Electronic copy available at: http://ssrn.com/abstract=1718692

1

Do local governments interact fiscally with their neighbors in a devolved health care system -- A spatial analysis of the Philippines? Abstract: The aim of this research is to better understand the effects of a decentralized health system of the Philippines on public health care expenditures by investigating determinants of local government public health spending for year 2007. Within the context of the Philippines’ decentralized health system and health externalities that arise at the individual-level and manifest at the regional level, particular emphasis is given to spatial spillovers and fiscal interactions among municipalities. The research addresses these issues in an empirical spatial econometric framework utilizing public finance local government data for the Philippines. Competition for health resources including doctors and drugs, thereby bidding up the costs of health inputs is one of the potential causes for positive fiscal spatial dependence cited in this paper. It is also anticipated that this could be an outcome of yardstick competition, where local agents strategically compete with each other by spending higher in order to be re-elected, 2007 being an election year. Introduction Historically, policy-making in developing countries was highly centralized. It is deemed that a centralized system emerged out of a strong colonial past or disenchantment of a military rule or dictatorships (in the case of Latin America) in some of these countries. In recent decades, however, decentralization has been at the center stage of policy experiments in many developing countries (Bardhan 2003, pp.1). Countries in Asia, Africa and Latin America are increasingly involving lower tiers of government in the delivery of public services that were earlier a responsibility of the central governments. Public services ranging from water, sanitation, education, and health are now provided by local governments in Uganda, Philippine, Nigeria and India. While local governments have limited revenue-raising capacities in developing countries as compared to the developed world, they enjoy substantial fiscal powers, specifically budgeting and spending.

To better understand the meaning of decentralization, a conceptual framework of three forms was developed, namely deconcentration, delegation and devolution. These forms differ in degree of discretion enjoyed by local bodies. A broad range of political, administrative and fiscal responsibilities are transferred from the central government to regional, provincial or municipal governments in decentralization. While deconcentration denotes transfer of administrative authority, delegation refers to transfer of authority over a specific set of tasks to local bodies such as state-owned enterprises. Devolution, on the other hand, is the transfer of authority to local governments that involves transfer of assets and rights, including fiscal powers of revenue raising and spending (Rondinelli and Cheema 1983).

Health service decision-making in developing countries like most public services was highly centralized with weak administrative capacity at the local government level (Mills 1990). In the last few decades (1980’s, and 1990’s) health care as a public responsibility has been devolved from national governments to sub-national governments in many developing countries.1 The economic rationale for decentralization is mainly to allocate 1 An example of deconcentration is a district-level office of Ministry of Health. The lower-level offices are bestowed clearly defined functions, powers or responsibility by the higher-level government that would

Electronic copy available at: http://ssrn.com/abstract=1718692

2

resources efficiently. Heterogeneous local needs are better satisfied by locally delivered services than when centrally provided. The reasons for devolving health care responsibility to local governments are similar. The theoretical arguments advanced for decentralizing health care include a) increasing local ownership and accountability b) improving community participation c) strengthening integration of services at the local level d) enhancing streamlining of services and e) promoting innovation and experimentation (Lakshminarayan 2003 p. 96). 2. Literature Review The term “pure public good” was first used by Samuelson (1954) to define collective consumption goods that people consume collectively such that an individual’s consumption does not lead to subtraction of the good for another individual’s consumption. They are characterized as “non-rival” and “non-exclusive”. A commonly used example of a public good is public defense provided to all citizens by the national government.

Samuelson (1958) however pointed out that not all goods provided by the government fall into this category of pure public goods. Buchanan (1965) filled this gap by defining “club goods” that share properties of “pure public goods” and “pure private goods”. While one person may consume a pure private good at a point of time, not all public goods may be consumed by an infinite number of people such as police, parks, and fire services. The latter category of public goods is called “club goods”. The utility of an additional person consuming a club good depends on the total number of people involved. These goods are therefore subject to congestion costs and the optimal number of consumers is determined at the point where marginal gain from an additional resident is equal to the marginal cost of congestion.

Under microeconomic principles, healthcare as a good may be classified as public or private based on its nature of consumption. Preventive health such as road safety and fluoridation of water, immunization, family planning and health education are examples of public health goods. On the other hand, consumers draw exclusive benefits from privately consumed health services such as hospital care. Musgrave (1959) used the term merit goods to describe public health goods such as family planning, and immunizations as those whose private benefit is lesser than the social benefit. These goods typically accrue externalities, and are under-supplied in a private market.

The theoretical background for this paper draws largely from the fiscal federalism literature, related to the provision of public goods by local governments and its implications. Proponents of decentralization believe that local provision of public goods is the most efficient way of allocating resources and achieving societal welfare in an economy (Tiebout 1956; Oates 1972; Musgrave 1959). Proximity to local governments helps consumers reveal their true preferences by “voting with their feet”, thus allowing for a more transparent and accountable system (Tiebout 1956). In stating the

enable the local officials to manage without constant reference to the Ministry. The second form of decentralization is the delegation of government functions to parastatal units, private sector groups or non-government organizations who act as agents of the national government. Devolution is that wherein the LGUs are given full autonomy and control over their assigned public services, subject only to oversight by the national government. In this case, the devolved health personnel answer directly to local elected officials, who are responsible to their constituents.

3

decentralization theorem, Oates points to the appropriateness of a decentralized system “in the absence of cost-saving from centralized provision of a good and of inter-jurisdictional externalities”, he further states that “the level of welfare will always be at least as high (and typically higher) if Pareto-efficient levels of consumption are provided in each jurisdiction than if any single, uniform level of consumption is maintained across all jurisdictions” (Oates 1972, pp.54).

Starett (1980) and Bewley (1981) illustrate non-Pareto states of equilibrium resulting from local provision of public goods in the presence of inter-jurisdictional spillovers and competition between regions. Local governments, ignore social benefits of providing a public service, leading to marginal benefit exceeding marginal cost of provision, and eventually an under-provision of a local good. Through compensation schemes between jurisdictions, Williams (1966) and Buchanan and Kaflogis (1963) show how local governments can achieve a socially optimal solution even in presence of external economies.2

Spatial interaction among local governments in the presence of externalities has been a major area of theoretical work in public economics. Fiscal interactions among local governments take place in a number of situations. First, regions may compete with each other in setting taxes and expenditures policies (Sinn 2003). For instance, in the “race to the bottom”, local officials may lower tax rates to attract more “businesses” and speed up economic development as compared to other regions. Second, fiscal interactions occur in the presence of spillovers. Local provision of public goods may cause positive or negative spillovers, i.e. accrue benefits or harm to citizens of regions, thus giving rise to interdependence among regions’ fiscal policies. Third, fiscal interactions may also be induced by inter-jurisdictional competition among local political actors in the form of “yardstick competition”. Residents of a region use their neighboring governments’ performance as the yardstick to measure and evaluate the performance of its own government and whether they should seek to replace their governments in the next election period (Shleifer 1985; Besley and Case 1995). Fiscal interactions, in addition to local governments of one tier, may also occur between two or more tiers of government. These are called vertical externalities.

Interdependence of fiscal policies between local governments, at one level or more may have serious distortions in achieving public sector equilibrium (Brueckner 1998). There are suggestions provided by scholars to remedy such deficiencies. Central government intervention or forms of revenue sharing are a few recommendations made (Oates 1999; Musgrave 1997). Tiebout (1956), Williams (1966), and Oates (1999) recommend interaction and coordinated decision-making between jurisdictions in order to minimize external economies and diseconomies. Contiguity facilitates integration of services and reduces diseconomies of scale, although not at the cost of reducing services or increasing costs. Individuals “vote with their feet” and join a community that provides a desired level and mix of public goods. Oates (1999) visualizes a Coasian process of

2 Buchanan and Kaflogis (1963) use the case of immunizations to show that when two parties yield

external economies to each other i.e. the externalities are reciprocal, cooperative purchase of the immunizations may lead to socially optimum position in which fewer immunizations are purchased than if each party acted independently. Thus, external economies need not result in an undersupply of aggregate resource inputs. Further, the output produced maybe lower or higher than in the case of individual adjustment.

4

integration, like Chesapeake Bay as a joint enterprise between Pennsylvania, Delaware, Maryland, Virginia, while Washington D.C. plays the role of the federal government.

The grounds for externalities arising in the health sector are based on altruistic theories that support interdependence of utility functions between individuals (Hochman and Rodgers 1969; Rosenstein and Wolpin 1983; Jacobsson et al. 2005). A great amount of theoretical evidence on altruism exists in the health care sector of developed countries. Hochman and Rodgers (1969) point to the importance of income redistribution in the form of fiscal transfers in order to achieve Pareto-efficiency. Income differential and transfer elasticity of the donor were stated as vital in determining desired amount of fiscal transfer. Spillovers of health benefits to neighboring regions arising from expenditures in a given region do not necessarily make its residents worse off.

