from devolution to consolidation: local health …nap.psa.gov.ph/ncs/11thncs/papers/invited...

20
11 th National Convention on Statistics (NCS) EDSA Shangri-La Hotel October 4-5, 2010 FROM DEVOLUTION TO CONSOLIDATION: LOCAL HEALTH SYSTEMS ARRANGEMENTS AND FACILITY- BASED DELIVERIES IN THE PHILIPPINES by Joseph J. Capuno and Marian Panganiban For additional information, please contact: Author’s name Dr. Joseph Capuno Designation Associate Professor Affiliation Univerity of the Philippines School of Economics Address Diliman, Quezon City Tel. no. +632- 9205460 E-mail [email protected] Co-author’s names Ms. Marian Panganiban Designation M.A. candidate Affiliation University of the Philippines School of Economics Address Diliman, Quezon City Tel. no. +632- 9279686 E-mail marian.[email protected]

Upload: ngodat

Post on 19-Aug-2018

219 views

Category:

Documents


0 download

TRANSCRIPT

11th National Convention on Statistics (NCS) EDSA Shangri-La Hotel

October 4-5, 2010

FROM DEVOLUTION TO CONSOLIDATION: LOCAL HEALTH SYSTEMS ARRANGEMENTS AND FACILITY- BASED DELIVERIES IN THE PHILIPPINES

by Joseph J. Capuno and Marian Panganiban

For additional information, please contact: Author’s name Dr. Joseph Capuno Designation Associate Professor Affiliation Univerity of the Philippines School of Economics Address Diliman, Quezon City Tel. no. +632- 9205460 E-mail [email protected] Co-author’s names Ms. Marian Panganiban Designation M.A. candidate Affiliation University of the Philippines School of Economics Address Diliman, Quezon City Tel. no. +632- 9279686 E-mail [email protected]

FROM DEVOLUTION TO CONSOLIDATION: LOCAL HEALTH SYSTEMS ARRANGEMENTS AND FACILITY-BASED DELIVERIES IN THE PHILIPPINES

by

Joseph J. Capuno and Marian Panganiban1

ABSTRACT

Under decentralization, many local governments in the Philippines have collaborated in their devolved health services, in some cases with Department of Health support. This paper investigates the impact of consolidated local health systems on facility-based deliveries. Using a panel of local government and household data for 2003 and 2008, the preliminary results indicate provincial health spending rises with consolidation but falls with city and municipal health spending. However, both provincial and municipal health outlays have positive but minimal effect of facility-based deliveries. To improve facility-based deliveries and thereby reduce further maternal mortality, greater coordination in local health spending is clearly necessary.

1 Introduction

Since 1991when the Philippines adopted a fiscal decentralization program, local government units (LGUs) have entered into various arrangements to improve their delivery and financing of health services. In some places, as few as two or three adjacent LGUs voluntarily consolidate their local health systems, while in other places all component LGUs participate in province-wide development undertakings in health. Arguably, the participating LGUs expect some benefits from consolidation, including possibly the economies of scale or scope, better coordination, and avoidance of externalities. That some LGUs do not join however also suggests some disadavantages, including possibly loss in flexibility to customize services according to local needs or conditions. How then do the various inter-LGUarrangements in health resolve the uneasy trade off between the losses in service specificity and the gains from, say, economies of scale or coordination? More specifically, does the consolidation of local health systems improve facility-based deliveries?” We explore these questions in this paper.

The research on the impact of local government arrangements on health outcomes has been scant, although there have been a number of studies describing the possible institutional arrangements to help achieve specific health sector goals (see for example Smith 1997). While there has been little investigation much less consensus on the optimal arrangement of local health systems, the literature on the theory for fiscal federalism provide some broad guidelines on the optimal assignment of expenditure functions to subnational governments (e.g., Boadway and Shah 2009).

One strand in the literature argues that in some circumstances smaller jurisdictions in

terms of geographic size and functional responsibilities can provide local officials with better information about their citizens’ preferences and can thus make them more responsive and directly accountable to the needs of their service clients (Lyons and Lowery 1989). The larger number of competing local governments will give individuals more choices in publicly-

1 We report here the preliminary results of an on-going study on the emerging arrangements of devolved health services and their consequences on maternal and child health outputs in the Philippines. We gratefully acknowledge the institutional support from the UPecon-Health Policy Development Program.

Page 1 of 19

provided services, and thereby will better accommodate differences in preferences for these services. Such competition may also constrain budget-maximizing bureaucrats and align their incentives towards promoting the voters’ welfare (Tiebout, 1956).

Another strand favors larger jurisdictions that yield benefits from economies of scale

while promoting consistency in service quality. They are also in a better position to internalize the costs and benefits of externalities generated by publicly-provided services. Moreover, the informational advantages associated with smaller arrangements can also be achieved in a consolidated set-up – a national government agency, for example can assign some functions of the central office to the local offices, which have direct and frequent contact with service clients (Bardhan 2002). Both factors - the economies of scale and customized local provision – seem to underlie the current arrangement of local health systems in the Philippines.

The emergence of the inter-LGU health arrangements is better appreciated within the

context of the country’s nearly twenty years of decentralzation experience. Relative to the expected efficiency gainsfrom a devolved health setup, the realized gains are found modest, partly because of the weaknesses in the design and implementation of the Local Government Code of 1991 (Lieberman, Capuno and Van Minh 2005). One purported gain is the increasing share LGUs in total health expenditures, which contributed to the rise in the share of health expenditures to GDP 2.9 percent in 1992 to about 3.3 percent in 2005 (Capuno and Kraft 2009). Per capita health spending has increased from PhP 334 in 1992 to about PhP 507 in 2005, representing a real 48 percent growth from 1992 levels2. Further evidence from a study by Racelis et al. (2005) on the Philippine National Health Accounts suggest that total health expenditures exhibit an increasing trend in both real and per capita terms after devolution, although minor setbacks were encountered in 1998, 2001 and 2002.

