VI. New Evidence

We are fortunate to have an extraordinarily suitable case-study of the transition from highly centralized public service provision to one that is highly decentralized, including large amounts of data concerning not only budgeting and real expenditures, but local preferences as well. This is the case of Bolivia, since 1953 one of the most highly centralized countries in Latin America, which underwent an extremely rapid, nearly overnight process of decentralization in 1994. Because the process was accomplished so quickly, and because the degrees of centralization/decentralization were so great, the Bolivian case comprises a sort of social experiment which we can use to test the ideas developed above.

The Popular Participation Law promulgated by the Bolivian government in April, 1994, and implemented as of July of the same year, brought about an enormous change in the balance of power between local and central government. The legal context in which decentralization took place was that of a unitary (as opposed to federal) state where municipal governments are independent, freely elected, and report to no one other than their constituencies. The core of the decentralization reform consists of four points: First, 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.^[10] Second, 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. Thirdly, Vigilance Committees 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 governments. 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 sides to resolve their dispute, relying on economic incentives to speed their agreement. Fourthly, municipalities were expanded to include suburbs and surrounding rural areas, to the point where the 311 municipalities exhaustively comprise the entire national territory.

In anticipation of decentralization, communities throughout Bolivia took part in a series of Participative Planning Exercises (PPEs) held at the provincial level (i.e. sub-departmental; Bolivia has nine departments), which led to the drawing up of Municipal Development Plans for some 150 municipalities so far. These seminars were convoked for three days each by facilitators from La Paz, included representatives from all sectors and strata of society, and were designed to discuss local problems and needs, suggest solutions, and eventually produce a list of projects for each municipality drawn up by consensus. While it is true that this methodology presents opportunities for the manipulation of opinion by the wealthy, educated, etc., there is no practical method for ascertaining information on needs other than asking the needy. Having observed one such complete exercise, I have every reason to believe that reasonable precautions were taken to ensure objectivity, and no reason to think that repeating the exercise would produce superior results.

In a year-long study of decentralization in Bolivia, I seek to combine the standard econometric analysis of a large, municipal-level database which I am building with more qualitative, deeper analysis of the social and institutional dynamics of municipal governments and municipal societies, which affect institutional performance but are not captured in the objective data. The large database includes substantial and very detailed data on the sources and uses of municipal funds, local preferences and priorities captured by the PPEs, a large amount of social, institutional and demographic data from the most recent census and integrated household surveys, and electoral data from the last two elections, among others. Most of this data has only become available during the past three years, as many of the types of information I am working with were not produced until quite recently, and financial and demographic data were not compiled at the municipal level before 1994. It is an additional stroke of luck that a national census was carried out in 1992, supported technically and financially by various aid donors. The main previous source of such demographic and social information were estimates based on the 1976 census, which by 1992 were woefully inaccurate. The qualitative research, by contrast, was carried out during 6 months of work in the field, most of which was spent in 9^[11] municipalities spread throughout the far reaches of Bolivia, selected to control for size, region, economic base, rural vs. urban setting, and cultural and ethnic makeup. This research involved extensive interviews with grass-roots representatives, elected and appointed government officials, and a variety of other business, labor, religious, and indigenous/ethnic leaders, as well as gathering planning, budgetary, and geographic data from local sources.

Does Decentralization Make A Difference?