Externalities in healthcare commonly occur among individuals in the consumption of public health goods such as immunizations, and control of infectious diseases. Vaccination not only has direct benefits to persons inoculated but also accrues indirect benefits to non-inoculated persons, as vaccinated people are less likely to transmit a disease. In economic terminology, marginal social benefit of vaccinations is higher than marginal private benefit, driving total production to be below the optimal level. Boulier et al. (2007), with the help of a Susceptible-Infectious-Removed (SIR) model analytically present vaccination externalities and show marginal social benefits and externalities to vary with the number of vaccinations, vaccine efficacy and infectiousness of the disease3.

Rosenzweig and Wolpin (1994) illustrate health externalities in subsidizing family planning programs and its effect on child health programs among low-income households in a Philippines’ province. Family control subsidies are shown to be Pareto-optimal when child health and family size are gross substitutes. The importance of incorporating health care externalities is also recognized in health economic evaluations.

Empirical Work Done on Fiscal Decentralization and Spillovers Several studies empirically test the presence of fiscal spillovers and policy inter-dependence among local governments in developed countries. Hanes (2002) studied spillover effects for Swedish local rescue services (fire protection) and identified strategic behavior among contiguous municipalities, causing an under-supply of local rescue services. Brueckner (1998) presented evidence of strategic interaction among cities of California in imposing growth control measures. In departure from the conventional definition of ‘geographical neighbors’, Case et al. (1993) showed interdependence between jurisdictions that are similar with respect to income and demography, instead of geographical proximity. People residing in states of similar demographic and economic characteristics are more likely to compete with each other in labor markets.

Decentralized health systems in the developed world that have recently been adopted by developing countries manifest externalities in provision of public health services such as family planning, child and maternal health and infant nutrition. 3 As a higher number of people are vaccinated, non-inoculated persons continue to derive indirect benefits of not contracting the disease. Second, degree of externality may depend largely on the effectiveness of the vaccine and infectiousness of the disease. For instance, if vaccine efficacy is low, chances of inoculated people transmitting the disease are higher, driving the non-inoculated people to get vaccinated. The externality will thus be small. Third, if the disease is less infectious, the cost of inoculation may be perceived to be higher than the chances of contracting the disease. In this case, non-inoculated people will free ride on the inoculated people, increasing the extent of the externality.

5

There has been considerable discussion of the potential negative effects of decentralization of public health services (Kolehmainen-Aitken and Newbrander 1997; Melgaard 1998; Feilden and Neilson 2001). Peyvand Khaleghian (2003) examined implications of political decentralized systems on the level of health services specifically immunizations among low and middle-income countries and find contrasting results between them. While the impact of decentralization on immunization coverage is positive for low-income countries it is negative for middle-income countries.4 In the Philippines, decentralization resulted in higher spending on curative services by local governments than preventive services. Under a system of local political control, communities were unaware of the benefits of preventive services such as immunization as compared to curative services that have more visible benefits.

The rationale for decentralization in developing countries is commonly justified on basis of the “proximity principle” such as involvement of local community in public decisions. Bardhan (2002) argues that the traditional fiscal federalism literature cannot be applied to decentralized systems in developing countries due to differences in their institutional design.5 Therefore, local governments in developing countries are likely to behave in a different manner as compared to those in developed countries due to differences in institutional (structure of incentives and organizations), administrative and managerial and political factors.

Nevertheless, decentralized provision of health services in developing countries raises questions of whether local governments compete with each other in financing health services. Do local governments alter their public health expenditures positively or inversely in response to those of its neighbors? A few recent studies (Granado et al. 2008; Caldeira et al. 2010) discuss inter-jurisdictional competition in developing countries.

Granado et al. (2008) discuss yardstick competition as a significant form of fiscal competition in Indonesia. Caldeira et al. (2010) apply fiscal federalism literature to a developing country in Africa, Benin and identify fiscal interactions among communes that are geographically close to each other or with a similar ethnic structure. A positive and significant strategic interaction was found among communes, particularly in election periods.

There is limited empirical evidence on interdependence of health expenditures of local governments. Although a few studies (Moscone et al. 2005; Joan and Pons-Novell and Pons-Novell and Pons-Novell 2005) investigate spatial dependence of health expenditures for developed countries, there is almost no similar work done for developing countries. Therefore, this dissertation aims at filling in this gap by using the Philippines, as a case study of a developing country that has devolved its health care responsibilities to its local governments.

This paper is divided into 7 sections. Section 3 provides an overview of the administrative and political structure of local governments in the Philippines. It also briefly introduces the reader to the range of fiscal responsibilities of local governments specifically in the health sector after the country implemented its decentralization code in 1991. The main research question is stated in Section 4, followed by a description of the

4 The differences were on account of different central-local relationships and patterns of community participation in countries. 5 For instance the Tiebout hypothesis will not hold true in developing countries because people are not so mobile to induce interjurisdictional competition among regions.

6

model and methodology used to address it. Data is analyzed with the help descriptive statistics in Section 5. Regression results and its interpretation are a part of Section 6. The final section concludes the paper with some policy implications. 3. Background 3.1. Philippines – Geography, Politics and Economy

The Philippines are a collection of 7000 islands that are divided into three groups specifically Luzon, Visayas and Mindanao. Its total population of 89 million (Census 2007) is comprised of people from diverse cultures and ethnicities. The country's population is predominantly young, with the 0-14 years age group representing 33.8 percent while the elderly make up about 4.4 percent of the total population (WHO).

The Philippines was under an authoritarian rule of Ferdinand Marcos from 1965 to 1986. At the end of the military rule, a new Constitution was ratified in 1987 that restored a pre-martial law constitutional system consisting of a President and Vice-President, bicameral legislature, and independent Supreme Court.6 Political institutions in the Philippines are described as weak and political relationships as “clientelist”(Abinales and Amoroso 2005).

Although a middle-income country with a GNI per capita income of $3690 (for 2007 based on PPP) it has large spatial variations in income distribution, poverty incidence and social indicators.7 While its progress in reducing infant and child mortality has been significant, it has not been very successful in improving maternal health. The Philippines’ double epidemiological burden of rising fertility and high incidence of communicable and non-communicable diseases makes public health a challenging issue for its government.89 3.2. Administrative structure under the Local Government Code of 1991 The Philippines is divided into 15 administrative regions and 3 special regions namely Metropolitan Manila Development Authority (MMDA), Autonomous Region of Muslim Mindanao (ARMM) and Cordillera Administrative Region (CAR).10 The nation-state is further divided into following political sub-divisions:

• 81 provinces • 120 cities • 1511 municipalities • 42,008 barangays or villages Under the Local Government Code of 1991 (LGC, 1991), the creation of

municipalities and cities was based on total income, population and land area (See Table 1 for the classification standards). Cities are classified into highly urbanized (33), independent component (5) and component cities (82). While they vary based on the

6 Its current President is Gloria Macapagal Arroyo from the LAKAS-CMD Party. 7 The GNI figure is taken from World Bank at http://data.worldbank.org/indicator/NY.GNP.PCAP.PP.CD. 8 Due to its geographical location, the country faces various natural disasters such as typhoons, landslides, volcanic eruptions and earthquakes. 9 The World Health Organization (WHO) classifies the Philippines as one of 22 high-burden countries, and ranks it 8th in the world in terms of estimated incidence of tuberculosis with 320 cases per 100,000 persons. Tuberculosis mortality in the Philippines is 54 per 100,000 persons. (WHO 2007). 10 In every regional capital, each of the 20 government departments has its regional offices.

7

criteria given in Table 1, highly urbanized areas and independent cities are independent of provinces. In addition, residents of independent component cities are prohibited from voting for provincial elective officials.11 Component cities, on the other hand are under the supervision of provinces they are geographically located in.12 Table 1: Classification of local government units Avg. Annual income

(accruing to General Fund)

Population Land area

Province P20,000,000 >=250,000 >=2000 sq kms Municipality P2,500,000 >=25,000 >=50 sq kms Component City P20,000,000 >=150,000 >=100 sq kms Highly urbanized city P50,000,000 >=200,000 3.3. Financial autonomy under the Local Government Code of 1991 3.3.1. Local government revenue and expenditures in the post-decentralization period

Following decentralization, local government units (now onwards will be referred to LGUs) assumed key role in the delivery of public services such as health, education, agriculture, highway and public works maintenance and construction in the Philippines. They enjoy fiscal autonomy and determine composition of local spending, taxing and borrowing to meet local development objectives within the limits of national government guidelines.

