In each case we have identified what it is that we should be looking at for each kind of efficiency. We have also built upon the work begun in Discussion Paper #3 on fiscal decentralization. Each of the kinds of efficiency is very complex and a full investigation of them would be almost never-ending. Our objective was to come up with several important observations that could quickly reveal progress (or lack thereof) on the three kinds of efficiency, without imposing impractical amounts of research time. Furthermore, we recognize that most empirical questions cannot be answered directly due to lack of empirical data. We realize that much of the information that we find in most developing countries is anecdotal. Therefore, we use economic theory to organize both the data and the anecdotal information, to try to build a case based on "the preponderance of the evidence." We make a compromise here between academic refinement and practical application.

As was pointed out in Discussion Paper #3, judgement about the effects of FD are unavoidable. No method exists to apply purely quantitative methods to evaluations of FD. However, some quantitative data (particularly on production efficiency) combined with observations on key non-quantitative factors can support the inevitable judgements. One of our objectives here is to make those judgements easier.

A short case study of the Philippines is presented to illustrate the application of the case study methods suggested here. The Philippines case is particularly useful for illustrative purposes. The passage in 1991 of a new Local Government Code (referred to here as "the Code") represents a major shift in local governance in the Philippines. It consolidated and amended the Local Government Code of 1983, the Local Tax Code (Presidential Decree 231) and the Real Property Tax Code (Presidential Decree 464). The Code includes far-reaching provisions affecting the assignment of functions across different levels of government, revenue sharing between the central and the local governments, the resource generation/utilization authorities of LGUs and the participation of civil society in various aspects of local governance. These provisions are aimed at providing the framework in support of increased local autonomy.

Because the Code has been fully in force since 1993, the Philippines case represents a fairly neat "before and after" picture. Furthermore, within the Philippines there are a number of people in government, academia, NGOs and the research community who have been involved in fiscal decentralization for many years. They represent an important resource for studies of this kind. Finally, in the Philippines there are more data available than one finds in most countries. This is not to say that the data are complete. They are not. But the data do allow basic comparisons to be made and some empirical statements tested.

The Philippines has a unitary form of government with a multi-tiered structure. It is a presidential republic with a bi-cameral legislature (composed of the Senate with 24 members and the House of Representatives with 240 members). At the top is the central government operating through some 20 departments/agencies. Administratively, the country is divided into 15 administrative regions and most departments maintain regional offices. In addition, there is one autonomous region, the Autonomous Region of Muslim Mindanao (ARMM). It should be emphasized, however, that the regions (with the exception of the ARMM) are just administrative sub-divisions and not regional governments with elected regional officials.

The second tier of government is composed of local government units (LGUs). The local government structure is composed of three layers. The first layer is composed of the province. In turn, the province is divided into municipalities and component cities, each of which is further subdivided into barangays, the smallest political unit. At the same time, highly urbanized cities exist at the same level as the provinces, i.e., they share the same functions and authorities. However, highly urbanized cities are partitioned directly into barangays (Figure 1).

At present, there are 77 provinces, 72 cities, 1,548 municipalities and some 42,000 barangays. Each level of LGU is headed by an elected chief executive (governor, mayor, barangay captain) and has a legislative body, or Sanggunian (composed of an elected vice-governor/mayor and council members). All elected officials have three-year terms of office and face term limits, normally three terms. To a large extent, each level of local government is autonomous although the higher level of government (e.g. provinces) exercise some degree of supervision over lower level governments (e.g., municipalities and component cities) in terms of budgeting and legislation.

The Code devolves to LGUs many functions previously discharged by central government. Prior to the implementation of the Code, the functions assigned to LGUs were limited to; the levy and collection of local taxes; the issuance and enforcement of regulations governing the operation of business activities in their jurisdictions; and the administration of certain services and facilities like garbage collection, public cemeteries, public markets and slaughterhouses. The central government carried the primary responsibility for agricultural planning and extension, construction and maintenance of local roads and public buildings and operation of high schools, hospitals/health services. In contrast, the Code transfers from national government agencies to LGUs the principal responsibility for the delivery of basic services and the operation of facilities in the following areas: agricultural extension and research, social forestry, environmental management and pollution control, primary health care, hospital care, social welfare services, repair and maintenance of infrastructure facilities, water supply and communal irrigation and land use planning (Table 1). The devolution is substantial not only in terms of the sheer number of functions that were shifted but more so in terms of number of personnel transferred. The Code specified that significant numbers of the personnel of national agencies would be transferred to and become employees of LGUs. Table 2 shows the numbers of devolved personnel and the corresponding reductions implied in the budgets of affected national government agencies appear in Table 3.

The Code also prescribes an LGU share in internal revenue taxes referred to as an internal revenue allotment or IRA. Under the Code, the aggregate internal revenue allotment is set at 40 percent of internal revenue tax collections three years prior to the current year. By comparison, the pre-Code regime also prescribed an IRA that was at most equal to 20 percent of internal revenue taxes and which also lagged three years. In addition, the Code provides that LGUs receive a 40 percent share in the proceeds from the utilization and development of national resources within their jurisdiction. The Code sets aside 40 percent of internal revenues, lagged three years, for transfer to LGUs. Twenty-three percent of the IRA is allocated to provinces, 23 percent to cities, 34 percent to municipalities, and 20 percent to barangays. The Code provides that the share of each particular LGU will be determined by a formula that takes into account population, land area and equal sharing. More specifically, 50 percent of the aggregate IRA share of each level of LGU is distributed according to each LGU’s share in total population, while 25 percent of the aggregate LGU share is allocated on the basis of land area, and the remaining 25 percent is allotted on an equal sharing basis.

Compared to pre-Code arrangements, the Code, increased transfers to LGUs, with barangays receiving the biggest proportional rate of increase, followed by cities and municipalities. In contrast, provinces registered the smallest incremental gain in the IRA. Hand-in-hand with the expansion in the IRA, additional expenditure responsibilities were also shifted to LGUs under the devolution program mandated under the Code. Provinces absorbed 45.6 percent of the total cost of devolved functions, municipalities 47.4 percent, cities 7.0 percent and barangays 0 percent. Taken together, the IRA is sufficient to cover the cost of devolved functions in the aggregate, but barangays and cities as a group are relative net gainers from the fiscal decentralization that was effected by the Code. Provinces and municipalities as a group are net losers with the latter being more adversely affected.