Under devolution, LGUs have also exhibited a capacity for innovation, particularly in

terms of establishing partnerships with the local communities, private sector, and non-government organizations or in integrating health objectives into other programs of the local government.3 Under the current setup, LGUs were also able to leverage existing systems (e.g. health insurance) to solve health problems specific in their areas.4 Some LGUs were able to hire more doctors and acquire more supplies that enabled them to improve health outcomes. Infant mortality rate also continued to decline since 1991, implying that some desired trends in health were sustained despite the weaknesses in the Local Government Code.

Notwithstanding the proliferation of innovations in local public services, the impact of devolution on health outcomes and outputs remains ambiguous at best. The financing constraints, political realities, differences in organizational capacity, and other institutional factors have led to insufficient services and deterioration in both the levels and quality of devolved health services. Differences in human resource capacities, as evidenced by the inability of some LGUs to hire doctors, nurses and other critical health personnel, and, more starkly, the unequal distribution of executive talents have led to widening variations in performance across LGUs. Some LGUs are “lucky” to have innovative local chief executives. While LGUs have become more experimental and innovative with devolution, the adoption of innovation is largely influenced by factors such as the local chief executive’s background, level 2 Figures in 1985 prices. 3 In Iloilo a province in Central Visayas, the town of Concepcion, and neighboring municipalities, were cited by the Galing Pook Foundation for its innovative health program of incorporating health into its population and environment endeavors and for adopting a multi-sector approach. The Galing Pook Foundation, which is a joint initiative of the Local Government Academy–Department of the Interior and Local Government, the Ford Foundation, and other individual advocates of good governance from the academe, civil society and the government, holds a yearly awards program for excellence and innovation in local governance. 4 An example is the Bindoy Social Health Insurance Indigency Program in Negros Oriental also a province in Visayas, which earned a Galing Pook Award in 2007.

Page 2 of 19

of electoral support, and time horizon, and other provincial characteristics (Capuno 2008). The greater autonomy conferred to LGUs under devolution was not automatically matched with the resources required to exercise it.

Moreover, it has been suggested that LGUs may have become more dependent on the

national government for their revenues since 1991. Under devolution, the LGUs receive a higher internal revenue allocation (IRA) as their share in the internal revenues of the national government, but which made them reluctant to raise revenues from local sources. Manasan (2005) finds that the mismatch between the revenue means and expenditure needs of various levels of local governments has worsened at all levels of local government with the implementation of the Local Government Code. Thus, even though evenness in national government spending across regions are observed, wide regional variations exist in LGU spending which implies spatial inequalities in the public provision of health services (Capuno and Kraft 2009).5

The problems of devolution described are very much apparent in the health sector –

particularly in the quality of maternal and child health care services. Although we observe an overall decline in the means of both IMR and MMR in the years 2000 to 2008, the standard deviation has been increasing among provinces.6 In spite of the downward trend, the Philippines still has a high MMR when compared with other countries. Some 230 women die in the Philippines for every 100,000 live births, compared with 110 in Thailand, 62 in Malaysia and 14 in Singapore.7 To address this problem, the DOH and UNICEF now assiduously promote facility-based delivery, in lieu of home-based delivery unless attended by skilled health professionals. This strategy is adopted given the observation that countries with modest per capita incomes but high proportions of facility-based deliveries have achieved relatively low mortality ratios (Dayaratna et al. 2000)8. Furthermore, the poor performance of the Philippines when it comes maternal and child mortality rates can be partly attributed to the low percentage of facility-based deliveries. The World Health Organization (2001) estimates that about 72 percent of deliveries in the Philippines occur outside of health facilities, frequently without the assistance of a skilled birth attendant, and many deliveries do not meet the minimal conditions for early essential newborn care.

Arguably, a greater reduction in IMR and MMR can be achieved with the concentrated

health interventions of local governments, namely the provinces and municipalities. There are both administrative and political reasons why these local governments will consolidate their devolved health services. Some of the reasons are explored here and examined for their impacts on frequency of facility-based delivery. Then, the institutional bases for consolidation are reviewed in section 2, and the conceptual framework for the analysis of its impact is presented in section 3.The empirical methodology and the data used are then discussed in sections 4 and 5, respectively. The prelimary results are then examined in section 6. The last section contains the tentative conclusions and policy implications of the paper.

5 In fact, we can associate the increase in health spending per capita the recent years to the increase in internal revenue allocation per capita received by the local government units. Correlation between health spending and internal revenue allocation is quite high, ranging between 0.80 and 0.86 for the years 2005 and 2008. 6 See Field Health Service Information System statistics from Department of Health. 7 Figures from Maternal Mortality Report 2005 by WHO, UNICEF, UNFPA and The World Bank. 8 The report also recommends that developing countries pursue the low-cost alternative of ensuring the presence of skilled birth attendants at birth.

Page 3 of 19

2 Institutional bases for consolidation

Geography, politics, administrative and legal factors condition in a complex way the emergence of consolidated local health systems in the Philippines. The Philippines is an archipelago of around 7,100 islands, divided into 17 administrative regions. Each region, except for the National Capital Region (NCR), is subdivided into provinces, and each province into municipalities and cities, and each of which into barangays (or villages). In 2009, there were already more than 1600 provinces, cities and municipalities, and around 40,000 barangays.

The constraints to cooperation that the geographical barriers pose are supposed to be offset by the political and administrative structures of local governments in the Philippines. Each province in governed by a governor, each municipality or city by a mayor, and each barangay by a barangay captan. A local chief executive is elected every three years, for a maximum of three consecutive terms, directly by his or her constituents. Administratively, each barangay captain is under the municipal or city mayor, and each mayor in turn is under the provincial governor. The provincial governor is aadministratively under the national government. This administrative setup within each province provides a de jure venue for consolidating the local government programs and projects. Region-wide consistency of provincial plans is attained during the regular meetings of the regional development councils comprising the local chief executives and key officials from the local branches of national government agencies.