The Local Government Code of 1991 stipulated broader taxing powers and increased the size of block grants (i.e. Internal Revenue Allocations). Prior to devolution, 20 percent of the internally raised revenue was distributed to local governments. Under the new decentralization code, 40 percent of the revenue is distributed based on population (50%), land area (25%) and an equal share (25%).13 The aggregate IRA is divided among different local government levels as follows: 23% to provinces, 23% to cities, 34% to municipalities, and 20% to barangays. The share is released directly to the provincial, city, municipal or barangay treasurer.14

In addition to the internal revenue allotment, local government units receive a share of 40 percent of the gross collections from mining taxes, royalties, forestry and fishery charges and other taxes, fees and charges. They also receive a share from the proceeds of government agencies or government-owned or controlled organization (For more details, See Appendix A). The Code also allows LGUs to enter into contracts or inter-local agreements on grants, loans, and subsidies. Additionally, LGUs exercise limited corporate powers of raising revenue by floating bonds in the private market.15

11 Residents of highly urbanized cities also remain excluded from the right to vote for elective provincial officials. However, qualified voters of cities who acquired the right to vote for elective provincial officials prior to the classification of cities as highly urbanized after the ratification of the Constitution and before the effectivity of the Code, shall continue to exercise such right (See Section 453). 12 If the component city is located within the boundaries of two or more provinces, it is considered a part of that province of which it used to be a municipality. 13 It is not weighted by “need” of each LGU, for instance mortality rate or socio-economic status. 14 The share is released on a quarterly basis within 5 days after the end of the each quarter. 15 Borrowing cannot exceed 20 percent of its own source revenue.

8

On the expenditure side, local governments exercise almost complete discretion in utilizing its internal revenue allotment, after setting aside 20 percent as development fund (ADB 2009). The total LGU spending doubled from 1.6 percent of GDP in 1985-1991 to 3.5 percent in 1992-2007. Similarly, the average share of LGUs in government expenditure net of debt servicing rose from an average of 11.3 percent before LGC to an average of 24.2 percent after (Shwartz et al., 2002).

3.3.2. Local Government Health Expenditures and its Composition

Among the LGUs devolved functions, between 1991 and 2007, the ratio of local government expenditure to GDP increased significantly for the social sector, particularly the health sector. While the total health expenditure constitutes 3 percent of the country’s GDP, the share of the local government aggregate expenditure on health rose from 0.08 percent in 1991 to 0.30 percent in 2007.

In 2007, around 80 percent of the total health care expenditure in the Philippines was on personal health care. The remainder was on public health and other costs related to research, training and general administrative and operating costs. It is quite clear from Figure 1 that private out-of-pocket expenses comprise a substantial share (67 percent) of total personal health costs. While the national and local governments including social insurance account for about 23 percent, private insurers, Health Maintenance Organizations (HMOs) and employer-based plans account for the remaining.

Figure 1: Health expenditure by source and use Source: National Health Accounts, 2007, Philippines.

The primary agents of public health spending in the Philippines are local government units (LGUs) that spend around 65 percent of the total expenditure, while 13 percent is contributed by the national government. About 18 percent more is spent in form of foreign contributions. Research and training costs are entirely covered by the national government. Local governments contribute approximately 40 percent of the total miscellaneous costs towards administration and operations.

9.66 13.19 22.470.0818.09 0.27

4.30

64.52

39.49

9.23

10.5767.10

9.6127.20

0.02 4.20

0.00

20.00

40.00

60.00

80.00

100.00

120.00

PERSONALHEALTHCARE

PUBLICHEALTHCARE

OTHERS

RESTOFTHEWORLD

OTHERPVTINSURANCE,HMO'SEMPLOYER­BASEDPLANSANDPRIVATESCHOOLS

PRIVATEOUT­OF­POCKETEXPENSES

SOCIALINSURANCE

9

Local public health expenditure for cities and municipalities for 2007 include expenditure on health programs such as medical, dental, health services, planning and administration of nutrition population and family planning programs. The main components of public expenditure are:

o Personal services (PS) that include salaries and wages, personnel benefits contributions).

o Maintenance and other operating expenses (MOOE) covering supplies and materials, utility expenses, advertising, repairs and maintenance, communication, traveling and professional services. And

o Capital outlays 3.2. Role of federal and local governments in the health sector under Local Government Code

Table 2 compares the degree of devolution in the health sector in the Philippines with a theoretical benchmark in a system of devolution. It is clear that while the local governments do not enjoy any statutory powers in passing legislations of national interest, setting regulations and health standards and enjoy minimal powers in training, they exercise significant powers in management, planning and resource allocation. They do not enjoy complete fiscal autonomy in raising their own revenue but enjoy significant latitude in expenditure decisions. Appendix B briefly states the key stakeholders involved in health policy-making at the local level in the Philippines. Table 2: Decentralization of functions Functions Theoretical

benchmark in a system of devolution

Philippines’ health system

Legislative ** --- Fiscal **

• Revenue-raising ** * Regulation ** --- Planning and Resource Allocation ** ** Management **

• Personnel ** ** • Budgeting and expenditure ** ** • Maintenance ** **

Intersectoral colloboration *** Interagency coordination ** Training ** * Notes: *** - Extensive responsibilities** -- Some responsibilities; * -- Limited responsibilities; -- No responsibilities Source: Mills Anne (1990)

Table 3 lists the functions performed by the local and central governments in the health

sector in the post-devolution period. The central government retains monopoly power in passing health legislations, setting regulations and health standards, providing health education and training medical personnel. It also provides personal care services in

10

specialized tertiary-care hospitals managed by the Department of Health (DOH)16. Provinces manage provincial health offices, provincial hospitals, and municipal and district hospitals that were previously managed by the central government. They provide primary and secondary health services.

Local governments including cities and municipalities provide primary health care services via primary health programs, maternal and child health, and communicable and non-communicable disease control services. They are responsible for procuring, storing and financing drugs, medical supplies and other equipment. Although vertical programs funded by international agencies are managed and controlled primarily by DOH, they are implemented with the help of municipalities and cities. Some recent donor-funded maternal and child health programs were directly administered by LGUs (Philippines Development Institute 2009). Table 3: Functions of central and local governments After devolution Type of care

Central government Health policy, regulation and quality assurance, training and technical supervision, Medical schools, Regional hospitals

Tertiary care

Local Government Units

Provinces Provincial/municipal or district hospitals Primary/Secondary care

Municipalities Rural health units Primary care/ Functions include delivery of maternal and child health programs, control of communicable diseases services, drug procurement, inventory, dispensing and financing

Cities City health offices/ Hospitals in Highly Urbanized Cities

Primary care/Secondary care

Barangays Barangay health stations Primary care

Source: Author compiled from several sources. 4. Research Design 4.1. Research Question Against the backdrop of the newly assigned responsibilities to local government units of the Philippines, this study aims at identifying any potential fiscal interactions in health spending for 2007 through a spatial econometric model. The main purpose of this paper is to answer the following research question:

• Is there an effect on public health spending of a region caused by fiscal spillovers from neighboring jurisdictions

o It is hypothesized that local government units, (in this case, municipalities/cities) alter their health spending in response to their neighbors’ spending?

16 DOH has retained operation of 47 hospitals (National Capital Region hospitals, Special hospitals, Regional hospitals, and Specialty hospitals) and has added 20 more by reclassifying some provincial hospitals into regional or national centers.

11

4.2. Theoretical Model The theoretical model used to answer the above research questions is borrowed from existing models based on consumer utility and government problem of indirect utility maximization in the presence of fiscal spillovers (Case et al, 1994; Albert Sole Ole, 2005, Joan and Pons-Novell, 2005).

Consumer utility is a function of private goods and public services consumed and tastes and preferences. In the presence of spillovers, consumer utility is additionally derived from public services provided in neighboring regions. Since local governments take these spillovers into consideration while supplying its residents’ services, consumer utility (V) is given to be: (V) = fi (xi, zi; zj, mi) - Eq. 1. where xi = set of private composite goods zi = public health good supplied by its own government zj = public health good supplied by the neighboring government mi = other structural exogenous variables such as demographic structure, individual preferences

A consumer’s private consumption is subject to its personal income, net of taxes and contributions. Consumer’s budget constraint is given as

Yi = xi – [(ti) bi - gi] - Eq. 2. xi = personal income ti = individual tax base bi = tax rate

As municipalities rely largely on inter-governmental transfers such as internal revenue allocations, the implicit tax price of a given public good is less the grant received from the central government.

Total public health expenditure of a given region may be expressed as a function of the total population of the given region (i) and the neighboring region (j); and per capita expenditures of region i and j.

(pi) = f (Ni*zi, (Nj*zj)β, mi) - Eq. 3. where zi and zj represent the per capita health expenditures incurred by a municipality and its neighboring municipalities respectively. The parameter β measures the magnitude of interaction between municipalities’ policies.

The budget constraint faced by the local government is expressed as Ri = bi(ti) + gi +c - Eq. 4. where gi is the size of the grant received from the central government, c include all other receipts.