The Code allows LGUs greater autonomy in mobilizing resources from local sources and in allocating said resources to meeting local needs. First, the Code expanded the tax base of LGUs to include products, activities and sectors that used to be outside the reach of local taxation. Second, it increased the maximum allowable rates at which most local taxes may be levied. Third, it markedly liberalized the scope for LGU credit financing. Central government approval prior to the issuance of LGU debt is no longer required even as greater flexibility as to source and type of credit is allowed. Fourth, the Code repealed some of the statutory requirements that limited the latitude of LGUs in allocating their budgets. For instance, mandatory contributions to the Philippine National Police (which used to be set at 18 percent of LGUs’ regular income in the general fund) and to hospitals operated by the Department of Health (which used to be equal to 3-5 percent of their regular income) were abolished.

Finally, the Code provides that LGUs shall promote the establishment and operation of people’s and non-governmental organizations (PVOs and NGOs) as active partners in the pursuit of local autonomy. In particular, the Code mandates the participation of PVOs and NGOs in local bodies like the local development council, local school board, local health board and local pre-qualification, bids and awards committee. This legal requirement that NGOs participate in local government is a unique feature of Philippine fiscal decentralization.

Four alternative indicators of fiscal decentralization are used in this section. First, the revenue decentralization ratio (RDR) measures the relative importance of local revenues in total general government revenue. It is defined as the share of LGUs in total general government revenue. Second, the expenditure decentralization ratio (EDR) measures the importance of local expenditures in general government expenditures. It is defined as the share of LGUs in aggregate general government expenditures. Third, the modified expenditure decentralization ratio (MEDR) takes into account the fact that some government activities like those on debt service are essentially the responsibility of the central government. The MEDR, thus, nets out debt service from total general government expenditure in calculating the expenditure decentralization ratio. Fourth, the financial autonomy ratio (FAR) provides an indication of local government independence from central government funding. It is computed as the ratio of locally raised revenues to total local expenditures.

Public sector finance in recent Philippine history remains largely concentrated at the center. Contrary to initial expectations, the increase in the RDR following the implementation of the Code was slight with the RDR rising from 4.9 percent in 1985-1991 to only 5.8 percent in 1992-1997 (Table 4). However, the degree of fiscal decentralization showed increasing intensity in the post-Code period when measured by the EDR and MEDR. The EDR doubled from 7.0 percent in 1985-1991 to 14.7 in 1992-1997. A similar picture was observed for the MEDR but the MEDR is 4-6 percentage points higher than the EDR in any case. Moreover, fiscal decentralization appeared to have deepened in the later years of Code implementation (1995-1997) relative to the early years (1992-1994) when measured by the RDR, EDR and MEDR. All three indicators registered substantial increments between the two post-Code sub-periods.

In contrast, the degree of fiscal decentralization appeared to have declined with the implementation of the Code when measured by the FAR. Thus, the FAR dropped from 51.6 percent in the pre-Code period to 36.7 percent in the post-Code period. This is to be expected given the step-wise increase in the IRA. Also, the FAR was observed to have been on the wane as one moves from the early post-Code period to the late post-Code period. It is apparent that local revenue generation is lagging revenue generated through the IRA. The FAR declined from 38.8 percent (1992-94) to 34.5 percent (1995-97).

Some variation in the FAR across different levels of LGUs is evident. Table 5 shows that city local government units ("cities") enjoy the greatest degree of fiscal autonomy. In 1985-1991, their FAR was highest at 66.4 percent compared to the municipalities’ 48.3 percent and the provinces 34.3 percent. With the implementation of the Code, the FAR of all levels of government dropped as the IRA increased. Cities continued to post higher FARs than municipalities and provinces. In 1992-1997, the FAR of cities was 49.8 percent compared to municipalities’ 33.4 percent and provinces’ 21.5 percent. More notable is that the FAR dropped for both municipalities and provinces over the 1992-97 period, but it did not drop for cities. This suggests that cities have a local tax base that is not available to provinces and municipalities. Indeed, the Code allocates a business tax to cities (and municipalities) but not to provinces.

The FAR as well as the RDR focus on the relative significance of locally sourced revenues. As such, these indicators may be misleading indicators of local autonomy in cases (like the Philippines) where taxes assigned to LGUs are rather limited and where transfers from the central government to local governments are mandated by law to be substantial. It should be noted that while Philippine LGUs have relatively less scope in raising own-source revenues, they do exercise considerable autonomy in deciding how to spend their total resources, including their share in internal revenue taxes collected by the central government. However, LGU-to-LGU variations in the RDR and FAR do measure the extent to which some LGUs are more (or less) financially autonomous relative to others. That is, LGUs that impose higher local taxes and/or collect local taxes more efficiently will score higher on the RDR and FAR than others. The EDR and the MEDR may provide a better picture of the degree of fiscal decentralization over time in the case of the Philippines. It captures well the shift in the expenditure responsibilities that devolution brought about. However, the RDR and the FAR are superior in focusing attention on how well local governments have performed relative to each other in utilizing their revenue raising powers to finance local needs.

The general picture that emerges is one where LGUs have relatively greater expenditure responsibilities, yet those expenditures are covered mainly by transfers from the central government. Locally raised revenue pays for a very small part of total expenditures, and LGU dependency on the central government as a source of revenue has increased substantially. Cities are more financially independent than municipalities and provinces, but then this would be expected where financial independence is based on a property tax and a business turnover tax, the two main taxes assigned to LGUs. Simply put, cities have more valuable property to tax and more business turnover than other governmental units.

An expanding IRA was the main feature of the Code’s implementation in its first few years. Total receipts of all LGUs in the aggregate were equal to 1.7 percent of GNP in 1985-1991 (Figure 2). This amount was divided almost equally between locally sourced revenue and externally sourced revenue. Externally sourced revenue is almost exclusively from the IRA. In 1992-1997, the share of income from external sources registered a marked increase as the Code raised the IRA for all LGUs in a stepwise fashion. External revenues expanded from 52.0 percent in the earlier period to 64.7 percent after 1992 (Figure 3). LGU income from external sources surged from 0.9 percent to 2.2 percent of GNP, while LGU local source revenues inched upwards from 0.8 percent to 1.2 percent of GNP (Figure 2).