Some LGUs however are outside the usual administrative setup. In the National Capital

Region (NCR), the 17 component cities and municipalities are each independent of one another and there is no provincial government to supervise them. In some provinces, there are independent cities that are outside the administrative conrtol of the provincial government. The constituents of these cities do not vote for the provincial governors, unlike the residents of component cities in some provinces. In the Autonomous Region of Muslim Mindanao (ARMM)9, a regional government is established and run by an elected regional governor that supervises directly the compornent provinces. All these LGUs – NCR cities and municipalities, indepenent cities, and ARMM regional government – are administratively within the ambit of the national government.10

Ostensibly, the devolution of the diffeent levels of health functions followed the

adminitrative hierarchy of levels of governments in the country. On the one hand, the Department of Health retained responsibility over the regional and specialty hospitals and public health programs. On the other hand, the provincial governments took charge of the devolved curative care services and most secondary and tertiary-level hospitals, while municipal and city governments handle primary care and the operation of smaller facilities such as the rural health units and barangay health stations. In addition to the formal administrative mechanisms for integrating the operations of the devolved health services, the Local Government Code further stipulates that LGUs can form special cooperative arrangements among themselves:

Section 33. Cooperative Undertakings Among Local Government Units. - Local government units may, through appropriate ordinances, group themselves, consolidate, or coordinate their efforts, services, and resources for purposes commonly beneficial to them. In support of such undertakings, the local government units involved may, upon approval by the sanggunian concerned

9ARMM is composed of Basilan, Maguindanao, Lanao del Sur, Sulu and Tawi-Tawi. The city of Isabela, although still politically part of Basilan, is not part of ARMM. 10 For this reasons, we treat both NCR and ARMM as provinces in our regression analysis.

Page 4 of 19

after a public hearing conducted for the purpose, contribute funds, real estate, equipment, and other kinds of property and appoint or assign personnel under such terms and conditions as may be agreed upon by the participating local units through Memoranda of Agreement. Consitent with this provision, some LGUs have banded togther, presumably because of

their common problems, shared goals or political alliances. For example, some LGUs have formed collaborative arrangements to tackle more specific development concerns in industry (e.g.Metro Iloilo-Guimaras Development Council), resource management (e.g. Northern Iloilo Alliance for Coastal Development), urbanization (e.g. Metro Cebu Development Council) and health (Inter-Local Health Zones). These configurations ranges from simple coordination in the planning and delivery of public services, to formal cooperation involving the joint use and financing of health services, to consolidation of devolved health services under the administration or supervision of the province or a single authority created for the purpose.

Apart from these voluntary arrangements, LGUs in many provinces have also

participated in the Province-wide Investment Plan for Health (PIPH), a strategy of the Department of Health to forge LGU partnerships to coordinate local health plans, secure resource commitments, and ensure the achievement of health goals in the province. Intended to be an integral part of the LGU development process, the PIPH deviates from the usual LGU planning practice wherein city or municipal development plans – which include local health plans – are reviewed and put together perfunctorily into one provincial development plan, with minimal regard for possible inconsistencies, duplications and potentials for possible economies of scale or externalities. To ensure LGU commitment, the plans under the PIPH are supposedly developed in a participatory manner, implemented with support from the DOH and donor institutions, and monitored and evaluated based on results.

Do these voluntary and DOH-led cooperative undertakings improve health outputs? If

yes, how exactly do they impact health outputs? Why would LGUs pursue cooperative undertakings in the first place? The next section provides the conceptual framework through which we will answer these questions. 3 Conceptual framework The theory of fiscal federalism emphasizes the tradeoff between the gains and losses of assigning an expenditure function to a subnational government unit (e.g., Boadway and Shah. 2009). The gains would come from the improved matching in the supply of and the demand for local public services. With superior information about local needs and conditions and their direct accounatbility to the constituents, the lowest level of local government is in the best position to customize service provisions. Thus, the welfare gains from service specificity increase, possibly at a decreasing rate, as the responsibility for the service is transferred from the central government to lower and lower levels of government. Examples of such customer-specific services are trust and confidence between patient and doctors and suitability of clinic hours to the schedule of working mothers or children.11

11 For example, there are no regular surgeons serving in barangay health stations because their capacities are under-utilized in this facility. Visits from patients who need medical attention from a surgeon are not very common. Services of surgeons are most needed in district hospitals where they can serve a broader group of patients who require their attention. However, having surgeons serve in district hospitals or other higher-level facilities implies a loss in service-specificity for patients who go to barangay health stations who may immediately need the services of a surgeon. Such instances may be rare, but the unavailability of the kind of service at the specific time and place needed by a patient constitutes a loss.

Page 5 of 19

However, the total costs of providing customized services would be greater than if the same generic service is provided to the same population by the central government instead. Unlike local governments, the central government can save on overhead expenses, easily deploy underutilized equipment or personel and access more readily available superior technology. Moreover, it avoids some of the coordination and enforcement costs that various independent decisionmakers face when they broker a cooperative arrangement. The total costs of service provision thus increases, possibly at an increasing rate, as the responsibility over the service is devolved to more and subnational governments.

The costs and benefit curves depicting the total costs of decentralization and the

benefits from service specificity with decentralized provision are depicted in Figure 1. The optimal assignment of the service is depicted by D* where the marginal cost (slope of the total cost curve) is equal to the marginal benefits (slope of the benefits curve). Now, the move towards consolidation presupposes that the responsibility over the service was decentralized beyond what is optimal (i.e., to the right of D*). The natural tendency then would be for LGUs to form alliances or cooperative agreements to take advantage of, say, the economies of scale, which outweight the loss welfare due to the provision less than fully customized services.

[Insert Figure 1 here.]

What could then be the sources of gains from consolidated provision? The gains in

cooperative arrangements relative to fragmented configurations can be due, but not limited, to the following: control of externalities (control of epidemics and disease outbreaks), resource-sharing (i.e., deployment of health personnel in neighboring LGUs), economies of scale (joint use of underutilized health resources or equipment),12 miinimizing information asymmetries (citizens are more likely to be aware of who governs them when there are fewer units of government, see Post 2002), and complementation of services in health and others (existing cooperation in providing one type of service can give rise to cooperation in providing related services)13, Furthermore, a consolidated arrangement reduced likelihood of service inequalities across local governments.