12

The indirect utility function (Eq.5)

W=Σ V {f (xi, zi; zj, mi)} - Eq. 5. is maximized subject to subject to the budget constraint of its consumers and the local government (Eq. 2 and 4)to obtain the optimal per capita public health expenditure, i.e.

€

zi* =

∂Vi(xi,zi,z j ,mi)∂zi

and

€

z j* =

∂Vj (xi,zi,z j ,mi)∂z j

4.3. Methodology The methods proposed in the field of applied spatial econometrics are widely used in the estimation of spatial dependence models. This section will take the reader through some important concepts central to the understanding of the estimation methods employed in this paper such as spatial weights, and model specifications before presenting the empirical model of this paper. 4.3.1. Spatial weights

Distance and relative location of a unit with its neighbors in space define a spatial relationship (Anselin 1988). The spatial weight (Wij) describes the relationship of region i with region j in space. It is a non-negative N x N matrix, whose rows and columns correspond to cross-section observations. Commonly, spatial weight matrices are row-standardized, such that its row elements sum up to one.17 The interaction in

€

z between region i and all its neighbors is obtained as a weighted average of interactions given by,

€

Wij∑ z j where every neighbor receives an equal weight. The specification of a weight matrix is generally arbitrary, based on some measures of

distance18. A range of matrices is discussed in the literature, based on geographic distance, economic and social proximity.

(i) Geographic distance matrices may be based on contiguity or distance (Anselin

1988). If two regions share a boundary, they are considered contiguous to each other and given a value of 1, and 0 otherwise. Alternately, neighbors sharing a common edge or common vertex analogous to the rook case, bishop case and queen case in chess game define contiguity.

(ii) The length of a shared boundary or a specific radius of distance from neighbors also describes proximity. Distance-based weight matrices of an inverse-distance weighting scheme are based on the concept of distance decay

17 This is not a requirement but facilitates interpretation of the model coefficients. It is not recommended that standardization be carried out automatically. As the weights are based on an inverse distance function that is theoretically justified, scaling the rows might result in loss of interpretation. 18 Revelli (2003) specifies that spatial weights should be defined based on a theoretical model and sample size. A parsimonious determination of the weights should be preferred such as binary weight matrix rather than more structured matrices.

13

where wij varies inversely with dij implying that a spatial relationship is weaker as distance between units increases. A cut-off distance Dij could also be specified to define this relationship.

(iii) Departing from the contiguity criterion, spatial weights may also be based on a socio-economic distance defined through similarities in terms of socio-economic characteristics. Interactions could also be facilitated by non-physical distance in the presence of well-developed transportation or telecommunications networks (Aguiar et al. 2001).

This paper uses a distance-based queen contiguity matrix to define neighborliness of local governments. The rationale behind the use of geographical distance is the nature of health as a good and the country’s topography. In developing countries where transportation networks are not well developed, residents especially those located in more scattered islands of the Philippines have limited mobility. Unless they have to seek specialized curative care, they may not choose to travel long distances but rather seek access from immediate neighbors.

An additional complexity faced in the use of spatial econometric models is an endogenous dependent variable. Endogeneity refers to the simultaneity in the relationship between spatial neighbors. The expenditure of neighbor i affects the expenditure of neighbor j and expenditure of neighbor j also affects expenditure of i. This results in the dependent variable being correlated with the error terms. In order to obtain consistent estimates, a simultaneous estimation procedure is recommended such as instrumental variables (IV) or non-parametric maximum likelihood estimation (MLE) approaches.

Consistent with previous studies (Revelli 2003), the spatial lag model is estimated using a two-stage least square estimation where the first stage involves identifying the relevant instruments. The use of spatially lagged exogenous variables as instruments is suggested in literature (Anselin 1988). 4.3.2 Spatial Models Anselin (1988) presents a framework of four traditional model specifications to test for spatial dependence for a cross-section of spatial units at a given point in time. Spatial dependence is tested for both, lagged dependent variable and error terms through spatial autoregressive models listed in Table 4. Table 4: Spatial model specifications in cross-section data Model Name Specification (in matrix notations) 1. Classical Linear Regression model with no spatial effects

€

zi = Xβ + e 2. Mixed Regressive-spatial autoregressive model

€

z = ρW1z + βX + e;e N(0,1) 3. Linear regression model with a spatial autoregressive model

€

z = βX + e;e = λW2e + uz = βX + (I − λW2)

−1u,u N(0,1)On substitution,

4. Mixed regressive spatial autoregressive model with a spatial autoregressive disturbance

€

z = ρW1z + βX + (I − λW2)−1u,u N(0,1)

5. Empirical Model and Data Analysis 5.1. Empirical Model

14

The empirical section in this paper focuses on estimating the slope of the reaction function

€

β in the spatial autoregressive model with a normally distributed error term (Model 2)(Brueckner 2003). Here, the interdependency in fiscal policies of regions i and j municipalities is represented through a mixed regressive spatial autoregressive model, where per capita public health expenditure (zi) function of a region is a function of zj (lagged per capita health expenditures).

€

zi =α + ρ W1∑ z j + βXi + e -- Eq. 6

If an increase in expenditure of jurisdiction j (

€

z j ) raises the marginal utility of

€

zi the reaction function will have a positive value, and the reaction function will be upward sloping. Conversely, if an increase in

€

zi decreases the marginal utility of

€

zi, the reaction function will be downward sloping. 5.2. Choice of control variables Health care is a core component of human capital investment and catalyst for economic progress of a country. Health care spending is associated with improved health outcomes, leading to higher workforce productivity and social welfare (Okunade and Murthy 2009).

The dependent variable in this analysis is per capita health expenditure. Traditionally, health expenditure was studied as a function of economic, political and demographic factors.19 Table 4 lists some of the previous empirical studies that have analyzed the determinants of public health expenditures. In addition, a few studies (Font and Novell; Moscone et al. 2005 and 2007) investigate the spatial component of expenditures.

Table 4: Determinants of public health expenditures in developed and developing countries

Costa-Font and Pons-Novell

Moscone et

al. Moscone

et al. Matteo &

Matteo Murthy & Okunade

Variables

2005 2005 2007 1998 2009 Dependent Variable Real Per Capita Health care expenditure

√ √ √ Mental health expenditure √ √ Explanatory Variables

Socioeconomic Per capita real GDP/SDP/Regional

Income

√ (*) √ (**) √ (**) √ (**) Unemployment Rate √ (**) Education level √(*) Marital status √ (**) Fiscal Variables Inter-governmental grants √ (**) Macro-economic Variables Per capita Foreign aid* √ (**) Demographic Variables Population √

Population by gender √

19 A few cross-country studies have attempted to study the relationship between technological progress and health care expenditure (Okunade and Murthy 2002; Di Matteo 2004).

15

Population density √ (**) √ (**) Percent of population aged over 65 years

√ (**) √ √ (**) √ (n.s) Ethnicity (Asian) √ (**) Percent of households headed by lone

parent

√ (**) Health status indicators

% people with long-term illness √ (***) Mortality Rate (Adult/Maternal) √ (**) √ (n.s)

Health Infrastructure Physicians per 1000 people √ (n.s) Use of Health inputs No. of doctors per capita √ (**) No. of hospital beds per capita √ (**) Political Variable

Regional and national incumbents of same party

√ (n.s.) Share of left-win MPs √ (n.s.) √ (**) Spatial Dependence √ (**) √ (**) √ (**)

Econometric method employed Spatial lag

Spatial lag and error

Pooled, RE,

Spatial Lag RE, Spatial

Error RE

Pooled time series

OLS, 2SLS, Robust LAE

Level of geographic aggregation Region-states Local Local councils

Provincial National

Region/Country of Analysis Spain

England England Canada African sub-

continent R2 75% 92% 84%

* Augments country’s GDP resource base and therefore crucial in case of developing countries

Income: Economists contend that income is an important determinant of total health care spending in a country (Newhouse 1977; Parkin et al. 1987; Okunade and Murthy 2009). The classical health economics literature renders estimation of elasticity of real per capita health expenditure as a method to determine whether health care at the margin is a necessary or a luxury good.20 Intuitively, although health care may seem more as a necessity than a luxury (Culyer 1988, pp. 20), there is mixed empirical evidence to support it.

In conformity with the Wagnerian hypothesis that expenditures increase with an increase in income, several empirical studies re-confirm the positive influence of regional per capita income on per capita health expenditures of states and local regions (Case et al., 1993; Matteo & Matteo, 1998; Costa-Font and Pons-Novell, 2005; Moscone et al. 2007). Due to unavailability of data at municipality-level, regional income per capita is used as a proxy for municipality or city income.