Code-mandated phased-in increments in the IRA peaked in 1994 (Figure 2 and Figure 3). In contrast, the ratio of locally sourced revenue to GNP (local revenue effort) increased during the post-Code period. However, the increasing local revenue effort in the post-Code period appeared to be confined to Cities alone since the local revenue effort of municipalities and provinces remained fairly stable during the same period (Figure 2).

There is a significant variation in the importance of externally sourced income in the total receipts of different levels of local governments. Provinces are the most dependent on non-local sources of income. In the post-Code period, externally sourced revenue averaged 79.5 percent for provinces, 68.7 percent for municipalities and 50.9 percent for cities. In all cases, the share of externally sourced revenue expanded in 1992-1997.

The bulk of locally sourced revenues of LGUs come from tax revenues rather than user fees. The share of tax revenues to total locally sourced revenues expanded from 66 percent in the pre-Code period to 74 percent in the post-Code period. Figure 4 presents the breakdown of locally generated revenues of LGUs. It shows that the share of taxes to aggregate locally generated revenues was consistently largest in cities and smallest in provinces in both the pre-Code and post-Code regimes, indicative perhaps of the limited taxing authorities of the latter.

Figure 5 differentiates between different types of local taxes. It suggests that while the real property tax (RPT) continues to be the single biggest source of revenue for LGUs, its contribution to total LGU tax revenue declined from 61 percent in 1985-1991 to 51 percent in 1992-1997. Provinces relied the most heavily on the RPT in contrast to cities, which were the least dependent on this tax source.

It was expected at the outset that local taxes would assume ever-increasing importance for LGU finance. This has not been the case. Table 9 presents buoyancy coefficients of the major sources of locally generated LGU revenue. Prior to the Code (1983-91), local source revenue of all LGUs in the aggregate was mildly inelastic, with a buoyancy coefficient of 0.97. This was mainly due to the inelasticity of the dominant revenue source: the real property tax. The buoyancy of the RPT with respect to GNP for all LGUs was 0.88 during this period, indicating that real property tax collections did not keep pace with GNP growth. Following the implementation of the Code, the overall buoyancy coefficient of aggregate local source revenue of all LGUs registered a dramatic improvement, with the coefficient rising to 2.33 in 1992-1997. The improvement in the overall buoyancy coefficient of non-RPT taxes was particularly hefty. Significant gains in the revenue performance of the RPT were also made during the period. Most of the gains were made in cities. Provinces and municipalities made some progress but this was much more modest than it is in cities.

The progress attained with respect to non-RPT taxes was primarily due to improvements in the rate elasticity (i.e., rate increases that were brought about by the adoption of the higher rates allowable under the Code). On the other hand, the advances made in the RPT arena were mostly attributable to improvements in the base buoyancy. This came about as most LGUs implemented a general revision of the schedule of assessed value of real property in 1993-94 and 1996-97. The Code mandated a revaluation of real properties in 1993 (to take effect in 1994) and then again in 1996 (to take effect in 1996). Increases in RPT buoyancy reflect these assessed value increases, not an improvement in RPT administration. Higher buoyancy increases for cities reflect that cities were aggressive in implementing the new assessments; provinces and municipalities lagged in implementation and this is reflected in lower RPT buoyancies.

Despite some post-code improvement, RPT revenue generation has not met expectations, and given the cost of collection the RPT represents a net loss for most LGUs. Table 12 illustrates the problem. Data in the table are taken from a pilot project initiated by the GOLD project to improve upon revenue generation from the RPT. The data refer to collection efficiency, measured as the percentage of potential revenue actually collected and cost of collection, represented by the cost in pesos of collecting one peso in revenue. The data are unweighted averages for the period 1993-97. The pilot project was conducted in four provinces and in several municipalities within those provinces. Data for two provinces (Nueva Vizcaya and Cotobato) distinguished information for each pilot municipality, while Capiz and Palawan Provinces offered data only on averages for the group of pilot municipalities. Three observations are important:

Over the period, collection efficiencies generally improved, but cost of collection did not. Given the fact that even 100 percent collection efficiency would not result in significant net revenue gains, one must question the role of the RPT as a contributor to LGU revenue independence.

Consistent with the devolution program, LGU spending on the social sectors rose markedly at all levels of government in the post-Code era. Total LGU expenditure was equal to 1.6 percent of GNP on the average in 1985-1991 (Figure 6). During this period, 42.8 percent of total LGU expenditure was on general public services, 32.9 percent on economic services and 20.5 percent on social services (Figure 7).

The mandated transfer to LGUs of functions previously discharged by national government agencies caused a major shift in the size and composition of LGU budgets. Aggregate LGU expenditure rose from an average of 1.6 percent of GNP in 1985-1991 to 3.3 percent of GNP in 1993-1997 (Figure 6). Most of the increment in LGU spending went to social services and general public services (both of whose budgets increased by 0.6 percent of GNP between these two periods). In contrast, LGU spending on economic services expanded by only 0.4 percent of GNP. Consequently, the share of social services to total LGU expenditure increased by 6.5 percentage points to 27.0 percent while that of economic services and general public services dropped by 6.4 percentage points and 3.4 percentage points to 26.5 and 39.5 percent, respectively (Figure 7).

The composition of LGU expenditures varies according to the level of local government. General public services captured the lion’s share of municipalities’ budgets while economic service sectors received the biggest share of total provincial outlays. In contrast, the expenditure of cities was more evenly distributed across sectors. In the period prior to the implementation of the Code, the ratio of social sector spending to total expenditure of provinces was highest at 21.1 percent, while that of municipalities was lowest at 14.7 percent.

There was an expansion in the budget share of the social service sectors relative to the economic and general public service sectors in all levels of local government. Consistent with the fact that provinces absorbed the bulk of social service functions, these expanded the most (from 21.1 percent in 1985-1991 to 35.1 percent in 1993-1997). Most significant among the social service functions assigned to provinces is the operation of tertiary hospitals, which is probably the single most costly service devolved to LGUs. Likewise, the social service expenditures of municipalities rose substantially (from 14.7 percent to 21.9 percent). In contrast, the share of social services in the aggregate city budgets remained stable in the vicinity of 26.7 percent (Figure 7).