Would D* correspond to an existing LGU or political jurisdiction? The existence of

various inter-LGU collaborations indicates that a higher level than municipalities and cities may be the better provider of the relevant local public service. Unlike in other countries however where there are counties, districts or metropolitan areas that serve as intermediate level between the province and municipalities, in the Philippines only the province comes higher than the municipalities or cities in most places. Further, the unavailability of a concrete measure of sub-provincial arrangements poses a serious data limitation. Moreover, the existence of a political constituency at the province level implies lower transaction costs than sub-provincial

12 In the province ofNegros Oriental, for example, then Governor Macias, installed health facilities in villages located in the mountains bordering Negros Oriental and Negros Occidental, to serve better the residents there. These health facilities were jointly supported by the provincial government of Negros Oriental and Negros Occidental and jointly used by their respective constituents living in the mountains. These health facilities would otherwise be under-utilized had they been available only for residents of Negros Oriental. 13 The experience of Alliance of Northern Iloilo for Health and Development (ANIHEAD) shows that cooperation in providing a specific service usually require complementary efforts in other public services. ANIHEAD is an alliance composed of the municipalities of Ajuy, Balasan, Batad, Carles, Concepcion, Estancia, Lemery, San Dionisio and Sara in the province of Iloilo. The alliance implements convergence strategic framework that integrates various population, health and environment projects into a Population, Health and Environment (PHE) program. The PHE aims to improve the town’s knowledge, attitude and skills related to family planning and reproductive health; quality, accessibility, and availability of family planning and reproductive health; as well as capacity on community-led coastal resource management. Its corresponding alliance for coastal development is the Northern Iloilo Alliance for Coastal Development (NIACDEV).

Page 6 of 19

arrangements in organizing LGUs at this level since the administrative costs of grouping and organizing LGUs are much lower. We thus test the hypothesis that a province-level arrangement leads to better than a fragmented setup in promoting facility-based deliveries, given its proven effectiveness in reducing maternal and infant mortality rates.

Further, the assumption of a province-level arrangement lends itself to an empirical test,

especially since the PIPH is a province-wide collaboration of all component LGUs a province. As in similar collaborations, the PIPH entails bargaining, coordination, monitoring and enforcement costs. But since the PIPH is DOH-led, the transaction costs borne by the cooperating LGUs are presumably lower than if they were to bear the costs alone. In places where the PIPH is not yet introduced, other mechanismsn must be in place to forge and sustain cooperation. Here we conjecture about politicaly-driven cooperation among local chief executives who may have genuine desire to serve their constituents or motivate more to reap the political payoffs from collaboration, The political motives among LCEs have to be sufficiently strong to counter the strong incentive to free-ride, lack of jurisdiction or power over other LCEs, and the high transaction costs of forming an alliance.14 The underlying model of incumbent behavior is one LCE whose desire to be re-elected motivates his or her allocation of fiscal resources, a standard public choice model (see, for example, Persson and Tabellini 2002; Solon, Fabella and Capuno 2009). 4 Empirical methodology A. Estimation model The following equations are used to estimate the impact of LGU consolidation on health outcomes. The first equation below links the health output Y, defined as the proportion of facility-based deliveries, to a host of explanatory variables, some endogenous (like health expenditures), i.e.,

(1) where the subscript it refers to the ith province in year t, H is a vector of health system variables, W is a vector of provincial characteristics, and X is an endogenous explanatory variables, u is the error term and the α and β’s are regression coefficients.

We use the instrumental variable to address the problems in estimation arising from the endogeneity of X:

(2)

where the π ’s are the regression coefficients, v is the error term and the rest of the variables are as defined in (1). We also estimate the following regression equation to check robustness of the relationship between proportion of facility-based deliveries and consolidation variables: 14 Such is the case in Northern Iloilo Alliance for Coastal Development (NIACDEV) an alliance of municipalities in Iloilo, a province in Central Visayas, Philippines. The loose cooperative arrangement of NIACDEV led to resignation of council chair over fishing and trading problems between two municipalities. In March 2008, then NIACDEV chairperson and Estancia Mayor Restituto Mosqueda resigned because of the prohibition set on Carles fishers to continue trading at Estancia, a member of NIACDEV. As alliance chair, Mosqueda considered the move of Carles, also member of the alliance, as a direct affront to his leadership particularly since no prior discussion nor notification was made on member town mayors.

Page 7 of 19

(3)

The hypothesis underlying equation (3) is that the instrumental variable (P) has no direct relationship to Y other than through X. B. Data Table 1 presents the descriptive statistics of key variables of interest15 for the years 2003 and 2008. These years were used because of the availability of representative figure for the proportion of facility-based deliveries from the National Demographic Health Survey (NDHS) conducted in those years. 16 To capture provincial-level consolidation, we use the following political measures: proportion of mayors belonging to same political party as governor, proportion of mayors re-elected, proportion of mayors in their first term, proportion of mayors in their last term, and dummy variables to capture if a governor is re-elected, if a governor is in his or her first term, and if a governor is in his or her last term. These measures were constructed from the certified list of winners from the Commission of Elections for the election years 2001 , 2004 and 2007. Political consolidation measures constructed from the 2001 election data were matched with 2003 NDHS health output variables while political consolidation measures constructed from 2007 election data were matched with 2008 NDHS health output variables. Another measure of provincial-level administrative consolidation is the existence of province-wide investment plan for health (PIPH) in the province. A dummy variable was created to take the value of 1 if the province has a PIPH and the 0 otherwise.

To control for factors which may confound the relationship between province-level cooperation and health outputs, the regression also includes health system variables such as the number of doctors, nurses, and midwives per 10,000 population, and socio-economic indicators such as poverty incidence, and proportion of women of reproductive age. Health system variables were derived from Field Health Service Information Systems (FHSIS), and thus captures only the number of public health personnel. Poverty incidence came from the National Statistical Coordination Board. Due to the unavailability of 2007 poverty incidence statistics, we use the available poverty data for 2006. The proportion of women of reproductive age came from the National Demographic Health Survey.