Total revenue per capita comprising of own-source revenue and central government transfers is a fiscal control variable used in the analysis. The total own-source revenue of municipalities and cities is derived from local and non-local sources. Own-source revenue comprises mainly of taxes and user fees charged to residents, while some cities

20 If measure of elasticity is between 0 and 1, it means health care is a normal good, if it is more than 1 it denotes health as a luxury good. Health care as a normal good would mean that it is meeting the basic physiological needs of the people while a luxury good (mostly in developed countries) suggests higher spending to be indicative of higher wealth or generous health benefit insurance coverage plans, shielding patients from incurring ‘full resource cost of care’ and thus leading to moral hazard in consumption, or high-cost technological change. (Murthy and Okunade, 2009).

16

raise additional revenue by floating bonds, debentures, and securities in the market, under the vigilance of the Central Bank. As compared to municipalities, cities also stipulate taxes at a much higher rate. Inter-governmental transfers: Federal transfers to local governments constitute a major component of total revenue with the primary purpose of meeting expenditure needs that cannot be met through own-source revenue of local governments. Generally, countries use three methods to establish vertical share entitlement in the form of unconditional or conditional transfers based on recommendations made by a Committee (India) or cost of provision. Another approach is of revenue- sharing, where the total national income is distributed among local governments based on a formula. Countries such as China, Russia, Mexico, Brazil, Argentina, Indonesia and Philippines, use the latter approach (Economic Review of Kerala, 2003).

In the Philippines, government transfers are of three types: (i) formula-based block grants (i.e. Internal Revenue Allocations (IRA)) (ii) origin-based share in national government revenues (share in national wealth and other taxes) and (iii) ad hoc conditional grants.21

Studies illustrate influence of inter-governmental transfer payments to local government health expenditure in decentralized settings (Matteo and Matteo, 1998; Costa-Font and Pons-Novell, 2005).22 This analysis controls for the effect of IRA as a percent of total income.

Demographic characteristics – Size of the population/ Composition of the population: Population size tests for economies of scale incurred while providing health care services. Previous studies show a decline in per capita local government expenditure in response to rising population, indicating presence of scale economies (Albert Sole Ole 2005; Revelli 2003). Densely populated regions may also accrue scale economies, along with congestion costs. Therefore population density is also controlled in this study. Composition of population: Whether level of health care spending is commensurate with needs or preferences of people is another question often raised in analyzing health care expenditures? A number of studies indicate that health care expenditure trends are associated with aging of the population (Culyer 1988; Matteo and Matteo, 1998). Further, certain segments of population such as women and infants are more susceptible to communicable diseases, medical complications and other illnesses. Healthcare expenditure, including medical or non-medical care expenditure thus may vary based on age structure of the population.23

21 Conditional grants may originate from lump-sum allocations under the central government budget, allocations by central government agencies from their own budget, and conditional or line item appropriations for pork barrel funds of legislators. 22 It is possible that local governments may strategically misrepresent their true needs and preferences in order to secure more than their fair share of resources from the central government. 23 The recent distribution of limited swine flu vaccines in response to the outbreak of swine flue in the United States from the federal government to the states was based on the total size of states’ population. Higher priority was given to more vulnerable sections of population such as pregnant women, people living with children younger than 6 months of age, healthcare and emergency medical services personnel, persons

17

One measure of population composition measured by percent of persons over the age of 65 is a control variable in one of the regression models. Due to unavailable data for all municipalities and cities, these variables are excluded from the spatial analyses.

Poverty Incidence: Poverty is a complex phenomenon with many dimensions, including insufficient access to housing, food, nutrition, health, education and leisure (Sen 1985). Traditionally, its statistical measure denotes the shortfall of income or consumption from a given threshold such as poverty line. The measure of poverty incidence used in this study is defined as the proportion of people living in an area with an average per capita expenditure below the poverty line.2425 Political variable: Romer and Rosenthal (1979) argue that public allocation process is a bargaining game between a politician-bureaucrat – a “typical” state-local official and resident voter. It is not just the preferences and a budget constraint that determine local fiscal allocations but bargaining between politicians and voters also matters.

Among western countries, it has been shown that voters may have an influence on government spending by organizing into interest groups. Mueller and Murrell (1990) show that although interest groups do not determine the size of the government per se, they may have an impact when the government does not have sufficient control over tax share determination.

There is limited literature on political ideology impacting health care spending. Navarro and Shi (2001), Navarro et al. (2006) contend that policies followed by social democratic and Christian democratic parties are aimed at reducing social inequalities and providing higher welfare and health services as compared to conservative parties.

Political parties in the Philippines are weak political structures and unable to perform efficiently as intermediate agents between state and society. Historically, small factions of elites, competing for the central government pool of resources, have dominated local politics in the Philippines.

A distinct characteristic of Philippine political parties is their shifting character of membership due to lack of ideological differences between parties. However, the church and religious groups play a dominant role in influencing political decisions of masses. The influence of church over policy issues such as family planning during Marcos-era, Aquino’s and Arroyo’s presidencies is well documented. Arroyo cut the DOH budgets in 2003 for purchasing artificial contraceptives and announced support to Catholic Church’s natural family planning methods (Abinales and Amoroso 2005).

Patronage through discretionary funds that politicians seek is the most visible form of corruption in the government in the Philippines. The quote by US Congressman “Tip” O’Neil “All politics in the Philippines is local” describes the locus of present-day congressional power in the Philippines. The President, Vice-President and Members of the Congress are elected with the help of a local network, often “[c]entered on the family in the age group of 6 months and 24 years old, and persons over 24 with chronic health disorders or compromised immune systems. (CDC, US) 24 Logged average per capita household income and expenditure were predicted by combing survey data with auxiliary data obtained from Family Income and Expenditure and Labor Surveys, and the poverty measure was derived from them (Small-area Local Poverty Estimates 2005). 25 Poverty line is estimated to represent monetary resources required to meet basic needs of members of households along with a non-food component.

18

and extended through their district by alliances and patronage” (Abinales and Amaroso, 2005 p. 15). Congressional members from provincial districts nurture these networks through “pork-barrel funds” disbursed by the President.26

Political alliances between the President and provincial governors have translated into benefits to the province due to reciprocity and granting of favors (personal and professional) between the two parties. For instance, in 2000 President Estrada conferred a higher share of tobacco excise tax to the Ilocos Sur governor, a crony of the latter in exchange of $8 million from illegal gambling (Abinales and Amoroso 2005). During election times, Presidential candidates rely on local politicians to garner local support.

In contrary to previous studies that measure the influence of political parties on health care spending, this paper does not consider political ideology due to weak ideological affiliation of politicians to political parties. Instead, it captures the influence of pork barrel funding by including a dummy variable that measures whether the municipality/city mayor and their congressional district members belong to the same party. It is denoted as 1 if they belong to the same party and 0 otherwise.2728

Spatial Interactions: The main variable of interest in this paper is the spatial variable that is meant to capture the influence of neighboring governments’ expenditures on that of a given municipality, after controlling for the above structural variables.

A number of country-level studies recognize the importance of spatial dependence to explain unobserved variation in health expenditures among countries. Consumption of health resources in a given country is found to be associated with unobserved population characteristics of its neighboring countries or diffusion of diffusion of technology across countries (Carrion-i-Silvestre 2005; Wang and Rettenmaier 2006; Chou 2007; Baltagi and Moscone 2010).

Spatial interactions are pointed to be an important consideration in the empirical models of health expenditure determination among local governments, particularly for developed countries (Costa-Font and Pons-Novell 2005; Moscone et al. 2005, 2007). Biased results from exclusion of spatial dependence were confirmed through a few empirical studies.

As discussed in the previous sections, in context of health care externalities fiscal spillovers in financing public health are most likely to occur among local governments while providing maternal and child health services like immunizations, family planning or control of infectious diseases like malaria, dengue fever. Although the central government retains control in the provision of specific donor-funded preventive reproductive health programs such as family planning, local governments administer these vertical programs. Despite “Comprehensive Healthcare Agreements” set by the central government, it is not always the case that local government complies with these provisions. Moreover, there is no punitive mechanism enacted against local governments that do not comply with these national standards. Second, there is evidence of local

26 A key characteristic of pre-modern Philippines’ politics was concentration of power of local leaders called datus. The political community was defined by personal relationships maintained through networks of personal loyalties and marriages, and not territorial boundaries (Abinales and Amaroso 2005). 27Gloria Macapagal Arroyo was the elected President in the 2004 Presidential election and remained in office during the entire 6-year period. 28 Congressional members are elected from each district of the province.

19

pressures (e.g. from the Catholic Church) that pose obstacles to the implementations of certain programs like family planning (Lakshminarayan 2003). Fiscal spillovers among governments may also arise in the provision of curative health services, specifically in presence of cross-border utilization of medical services at hospitals run by provinces due to inter-provincial variations in quality of services. In a system of competitive local governments, the latter may indirectly influence long-term decisions of people by discouraging patients with chronic care from settling in their jurisdictions through provision of poor facilities, difficult access or poor reported outcomes.