The lessons which we can draw from this research are not yet final, as much work remains to be done. But a large portion of the data has been carefully analyzed and there are already a number of robust conclusions which we can discuss. To begin with, aggregate public spending patterns do change when resources and power are devolved to lower levels of government. And they do so in ways which this researcher at least finds positive and heartening, although we do not wish to make too much of this as we will examine spending in the light of objective needs below. A comparison of annexes 2 and 3 shows that the share of Education, Civil Works, and Water & Sanitation in total national investment rose dramatically after decentralization, whereas investment in Multisectoral, Industry & Tourism, and Communications fell. And if we disaggregate local from central government figures, we see that municipal spending in these first three categories was proportionally much higher than that by the center. This is further bolstered by examining average municipal investments (Annex 3),^[12] where we see that in both percent and per-capita terms it is the smaller municipalities which place a higher priority on Education and Health. A much coarser measure - the number of municipalities which spend (any amount) in each sector (Annex 4) - reveals a similar picture. We must conclude that decentralization is not a policy-neutral measure, and does indeed have significant implications for the structure of public spending and investment. As the theory above predicts, local priorities are different from those at the center, and shifting power and money into the hands of the community^[13] produces different outcomes. Whereas the national government prioritized investment in Hydrocarbons and Transport, local governments prefer to invest in human capital and social services. The latter, public goods with significant externalities, appear - to this researcher at least - to be preferable to items such as Hydrocarbons, or Industry & Tourism, sectors from which local governments should almost certainly stay away. But we do not wish to emphasize subjective assessments. We turn now to demographic data to investigate whether or not municipalities spend according to objective needs.

Allocative efficiency, a theme so central to any discussion of decentralization, can be investigated quite carefully for Bolivia by comparing public spending patterns before and after decentralization to (a) the demand for public investment, for which we can use the results of the PPEs, and (b) health and educational indicators, and other demographic data which indicates objective need for given services. Part (a) of the analysis is unfortunately not ready, as data on the results of the PPEs have only recently become available. Once we have these results, however, they will by their nature be an especially powerful indicator of whether or not local governments respond to demand. Part (b), on the other hand, is far advanced, and rewards scrutiny. The tables of Annex 5 list t-statistics^[14] from a series of regressions on different measures of educational budgeting (ex-ante) and investment (ex-post) from 1994-5, for all 311 of Bolivia’s municipalities. They tell a provocative story of how these decisions are made. We see from these results that positive educational attainment levels are significant and negative terms in the models, whereas no attainment (no schooling) and literacy are significant and positive. This holds for both budgeting and real expenditures, and across different measures of spending (percent of budget, per capita, and per student enrolled).^[15] School attendance rates and the number of classrooms per student are also negative and significant. These results indicate that municipalities effectively treat their education budgets as investments in human capital, with familiar characteristics of decreasing returns, and not only as simple service provision. While education is a widely favored investment, and those who are literate prefer it more than those who are not, the surprising result, controlling for this effect, is that education spending rises as educational attainment falls. As the stock of educational capital decreases, and thus the return to education rises, municipalities choose to invest more in education. As their never-schooled population rises they spend more as well. Education spending also rises with the Unsatisfied Basic Needs indicator (which rises with greater levels of deprivation), and falls with increased values of the Human Development Index (which falls as human development improves), supporting the human capital model of investment decisions.

The story for Water & Sanitation and Transport, though so far less detailed, is similar (Annexes 6 & 7), though with an interesting twist. Investment in Sanitation rises as fewer households have private sewerage service. This agrees with the above interpretation. As more households have public standpipe service, however, water investments increase, and the "no service" term for Water & Sanitation is not significant. This implies that in the water sector a logic similar to that in Education is at work, but only among those with some level of existing service. Hence investment is greatest where there exists a low level of service and decreases as service improves, but the poorest of the poor are ignored. By this interpretation, attention is skewed toward the "bourgeois poor". Another explanation is that those who do not have water service may not realize its value and not demand it, whereas those who have tasted the benefits of a low level of service want more, and in this case get it. The Transportation equations suffer from a lack of viable indicators of need. The notable term here is peasant communities, which is positive and (barely) significant. Thus, where a given municipality has more peasant communities,^[16] which tend to be dispersed across rural areas, transportation spending is higher.