We observe that between 2003 and 2008, the mean proportion of facility-based

deliveries declined from 30 percent to 14 percent. We find little change in the proportion of women of reproductive age and poverty incidence. The number of doctors has decreased between 2003 and 2008 while the increase in the number of midwives is minimal. In 2003, only five percent of provinces have PIPH, by 2008, the percentage has risen to 41. Provincial health spending per capita has declined while component city health spending per capita has markedly increased; the opposite holds true for internal revenue allocation per cpita. Little has changed in the proportion of mayors belonging to same party as the governor, although the proportion of mayors of re-elected has declined from 0.47 to 0.33. The percentage of governors re-elected though has increased from 0.47 to 0.64. Governors in their first term also declined between 2003 and 2008 as governors in their last term increase from 0.18 to 0.51. The proportion of mayors in their first term has increased very slightly from 0.48 to 0.50. The change in the proportion of mayors in their last term is more noticeable as it rose from 0.16 to 0.22.

15 As independent cities are jurisdictions unto themselves, they were excluded in the construction of the variables in the analysis. 16 Note that these figure includes deliveries in both public and private facilities.

Page 8 of 19

C. Endogenous variables and choice of instruments We treat provincial health spending per capita as endogenous, and the presence of PIPH as exogenous. . To reduce these ambiguities, we use instruments to tease out the independent variations in these variables that are not correlated with the proportion of facility-based deliveries. To futher control for possible unobserved fixed effects, we use panel data estimation techniques. The results reported in the next section are calculated using xtivreg and xtivreg2 command in STATA.

While it can be argued that the presence of PIPH is also endogenoues since the provinces where the PIPH was initiated were purposely selected by the DOH for their relatively low heir health outputs and outcomes. Moreover, provinces were picked based on the willingness of the provincial government to participate. In places where the governor who felt confident of the support of the mayors (possibly because of political consolidation) may have convinced the DOH to select his or her province. However, we do not investigate the endogeneity of PIPH here, largely because using the same xtivreg command in STATA, which assumes a continuous endogenous regressor, would produce inconsistent estimates in the case of binary endogenous regressor.

The PIPH is an attempt by the DOH to forge an administrative consolidation of local

health systems. Arguably, such consolidation occurs even before or outside PIPH, possibly because of political reasons. We posit that such politically motivated consolidations are likely to occur in provinces with high proportion of LCEs (mayors and governors) that belong to the same political party.17 Political consolidation is also assumed likely in provinces with high proportion of re-elected LCEs owing to, among others, greater familiarity among them. Term limits as well would have greater bearing on health expenditures than on facility-based deliveries, if only because the fiscal resources would determine the availability of services with which to promote desired health outputs. These hypotheses and assumptions are tested below. 5 Analysis of results A. Main results Table 2 shows the results of the instrument variable fixed effects panel regression18 when provincial health spending is endogenized. Panel A contains the first-stage regression results while panel B provides the first-stage regression results. There are five models presented, with each model using different instruments for provincial health spending per capita. Model 1 uses party affiliation as instrument for provincial health spending, while model 2 uses party affiliation and re-election of mayors. Model 3 adds a dummy variable to represent the re-election of governor. Model 4 includes a dummy variable for governors in their first term and the proportion of mayors in their first term. Last, model 5 then adds another dummy variable for governors in

17 This is not to say that they always belong to the same party in each election. Party switching among politicians is common in the Philippines. The general observation though is that most local politicains swithc to the party of the incumbent president or to the dominant presidential candidate, but they stay in the same party in between elections. 18Results from our Breusch and Pagan Langaranian multiplier test using xttest0 in Stata 10 show that panel regression techniques capture variations in the proportion of facility-based deliveries better than pooled regression. P-value (0.03) from the test allows us to reject the null hypothesis that the var (u)=0 across time and across observations. We also conduct a Hausman test to verify whether fixed-effects or random-effects panel regression should be used. P-value from the test shows that we cannot reject the null hypothesis that coefficients from fixed-effects and random-effects panel regression are not systematic. We use fixed-effects panel regression to account for possible omitted variable bias.

Page 9 of 19

their last term and the proportion of mayors in the last term to the instruments of provincial health spending.

[Insert Table 2 here.]

From panel A, we observe that component city health spending per capita is consistently positive and statistically significant with proportion of facility-based deliveries across all models, except for the first one, if we assume a p-value less than 0.10. The magnitude of its effect on facility-based deliveries is considerably small – almost zero. Similarly, we also find the number of midwives per 10,000 population is significant and negative with facility-based deliveries in the last four models if we allow a p-value of less than 0.10. Its effect on facility-based deliveries ranges from 0.08 to 0.12. Provincial health spending per capita has a significant and positive effect with facility-based deliveries in models 2, 3 and 5. The magnitude of its effect on facility-based deliveries − around 0.002 − is larger than that of component city health spending per capita, but smaller than the impact of the number of midwives. The dummy variable with PIPH meanwhile is significant and assumes a negative sign in models 3 to 5, with coefficients ranging from 0.10 to 0.13. The negative impact largely reflects the selection criteria of the DOH that initially targeted the provinces with relatively low health indicators. The number of doctors, the proportion of women of reproductive age, and poverty incidence are statistically insignificant determinants of facility-based deliveries.

Results in panel B show the first-stage results of panel A’s instrumental variable

regression. The effect of component city health spending per capita on provincial health spending per capita is negative is negative in all models, and statitically significant in the last four (models 2-5). Consistently in all five models, the number of midwives has a significant and positive relationship with provincial health spending. Of the political variables used as instruments, proportion of mayors re-elected is positively significant with provincial health spending in all the models it is introduced. The other political variables have no significant relationship with health spending by the province. The adoption of PIPH has no statistically detectable effect on provincial health spending, although its sign is negative. This means that the PIPH has no influence yet presumably because the DOH and donors have not yet able to leverage their resources for greater local health spending during the years covered.

B. Robustness checks The validity of the instrumental variable regression results depends on the assumption that political consolidation has no direct effect on facility-based deliveries. Although this assumption may be questionable, we check its validity by controlling for the variables that may be simultaneously correlated with acility-based deliveries and political consolidation.