In addition, competition between provinces may occur if some provinces have their own insurance or health delivery program.29 Malagag in Davao del Sur and Sebaste in Antique started their own programs of health service delivery in their respective municipalities. Innovative local government initiatives in health service delivery included charging graduated user fees to users based on their income levels and transforming a rural health center into a community clinic by tapping external sources of funding. Besides benefiting its own residents, the community clinic in Sebaste also served residents of neighboring municipalities.

Differences between regions in quality of health services may partly arise on account of the system of drug procurement, inventory, dispensing and financing of each local government unit. This may translate into differences in quality, price and supply of locally procured drugs and in turn level of health service (Lieberman et al.). 5.3. Data Table 5 lists the dependent variable, independent variables, and expected signs of coefficients.

Based on theory, the influence of total revenue per capita and IRA are expected to be positive. Larger amounts of local resources at the disposal of a LGU are likely to drive per capita health expenditures to increase. Similarly, higher per capita expenditures could be the result of larger grants received from the central government.

The influence of the need-based variable specifically percent of old people in the total population is expected to be positive. Lower per capita health expenditures in more populous LGUs implies economies of scale. However, depending on congestion costs, the sign of population density on expenditure maybe either positive or negative.

Theoretical and empirical evidence shows that regional income is more likely to have a positive influence on health expenditures. In the presence of inter-governmental grants, poorer regions associated with larger health needs but limited resources may either have a positive or negative relationship with health expenditures. Additionally, access to pork-barrel funds is likely to have a positive influence on the latter.

Theoretical essays on fiscal decentralization and its impact on health care expenditures, mostly in context of developed countries discuss the potential negative spillovers while empirical studies show evidence of positive spatial dependence. Previous studies of fiscal spillovers in general public services, specifically in developing countries identify positive spatial dependence among neighbors.

29 Bukidnon and Guimaras provinces have their own insurance programs. Some other cases include health card system of Paranaque city and Community Primary Hospital Program of Negros Oriental (Lieberman et al.).

20

Table 5: Data Sources Variables Data Sources Expected signs of

coefficients Dependent variable Per Capita health

expenditure Bureau of Local Government Finance

Independent variables Fiscal Variables Bureau of Local

Government Finance

Revenue per capita + % of IRA in total income + Socio-economic variables Census, 2007 Local Poverty Index Small-area Local

Poverty Estimates Survey, NSCB, 2005

+/-

Regional income + Population - Population density -/+ Percent of population over

age of 60 +

Political Variable Pork-barrel measure COMELEC + Spatial variables Municipality and City

Shapefiles • PhilGIS30 +/-

The primary source of fiscal data is the Bureau of Local Government Finance Report

for year 2007. The remaining data on socio-economic variables are obtained from several sources including the Bureau of Census and National Statistical Office, Philippines. The political parties’ information is obtained from results published by Commission of Election (COMELEC) for year 2004. 5.4. Descriptive Statistics Table 5 presents the descriptive statistics of all variables used in this study. The average per capita expenditure on health by a municipality and city is P127 Peso and P185 Peso respectively.

While it is clear that the average own-source revenue of municipalities is as low as 13 percent, it is approximately 40 percent for cities. IRA makes up a significant share of the total revenue of municipalities and cities (84 and 61 respectively) percent.

The average regional per capita income of cities (P 11,369) is much higher than that of municipalities (P 16,091). As discussed earlier, the country has a low life expectancy. Its mean share of population over 65 years of age is quite low.

Figures 2,3 and 4, map the geographic distribution of per capita health expenditures for municipalities through choropleth maps. It is evident that the variable tends to distribute in clusters with the highest concentration in the top-most Luzon region and the southern-most Mindanao region, the former including the National Capital Region clearly displaying much higher average per capita expenditures than the Visayas (middle) and

30 http://www.philgis.org/freegisdata.htm. Some other sources are http://www.asian.gu.edu.au/price.html#thephil; National Mapping and Resource Information Authority

21

Mindanao regions. The spatial variation in the distribution of poverty incidence is stark, with the Mindanao and Visayas regions lagging far behind the Luzon region. Table 5: Descriptive Statistics Variables Obs Mean Std. Dev. Min Max Health expenditure pc

Municipality 1,479 127 105 0 1,328 City 119 185 132 18 1,060 Total revenue pc Municipality 1,479 1,930 5,763 0 218,071 City 119 3,234 1,589 1,136 14,275 Regional income Municipality 1,479 16,091 4,115 3,582 40,252 City 119 11,369 10,018 3,582 40,252 Percent of old Municipality 605 8 2 3 15 City 54 6 1 3 9 Poverty incidence Municipality 1,479 24 16 0 0.89 City 119 47 16 0 0.60 Percent of IRA in total revenue

Municipality 1,479 84 14 0 100 City 119 61 23 7 95 Population Municipality 1,479 37,996 35,714 114 556,330 City 119 265,375 340,383 32,790 2,679,450 Population density Municipality 1,479 967 1,762 0 229 City 119 10,442 22,260 30,349 172,188

22

Figure 1. Spatial variation in per capita health expenditures

Figure 2. Spatial variation in grant sizes

23

Figure 3. Spatial variation in poverty incidence Note: ARCGIS software was used to make the thematic maps. 6. Regression Results A total of 1466 observations including 1346 municipalities and 120 cities comprise data for the regressions. The dataset excludes the stand-alone municipalities without any neighbors. To induce normality, all variables are transformed into logarithmic forms and therefore the coefficients are interpreted as elasticities (except Model 3). Model 1 presents OLS regression results of a sub-sample of 634 observations, for which data on elderly population was available. Share of elderly population as an independent variable was positively associated with health expenditures, but not statistically significant.31

As a first step, an OLS regression is run to identify any spatial dependence in error terms. The spatial statistics observed were Moran’s I, Robust Lagrange Multiplier (LM) lag and error estimates. The Moran’s I estimate of 7.6519 was positive and significant, indicating a strong spatial positive autocorrelation of residuals. The robust LM spatial lag (53.4842) and error (99.3624) estimates were also both positive and statistically significant. 31 The coefficients for women and children were negative and positive respectively, tested for a restricted sample. There were not statistically significant. The coefficient signs may vary depending on differences in health services offered by municipalities or cities and provinces. The former provide primary health services while the latter (provinces) hospital services, thus catering more to women or children than older population. Therefore in the municipality analysis, the coefficient for women or children may be expected to be higher than for older population.

24

Fiscal interactions in a decentralized system are well documented in the fiscal federalism and health economics literature. Therefore, the choice of a spatial autoregressive model with a spatially lagged dependent variable may be justified based on economic theory and intuition in this analysis.

The regression results, correcting for endogeneity appear under Models 3 and 4. The reaction coefficient

€

ρ is consistently positive and highly significant in Models 3 and 4, although its strength reduces from 0.3256 to 0.1541.32 The interaction coefficient is consistent and positive, after controlling for any common provincial effects in Model 4. The Anderson cannon LR and Hansen J statistics indicate that the instruments used in the first-stage regression of the two-stage least square estimation were appropriate and the model is identified.

As for the remaining regressors, the signs of coefficients were as expected. Higher population is associated with lower per capita expenditures, implying economies of scale.33 Population density is positive but insignificant. All the fiscal variables have a positive and significant impact on per capita health expenditures. Poverty incidence is negatively associated with expenditure but regional income influences expenditures positively. The political variable that measures the pork-barrel effect is positive and significant. This may imply that political actors, at the local and national levels may influence the availability and use of funds. Table 5: Regression Results

Dependent/Independent variable

Model 1 (OLS)

Model 2 (OLS)

Model 3 (non-log form)

Model 4 (IV)

Regionalincome 0.00379** 0.1681*** 0.00204*** 0.21114*** Poverty incidence -0.6408* -0.0480** -27.222 -0.8057** Logpop -0.04386* -0.0515*** -8.4843** -0.0714*** logpopdensity 0.0229 5.9790 0.01321 Log elderly 0.0522 Logtotrevpc 0.7515*** 0.6882*** 142.6287*** 0.7354*** IRApertot/logIRApertot 0.2381** 0.2210** 0.7576*** 0.21898*** Mayeqscong 0.0605** 0.0864** 9.0316*** 0.0932***

€

ρ 0.1120*** 0.3256*** 0.29834*** 0.15416*** Province effects clustered

No Yes

Sample size 634/437 1466 1539 1466 R2 0.6211 0.5240 0.5931 0.51 Moran’s I 7.6519*** Robust LM test lag 53.4842*** Robust LM test error 99.3624*** Anderson canon LR statistic

1062.830*** 797.197***

Hansen J statistic 2.343 (0.8856)

6.635 (0.35)

32 Spatial interaction was interacted with fiscal autonomy to test whether spatial interactions are higher among those LGUs with more fiscal autonomy. And there was no evidence in support of it. 33 An alternative explanation for the negative coefficient of population could be that the revenues to do rise commensurate with population or fiscal need, thus indicative of a fiscal gap (Manasan 1998).