These results point consistently to a highly rational model of local decision-making where the fundamental criterion is need, and governmental outputs are conceived of as the equivalent of productive capital, with the marginal investment going to those municipalities where the return is highest. This finding is both robust and surprising, especially when we consider that most of the municipalities which account for it suffer high degrees of deprivation, with poor, badly educated voters and governing officials generally unprepared for their posts. And yet the decisions they make are not only rational, but more rational - the evidence above suggests - and more in line with local needs than decisions made at the center. Powerful supporting evidence comes from comprehensive interviews from the municipal case studies. Out of over 200 interviews, a mere 4 responded being less satisfied with local government than with government from the center; another 41 opined that the two were roughly similar; and the remaining 167 reported significant or dramatic improvements in public sector responsiveness. It is true that these were subjective interviews, with some risk of insinuating responses or misinterpreting answers.^[17] But the magnitude of this trend cannot be ignored. We must conclude that allocative efficiency, public service provision, and indeed popular satisfaction with government have all improved with decentralization.

This directly contradicts claims that local government is too poor, too ignorant, or too prone to interest-group capture to operate efficiently, necessitating the guiding hand of national government which is technocratic, capable, and generally knows what to do. Here we begin to see evidence of the opposite: local government has a deep understanding of its task, and has the capability and the incentive structure to produce the public outputs that people want. It is worth reiterating that these results are not final, and there is much (very) new data that must still be incorporated into the analysis. But we also note that the results so far are quite robust, all point in the same direction, and are most unlikely to be overturned by new data.

Local Government Effectiveness

Having established that decentralization does work, we turn to the question of how, and to the ultimately similar question of where, it works. And we begin noting that regression analysis obscures the existence of enormous diversity in municipal performance and local responses to the challenges of decentralization. From huge, urban, complex La Paz strung off the edge of the altiplano to tiny Baures deep in the bowels of the Amazon, the 311 Bolivian municipalities vary immensely in their levels of human capital, institutional capacity, natural resources, size, populations, and many other factors that affect in differing ways the quality of government they achieve. Close inspection of their spending patterns, action plans, and results obtained thus far reveals a similarly broad range of local-government effectiveness and efficiencies. What variables explain this variation? Why does local government succeed in some localities and fail in others? Are the determining factors mainly economic, geographic, social or institutional? Are they susceptible to policy interventions, or are some municipalities simply destined to corruption and failure?

Additional econometric work will undoubtedly shed much light on this question. But we turn now to qualitative research from the case studies, which offers insights with a richness of depth and detail that econometric approaches cannot match. Annex 8 shows descriptive and selected administrative data on the municipalities chosen. We see here how these cases varied in size, region, language and culture, environmental and social diversity, and economic base;^[18] insofar as possible, we tried to control for local government success to date as well. The following results come from over 350 hours worth of semi-structured and unstructured interviews carried out at all levels of society, in several different languages, and under startlingly different conditions. The semi-structured interviews began with investment projects currently in execution, focusing on how they were planned and executed, and if they responded to real local needs. The interviews then broadened to include a range of questions concerning local government performance, interest-group capture and representation, and satisfaction with municipal outputs, trying always to establish why subjects responded as they did. Unstructured interviews were much more far-ranging, according to subjects’ interests and willingness, but were mainly concerned with social and political conflict and their effects on the quality of local government. The results from these interviews were characterized and systematized into a series of discrete variables ranging in value from 1 to 100. These were then classed as dependent and independent variables, and used to create the graphs in Annex 9, which explore the causes of local government effectiveness. The following is not the only, or even the main, way in which findings from the cases will be ultimately used; but it is a very efficient way in which to present a vast amount of information.

We see in the first two scatter plots that Mayor’s Effectiveness tracks Local Government Effectiveness^[19] (LGE) very well, whereas Municipal Council Effectiveness seems to have less of an effect. This resonates with the general impression that strong mayors have a large effect - for good or ill - on the municipalities over which they preside, and might be explained by the persistence of patriarchy and paternalism in large parts of Bolivian society. But as a theoretical explanation and policy prescription it is not very helpful: "To have good local government we need good mayors". We must look deeper.