Instrumental variable regression requires that there be at least as many instruments as there are troublesome explanators. We first test whether we are using too few instruments through an under- identification test. The general rule is to have at least as many instruments as troublesome explanators in order to have an equation exactly identified (Murray 2006). We thus want our instruments for provincial health spending sufficient to capture its possible endogeneity. Panel C of Tables 2 shows the results of our under-identification test. We test the null hypothesis that the model is under-identified. In the table, using only the proportion of mayors belonging to same political party as governor as instrument does not allow us to reject the null hypothesis. Only with the inclusion of proportion of mayors re-elected and variables capturing the presence of mayors and governors in their first term can we confidently assume that the estimation model is exactly identified.

Page 10 of 19

According to the model estimated in Table 2, political consolidation affects provincial

health spending, which in turn affects facility-based deliveries. We can test whether any of the political consolidation variables use has a direct effect on facility-based deliveries by conducting an overidentification test. Our over-identification test presumes that our model with instruments used is exactly identified, i.e., that exogenous regressors have properly been treated as exogenous and have not been used as instruments.19. P-values from our identification test show that we could not reject this null hypothesis even at the 10 percent level of significance. The over-identification test results lead us to the same conclusion: even with the use of all political variables as instruments, the model remains properly identified. More concretely, the re-election and election terms of mayors and governors and party affiliation have no direct impact on facility-based deliveries, and work indirectly through provincial health spending.

Table 3 presents an easy-to-interpret version of the over-identification test. It adds

political variables as exogenous variables in a single-stage regression with facility-based delivery. If these variables have a direct impact on facility-based deliveries, we would expect them to be statistically significant in this regression. None of them are found to have a significant relationship with facility-based deliveries at a five percent level of significance. Moreover, the inclusion of these political variables does not significantly change the coefficient and standard errors of all other coefficients and the model’s R-squared. This result further substantiates our findings from the over-identification test.

[Insert Table 3 here.]

C. Discussion

We focus our analysis on models 4 and 5 in Table 2 since they best satisfy the identification requirements of instrumental variable regresssion. In the table, the coefficient of provincial health spending per capita with facility-based deliveries is approximately 0.002. The first stage regression results give us an idea of what can increase or decrease provincial health spending. Component city health spending is negatively associated with provincial health spending. This suggests that there may be possible substitution effect between the two. Cities and municipalities may opt to reduce their health budgets in response to their provincial government’s large health expenditures. The number of midwives increases with provincial health spending. Of the political variables, only proportion of mayors re-elected are significant with provincial health spending, and its effect on health spending is highest compared with all significant variables in the first stage regression. This can be interpreted in two ways: mayors who are re-elected are more likely to be familiar with each other, with their political positions, and may have even worked together before which should make it easier for them to lobby collectively for higher provincil health spending. Alternatively, they may also be familiar with the procedures and people in the provincial government, which may make them more adept at negotiating for increases in health spending. In summary, the results in Table 2 imply that political consolidation has a significant effect on health spending while health spending has significant but very small impact on facility-based deliveries.

19 The results though have to be interpreted with caution. We may be unable to reject the null hypothesis in cases when the instruments used are invalid but still highly-correlated with each other.

Page 11 of 19

6 Conclusion Following the implementation of Local Government Code of 1991, there have been wide variances in the performances of LGUs, both in terms of achieving health outputs and outcomes. This has been especially apparent in maternal mortality rate, whose overall mean has been declining but whose standard deviation among provinces has been rising post-devolution. This trend is closelytied with the continued low proportion of facility-based deliveries in the country.

To address this problem and to exploit possible gains from banding with others, there have been efforts by LGUs themselves or by the DOH and donor institutions to coordinate local health systems. Some LGUs for example have formed alliances and other collaborative arrangements to improve their delivery and financing of their health services. Many of these arrangements were strengthened through DOH’s introduction of PIPH. The adoption of the PIPH across provinces and the various political collaborative formations among LGUs within provinces allowed us to test whether consolidation has indeed led to higher facility-based deliveries.

Using fixed-effects instrumental variable panel regression techniques, we find that

collaboration or coordination could help improve as facility-based deliveries but only indirectly. Political collaboration can help direct more resources in terms of provincial health spending towards health, which in turn slightly improves facility-based deliveries. We also observe a substitution effect between component city and municipality health spending and provincial health spending as evidenced by their negative sign in the first-stage regressions.

While preliminary, the results reported here suggest a few inputs to policy.First, there is

need to further facilitate and strengthen such collaborative arrangements. In health, this means that the PIPH should be opportunistically introduced where such arrangement already exist. Where such an arrangement does not yet exist, the PIPH could lay the necessary foundation. The second policy implication is that sub-provincial inter-LGU arrangements could be counterproductive since the municipal and city health spending is found to have negative effects on provincial health spending. This could mean either that the former free rides on the latter, or that they provide effectively substitute health services. This should be avoided. Given the deteriorating health outputs in some places, more and complementary service provisions are clearly needed. Already, the provincial health spending has only minimal effect on facility based delivery. Third, LCEs should be encouraged to take a longer view of health, with greater focus on outputs and outcomes while maintaing the same concern about health financing. This could be achieved for example by engaging more stakeholders in the PIPH so that the LCEs could take political credit for the lagged effects of health investments made in their terms.

Page 12 of 19

References Bardhan, P. (2002), ‘Decentralization of governance and development’, Journal of Economic

Perspectives 16(4): 185–205. Boadway, R. and A. Shah (2009). Fiscal Federalism: Principles and Practices of Multiorder

Governance. Cambridge, UK: Cambridge University Press. Capuno, J. J. and A.D. Kraft. (2009). Equity in the Delivery of Health and Education Services in

the Philippines: Background Country Report. Report Submitted to the Asian Development Bank.

Dayaratna V., W. Winfrey, K.. Hardee, J. Smith, E. Mumford, W. McGreevey, J. Sine and R.

