rises in all the variables, especially in the more general measures. We interpret it as a proxy for the strength of local civil institutions. Private Sector: This PCV rises in the number of private businesses registered locally. We construe it as an indicator of the dynamism of the local private sector. Training: This variable rises in categories of training (i.e., institutional strengthening) received by the municipality and falls in those requested but not yet received. Hence we interpret it as a measure of the intensity of capacity-building efforts undertaken by/for local government. Information Technology: This PCV rises in the IT systems - hardware and software (especially software) - at the disposal of each municipality. Project Planning: This PCV loads positively where municipalities use information on education and health when planning projects, where sectoral regulations are followed in water & sanitation, where a Municipal Development Plan exists, and where councilmen and oversight committees identify investment projects using the MDP and urban cadaster. It loads negatively where the mayor is the one who identifies investment projects, and where problems arise with the Annual Operating Plan. This is thus a straightforward indicator of informed project planning which follows consensual and open procedures. 35 Figure A1.1 CIVIL INSTITUTIONS TRAINING Eigenvectors Eigenvectors Variable 1 Variable 1 cv 0.09745 capadpe 0.28556 indig2 0.01988 capcil 0.30671 jvec2 0.29229 capci2 0.2612 otbregi 0.4194 capdis 0.2793 otbregi2 0.43286 caplemu 0.34451 otbs_e 0.42137 caporad 0.38803 otbs_pj 0.42934 capprin 0.37869 otbsoli 0.42372 capprop 0.34559 temacz -0.14204 temadis -0.20036 PROJECT PLANNING temaorad -0.22559 Eigenvectors temaprop -0.18667 Variable 1 catastur 0.04701 dpoacoor -0.00839 INFORMATION TECHNOLOGY dpoaotro -0.07581 Eigenvectors epoaham 0.00306 Variable 1 evalres 0.07426 sitotal 0.51744 idenalc -0.00973 siotro 0.36119 idencons 0.0145 sisin_ad 0.42748 idencv 0.09214 sisin_ai -0.27289 idenpdm 0.14818 sisinidp 0.28173 info_ed 0.53349 sicom 0.38812 info_sa 0.51649 impresor 0.3385 pdm94 0.14019 plan_sye 0.56911 reconu_a 0.24654 PRIVATE SECTOR Eigenvectors Variable 1 eereg_cm 0.61675 eereg_ea 0.56212 eereg_fi 0.55103 36 FguAreAI.2: Sunmuy of PEidinpal Conponent Variables, PCV Constituents, and Needs Variables Variabie I Obs 1Van St Dev. Mn Max Vaiiable Obs nan SkI. 1lv. Mn Max fPindpal Copionent Variables pctlr 310 -5.4000E409 1.6762 -28227 4.2889 pcpsl 302 -3.2400E-09 1.5298 -0.3015 18.0787 cv4*e 310 0.2516 0.4346 0 1 eeregcmn 306 202.7255 1229.8060 0 14117 cqxil 310 0.2 0.4006 0 1 eerg ea 306 0.5556 20973 0 30 cax2 310 0.5710 0.4957 0 1 eeregfi 310 2.6097 26.7243 0 454 oWdis 310 0.4871 0.5006 0 1 pcppl 310 2.36001309 1.5915 -2.7175 2.2313 cqiemu 310 0.3452 0.4762 0 1 catutr 310 0.1581 0.3654 0 1 a4txrad 310 0.3 0.4590 0 1 poxoor 310 0.8548 0.9991 0 4 cappin 310 0.3613 0.4812 0 1 doxxhro 310 0.6968 1.1790 0 4 c07pop 310 0.3903 0.4886 0 1 epTa-An 310 0.8355 0.3713 0 1 temxz 310 0.5194 0.5004 0 1 evalres 310 0.8226 0.3826 0 1 ten-ds 310 0.3161 0.4657 0 1 idenac 310 0.7968 0.4030 0 1 ternorad 310 0.5Q65 0.5008 0 1 idcons 310 0.4129 0.4932 0 1 teanpop 310 0.4290 0.4957 0 1 iwNcv 310 0.7323 0.4435 0 1 pritl 310 1.6400O-08 1.5235 -1.5591 5.0864 idenpdn 310 0.3742 0.4847 0 1 sitotal 310 0.4355 0.4966 0 1 nfo_ed 310 0.5581 0.4974 0 1 siotro 310 0.2226 0.4167 0 1 ifo sa 310 0.5839 0.4937 0 1 sisin ad 310 0.1548 0.3623 0 1 pdm94 310 0.3032 0.4604 0 1 sisin ai 310 06968 0.4604 0 1 plan sye 310 0.5839 0.4937 0 1 sisinidp 310 0.3258 0.4694 0 1 reccmu