The next two graphs reject prominent tenets of the conventional wisdom on decentralization. Defenders of centralism claim that poor, rural, distant locales lack the human, technical and economic resources for successful local government, and that ignorance and poverty mix with tradition to create a dangerous brew of domination and oppression. If this is true, we would expect LGE to fall as the index of Unsatisfied Basic Needs rises (where 1.0 is maximum deprivation) along the diagonal line. Likewise, this view holds that government will be more effective in large, urban populations than in small, rural ones. But our evidence indicates the opposite. Those populations where unsatisfied needs are greatest achieve the highest LGE values; if there is a systematic relationship, it is upward, not downward, sloping. And LGE shows no strong relationship with municipal urban-ness, although, again, if there is a relationship it is the opposite of what conventional wisdom suggests.

The last three graphs also contradict the centralist thesis, and offer reasons for why it is wrong. We see in the first that LGE decreases with employees per municipal population, whereas one might expect government to be better where it is more fully staffed. Likewise, as local tax revenues increase a municipality has more resources to work with, and so might fairly be expected to offer better services.^[20] But the opposite is true: LGE increases as the local tax base shrinks, rising quite quickly as taxes approach zero. This is a surprising conclusion - municipalities which have fewer employees per capita and less money provide better local government. More careful examination of the data reveals that it is the smaller, poorer municipalities farther removed from urban centers which produce better government and higher levels of citizen satisfaction, whereas cities score low on distributional equity and popular satisfaction with outcomes. Data on aggregate spending totals discussed above confirm this result: average municipal spending patterns, which favor the many smaller, rural municipalities in Bolivia, are skewed much more towards primary needs and the social sectors than national totals, in which large cities are overrepresented. The last graph in Annex 9 offers the beginning of an explanation: it shows a decreasing quadratic relationship between Mayor’s Effectiveness and total local taxes. Thus, as local taxes fall to zero, and especially in the region near zero, mayoral ability increases dramatically. This suggests - as per above - that the best mayors are in the smallest, poorest locales, and the worst are ensconced in the cities.

To understand this result, we must consider that the Bolivian decentralization reform was launched with much fanfare and a publicity assault which attempted to communicate to everyone the simple per-capita formula of revenue sharing, and urged communities to demand of local government their rightful share. My research indicates that this campaign was largely successful, and throughout the country the pressure on mayors to deliver is significant. Own resources and the local tax base, on the other hand, are mysterious topics in municipalities that have them. Interviews in such municipalities consistently showed that only the Mayor and his Financial Officer had information on local revenues. Vigilance Committee members, grass-roots leaders, business executives, and even Municipal Council members were typically ignorant of how much was raised, let alone who paid it and - most importantly - how it was used.

This leads us to propose the following, incentives-based theory of local leadership, which is consistent with the political economy model of local government discussed above. In a context of many municipalities of varying size and resources, and many political entrepreneurs who are mobile, corrupt politicians will seek office in large, complex, relatively resource-rich localities, and not in small, rural municipalities with no tax base. Thus, government of the latter will be left to relatively virtuous politicians with a public service ethic, or those who are cajoled into office. This is because: (a) small municipalities with no tax base are financed entirely by revenue-sharing, which is effectively watched over by the entire community as a result of the initial public relations campaign; (b) local, spontaneous forms of supervision and control are much stronger in small, rural places where anonymity is rare and word-of-mouth suffices to disseminate information throughout the community;^[21] and (c) the social fabric is stronger, and the costs to social organization and mobilization lower, thus facilitating corrective action at the grass-roots when politicians go astray.

In cities, by contrast: (a) local resources are treated as a municipal secret, providing politicians with numerous opportunities for corruption; (b) spontaneous forms of social control are much weaker due to the high degrees of anonymity and complexity which characterize social relations in urban areas; (c) the fabric of city society is much weaker, as multilayered forms of organization based on activity or purpose, and not simple geographic location, prevail; and (d) urban anonymity and complexity significantly raise the costs of grass-roots mobilization against bad politicians. The burden of supervision and review thus falls upon the legal system, which in many countries like Bolivia is too weak to serve as an effective counterweight to abuses of municipal power. Thus, political entrepreneurs looking for opportunities to capture rents will naturally drift toward cities, and will actively avoid rural municipalities, which by default will be left to politicians who may be ignorant or hapless, but at least are honest.