Berg. (2000). Reproductive Health interventions: which ones work and what do they cost? Occcasional Paper No. 5. Washington, DC: POLICY Project. Available at http://www.policyproject.com/pubs/occasional/op-05.pdf.

Galing Pook Foundation. (2009). Kalusugan Muna! Policy Forum on Health. Retrieved at

http://www.fes.org.ph/uploads/documents/Documentation%20Health%20Forum.pdf Gonzales, G. (2004). Metro Cebu: A Metropolitan Area in Need of Coordinative Body. PIDS

Discussion Paper 2004-49, Phlippine Institute of Development Studies, Makati Philippines.

Hiblonada, F., “Mosqueda resigns as head of NIACDEV”, The News Today, 4 March 2008. Juan, E.D., ed. (1999). Excellence in local governance: Innovative practices in human resource

management. Local Government Academy, Development Academy of the Philippines, Pasig City.

Lieberman, S.S., J.J. Capuno and H. Van Minh (2005), ‘Health decentralization: Some lessons

from Indonesia, the Philippines and Vietnam’, pp. 155–78 in R. White and P. Smoke (eds), East Asia Decentralizes: Making Local Government Work, Washington DC: World Bank.

Lowery, D. and Lyons, W.E. (1989). The impact of jurisdictional boundaries: An individual- level

test of the Tiebout model. Journal of Politics 51: 73-97, Cambridge University Press. Manasan, R.G. (2005). Local Public Finance in the Philippines: Lessons in Autonomy and

Accountability. Philippine Journal of Development, 60 (1): 31-102. Murray, M. (2006). Avoiding Invalid Instruments and Coping with Weak Instruments. Journal of

Economic Perspectives 20(4): 111–132. Persson, T. and G. Tabellini (2002). Political Economics: Explaining Economic Policy.

Cambridge, MA: The MIT Press. Post, Stephanie. (2002) "Local Government Cooperation: The Relationship Between

Metropolitan Area Government Geography and Service Provision" Paper presented at the annual meeting of the American Political Science Association, Boston Marriott Copley Place, Sheraton Boston & Hynes Convention Center, Boston, Massachusetts,. Available at URL: http://www.allacademic.com/meta/p66016_index.html.

Page 13 of 19

Racelis, R, F. Dy-Liacco, R. Sabenano, M. Beltran, and T. Manaog. (2005). The National Health Accounts of the Philippines: Continuing Development and New Findings. Philippine Journal of Development, 62 (1&2):179-210.

Smith, B.C. (1997). The decentralization of health care in developing countries: organizational

options. Public Administration and Development, 17: 399-412. Solon, O., R. V. Fabella and J. J.Capuno (2009). Is local development good politics? Local

development expenditures and the re-election of governors in the Philippines in the 1990s. Asian Journal of Political Science 17(3): 265-284.

Tiebout, C. (1956). A Pure Theory of Local Expenditures. Journal of Political Economy 64 (5):

416–424. World Health Organization, UNICEF, UNFPA, and The World Bank. (2007). Maternal mortality

in 2005: estimates developed by WHO, UNICEF, UNFPA and The World Bank. Available from URL: http://www.unfpa.org/public/site/global/lang/en/pid/389

Page 14 of 19

Figure 1: Trade-off between benefits from service-specificity and total costs of decentralized provision D*

Centralized Decentralized

Benefitsfrom

servicespecificity,

Totalcosts

Level ofgovernment

Total costs

Service specificity

Page 15 of 19

Table 1: Summary statistics of panel variables 2003 2008

Variable Mean

Standard Deviation Min Max Mean

Standard Deviation Min Max

Dependent variable Proportion of facility-based deliveries 0.30 0.18 0.00 0.76 0.15 0.06 0.05 0.46 Explanatory variables Proportion of women of reproductive age 0.23 0.03 0.15 0.31 0.23 0.03 0.16 0.30Poverty incidence 31.72 12.90 3.40 64.60 33.91 13.08 6.40 63.00Doctors per 10,000 population 0.42 0.59 0.00 4.52 0.36 0.35 0.01 2.80Midwives per 10,000 population 2.51 1.36 0.00 8.47 2.88 3.20 0.07 21.41 Province has PIPH (%) 0.05 0.23 0.00 1.00 0.41 0.50 0.00 1.00

Health spending per capita of provincial government 163.35 135.90 0.00 910.68 138.90 161.98 3.38 979.02Health spending per capita of component cities and municipalities 113.62 50.19 38.41 296.16 367.00 1602.46 1.00 12805.93Internal revenue allocation per capita, provincial goverment 744.56 729.99 188.33 6020.68 784.36 1017.49 24.62 5655.28Internal revenue allocation per capita, component cities and municipalities 119.24 254.59 5.89 1791.33 87.54 372.18 0.00 2851.51

Proportion of mayors belonging to same party as governor 0.38 0.27 0.00 1.00 0.36 0.25 0.00 1.00Proportion of mayors re-elected 0.47 0.19 0.00 1.00 0.33 0.19 0.00 1.00Governor is re-eleected 0.47 0.50 0.00 1.00 0.64 0.48 0.00 1.00Governor in first term 0.49 0.50 0.00 1.00 0.27 0.45 0.00 1.00Governor in last term 0.18 0.39 0.00 1.00 0.51 0.50 0.00 1.00Proportion of mayors in their first term 0.48 0.17 0.00 0.91 0.50 0.17 0.00 0.91Proportion of mayors in their last term 0.16 0.11 0.00 0.50 0.22 0.14 0.00 0.64

Page 16 of 19

Table 2: IV fixed-effects panel regression, dependent variable is proportion of facility-based deliveries, endogenous variable is provincial health spending per capita Panel A: IV Fixed-effects panel regression, dependent variable is proportion of facility-based deliveries in all facilities with endogenous provincial health spending per capita (1) (2) (3) (4) (5) Provincial health spending per capita

0.00 (0.94)

0.003 (0.09)

0.002 (0.13)

0.002 (0.05)

0.002 (0.06)

Component city health spending per capita

0.000 (0.99)

0.0008 (0.08)