a 310 0.6839 0.4657 0 1 sicom 310 0.2806 0.4500 0 1 pecil 303 2.4000E09 2.2150 -2.1130 14.5313 inpresor 310 0.2903 0.8737 0 10 cv 310 0.6419 0.4802 0 1 Need Vaniables indig2 310 0.6290 3.5208 0 51 sa _insa 310 32.0264 20.0876 0 85.5147 jvec2 310 8.9548 26.2524 0 247 sa_oto 310 4.3985 7.4206 0 65.2706 otbregi 308 34.25 41.3093 0 299 desmod 294 8.2202 4.4993 0 26.2548 otbregl2 310 46.9226 49.6351 0 339 dilos 310 0.9161 0.2776 0 1 oths_e 307 50.2280 59.0375 0 520 analf 310 30.4638 15.8231 5.5 78.7 othspj 305 43.8557 52.5067 0 416 ed_ alfa 310 69.0462 15.9098 21.2128 94.5433 otbsoli 308 40 43.9176 0 323 edaia6 310 26.5292 13.1925 6.3780 69.7183 dile 310 0.5032 0.5008 0 1 sin alca 310 76.1424 21.8893 14.6586 100 sinL agu 310 74.3487 21.1723 17.9204 100 n=a4pc 304 0.0014 0.0108 0 0.1517 infot4pc 286 6.0100Er05 0.0006 0 0.0095 deslevh 294 23.0698 7.2684 0 57.1429 sin luz 310 76.0124 25.4209 5.9936 100 aguanr 310 67.6176 23.3971 10.4521 100 alca sin 310 76.2768 21.8418 14.6586 100 alca ot 310 16.1283 16.3147 0 64.1026 aguadv 310 &9680 10.3644 0 56.4501 agua fv 310 16.7037 13.7505 0 65.9341 agpaft 310 6.7107 7.1615 0 48.2235 tea4pe 304 2.8300E-05 8.34OOE-05 0 0.0007 37 Annex 2. The Demand for Water Management Services We can understand the results for water management better by noting that, in terms of demand, water & sanitation and water management are different from other sectors in that they are the aggregation of two public services - water and sewerage - which are technically and economically distinct. Indeed, water and sewerage are additive, sequenced services where water is generally the higher priority and one must have the first before having access to the second. Improvements to local water systems will tend to follow a progressive pattern as communities become wealthier and more is invested in this sector, resulting in a service escalation roughly as follows: 1. No water, no sanitation 2. Public standpipes, no sanitation 3. Public standpipes, open sewers 4. Private standpipes, open sewers 5. []... 6. Internal plumbing, municipal sewerage This pattern holds both across time, and cross-sectionally across a range of communities differing in wealth. In reality such an escalation occurs not in discrete steps but more-or-less continuously in terms of access rates of the population to each level of service. The point is that sequencing occurs, and not the precise path that the sequence takes. Our analysis of needs variables must take this into account, and accordingly we posit a demand curve for this sector which looks like figure A2 below. Figure A2: Marginal Demand for Water & Sanitation Investment in Water & Sanitation O Water & Sanitation S o i a w& > ~~~~~~~~~~~~~~~~Water Infr. O 0 loW Sanitation 4 W& Infra. 38 If demand for water & sanitation were linear in nature and progressive in terms of needs (i.e., existing infrastructure), it would look like the downward-sloping line in the upper half of the graph. Common sense and our data suggest that demand will not vary uniformly over different levels of service provision, however, nor equally over each type of service. It stands to reason that populations with low levels of water and sanitation will demand more new water infrastructure than communities where service levels are high. The lower curves in figure A2 are implied demand curves for water and sanitation respectively, assuming that investment in this sector goes first to water (public standpipes) and only then to sewerage (initially rudimentary), as water service is simultaneously improved. Hence the origin of the sanitation axis is