25

A strong and positive spatial dependence in public health expenditures is a key finding of this study. An increase of one percent of neighboring local government expenditure causes an increase of 0.154 percent in the level of health spending of a given region. This finding is consistent with the previous studies of spatial dependence in general public expenditures found among local governments in developing countries such as Indonesia and Benin.

Several explanations can be given to justify the above result.34 One potential causes of positive spatial dependence is bidding by municipalities for medical supplies such as drugs and specialized personnel like doctors. Competition between municipalities drives cost of health inputs to rise and induces strategic interaction in spending. Municipalities in the Philippines are responsible for procuring, storing and financing their drugs and other medical supplies. Lieberman et al. (2005) discuss the corrupt bidding practices prevalent in the local drug procurement system. “Bids are rigged, qualified bidders are pre-identified, and bidders connive” (p. 15).35 This causes stock shortages, poor quality of drugs and purchase prices to be much higher than private pharmacies.

They also point out to the difficulties faced by local governments in hiring physicians, nurses and medical technicians that are in great demand in foreign markets. Under the decentralized system, the central government requires local governments to provide additional compensation to health personnel including physicians. Local governments may be compelled to out-perform other LGUs in attracting physicians by offering them attractive salaries, honoraria such as free board and lodging.

Positive fiscal interactions, as discussed in literature can be on account of be yardstick competition between local governments. However existence of yardstick competition is debatable in young democracies such as the Philippines (Caldeira et al. 2010). Caldeira et al. (2010) find evidence of opportunistic behavior among local governments in the form of positive fiscal interactions, especially during election times. Drawing from their findings, one of the possible reasons for spatial dependence in the Philippines could be that the given year of analysis (i.e. 2007) was an election year for local governments. So in order to be re-elected to power, it is possible that they strategically compete with each other by providing higher public health services to its residents than its neighbors. 7. Policy Implications This chapter has sought to highlight the key implications of decentralizing health care that may be relevant to policy-makers seeking to identify optimal decentralization policies.

Due to data limitations, an association between health expenditures and outcomes was tested for cities only and was not found to be significant.36 If higher health expenditures

34 Overall health outputs produced in presence of negative spillovers may be optimal due to the nature of health externalities based on altruistic theories. An improvement in health outcomes of neighboring governments (that may be caused by higher health spending in a given region) does not necessarily make residents of a given region worse off. Despite the possibility of free riding by neighboring regions a given region may not alter its health spending. 35 Local governments in decentralized systems seek their estimated supply of vaccines, transportation and other supplies from the central governments that in turn allow local health providers to bid for the vaccines and other supplies. 36 A relevant argument in the Philippines is the control of local governments by the elite that may cause spending to be inefficient and inequitable.

26

do not translate into improved health outcomes, this may suggest that competition between regions is more driven by supply than demand-related factors such as limited resources and infrastructure. This indicates that it may be necessary to empower the LGUs to raise more revenue on their own and have better access to infrastructure. The responsibility of compensating doctors of local governments may require that local governments have either opportunities of raising more revenue or access to larger grants.

There have been recommendations made in the House and Senate about revisiting the formula for allocating IRAs and incorporating factors of fiscal capacity and needs of regions to ensure more efficient use of transfers (Llanto 2009).

Although DOH has created a Local Government Assistance and Monitoring Service to manage transition programs, more stringent enforcement and compliance to national standards may be necessary. There is a need to regulate supply of drugs and medical supplies that is completely driven by local markets. Bibliography Abinales PN, DJ Amoroso (2005), “State and Society in the Philippines”, Rowman and Littlefield Publishers, Inc. Anselin (1988), “Spatial Econometrics: Methods and Models”, Studies in Operational Regional Science, Kluwer Academic Publishers. Asian Development Bank (2009), Report and Recommendations of the President to the Board of Directors, “Proposed Loan for Subprogram 2 Republic of the Philippines: Local Government Financing and Budget Reform Program”. Albert Solé Ollé (2005), “Expenditure Spillovers and Fiscal Interactions: Empirical Evidence from Local Governments in Spain” Bardhan (2003), “Decentralization of governance and development”, Journal of Economic Perspectives. Besley and Case (1995), “Incumbent behavior: Vote-seeking, tax-setting and yardstick competition”, American Economic Review, 85/1, pp. 25-45. Boulier et al. (2007), “Vaccination Externalities”, The B.E. Journal of Economic Policy and Analysis”, 7/1, 23. Brueckner JK (1998), “Testing for Strategic Interaction Among Local Governments: The Case of Growth Controls”, Journal of Urban Economics, 44, pp. 438-467 Brueckner JK (2003), “Strategic Interaction Among Governments: An Overview of Empirical Studies”, International Regional Science Review, Vol. 26, No. 2, 175-188 Buchanan JM (1965), “An Economic Theory of Clubs”, Economica, 32:1-14 Buchanan JM (1949), “A Pure Theory of Government Finance”, The Journal of Political Economy, 57/6, pp 496-505 Buchanan JM and Craig Stubblebine (1962) “Externality”, Economica, New Series, 29/116, pp 371-384 Buchanan and Kafoglis (1963), “A Note on Public Goods Supply”, The American Economic Review, 53/3, pp. 403-414 Bureau of Local Government Finance, Philippines Caldeira E, Foucault M, Rota-Graiosi G (2010), “Decentralization in Africa and Fiscal Competition” Evidence from Benin”, Scientific Series, CIRANO, Quebec, Montreal. Campbell RJ (2004), “Leviathan and Fiscal Illusion in overlapping jurisdictions”, Public Choice 120: pp. 301-329

27

Capuno, J. J. and O. Solon (1996). “The impact of devolution on local health expenditures: Anecdotes and some estimates from the Philippines”, Philippine Review of Economics and Business 33(2): 283-318. Economic Review of Kerala (2003) Case et al. (1993), “Budget Spillovers and fiscal policy interdependence”, Journal of Public Economics, 52, pp. 285-307 Cheema GS, DA Rondinelli (eds) (1983), Decentralization and Development: Policy Implementation in Developing Countries. Beverly Hills, California: Sage Publications. Daniels, Norman (1981), “Health-care Needs and Distributive Justice”, Philosophy and Public Affairs, 10/2, pp. 146-179 Epple Denis and Allan Zelenitz (1981) “The Implications of Competition among Jurisdictions: Does Tiebout Need Politics”? Journal of Political Economy, 89/6 Feilden, R., and F. Nielson. 2001. "Immunization and health reform: making reforms work for immunization. A reference guidc." WHO/V&B/01.44. World Health Organization, Geneva. Levaggi R, P Smith (1994), "On the Intergovernmental Fiscal Game", Public Finance-Finances Publiques. 49(1), pp.72-86. Lieberman SS, JJ Capuno, HV Minh (), “ Decentralizing Health: Lessons from Indonesia, The Philippines and Vietnam” East Asia Decentralizes. Llanto GM (2009), “Fiscal Decentralization and Local Finance Reforms in the Philippines”, Philippines Institute for Development Studies. Matteo LD, RD Matteo (1998), “Evidence on the determinants of Canadian provincial government health expenditures: 1965 --1991”, Journal of Health Economics 17, 211 – 228 Mills Anne et al. (1990), “ Health system decentralization: Concepts, issues and country experiences”, World Health Organization. Murthy VNR, AA Okunade (2009), “The core determinants of health expenditures in the African context: Some econometric evidence for policy”, Health Policy 91, 57–62 Niklas Hanes (2002), “Spatial Spillover Effects in the Swedish Local Rescue Services”, Regional Studies, 36/5, pp. 531-539 Jacobsson F et al. (2005), “Caring externalities in health economic evaluation: how are they related to severity of illness?” Health Policy 73, pp. 172–182 Costa-Font and Jordi Pons-Novell (2005), “Public health expenditure and spatial interactions in a decentralized national health system”, 16/3, pp 291 – 306, John Wiley & Sons, Ltd. Kolehmainen-Aitken RL and Newbrander W (1997), "Decentralizing the management of health and family planning programmes." Management Sciences for Health. Boston, Mass. Processed. Lakshminarayan (2003), “Decentralization and its Implications for Reproductive Health: The Philippines Experience”, Reproductive Health Matters, 11/21, pp. 96-107. Lynn MacDonald (2008) “The Impact of government structure on local public expenditures”, Public Choice 136:3-4, 457-473 Madies T, Paty S, Rocaboy Y (2004), “Horizontal and vertical externalities: An overview of theoretical and empirical studies”, Urban Public Economics Review. Melgaard B (1998) "Immunization and health reform: The implications of decentralization." Paper presented to the Scientific Advisory Group of Experts (SAGE),