The increase in aggregate LGU expenditure on social services between 1985-1991 and 1993-1997 went to health, education, housing/community development and social welfare, in that order. This development is largely due to the fact that the cost of devolved health functions accounted for more than half of the total cost of all devolved functions. At the same time, the cost of devolved social welfare services, although not as large, was also substantial. In other words, higher LGU spending on health and social welfare services in 1993-1997 was less a product of conscious policy choices by LGUs and more a reflection of inherited obligations. In contrast, higher LGU expenditures on education and housing/community development do reflect the higher priority that local officials assign to these sectors in the more decentralized regime, since the direct impact of the devolution program on these sectors was nil.

The previous section established that there has been a dramatic increase in local government spending in nominal and relative terms between 1985-1991 and 1993-1997. But this is partly explained by the transfer of expenditure responsibilities to LGUs from the national government agencies, whose budgets were substantially reduced (at least at the outset). It is not as obvious whether the larger LGU outlays in 1993-1997 maintained or augmented the level of services enjoyed by local communities. To assess whether indeed LGUs have increased (or decreased) spending on the various sectors, the analysis must consider the cost of devolved functions.

What would the LGU expenditure level in 1993-1994 have been if LGUs continued to spend what they did in 1991 and, in addition, maintained the level of spending by the national government agencies on devolved functions? Table 6 estimates what is needed to preserve general government expenditures at their 1991 levels in real per capita terms (i.e., after adjusting for inflation and population growth). It is this that is compared with actual levels of LGU expenditures in 1993-1994.

The results show that actual aggregate LGU spending in 1993-1994 was more than enough to maintain the spending level of 1991, even after taking inflation and population growth into account. In particular, actual LGU spending on general public services and on social services greatly exceeded 1991 levels but focus within the social sector shifted. In education, LGUs in 1994 spent almost four times what was needed to cope with both inflation and population growth. Provinces, municipalities and cities all gave education high priority. Total LGU spending on housing and community development in 1994 was more than twice the amount required to preserve the 1991 level in real per capita terms. Housing and community expenditures of municipalities and cities exhibited significant growth in real per capita terms unlike those of provincial governments, which barely kept up with inflation and population growth.

In contrast, actual 1993-1994 expenditure on social welfare by all types of LGUs (except provinces) fell short of 1991 levels and did not even keep up with inflation. While the 1993 spending of LGUs in all levels on health (with the exception of cities) was also below the amount needed to sustain previous levels in real per capita terms, 1994 health expenditures of LGUs in all levels was sufficient to maintain previous levels in real per capita terms.

LGU spending on economic services in 1993 did not even keep up with prices, let alone population growth while 1994 spending levels on these sectors was enough to sustain 1991 expenditure levels in real per capita terms. Within the economic sectors, agriculture was badly hit with both 1993 and 1994 LGU spending levels falling short of the amount required to maintain 1991 levels in real per capita terms. Although LGU spending on the transportation/communication and other economic services in 1993 was not enough to preserve the 1991 levels after adjusting for inflation and population growth, the situation was reversed in 1994.

In general, therefore, one must conclude that LGUs "underspent" on health, social welfare and agriculture services relative to levels prevailing before devolution. It would seem a puzzle and a cause for concern as to why LGUs have allocated less to certain types of services and more to others. LGUs as a whole have, after all, received more than sufficient amounts to continue providing the services devolved from national government agencies. There have been specific cases, of course, where the devolution formula was unmistakably deficient, so that IRAs were insufficient to even cover the cost of devolved functions. Yet this cannot have been the main reason for LGU underspending, for even provincial governments with sufficiently large social services budgets chose to allocate these differently, spending more, say, in the direction of education. What this suggests is that after devolution, many LGUs have essentially chosen to discontinue the pattern of budget allocation previously implemented by national government agencies, most likely because these do not accord with their own priorities.

There are three possibilities. First, the devolved functions themselves may be deemed unresponsive or unimportant to actual local needs. For provinces, in particular, a large part of devolved health services consists of huge outlays for tertiary hospitals, which are nonetheless ill equipped and unresponsive to community needs. It is not surprising, therefore, that some LGUs choose to discontinue their support for such services. Second, if LGUs are given expenditure responsibilities with significant spillover effects (i.e., responsibilities, like public health services, whose benefits not exclusively enjoyed by their constituents) then it is expected that LGUs will underprovide these services without additional financial support from the central government, perhaps in the form of matching grants. If the benefits are confined to a well-defined sub-national area, then cost sharing among LGUs that benefit from the service may be the more appropriate arrangement. There is a third possibility that LGU priorities themselves may be misplaced or badly informed, placing expedient showcase projects ahead of continuing programs with long term effects.

Section 4 summarized the factors that one would want to observe to make judgements about efficiency changes associated with fiscal decentralization. That discussion leads to a focus on three dimensions of efficiency, production, allocation and fiscal efficiencies. Section 6 summarized salient features of fiscal decentralization in the Philippines. We now combine these to comment on likely efficiency changes from fiscal decentralization in the Philippines case.

Production efficiency has three main components. The first is conceptually the simplest and deals with the cost of production differences between LGUs and centralized national government agencies. Second, structural changes can promote the quest among LGUs to seek cost minimization. Third, local capabilities will influence how well LGUs are able to perform.

It is extremely difficult to obtain comparable and aggregate cost estimates of various goods and services delivered or facilities operated by LGUs and the central government alike. In most cases LGUs and the central government provide different things, so no comparison is possible. In other cases where they do provide the same things, there may be significant qualitative differences. LGUs may choose to build smaller, less elaborate classrooms than the central government. LGUs may choose to build roads that are narrower or less able to carry heavy loads. LGU concepts of what constitutes health care may differ from those of the central government. The best one can achieve in the Philippines, and indeed in any developing country, is to assemble the great mass of anecdotal evidence that LGUs do it better/cheaper, or at least differently when compared to the central government.

School buildings. Experience under the on-going Third Elementary Education Project (TEEP) of the World Bank indicates that LGUs are able to construct school buildings at a lower cost than the Department of Public Works and Highways (DPWH), which implements the School Building Program (SBP) of the Department of Education. The World Bank (1998) noted that LGUs’ cost in constructing one classroom is P180,000 while data from San Carlos City (in Western Visayas) indicated that its unit cost to build one classroom is P235,000. Compare these figures with the DPWH’s unit cost of P305,000.