shifted to the right of that of the water axis. The curve in boldface is the vertical sum of the two independent curves. The negative range at the left of the graph, where many people have no potable water at all, can be interpreted as a poverty trap. Here a lack of water & sanitation infrastructure leads to ignorance about its benefits, and hence to low demand. As service levels rise, however, people witness its value to themselves or their neighbors, and demand - and hence investment - increases. At high levels of service provision, where water infrastructure is abundant, investment falls again. Our results support this interpretation. Indeed the graph above is constructed from the results in figure 14, as well as additional regressions using variables for many levels of infrastructure stock. 39 Annex 3. Sectors Where State Variables Are Not Significantly Different (i.e., rlm = P2m). Industry & Tourism Figure A3.1 Test 3: r2m-3pi. = QSm + liZim - .** + fSZ5m + Em Model* Independent Variable Il c o il Private Sector PC a1 l i f .00021 0.00029 9saE-tv 0.66 1.046 0.42 municipaities 1Tnce we canot draw co0usosfo0hi0 eto.Ufo0naeywehv 0.418 infornalonTecholoy PV I 0(066 1.312 no good,Comrhnsivet iniaoofneoridsr 02 touis. Th0a2al e a s -1 .562 -l1.583 -L158 * te nregreso s r portea wmic roust stan dar errors o-stats in parenthesesi PCsVI = Imst pricipal component i ane iable Our models in this sector are all isnsignificant. We saw anove , indiating are investicantly different on average (i.e., test i ), but indivfdually different only for 7 municipalities. Hence we cannot diraw conc;lusions for this sector. Unfortunately we have no good, compreheiisive indicator of neced for industry & toun'isn. The vaniable we can use is the nurnber of theaters per capita - mimicipalities trying to develop their tourist potential often invest in theatcrs, spoils staditinis and similar projects in the hopes of attracting internal toun'sm (see Faguet 200()b). This variable is significant and positive, indicating that investment rose under decentralizationi where such infrastructure was in greatest abundance. No other variable is significant in our models, including surprisingly the number of private sector firms, which we would expect to benefit most from such investment. The evidence thus weakly suggests that post-decentralization investment in this sector was regressive in terms of need. 40 Transport & Communication Transport Figure A3.2 Test 3: S2m im = +Sm ± . + Tl5z5m + -ll Model* Independent Variable I II III Private Sector PCVlI -0.00083 0.00573 -0.00143 -G0.102 0.626 -0.25 5 0.437 Information Technology PCV1 0.023t)5 0.666 Prob>F 0.9281 0.9404 0.8689 OLS regressions reported with rob ust standard errors t-stats in parentheses; PCV 1 = 1 st pricipal componient variable Communication Figure A3.3 Test 3: 12Sm Im + 1Z + + m 5ZI + FmSSm Mod el* Independent Variable I I c III Privat iSecorPv a t -0.00044 -.00054 -000049 -0.746 - 1.1l32 -0.984 -0.474 Infomatln Tehnolgy PV 1 0.00013 0,109 _constant -0.00014 -0.00183 -0.00251i -0.034-0.417 -0.492 uiNregressions reportec with robust stnaders t-stats in parentheses; PCV I = I st pricipal component variable 41 All of our models in these sectors are insignificant. Additionally, though both sectors pass test 1, they fail test 2 (above) for a significant difference between state variables. Hence we cannot draw conclusions from our results. As it happens there are no conclusions to draw anyway, as none of the variables we use are significant. Our models suffer from the absence of a needs-related indicator for transport, and