28

Geneva, pp. 9-11 June 1998. Document reference GPV-CVI/SAGE.98/WP. 13. Geneva: World Health Organization Moscone et al. (2005), “Exploring the Spatial Pattern of Mental Health Expenditures”, The Journal of Mental Health Policy and Economics, 8, pp. 205-217. Moscone et al. (2007), “Mental health expenditure in England: A spatial panel approach”, Journal of Health Economics, 26, pp. 842-864. Musgrave Richard (1959), “The Theory of Public Finance”, New York, McGraw Hill National Statistical Coordination Board (2005), “Estimation of Local Poverty in the Philippines”. Musgrave, Richard (1997), “Reconsidering the Fiscal Role of Government”, The American Economic Review, 87/2, pp. 156-159. Navarro V, Shi L (2001) The political context of social inequalities and health. International J Health Services, 31: 1–21. Navarro et al. (2006), “Politics and health outcomes,” Lancet, 368: pp. 1033–37 Newhouse JP (1977) “Medical care expenditure: A cross-national survey” Journal of Human Resources, 12, pp. 115-125 National Health Accounts (2007), National Statistical Organization, Philippines. Oates, Wallace (1968) “The Theory of Public Finance”, The Canadian Journal of Economics, 1/1, pp. 37-54 Oates Wallace (1972), “Fiscal Federalism”, NY, Harcourt Brace Jovanovich, New York (1972) Oates Wallace (1999), “An Essay on Fiscal Federalism”, Journal of Economic Literature, 37, pp. 1120-1149 Parkin et al. (1987), “Aggregate Health care Expenditures and National Income”, Journal of Health Economics, 6, pp.109-127 Peyvand Khaleghian (2003), Decentralization and Public services: The Case of Immunization”, Policy Research Working Paper 2989, World Bank. Philippines Institute for Development Studies (2009), “Improving Local Service Delivery for the MDGs in Asia – The Case of the Philippines, Discussion Paper Series No. 2009-34. Romer and Rosenthal (1979) “Education, Welfare and the New Federalism: State Budgeting in a Federalist Public Economy”, Studies in State and Local Public Finance, The University of Chicago Press, pp 187-227 Samuelson Paul (1954), “The Pure Theory of Public Expenditures”, The Review of Economics and Statistics, 36/4,pp. 387-389 Samuelson, Paul (1955), “Diagrammatic Exposition of a Theory of Public Expenditure”, The Review of Economics and Statistics, 37/4, pp. 350-356 Samuelson, Paul (1958), “Aspects of Public Expenditure Theories”, The Review of Economics and Statistics, 40/4, pp. 332-338 Shleifer A (1985), “A theory of yardstick competition”, Rand Journal of Economics 16(3), pp. 319-327. Shwartz et al. (2002), “Decentralization and Local Government Health Expenditures in the Philippines, Measure Evaluation Report. Sinn HW (2003) “The new systems competition”, Blackwell. Oxford” Tiebout: A Perspective”, National Bureau of Economic Research, Working Paper No. 954

29

Starett David (1980), “Measuring Externalities and Second Best Distortions in the theory of Local Public Goods”, Econometrica, 48/3, pp.627-642 Tiebout Charles (1956) “A Pure Theory of Local Expenditures”, The Journal of Political Economy, Vol. 64, no. 5, pp. 416-424 Truman, B (1981), “A Critique of Tiebout’s Theory of Local Public Expenditures”, Econometrica, 49/3, pp. 713-740. Williams A (1966), “The Optimal Provision of Public Goods in a System of Local Government”, Journal of Political Economy, 74/1, pp. 18-33 World Bank Report (1987), “Financing Health Services in Developing Countries”, A World Bank Policy Study WHO Budget Briefer (2007), “Closing the Health Inequalities Gap”, Congressional Planning and Budget Department.

30

Appendix A Table A.1. GENERAL FUND

Tax on Business Provincial/City Impositions

• Business of printing and publications

• Franchise tax • Tax on sand, gravel, and

quarry resources* • Amusement tax on admission • Tax on amusement places* • Annual fixed truck on

delivery trucks or vans * Municipalities/barangays have shares in the proceeds

Other taxes Provincial/City Impositions

• Tax on transfer of real property ownership*

• Professional tax • Other impositions

TAX REVENUE

Provincial/City Impositions Real Property Tax

• Basic Tax** • Special levy

** Proceeds are distributed between cities, municipalities and barangays within the Metropolitan Manila Area (Section 271).

City/Municipal Impositions • Manufacturers, assemblers • Wholesalers, distributors • Exporters or manufacturers,

dealers of essential commodities

• Retailers • Contractors and other

independent contractors • Banks and other financial

institutions • On peddlers • Other business taxes

City/Municipal Impositions

• Community tax • Other impositions

NON-TAX REVENUE

Regulatory Fees • Mayor’s Permit • Permit Fees under

the Building Code • Zonal/Locational

Clearance fees • Fees on weights and

measures • Motorized Tricycle

Operator’s Permit • Cattle Registration

Fees • Civil Registration

Fees • Slaughter Permit

Fee • Other Regulatory

Fees

Service/User Charges • Secretary’s Fees • Garbage Collection Fees • Parking Fees • Other Receipt/User charges

(Sanitary inspection fees, health examination fees)

Receipt from Economic Enterprises

• Receipts from Markets

• Receipts from Slaughter houses

• Receipts from cemeteries

• Receipts from Bus terminal

• Rentals • Receipts from

waterworks systems

• Other receipts from economic enterprises

NON-TAX REVENUE

Toll Fees Other Receipts • Fishery Rentals • Sales of assets • Miscellaneous Receipts

SHARES National Tax Collections Extraordinary Receipts/Grants/Aids Loans and Borrowing

31

FROM NATIONAL TAX

• Internal Revenue Allotment (IRA)

• Local Government Stabilization and Equalization Fund (LGSEF)

• Local Affirmative Action Project Fund (LAAPF)

• Priority Development Assistance Fund (PDAF)***

• Share in National Wealth

• Share in Tobacco Excise Tax

• Share in Expanded Value-added tax

• Share from Economic Zones

*** Allocated to Congress members for discretionary spending in LGU’s.

• Grants^– Foreign and Domestic 1. Calamity Fund 2. Municipal

Development Fund 3. Local Government

Empowerment Fund 4. Countryside

development fund 5. DECS School Building

Program • National Aids • Share from Lotto • Rebates on MMDA

Contributions • Other extraordinary receipts

^-Conditional cash grants are given to extremely poor households a Poverty reduction and social development program called Pantawid Pamilyang Pilipino Program (4Ps).

• Foreign • Domestic • Bond Floatation

SHARES FROM NATIONAL TAX

Inter-local transfers

Appendix B: Key Stakeholders in the Local Health Policy-making process in the Philippines At every level of local government, the key stakeholders involved in the policy-making for health care financing by local government units include the

• Local Health Board (Local health authority) comprising of some elected and some appointed members. They enjoy advisory powers, planning authority and responsibility for health services.

• Local Finance Committee • Local Treasurer • Local Legislative body and the Local Chief Executive

The Local Chief Executive (provincial governor/ city or municipal mayor), upon receipt of the actual (and estimated) income and expenditure statements of the preceding and current years (first two terms) from the local treasurer, recommendations on the level and ceilings of health expenditures made by the local finance committee based on approved local development plans, budget proposals made by heads of departments and offices prepares the annual health budget (Local Government Code 1991).

32

Table B.1. Legislati

ve Body Local Chief Executive

Presiding Officer

Members of Local health Boards

Other members of local health boards

Local Finance Committee

Province sangguniang panlalawigan

Governor Vice-governor

Governor (chairman), Provincial health officer (vice-chairman), chairman of the Committee on health of the sangguniang panlalawigan

Member from DOH, representative from the private sector and non-governmental organizations involved in health services

Provincial Planning and development Officer, Provincial Treasurer and Provincial Budget officer

City sangguniang panlungsod

City Mayor Vice-mayor City Mayor (chairman), city health officer (vice-chairman), chairman of the Committee on health of the sangguniang panlungsod

Member from DOH, representative from the private sector and non-governmental organizations involved in health services

City Planning and development Officer, City Treasurer and City Budget officer

Municipality

sangguniang bayan

Municipal Mayor

Vice-mayor Municipal Mayor (chairman), municipal health officer (vice-chairman), chairman of the Committee on health of the sangguniang bayan

Member from DOH, representative from the private sector and non-governmental organizations involved in health services

Municipal Planning and development Officer, Municipal Treasurer and Municipal Budget officer

Barangay sangguniang barangay

punong barangay