It should be emphasized that the lower cost incurred by LGUs in the construction of school buildings is a theme that is repeated heard in interviews with LGU officials from Benguet and Ifugao in the northern part of the country to Agusan del Sur in the south. Alonzo (1998) also reports that many local government officials pride themselves in being able to construct three classrooms for the DPWH price for two.

Part of the explanation for this phenomenon appears to stem from the fact that the DPWH cost estimate includes an allowance for management overhead which LGUs’ costings do not include. LGUs also suggest that they are able to use lower cost materials that are locally available (say, sand/gravel from LGU-operated quarries). Also, some LGUs encourage local communities to contribute labor inputs.

Day Care Centers. Data from San Carlos City indicates that its unit cost in constructing a 64 sq. meter day care center in 1998 is P285,000. The comparable figure used by the Department of Social Welfare in costing the day care center portion of its Early Childhood Education Project is P304,000-P338,789. Most barangays have initiated day care facilities in the post-Code period. Most of these rely on donated space, space that is normally used for other purposes or facilities rented at very modest cost. If the objective is the provision of day care rather than the establishment of a day care center, then the barangay programs’ costs come in far below those of the central government. Admittedly, the barangay programs may appear rather "rustic" by comparison, but provide the service nevertheless.

Road Construction. Data from San Carlos City show that the city was able to construct the 42 km Translink Highway at a cost of P2 million per kilometer. In comparison, information obtained from the DPWH suggests that the cost of a similar road construction activity is P2.2 million per kilometer. Alonzo (1998) reports that barangay roads are constructed by some LGUs at the two-thirds the cost of similar roads constructed by the DPWH.

Table 7 and Table 8 document the ex-post assignment of expenditure responsibilities across levels of government before devolution, in the transition phase of the devolution program and in recent years. Table 7 confirms the shift of more functions from the central government to LGUs. The transfer was dramatic in the areas of health, agriculture and social welfare. While Table 7 suggests that many responsibilities are shared by LGUs with the central government, the Code actually provides an explicit and clear delineation of functions across levels of governments except perhaps in the area of environment and natural resource management.

The devolution of expenditure responsibilities to LGUs, with a few important exceptions, is consistent with the decentralization theorem. For the most part, these are activities that can be provided at low levels of government. The activities identified have few important spillovers to a broader community to indicate that they should be provided by higher levels of government. The functioning of the decentralization theorem is enhanced by the Philippine tendency for LGUs to regroup into larger cooperative units when important spillovers occur. Generally, significant expenditure responsibilities are devolved to very small LGUs, such as the barangays. Even most municipalities are very small, both geographically and in population. Frequently, smaller LGUs form joint efforts to carry out their responsibilities when there are economies in so doing and when inter-LGU spillovers are present. Many examples exist of contiguous LGUs combining their resources to deal with watershed problems, coastal and local fishing problems, solid waste disposal and others

There are two important exceptions to application of the decentralization theorem. Most education and law enforcement are still provided at the national level. In 1997 central government agencies accounted for 92 percent of all education spending and 98 percent of all law enforcement. School construction and maintenance are devolved to LGUs and indeed this reaches low levels of decision making. Both are controlled by local school boards and are funded by a property tax "add-on" called the Special Education Fund (SEF). On the margin, school construction and maintenance comes from the SEF. Teachers salaries and overall education policy is made at the national level, and all teachers are national employees of the Department of Education, Culture and Sports (DECS). Large parts of both education and law enforcement could be devolved to LGUs and such a move would probably improve upon production efficiency.

Expenditure assignments are generally clear and well understood by LGUs, but the national government does not always "allow" LGUs to follow their own predilections in areas that have been assigned to them. Our interviews revealed many cases where national government agencies try to dictate to LGUs how they should be carrying out their expenditure responsibilities. This interference in LGU affairs by national agencies comes from several sources:

Apparently, national government agencies have not learned how to work well with LGUs and consider it their responsibility to impose national standards and programs. National agencies provide no back-up to LGUs in carrying out their devolved functions. This problem may be more apparent than real. While LGUs report many instances where national government agencies try to tell them what to do, few consider it a problem. Most LGUs simply ignore those national agencies that do not serve LGU interests. Nevertheless, LGUs report that they sometimes go along with National agencies because they control resources that may be lost to the LGU if they were uncooperative.

One can only surmise that the devolved national government agencies (NGAs) have not fully given up on their devolved functions in the interest of turf-maximization. Section 17 of the Code (which allows for central government augmentation or direct provisioning of basic services assigned to lower levels of government) provides NGAs the legal cover to do this. Moreover, the Code and its Implementing Rules and Regulations failed to define the mechanisms through which national government can direct such assistance to LGUs. At the same time, Executive Order (EO) 53 mandates that NGAs retain management control over all foreign-assisted projects and/or nationally funded projects even if the same involve devolved activities. On the other hand, many of the so-called devolved NGAs are made accountable for the overall outcome in their respective areas and as such they deem it their responsibility to direct LGU behavior in support of national objective. Thus, most of them tend to make full use of the provisions of the Code and EO 53 regarding augmentation. Consequently, the existing regulatory framework effectively permits, even encourages, the existence of a two-track delivery system, where both NGAs and LGUs can initiate devolved activities (Gonzalez, 1996).

In 1996 the Investment Coordinating Committee (ICC), a national government agency, laid down the policy framework for LGU access to official development assistance (ODA). Central to the procedural guide, prescribed by the ICC, for the design of co-financing programs between a NGA and one or more LGUs is the "lead agency" requirement. This means that a national government agency has to take the initiative in identifying and developing a co-financing program within the context of a sector program. The lead agency also has the responsibility for securing ICC approval, and subsequently, setting up the project management office. The lead agency requirement, by its very nature, makes both project design and implementation highly centralized. As a result, central agencies tend to retain some activities (like implementation oversight, monitoring and evaluation, and procurement and service delivery) that are better performed by lower level governments. Consequently, it further reinforces the two-track delivery system referred to earlier.