an unsatisfying one for communication. Energy Figure A3.4 Test 3: P32mi13imP = ;Sm + ii1ZIm + *+ ± 15Zm + ±m Model* Independent Variable I II III Private Sector PCVl -0.00464 -0.00341 -0.00468 -1.283 -1.794 -2.311 CIVi Institutons PCV1 0. OIY 0.422 nomton Tehnlgy PC1 10.00863 1 ~~~~~1.466 0.895 1.01 0.016 regressions reporte wit ro ust standr errors t-stats in parentheses; PCV1 = 1st pricipal component variable Investment in energy did not change significantly before and after decentralization, as we saw above, and hence this model is included here for the sake of completeness. Our variable of need, the share of population without electricity, is not significant in any of our models. Participative planning methodologies are significant, and negative, suggesting that the use of such processes led to a decrease in post-decentralization investment in energy. The private sector is significant and negative in two models and insignificant in the third, implying that this result is sensitive to specification. Civil institutions, IT and local training activities are insignificant in all the models. But in sum we make no claims about energy investment. 42 Bibliography Andersson, A., Harsman, B., Quigley, J., eds. Governmentfor the Future: Unification, Fragmentation and Regionalism. Contributions to Economic Analysis, vol. 238. Amsterdam; New York and Oxford: Elsevier Science, North-Holland, 1997. Atkinson, A.B. & J.E. Stiglitz,Lectures on Public Economics, London: McGraw-Hill, 1980 Bardhan, P. and D. Mookheijee, "Expenditure Decentralization and the Delivery of Public Services in Developing Countries," Working Paper No. C98-104, Center for International and Development Economics Research, Berkeley, 1998. Bennet, R.J., Decentralization: Local Governments and Markets, Clarendon Press, 1990. Bennet, R.J., Developments in Decentralization to Local Governments and Markets: Recent Experiences in OECD, Forner Communist and Developing Countries, UN UP 1993. Bergstrom, T. and R. Goodman, "Private Demand for Public Goods," American Economic Review, 1973. Besley, T. and A. Case, "Does Electoral Accountability Affect Economic Policy Choices? Evidence from Gubernatorial Term Limits," Quarterly Journal ofEconomics, 110, 1995. Besley, T. and A. Case, "Incumbent Behavior: Vote-Seeking, Tax-Setting, and Yardstick Competition," American Economic Review, 85, 1995. Besley, T. and S. Coate, "Centralized versus Decentralized Provision of Local Public Goods: A Political Economy Analysis," NBER Working Paper No.W7084, 1999. Besley, T., and S. Coate, "Public Provision of Private Goods and the Redistribution of Income," American Economic Review, 77, 1991. Besley, T., and S. Coate, "An Economic Model of Representative Democracy," 1995, manuscript. Besley, T. and S. Coate, "Efficient Policy Choice in a Representative Democracy: A Dynamic Analysis," Working Paper: 95/10, Center for Analytic Research in Economics and Social Science (CARESS), University of Pennsylvania, 1995. Besley, T., and S. Coate, "Legislative Elections With Citizen Candidates" manuscript. Cheema, G.S. and D. Rondinelli, Decentralization and Development: PolicyImplementation in Developing Countries, Beverly Hills; London: Sage Publications, 1983. Coase, R., 1960, "The Problem of Social Cost" Journal of Law and Economics, 3, Oct, 1-44. Crook, R.C. and A.S.Sverrisson, To What Extent Can Decentralized Forms of Government Enhance the Development of Pro-Poor Policies and Improve Poverty-Alleviation Outcomes?, typescript, 1999. Downs, A., An Economic Theory ofDemocracy, New York: Harper and Row, 