0.00007 (0.07)

0.00007 (0.03)

0.00007 (0.01)

With PIPH -0.17 (0.56)

-0.08 (0.15)

-0.10 (0.05)

-0.10 (0.03)

-0.13 (0.01)

Doctors per 10,000 population

-0.04 (0.97)

0.26 (0.37)

0.21 (0.41)

0.20 (0.37)

0.18 (0.36)

Midwives per 10,000 population

0.00 (1.00)

-0.12 (0.06)

-0.09 (0.07)

-0.10 (0.02)

-0.08 (0.009)

Proportion of women of reproductive age

0.30 (0.92)

1.25 (0.11)

1.07 (0.14)

1.04 (0.13)

0.73 (0.38)

Poverty incidence -0.01 (0.85)

-0.003 (0.52)

0.001 (0.71)

0.001 (0.72)

-0.0000 (0.98)

R2 0.13 -0.65 -0.24 -0.17 0.04 Number of groups 72 72 72 72 72 F- statistic (0.65) (0.10) (0.03) (0.01) (0.00) Panel B: First-stage for provincial health spending per capita Component city health spending

-21.45 (0.18)

-0.018 (0.03)

-0.018 (0.05)

-0.012 (0.05)

-0.02 (0.01)

With PIPH -0.02 (-0.023)

-15.72 (0.29)

-16.08 (0.30)

-4.37 (0.80)

-8.11 (0.71)

Doctors per 10,000 population

-78.73 (0.26)

-100.71 (0.12)

-116.38 (0.12)

-106.68 (0.12)

-95.70 (0.22)

Midwives per 10,000 population

29.92 (0.00)

30.64 (0.00)

30.90 (0.00)

30.56 (0.00)

31.81 (0.00)

Proportion of women of reproductive age

-230.89 (0.22)

-150.27 (0.38)

-155.14 (0.40)

-158.52 (0.39)

-133.76 (0.58)

Poverty incidence -2.09 (0.15)

-2.08 (0.09)

-2.05 (0.12)

-2.24 (0.07)

-1.68 (0.27)

Proportion of mayors belonging to same political party as governor

-5.98 (0.77)

-14.68 (0.47)

-13.52 (0.52)

-18.22 (0.36)

-15.33 (0.51)

Proportion of mayors re-elected

65.50 (0.02)

64.92 (0.03)

93.53 (0.01)

87.54 (0.004)

Governor is re- -8.84 -3.92 0.38

17

elected (0.58) (0.79) (0.98) Governor is in first term

22.95 (0.08)

23.95 (0.10)

Proportion of mayors in their first term

64.83 (0.11)

33.21 (0.41)

Governor is in last term

-19.40 (0.17)

Proportion of mayors in their last term

-73.47 (0.33)

R2 0.48 0.52 0.52 0.55 0.58 Number of groups 72 72 72 72 72 F-statistic (0) (0) (0) (0) (0) Panel C: Under-identification tests p-value from χ2- test20

0.75 0.07 0.11 0.02 0.06

Panel D: Over-identification tests p-value from χ2- test21

_ 0.75 0.29 0.61 0.56

p-values in parentheses

20 The under-identification test is based on Kleibergen-Paap rk-LM statistic which checks whether the equation is identified, i.e., that the excluded instruments are correlated with the endogenous regressors. A rejection of the null hypothesis indicates that the matrix is full column rank, i.e. model is identified. 21 This is based on the Sargan-Hansen test is a test of over-identifying restrictions. The joint null hypothesis is that the instruments are valid instruments, i.e., uncorrelated with the error term, and that the excluded instruments are correctly excluded from the estimated equation.

18

Table 3: Ordinary least squares panel regression, dependent variable is facility-based deliveries (1) (2) (3) (4) (5) (6)Province has PIPH -0.15 -0.15 -0.13 -0.13 -0.11 -0.11 (0.014)

-0.03

(0.013) (0.038) (0.034) (0.120) (0.192)Provincial health expenditure per capita

0.00 (0.718)

0.00 (0.718)

0.00 (0.745)

0.00 (0.842)

0.00 (0.615)

0.00 (0.568)

Component city health expenditure per capita

0.00 (0.060)

0.00 (0.064)

0.00 (0.050)

0.00 (0.090)

0.00 (0.109)

0.00 (0.373)

Doctors per 10,000 population

0.04 (0.878)

0.04 (0.873)

-0.05 (0.866)

0.03 (0.911)

0.04 (0.878)

0.08 (0.784)

Midwives per 10,000 population

-0.03

(0.047) -0.03

(0.050) -0.02

(0.087) -0.03

(0.098) -0.02

(0.164) -0.02

(0.372) Poverty incidence

-0.00 -0.00 -0.00 -0.00 -0.00 -0.00(0.457) (0.501) (0.330) (0.315) (0.316) (0.363)

Proportion of women of reproductive age

0.55 (0.610)

0.54 (0.626)

0.76 (0.461)

0.78 (0.441)

0.76 (0.453)

0.55 (0.659)

Proportion of mayors belonging to same party as governor

0.01(0.954)

-0.02 (0.835)

-0.03 (0.796)

-0.04 (0.716)

-0.03 (0.847)

Proportion of mayors in province re-elected

0.21(0.090)

0.21 (0.091)

0.27 (0.089)

0.27 (0.108)

Governor re-elected 0.04(0.359)

0.05 (0.274)

0.06 (0.274)

Governor is in first term 0.06 (0.458)

0.03 (0.733)

Proportion of mayors in first term

0.11(0.446)

0.10 (0.543)

Governor is in last term -0.04 (0.578)

Proportion of mayors in last term (0.867) Constant

0.27 0.27 0.20 0.15 0.06 0.12(0.374) (0.385) (0.472) (0.600) (0.820) (0.732)

Observations 144 144 144 144 144 139Number of groups 72.00 72.00 72.00 72.00 72.00 72.00R2 0.25 0.25 0.33 0.34 0.37 0.40F-statistic (2.69) (2.37) (4.18) (2.73) (2.10) (1.61)p-values in parentheses

19