Some signs of "creeping re-nationalization" of devolved activities can be observed in Table 8. As one would expect, national government (non-debt service) spending during the Code implementation period (1992-94) dropped from 11.6 percent of GNP to 11.4 percent. It is surprising how slight this drop is during a major shift to fiscal decentralization. Since 1994 the percentage of central government spending rose to 13.2 percent of GNP. Growth rates in components of the national budget, where the national government shares responsibilities with LGUs indicates very rapid growth for centralized programs. During 1994-97 there has been a pronounced increase in health, environment and natural resources, agriculture and social welfare. As a percentage of all non-debt service spending, the central government’s share fell from 88.3 percent in 1992 to 77.9 percent in 1994, but then rose to 78.7 percent in 1997.

Revenue assignments are an important feature of the structure within which LGUs operate. They are of three kinds:

The three sources of revenue to LGUs suffer from important flaws. Ideally, one would want sources that are dependable, allow multi-period planning and funding for projects that span several years. The IRA is most dependable, but the inflation tax built into it and its apparent vulnerability to political exigencies create considerable uncertainty. The share of natural resource revenues is totally unreliable. As we saw above, local revenue sources have not kept up with LGU spending requirements.

LGUs are not very constrained in the way that they spend their IRAs and other revenues. The Code specifies that 20 percent of the IRA must be earmarked for a "Development Fund" and used only for development purposes. Within that 20 percent, 5 percent is earmarked for disaster relief. Other than that, LGUs can determine how to meet their expenditure responsibilities without constraint. However, LGUs report that many of the expenditures they make anyway can be considered "development" so they have no problem satisfying the requirement without compromising their independence. The disaster fund can be held over to subsequent periods and used in any way LGUs want if there is no disaster. Most consider the disaster fund an advantage because they are able to act quickly in the event of disaster, but suffer no revenue loss if none occurs.

Initially, LGUs had difficulty mustering the skills needed to carry out their devolved responsibilities. Over time these problems seem to have declined. Several important incentive problems exist that have made it difficult for LGUs to carry out their responsibilities. These include:

Despite these incentive problems, some LGUs are now considered better places to work than they were before devolution. Several people interviewed agreed that they preferred to work for their LGU because LGUs could now provide career opportunities, the work environment was better than in national agencies and the pay was the same as it is at national agencies. Furthermore, smaller cities were becoming better places to live as local services improved.

Innovation by LGUs can stretch the revenue and skills available to them in ways that are probably not available to the national government. Capabilities of LGUs to deliver services are enhanced by their abilities to mobilize local resources and people and combine that with modest amounts of revenue. Examples abound. Pineda (1998) identifies a great number of innovative ways that LGUs have used to extend health care such as organizing volunteers, placing barangay health stations in private homes of concerned citizens, hiring part-time workers and soliciting help from the local private sector among others. Mercado (1998) documents many innovations in environmental services such as coastal protection and solid waste management by way of community involvement, often including input from NGOs. The Galing Pook Awards discussed in Miller (1999) offer many innovative examples. Interviews indicate that innovations of these kind are a growing trend as experience accumulates.

In summary, production efficiency has probably been improved. However, efficiency improvements must have been modest since LGUs only spend about 3.7 percent of GDP (in 1997) compared to about 1.9 percent in the immediate pre-Code period. Cost comparisons seem to favor LGUs over national agencies. Structural changes, with a few exceptions, follow the decentralization theorem. Lack of local capability has not been a major constraint though the record there is uneven. Capabilities vary considerably from one LGU to the next. Innovation has extended the productivity of revenue. Improvements in production efficiency could probably be gained by devolving additional services that are justified by the decentralization theorem, such as education and law enforcement. Greater attention to equitable and adequate pay, and to the career development paths of employees would probably yield improvements in efficiency. Effort should also be made to bring national agencies on board as supporters of LGU activities. An important problem is related to the shared resources that are supposed to go to LGUs. Inflation cuts into the IRA before revenues are delivered, shared revenue from natural resource exploitation is rarely forthcoming, and administrative decisions can cut the IRA to meet national objectives.

To gain improvements in allocative efficiency, LGUs reallocate resources from what had been provided by centralized government toward provision of a set of goods and services that better match local demand and local willingness to pay. In section 6.1.4 we compared expenditures before and after the Code went into effect. Comparisons adjusted for inflation and population change asked the question "did LGUs maintain the same level of real per capita expenditures after the Code went into effect as had been provide by the central government before the Code was implemented?" That exercise revealed reallocations did occur.

Reallocations of resources among expenditure categories have not been large. In most cases LGUs received IRA allocations that were adequate to cover the costs of the services that were devolved on them. However, in about 25 percent of cases where the IRA was less than the cost of devolved services, little or no room for reallocation existed. Even in those LGUs that had "excess revenues" from the IRA, little reallocation occurred. This may be in some measure inertia, since it takes some time for LGUs to test the limits of their abilities and the limits of the law. Nevertheless, upon devolution some LGUs have reallocated resources to better match local preferences. There are many ways in which reallocation has occurred, and while the list of reallocations may seem rather ad hoc, the point remains that LGUs now have greater flexibility in satisfying local demand and they have exercised that power.

To best allocate expenditures to match local demand, some mechanism to detect local demand must exist. LGU officials encounter a number of indicators of local demand for LGU services.

In general local governments in the Philippines have many mechanisms to connect decision-making to policy. Furthermore, LGU officials have an incentive to respond to local demand.

It is important for allocative efficiency that LGUs raise significant revenue from their own sources. In this way local citizens "feel" the cost of public expenditures because some of it comes out of their own pockets. On the margin increases in expenditure imply increased local taxes. Faced with this, local citizens are best able to match marginal costs and benefits, thereby enabling them to approach greater allocation efficiency. Section 6.1.2 above provides figures showing the evolution of the structure of LGU local source revenue.