1957. Dunkerley, J. Rebellion in the Veins: Political Struggle in Bolivia, 1952-82.London: Verso, 1984. Dunkerley, J. Political Transition and Economic Stabilization : Bolivia, 1982-1989 London: Institute of Latin American Studies, 1990. Faguet, J.P., "Institutions, Political Competition and Local Government Decision-Making: Decentralization and the Provision of Public Services in Bolivia" (Paper 2) manuscript, 2000a Faguet, J.P., 'The Micro-Political Foundations of Local Govenmment in NineBolivian .Municipalities: Organizations, Social Heterogeneity and the Control of Public Resources" (Paper 3), manuscript, 2000b. Faguet, J.P., Decentralizing the Provision of Public Services in Bolivia: Institutions, Political Competition and the Effectiveness of Local Governmen4 manuscript, 2000c. Greene, W. Econometric Analysis. Upper Saddle River: Prentice Hall, 1997. 43 Heckman, J. and J. Snyder, "Linear Probability Models of the Demand for Attributes With An Empirical Application to Estimating the Preferences of Legislators,"RAND Journal of Economics, Vol.28, No.0 (1997), pp.S 142-S 189. Jackson, J.E., A User's Guide to Principal Components, New York: John Wiley & Sons, 1991. Klein, H. Historia de Bolivia.. La Paz: Libreria-Editorial Juventud, 1993. Lea Plaza, M., interview, La Paz, 29 September, 1997. Mr Lea Plaza was then Director of Participative Planning in the National Secretariat of Popular Participation, Ministry of Human Development. Ley de Participaci6n Popular, Reglamento de las Organizaciones Territoriales de Base, Secretaria Nacional de Participaci6n Popular, Ministerio de Desarrollo Sostenible y Medio Ambiente, La Paz, 1994. Maddala, G. S., Econometrics. New York, London: McGraw-Hill, 1977. Manor, J., The Political Economy of Democratic Decentralization, Washington, D.C.: The World Bank, 1999. Mueller, D., Public Choice II, Cambridge: Cambridge University Press, 1989. Oates, W., Fiscal Federalism, Harcourt Brace: New York, 1972. Persson, T. and G. Tabellini, "Federal Fiscal Constitutions: Risk Sharing and Redistribution," Journal of Political Economy, 1996, vol. 104, no.5: 979-1009. Persson, T., G. Roland and G.Tabellini, "Separation of Powers and Political Accountability," Working Paper No.136, Center for Economic Studies, University of Munich, 1997. Rondinelli, D., "Government Decentralization in Comparative Perspective: Theory and Practice in Developing Countries," International Review ofAdministrative Science, 47(2), 1981. Rondinelli, D., G.S. Cheema, and J. Nellis, Decentralization in Developing Countries: A Review of Recent Experience, World Bank Staff Working Papers No.581, 1984. Rubinfeld, D., "The Economics of the Local Public Sector," chapter in Handbook of Public Economics, Auerbach, A., and M. Feldstein, eds, Oxford: North-Holland, 1987. Rubinfeld, D., P. Shapiro and J. Roberts, "Tiebout Bias and the Demand for Local Public Schooling," Review of Economics and Statistics, 69(3), August 1987, pages 426-37. Seabright, P., "Accountability and Decentralization in Government An Incomplete Contracts Model," 1995, mimeo. Smith, B.C., Decentralization: The Territorial Dimension of the State, London: George Allen & Unwin, 1985. Tiebout, C.M., "A Pure Theory of Local Expenditures," Journal of Political Economy, 64, 416-424. van Zyl, Johan, Analyzing Capacity of Local Government: A Case Study of Municipalities in North East Brazil, The World Bank, 1995. Varian, H., Microeconomic Analysis (Third Edition) New York: W.W.Norton & Co., Inc., 1992. Veira, 1967, in Rondinelli, D., G.S. Cheema, and J. Nellis,Decentralization in Developing Countries: A Review ofRecent Experience, World Bank Staff Working Papers No.581, 1984. Williamson, O., Institutions and Economic Organization: The Governance Perspective, Annual Bank Conference on Development Economics Paper, The World Bank, 1994. 