Local Real Property Tax (RPT): The real property tax was intended to become the main source of own revenue for the LGUs but it has not turned out to be. The RPT is set at 1 percent of assessed value (called the "basic" RPT). However, another 1 percent is taxed and is earmarked for a "Special Education Fund" (SEF). Technically, these two taxes are separate though they are administered by the same procedures and are collected together. As a source of revenue to LGUs the RPT suffers from important problems such as:

Incentives to collect the tax are weak. RPT collection is a responsibility of the municipalities, but the municipalities are required to share the revenue with the provinces and barangays. Of the basic tax revenues collected, the municipality is required to deliver 35 percent to the province and 25 percent to the barangays. It keeps only 40 percent for itself. Furthermore, assuming the SEF portion of the tax is collected, all of it goes into the SEF, which is under the control of local school boards. Thus, of revenue collected by the municipality as little as 20 percent goes into its own coffers, unearmarked. Yet the Municipalities bear most of the cost of administration and collection.

RPT delinquencies are a great problem. Despite sufficient powers under the law to enforce collection from delinquent accounts, the collection rate has rarely exceeded 60 percent in the period between 1983 and 1997 (Table 10). Worse, the collection efficiency for the RPT declined from 56.7 percent in the pre-Code regime to 53.8 percent in the post-Code regime. Moreover, it is also shown that the RPT collection efficiency in both provinces and cities deteriorated during the period under study, with former suffering a bigger cut than the latter.

If problems were solved, the RPT has the potential to be a significant revenue source. RPT revenue is determined by the tax rate, the collection rate (the ratio of actual tax revenue to tax liability), the assessment ratio (the ratio of the assessed value to the fair market value established for RPT purposes) and the assessment efficiency (the ratio of the fair market value to the true market value). Based on counterfactual simulations, Manasan (1995) shows that if the collection rate is increased to 100 percent (from 50 percent) while other variables are held constant, RPT revenue will increase by 100 percent. Doubling the statutory rate will yield the same proportional increment in RPT revenues. On the other hand, if ceteris paribus the assessment efficiency is increased to 1 (from 0.33), RPT revenue will increase by 200 percent. In like manner, if both the collection rate and the assessment efficiency are raised to 1, RPT revenue will increase by 500 percent. These results imply that: (1) correction of the undervaluation problem by adjusting the schedule of fair market value to the level of the true market value has the greatest potential in increasing RPT revenue; (2) improvements in collection efficiency will lead to substantial growth in RPT revenue quite independently of the undervaluation problem; and (3) RPT tax capacity, given present tax rates and assessment levels, is not negligible.

Department of Finance guidelines instruct local assessors to use: (1) sales data from the Register of Deeds; (2) sworn statements of real property owners filed every time a general revaluation is undertaken or when a transfer or construction is made; and (3) expert opinion of real estate agents and professional appraisers. The first two of these sources are inherently subject to under-estimation. However, local assessors rely on them, often to the exclusion of the third, because somehow the latter is viewed as "unofficial" and therefore, indefensible if challenged either by local taxpayers and/or local elected officials. In this regard, the use of zonal values of real property (which the Bureau of Internal Revenue uses as the basis for the capital gains tax) should be considered as an alternative source of information in computing the fair market value for RPT purposes.

Problems with other local taxes are also apparent. The IRA is negatively related to local business tax effort in statistical studies. Local user fees rarely cover the cost services provided. As a result, there has been little increase in the share of total revenues originating from local taxes despite the assignment of new taxation power to LGUs. On average, before devolution (1985-91) LGU own revenue was 4.9 percent of total government revenue; after devolution this increased by only 0.8 percent to 5.7 percent (1992-96) (Manasan, 1998). The increase in the proportion of local revenues in total revenues occurred entirely in cities. Own revenues for cities have averaged 2.1 percent of total government revenues prior to 1991 and 2.9 percent since. For other units of local government there has been no proportional increase in own revenue generation.

3. Allocative efficiency: accountability

There is a consensus that democracy has taken hold and strengthened in the Philippines, and this is the main source of accountability for local officials. Reports are that there is an increasing frequency of contested elections at the local level rather than the simple transfer of power within dynasties. As mentioned above, there is increased involvement of NGOs and peoples organizations in local spending/taxing decisions. Furthermore, that involvement has become more well accepted by local officials than was the case in the immediate post-Code era. (see Miller, 1998). From al reports, accountability of LGU officials to their constituents has improved and in most cases is already at acceptable levels.

In summary, allocative efficiency has probably improved, but any improvements have been slight. The mix of expenditures has changed slightly since 1992, and LGUs’ preferences are better reflected in expenditure patterns. Many have used innovative ways to leverage revenue by mustering local, in-kind resources. However, changes have been small. Many LGUs are not able to meet devolved expenditure responsibilities and create new expenditures out of the IRA. Very few LGUs have exploited their own sources of revenue. Yet, all face hard budget constraints. Ultimately, allocative efficiency will be improved for those jurisdictions that improve upon revenue generation through local taxes such as the property and business taxes.

Fiscal efficiency requires that there be correspondence between taxes paid and services received and that the intergovernmental grant structure be non-distorting of local tastes and preferences. It also requires that LGU spending and fiscal management be consistent with national macro-economic stability.

There are primarily three local taxes, the local business tax, the RPT and user fees. Of these we have already discussed the RPT, which is not effective in bringing taxation to the level where local services are found. This is probably not a function of FD per se, but a result of an underdeveloped property tax system. User charges too have been found to be defective. Manasan’s (1998) examination of user fees levied by cities found that in no case did the fees cover the cost of services provided by cities. Surveys by the GOLD project have found the same. This is understandable for services such as hospital care which is widely considered a "public good," but it is not so acceptable for services such as public markets, slaughterhouses, etc. This leaves the local business tax as the only effective connection between local services and local taxation. If the RPT costs more to collect than it yields, then it is a net drain on resources, and if user fees do not cover the cost of service, then the LBT is the only one in the LGU arsenal of taxes that makes the required connection. Unfortunately, Manasan (1998) has also shown that the IRA formula acts as a disincentive to collect the LBT.