44 Policy Research Working Paper Series Contact Title Author Date for paper WPS2493 Annuity Markets in Comparative Estelle James November 2000 A. Yaptenco Perspective: Do Consumers Get Dimitri Vittas 31823 Their Money's Worth? WPS2494 The Relevance of Index Funds for Ajay Shah November 2000 A. Yaptenco Pension Investment in Equities Kshama Fernandes 31823 WPS2495 The Australian Annuity Market David M. Knox November 2000 A. Yaptenco 31823 WPS2496 Perspectives on the Sources of Somik V. Lall November 2000 R. Yazigi Heterogeneity in Indian Industry G. Chris Rodrigo 37176 WPS2497 State Policies and Women's Monica Das Gupta November 2000 M. Das GLupta Autonomy in China, India, and the Sunhwa Lee 31983 Republic of Korea, 1950-2000: Patricia Uberoi Lessons from Contrasting Danning Wang Experiences Lihong Wang Xiaodan Zhang WPS2498 Sustaining Economic Welfare: Kirk Hamilton November 2000 L. Rivera Estimating Changes in Per Capita 82819 Wealth WPS2499 The Treatment of Non-Essential Isidro Soloaga December 2000 L. Tabada Inputs in a Cobb-Douglas Technology: 36896 An Application to Mexican Rural Household-Level Data WPS2500 Investigating Corruption Canice Prendergast December 2000 H. Sladovich 37698 WPS2501 Anti-Corruption Policies and Jeff Huther December 2000 A. Sanitos Programs: A Framework for Anwar Shah 31675 Evaluation WPS2502 Implications of the Currency Crisis Masahiro Kawai December 2000 A. Azarcon for Exchange Rate Arrangements Shigeru Akiyama 36049 in Emerging East Asia WPS2503 Proposed Strategy for a Regional Masahiro Kawai December 2000 A. Azarcon Exchange Rate Arrangement in Shinji Takagi 36049 Post-Crisis East Asia WPS2504 Social Capital and the Quality of Stephen Knack December 2000 P. Sintirm-,/oagye Government: Evidence from the 38525 United States Policy Research Working Paper Series Contact Title Author Date for paper WPS2505 Family Altruism and Incentives Roberta Gatti December 2000 R. Bonfield 31248 WPS2506 Ethnicity and Wage Determination Abigail Barr December 2000 R. Bonfield in Ghana Abena Oduro 31248 WPS2507 Public Expenditures and Gunnar S. Eskeland December 2000 H. Sladovich Environmental Protection: When Is 37698 the Cost of Funds Irrelevant? WPS2508 Sources of Financial Assistance for Mattias Lundberg December 2000 V. Soukhanov Households Suffering an Adult Mead Over 35271 Death in Kagera, Tanzania Phare Mujinja WPS2509 How Tax Policy and Incentives Jacques Morisset December 2000 N. Busjeet Affect Foreign Direct Investment: Neda Pirnia 33997 A Review WPS2510 Environmental Protection and Gunnar S. Eskeland December 2000 H. Sladovich Optimal Taxation 37698 WPS2511 Monetary Policy under Flexible Pierre-Richard Ag6nor December 2000 M. Gosiengfiao Exchange Rates: An Introduction to 33363 Inflation Targeting WPS2512 Quantifying the Impact of Technical Keith E. Maskus December2000 L. Tabada Barriers to Trade: A Framework for John S. Wilson 36896 Analysis Tsunehiro Otsuki WPS2513 Do State Holding Companies Facilitate Michel Kerf December 2000 M. Leon Private Participation in the Water Sector? 36151 Evidence from C6te dIlvoire, the Gambia, Guinea, and Senegal WPS2514 Intersectoral Dynamics and Economic Norbert M. Fiess January 2001 A. Pillay Growth in Ecuador Dorte Verner 88046 WPS2515 Firm-Level Survey Provides Data on Mary Hallward-Driemeier January 2001 E. Khine Asia's Corporate Crisis and Recovery 37471