It is possible that the IRA formula delivers revenues to the LGUs in a way that acts as a disincentive to local tax efforts. The IRA is by far the largest source of revenue to most LGUs. Indeed, the IRA more than covers the cost of devolved expenditures for most LGUs. In addition, if local revenues come at considerable cost to LGUs then it is logical that LGUs would rely upon the IRA for revenue and not exert a vigorous local tax effort. The empirical evidence seems to support this view. Manasan (1998) and Llanto, et al. (1998), show via regression analysis that there is a negative correlation between local tax effort and IRA per capita. Manasan (1995) shows that while intergovernmental transfers had a neutral effect on local revenue performance in 1985, it substituted for local tax revenue in all levels of local governments in 1992 and 1993. While the relationship between per capita local tax revenue and per capita IRA is weak for provinces (where a P1 increase in per capita IRA leads to a P0.01 reduction in per capita local tax revenue in 1992-1993), it is not negligible in the case of cities (where a P1 increase in per capita IRA is predicted to lead to a P0.43 decrease in per capita local tax revenue) and municipalities (where a P1 increase in per capita IRA results in a P0.09 decline in per capita local tax revenue). Moreover, Manasan (1998) reports similar results for per capita local business tax revenue of cities in 1994 and 1995.

One is tempted to recommend that the IRA be altered to better provide incentives for local tax effort and perhaps to meet other objectives as well. However, there are advantages to the IRA as it exists:

Despite these advantages, there remains a mismatch between the financial resources and the expenditure responsibilities that have been transferred to LGUs. The increase in the IRA share of some LGUs is not enough to finance the functions devolved to them; others have received resources beyond the expenditures devolved on them. In 1993 the per capita net resource transfer (i.e., per capita 1993 IRA, less per capita 1992 IRA, less per capita cost of devolved functions, adjusted for inflation) was negative in 37 out the 66 provinces for which data were available. This figure was down to three in 1994. However, the situation is likely to have worsened since then given the increasing number of unfunded mandates.

Unfunded mandates have not been effectively dealt with in the Philippines. No extra transfers are available to cover unfunded mandates, nor is there a system of equalization grants to assist poor or otherwise disadvantaged LGUs to serve devolved expenditure responsibilities. Examples of unfunded mandates include:

Anecdotal information indicates that as a solution to unfunded mandates, LGUs simply ignore them. In the cases of salary increases laws have included language that allows LGUs to not pay the salaries if funds are not available. However, this puts LGUs in a very inflexible position, since there is an automatic claim on additional resources by employees who are not paid up to the legal scale. This not only reduces an LGU’s flexibility to reallocate resources but is bad for worker morale and may act as a disincentive for a vigorous local tax effort.

The total amount of resources commanded by LGUs is small. Collectively, they spend about 3.5 percent of GNP, receiving about 2.3 percent of GNP via the IRA. Furthermore, borrowing by LGUs is very limited (see GOLD, 1996, p. 6). The Code allows LGUs to borrow, but they are limited in the debt service burden that they can take on. Debt service is limited to 20 percent of their "regular income," which is taken to mean 20 percent of their IRA. Yet LGUs are not allowed to pledge their IRA as security for loans. There has been a way around these requirements because LGUs can arrange an IRA "intercept." By keeping their IRA deposits with banks that have extended them credit, those banks can be allowed to "intercept" IRA income to cover debt service before it is released to the LGUs. However, the Code only allows LGUs to keep deposits with state-owned banks, so this facility is not available with private banks. Furthermore, the requirement to keep deposits only with state-owned banks does not allow LGUs to develop the relationships with private banks that may lead to private financing for LGUs. The only lenders in effect are state-owned banks, and almost all of that comes from the Municipal Development Fund. (See Llanto, et al., 1998, for details)

The Code prohibits LGUs from operating at a deficit. Balanced budgets are required. Even in cases where unfunded mandates are imposed, LGUs cannot run deficits and balance their accounts by simply ignoring mandates. The balanced budget constraint is a double one. Since there is no substantial credit available to them, most LGUs could not run a deficit if they wanted to.

There is very little chance that the fiscal decentralization scheme in the Philippines could precipitate significant macro-economic instability. Budget constraints are hard and resources committed to LGUs are small. Without a practical way to borrow, LGUs cannot get into the debt problems that have caused difficulties in other countries. Also, in the Philippines the problem of central government "bail-outs" for LGUs that overspent and over borrowed simply does not arise.

In summary, fiscal efficiency could be improved upon. Local taxes already exist that would make the correspondence between taxation and the economic activity that is supported by local expenditures. But local taxes are ineffective in making this connection. Macroeconomic stability does not seem threatened at all by the form of fiscal decentralization practiced in the Philippines. LGUs cannot run operating deficits and are extremely limited in their ability to borrow. Thus, the instabillity observed in other countries, where national governments implicitly (and in some cases explicitly) guarantee local debt, is not present in the Philippines.

This study is a follow-on to work begun earlier in the CAER project on measuring and assessing the effect of fiscal decentralization in developing countries. In CAER II Discussion Paper #3 (Guess, Loehr and Martinez, Fiscal Decentralization: A Methodology for Case Studies, March 1997) a general format for doing case studies was created. That paper did not include any case studies. It was comprehensive, dealing with all of the most important economic factors affecting the fiscal decentralization process. This paper does two things. First it recommends an approach to assessing the economic efficiency impacts of fiscal decentralization and second, it does a brief case study of the Philippines to illustrate the method.

The methods recommended recognize the difficulty of doing economic analysis in developing countries. Data are scarce and other information incomplete. Nevertheless, assessment of what are basically economic factors requires that observations and analysis follow the logic of economic theory, but purely quantitative methods cannot be used exclusively. Judgement is required. Just as economic theory organizes empirical investigation, it also organizes the way we think about problems. If information about those problems is judgmental and anecdotal, theory organizes whatever the available information is. It is normal in developing countries (indeed, almost everywhere) that information is incomplete. In the end, economic theory can be used much as rules of evidence are used in a court case, to build a picture of "the preponderance of the evidence," even where some of the evidence is anecdotal.

The objectives of this project are:

The main effect of FD is to provide improved resource allocation and therefore improved economic efficiency. Improved efficiency is the main positive economic factor to come from FD. Efficiency gains rest on the presumption that local governments are much better in identifying and fulfilling the needs of households, since they are closer to them, and in mobilizing and using local resources to pay for goods and services having purely local impacts. Improved efficiency is defined as getting greater consumer (taxpayer) welfare from fixed resources. Not all gains in consumer welfare can be measured as increases in output. When FD is successful, resources have been reallocated so that marginal costs of publicly provided goods and services more closely match marginal benefits. Relative allocation questions come into play.

We recommend that attention be focused on three dimensions of economic efficiency: