20498 vol. 2 May 31, 2001 PHUIPPINTES POVE RTY AsSESSMENT Volume II: METHODOLOGY The World Bank X X 1 2 e ~~~~~~~~j i PHILIPPINES POVERTY ASSESSMENT Volume II: Methodology May 31, 2001 Poverty Reduction and Economic Management Unit East Asia and Pacific Region The World Bank Currency Equivalents (As of May 31, 2001) Currency Unit Peso $1.00 50.58 Pesos 1.00 peso = $0.019 Fiscal Year January 1 - December 31 Jemal-ud-din Kassum Vice President, EAP Vinay Bhargava, EACPF Country Director Homi Kharas, EASPR Sector Director Tamar Manuelyan Atinc, EASPR Task Manager ii Abbreviations and Acronyms ADB Asian Development Bank APIS Annual Poverty Indicators Survey ASEM Asia-Europe Summit ARMM Autonomous Region of Muslim Mindanao CARL Comprehensive Agrarian Reform Law CARP Comprehensive Agrarian Reform Program CBPIMS Community-Based Poverty Indicators Monitoring System CDA Cooperative Development Authority CDF Cumulative Distribution Function CHED Commission on Higher Education CIDSS Comprehensive and Integrated Development of Social Services CMP Community Mortgage Program CPI Consumer Price Index DAR Department of Agrarian Reform DECS Department of Education, Culture and Sports DENR Department of Environment and Natural Resources DHS Demographic and Health Survey DILG Department of Interior and Local Government DOH Department of Health EGS Employment Guarantee Scheme ERAP Enhanced Retail Access for the Poor FGT Foster-Greer-Thorbecke FIES Family Income and Expenditure Survey GASTPE Government Assistance to Students for Private Education GDP Gross Domestic Product GNP Gross National Product HDI Human Development Index IMF International Monetary Fund IRA Internal Revenue Allotment LG Livelihood Groups LGC Local Government Code LFS Labor Force Survey LGU Local Government Unit MBN Minimum Basic Needs MBN-CBIS Minimum Basic Needs-Community-Based Indicator System MOOE Maintenance and Other Operating Expenditures MTPDP Medium-Term Philippine Development Plan hii ABBREVIA TIONSANDACRONYM5 NA National Accounts NAPC National Anti-Poverty Commission NCR National Capital Region NEAT National Elementary Achievement Test NEDA National Economic and Development Authority NFA National Food Authority NGO Non-Government Organization NPR National Protection Rate NSAT National Secondary Achievement Test NSCB National Statistical Coordination Board NSO National Statistics Office OL Operation Leasehold OLS Ordinary Least Square OLT Operation Land Transfer PD Presidential Decree PO People's Organizations GR Quantitative Restrictions SIR Slum Improvement and Resettlement SRA Social Reform Agenda SUC State Universities and Colleges SWS Social Weather Stations UNDP United Nations Development Programme US United States WTO World Trade Organization iv Table of Contents PROFILE OF THE POOR 1 Who Are the Poor 4 Household characteristics 4 Determinants of living standards 8 Regional dimension and social indicators 9 What Does It Mean To Be Poor 9 Access to Public Services 11 Endnotes 13 GROWING OUT OF POVERTY? 15 Consumption Poverty 16 Growth, Inequality and Trickle down? 17 Sustained Poverty Reduction? 18 Income vs. Consumption Poverty 19 Social Indicators and Self-rated Poverty 20 Rural-Urban Trends 22 Sectoral Composition of Poverty 24 Regional Disparities 25 Endnotes 31 SOCIAL SERVICES AND THE POOR 33 Education, Labor Markets, and Poverty 33 Patterns of expenditure on education 34 Educationalattainment 36 Labormarket outcomes 39 Health Services and the Poor 45 Endnotes 49 TARGETEDPOLICIESANDTHE POOR 54 Rice Policy 55 Agricultural protection 55 NFA domestic market operations 56 Land Reform and Housing 57 Intergovernmental Transfers 64 Endnotes 70 THE CRISIS AND THE POOR 72 Welfare Impact of the Crisis 73 Government and Household Response 79 Endnotes 86 ANNEX A. CONSTRUCTING ABSOLUTE POVERTY LINES: ADISCUSSION OF METHODOLOGY 90 Household Surveys 90 Poverty Lines 91 Food thresholds 92 Non-food component 93 Real Expenditures and Cost-of-Living Indices 94 ANNEXB. POVERTYAND SOCIAL INDICATORS 100 v TABLE OFCONTENTS TABLES Table 1.1: Poverty in the Philippines 2 Table 1.2: Who are the poor? 4 Table 1.3: Poverty by class of worker, 1 997 4 Table 1.4: Poverty by sector of employment, 1997 5 Table 1.5: Poverty profile and household characteristics, 1997 7 Table 1.6: Spatial and household characteristics and living standards, 1997 8 Table 1.7: Pattern of spending and behavior for the poor 10 Table 1.8: Housing condition by expenditure decile, 1997 10 Table 1.9: Households and transfers, 1997 10 Table 1.10 : Private transfers benefit the poor, 1 997 11 Table 1.11: Public services and the poor, 1998 11 Table 1.12: Distribution of public infrastructure, barangay level information, 1 998 12 Table Al.1: Definition of variable used in the log consumption model (Table 1.6) 12 Table 2.1: Average living standards, poverty and inequality, 1985-1997 16 Table 2.2: Evolution of select social indicators 21 Table 2.3: Living standard, poverty, and inequality by locality, 1985-1997 23 Table 2.4: Decomposition of inequality in per capita consumption, 1988-97 26 Table 2.5: Convergence of poverty and living standards across provinces, 1988-97? 27 Table A2.1: Trend and seasonal components of variables in the analysis 28 Table A2.2: Correlates of self-rated poverty 28 Table A2.3: Correlates of self-rated poverty lines 29 Table A2.4: Sectoral poverty profile, 1985-1997 30 Table 3.1: Government spending in education, 1997 34 Table 3.2: Direct private cost of education, by education level and income quintile 38 Table 3.3: Rates of return to education and experience 39 Table 3.4: Rates of return to education and experience, by gender 41 Table 3.5: Total government spending on health 45 Table 3.6: Health status and the poor 46 Table 4.1: Nominal protection rates of major agricultural commodities 55 Table 4.2: NFA rice procurement and distribution 56 Table 4.3: NFA rice distribution and poverty across regions 57 Table 4.4: Land distribution status by land type and mode of coverage, 1972-1997 60 Table 4.5. Original tenure status of households who benefited from land reform, 1972-1988 61 Table 4.6: Determinants of income, 1985 and 1998 61 Table 4.7A: Transition matrix for movement in land tenure status between 1971 and 1988 63 Table 4.7B: Transition matrix for movement in land tenure status between 1988 and 1998 64 Table A4.1: Socio-economic characteristics and land tenure status in sample villages across three generations 67 Table A4.2: Household income, assets, and production structure, in sample villages, 1985 and 1998 68 Table A4.3: Impact of the land reform on human capital formation, asset accumulation, and long-term productivity and income 69 i TABLE OFCONTENTS Table 5.1: Indicators of economic activity and household impact during the crisis 73 Table 5.2: The incidence of crisis-related economic shocks 74 Table 5.3: Impact of the crisis on consumption poverty and inequality 76 Table 5.4: Impact of economic crisis and El Nino 77 Table 5.5: Impact of the crisis on income poverty and inequality 78 Table 5.6: Timeline of the crisis and response 79 Table 5.7: Social sector expenditures, 1997-98 (in 1995 million pesos) 80 Table 5.8: Household responses to crisis 81 Table 5.9: Employment impact of the crisis 82 Table A5.1: Descriptive statistics of model variables (1998 APIS) 84 Table A5.2: The estimated consumption and income models (1998 APIS) 85 Table Annex A.1: Estimates of food thresholds and poverty Lines: Absolute cost-of-basic-needs approach (1997, pesos per capita per year) 95 Table Annex A.2: Mean expenditure, cost-of-living index and living standard, by province (1997, pesos per capita per year) 97 Table Annex A.3: Regional cost-of-living indices (NCR 1997 = 100) 99 Table Annex B.1: Consumption distribution in the Philippines, 1985-1997 101 Table Annex B.2: Income distribution in the Philippines, 1985-1997 102 Table Annex B.3: Provincial living standard and poverty, 1997 1 03 FIGURES Figure 1.1: Poverty and household size in the Philippines, 1997 6 Figure 1.2: Poverty thresholds and household size 6 Figure 1.3: Poverty and social indicators, 1997 9 Figure 2.1: Annual rate of decline in poverty incidence in East Asia, mid-1 980s to the mid-i 990s 16 Figure 2.2: Pro-poor growth in East Asia (mid-1980s to the mid-1990s)? 17 Figure 2.3: Growth and poverty reduction,1988-97 18 Figure 2.4: Income and consumption inequality, mid-80s and mid-90s 18 Figure 2.5: Evolution of growth, inequality and poverty in the Philippines: 1985-97 19 Figure 2.6: Pro-poor growth in the Philippines,1985-97? 19 Figure 2.7: Income vs. consumption poverty: does it matter? 20 Figure 2.8: Income vs. consumption growth and inequality: does it matter? 20 Figure 2.9: Subjective poverty and seasonally adjusted real GDP per capita 21 Figure 2.10 Subjective poverty lines and real GDP per capita 22 vWi TABLE OF CONTENTS Figure 2.11: Urban-rural disparities: mean living standards, inequality and poverty, 1985-97 23 Figure 2.12: Sectoral shares in the number of poor and the aggregate poverty gap, 1985-97 24 Figure 2.13: Sectoral development of the economy,1984-98 25 Figure 2.14: Change in poverty incidence across provinces, 1988-97 26 Figure 2.15: Evolution of poverty and living standards across regions, 1988-97 27 Figure 3.1: Public spending on education 35 Figure 3.2: Tertiary scholarships 35 Figure 3.3: Gross secondary enrollment and income, various countries 36 Figure 3.4: Educational attainment, ages 15-1 9 Figure 3.5: Educational attainment by birth cohort 36 Figure 3.6: Educational attainment, ages 15-19, by gender and place of residence, 1998 37 Figure 3.7: Proportion currently enrolled, ages 6-14, by wealth group, 1998 37 Figure 3.8: Average years of schooling attained, by age and income quintile, 1 998 37 Figure 3.9: Mean log hourly wages, by education and experience 43 Figure 3.10: Net present value of lifetime wage earnings, by educational attainment and discount rate 44 Figure 3.11: Public health spending benefits the poor overall, 1998 46 Figure 4.1: Share of rice in total consumption and net consumption of rice, by deciles, 1997 55 Figure 4.2: Housing and land reform: the tale of two 'targeted' programs 58 Figure 4.3: Land redistribution without extension services ? 62 Figure 4.4: Cultivated versus owned-land, 1985 and 1998 63 Figure 4.5: IRA allocations are unrelated to poverty 65 Figure 4.6. Poverty and economic classification of provinces 65 Figure 5.1: Change in the cumulative distribution function due to the crisis 77 Figure 5.2: Percentage change in the cumulative distribution function due to the crisis 77 Figure 5.3: The relative magnitudes of income and consumption shocks 79 BOXES Box 1.1: Choosing a welfare measure 2 Box 3.1: Decentralization, redistribution and the provision of social services 48 Box 4.1. Land reform in the Philippines 59 Box 4.2: The design of a system of intergovernmental grants 66 viii PROFILE OF THE P 1.1 A quarter of the Filipino populatio about 18.2 mnillion Filipinos, were deemed to b in 1997 (Table 1. 1). The rural povertyincidence at 37 percent while the urban incidence is about 12 pr cent. The national poverty gap index is estim-ated at 6.4 percent which, with a headcount index of 25 percent, implies that on average the consumption of the poor is about three- quarters of the poverty line. 1.2 These poverty estimates are based on data from the 1997 Family Income and Expenditure Sur- vey (FIES), and are constructed using consumption per person as a measure of welfare (see Box 1.1). The poverty line is based on a given standard of living, which is fixed across space and time. Provin- cial povertylines are constructed following the widely used cost of basic needs approach. This involves: (i) setting a food bundle in each province based on the average consumption pattern of a reference group of poor households fixed ntionally in terms of their . per capita consumption; (ui) adjusting this bundle to satisfythe minimumnutritional requirement of 2,000 calories per person per day, (iii) valuing the adjusted bundle at consumer prices prevailing in each prov- inceto derive provncial foodpovertylines;(iv)con- structing the non-food poverty line for ec rv ince as the non-food spending of those hueo4 whose toual spending is in the neighborhodfth food threshold; and, (v) finally deriving the provin- cial poverty lines as the sum of food and non-food - \ poverty lines. The approach fixes the standard of living used for provincial compan'sons but not the composition of the consumption bundle used in each X province. Differences in composition may arise as a result of spatial differences in tastes or relative prices faced by households.' PHILIPPINES POVERTY ASSESSMENT Table 1.1: Poverty in the Philippines 1.3 The estimates of poverty based on this ap- Headcount Poverty Squared poverty proach are substantiallylower than the official esti- index gap index gap index mates. For instance, the official incidence of pov- (Incidence) (Depth) (Severity) erty for 1997 is 36.8 percent (NSCB, 1999a). The 1997 25.1 6.4 2.3 difference is mostly due to the higher official pov- Rural 36.9 10.0 3.6 erty lines, rather than the use of per capita income Urban 11.9 2.6 0:9 as the measure of welfare in the official estimates. 1998 (projection)* 27.8 7.6 2.9 For instance, the application of basic needs pov- 1999 (projection)* 26.3 6.9 2.5 erty lines (underlying Table 1.1) to per capita in- Note: rhehead ountindexmeasuresthe share opopulation wthconsumption (income) coe ields a headcount of 22.1 percent (Balisacan, levels belowthepovertythireshold Thismeasure gives the prevalence of poverty butis c e silent on howpoorthepoorare Thepovertygapiidexisdefinedasthe meandistance 1999a). On the other hand, the official regional pov- belowthepoveitylineasapropoition of thatline (where the non-poorare countedashaving zerpovedrtygaps),andgivesameasureofthe 'depth'ofpoveityThesquaredpovertf,gap erty lines are 15-68 percent higher than the basic indexis defined asfthemean of the squaredproportionate poveitygaps, is sensitive to used thisReport. higher astfnbutionbelowthepovertfyihe, andretiectsthe 'severityofpoverfyAllpoverty,mea- needs povertyimes uR The sures are exprssedaspercentages 'Projectionsbasedon I997FIES data andgrowth official poverty lines reflect the use of a relatively rates foragriculture, industry and services sectors for 1998 and t999 taken from the national accounts. The projections assume no change in relative inequalities within more expensive food bundle, which corresponds Source Staffestimatesbasedon 1997FlESdataandnatonalaxcountsdatafromNSCB. to the aTerag food bundle within a region rather Box 1.1: Choosing a Welfare Measure Identification of the poor requires the use of an indicator of a household or individual's standard of living. Several such indicators are used in the literature on poverty. One popular approach uses consumption or income as the relevant indicator; this so called 'money metric utility' approach conceives well-being as the ability to fulfill certain preferences which are represented in terms of equivalent income or consumption. Much of the analysis in this report is based on this approach. Income or Consumption? Both measures are used in the literature and by various countries in tracking poverty. There are some conceptual as well as practical reasons to favor a consumption-based standard over one using income. First, income may overestimate or underestimate living standards. If a person can borrow or use his savings, his level of living is not constrained by current income. Even in underdeveloped regions, households typically have some ability to buffer their welfare against temporary variations in income thorough saving or borrowing. Second, a strong case can be made for preferring consumption over income as a broad indicator of welfare based on practical data considerations. It is more difficult to acquire accurate information on income than on consumption. For example, one has to undertake multiple household visits or use recall data to obtain reasonably accurate estimate of annual income whereas consumption over, say, the previous few weeks can provide a satisfactory measure of welfare. Moreover, households may understate their incomes to avoid future problems with tax agencies - a common practice especially among self-employed professionals. The difficulty also extends to imputing "incomes" of households, which consume part of their production, such as the case for the large majority of the farming population. Owing partly to cost considerations, the survey instrument used by statistical agencies to acquire information on households is often short on details needed to accurately estimate "net income" from own-production activities, especially farming. It is useful to supplement income/consumption based measures of welfare with others which are derived from a capabilities- based approach to welfare; the most commonly used ones relate to educational attainment and measures of health. In this framework, education and health are important not only because they influence well-being in the income space but also because they are desirable in their own right. An individual who dies prematurely or suffers from severe ill health is thus deemed poor in the capabilities space. This report also makes use of non-income indicators of well being. Other approaches to welfare emphasize the ability to participate in activities and enjoy living standards that are customary or widely accepted in society (social exclusion approach) and are akin to the concept of relative deprivation. Further, the poor themselves appear to focus prominently on social deprivation, as exemplified by the lack of dignity, self respect, security and justice, in defining poverty. In a participatory study conducted at 486 sites in 23 countries, poor people were asked to analyze and share their ideas of well-being and ill-being. Well-being was variously described as happiness, harmony, peace, freedom from anxiety, and peace of mind whereas people described ill-being as lack of material things, as bad experience, and as bad feeling about the self. The self-rated poverty indices analyzed in this report come closest to measuring this aspect of deprivation. Moreover, the report's discussion of inequality with respect to assets and their returns is also grounded in part in a concern over relative deprivation. The analysis of these wider dimensions of well being in this report is nonetheless limited, for want of readily available and reliable data. Source: WodBanOhk20O0f), Balisacanwf.999b). 2 1: PROFIL E OF THE POOR than the food bundle of the poor in the region.' lute poverty, the poverty threshold vanies from Similarly, the differential between the official and household to household based on their own per- basic needs poverty lines tends to be greater for ception of their povertystatus. The main reasonwhy the richer regions,3 again reflecting the relatively self-rated poverty incidence is so much higher than more expensive food bundles typically consumed the income or consumption-based poverty is that in the richer regions. Since the allowance for basic the self-rated poverty thresholds are generally con- non-food expenditure (i.e., the non-food poverty siderably higher than the thresholds used for esti- line) is tied to the food threshold itself, the higher mating income/consumption poverty Thus, for in- food poverty line under the official approach fur- stance, for the National Capital Region (NCR), the ther translates into a higher overall poverty thresh- median self-rated poverty threshold for June-Sep- old. tember 1997 was PhP10,000 per month, while the basic needs and the official 1997 poverty lines for 1.4 1997 is the most recent year for which the NCR were PhP4,495 and PhP6,077 per month re- FIES data are available.4 However, projections for spectively 1998 and 1999 can be made using these data and the sectoral growth rates available from the national 1.6 There are also very large urban-rural dif- accounts. Assuming that relative inequalities within ferentials in the self-rated poverty thresholds. For sectors do not change and that average consump- June-September 1997, the median urban threshold tion growth for households in a given sector mir- for NCR (Metro Manila) was twice as high as the rors the growth of value added within that sector, median threshold in each of Luzon, Visayas and the projected levels of poverty for 1998 and 1999 Mindanao regions. Since there are urban areas are shown in Table 1.1. The poverty estimnates indi- within these regions, the typical urban-rural differ- cate that with the contraction due to the crisis (a 2.7 ential would be even higher. This differential seems percent decline in per capita GDP in 1998), the inci- to be significantly higher than the cost of living dence of poverty increased in 1998 to 27.8 percent,5 differential implied bythe official or the basic needs while subsequent recovery (a 1.1 percent increase in poverty lines. The official NCR poverty line was per capita GDP during 1999) helped bring the pov- 24 percent higher than Luzon poverty line, 53 per- erty incidence down though still somewhat higher cent higher than the Visayas line and 35 percent than its pre-crisis level. The agricultural sector played higher than the Mindanao line. Similarly with the an important role in both the contraction and the basic needs poverty lines, these differentials were subsequent recovery The growth rates of value 21, 42 and 42 percent, respectively Thus, the ur- added in agricultural, industrial and service sectors ban-rural differential in self-rated poverty thresh- were -6.8, -1.9, and 3.5 percent, respectively during olds is two to five times higher than the cost of 1998, and 6.7, 0.5, and 3.9 during 1999. living differential, and it almost certainlyreflects the higher aspirations of city dwellers. Some of these 1.5 Boththe povertyestirates reported in Table higher aspirations may also be fueled by the rela- 1.1 and the official povertyestimates in turn are sub- tively higher levels of inequality in urban areas; ur- stantially below the estimates of self-rated poverty ban Gini indices of per capita consumption in 1997 based on surveys conducted by the Social Weather were about 30 percent higher than the Ginis in ru- Stations (SWS) .6 The average rates of self-rated pov- ral areas. If urban people aspire for higher living ertyincidence for the years 1997 through 1999 were standards, it is not surprising then to find that the 59.3, 61.0 and 61.7 percent, respectively. Self-rated rural-urban differentials in self-rated poverty rates poverty is based on households responses to the are considerablynarrower (55 percent urban against question as to where they would place their fanily 70 percent rural during March-October 1999) than on a card mnarked "not poor," "on the line," and those implied by the non-self-rated poverty mea- "poor" (SWS, 1999). This is not a rneasure of abso- sures. 3 PHILIPPINES POVERTY ASSESSMENT WHO ARE THE POOR? educational attainment and experience, sex, civil sta- tus, and economic sector of employrnent, at least in 1.7 Povertyinthe Philippines is stillalargelyrural the short term. But the educational attainment of phenomenon despite rapid urbanization in recent the household head is the single most important con- years. The rural poor account for about 77 percent tributor to the observed variation in household wel- of the poor (Table 1.2). Other poverty measures fare. show the same order of magnitude. Within rural areas, povertyis largelyagriculture-driven. Xile ag- 1.10 The traditional characterization of the poor riculture-dependent households represent only 40 is that the poorest of them are the landless and those percent of the total population, the sector accounts dependent mainly on wage incomes (see for ex- for over two-thirds of the poor, simply because ample, Hayami, et. al., 1990). Surprisingly Table 1.3 povertyincidence is higher in agriculture than in any shows that the depth and severity of poverty among other sector of the economy Agriculture accounts the self-employed are at least as high as "wage" for an even higher share of the depth and severity households. In agriculture, the poor self-employed of poverty (Table 1.2), heads of households include primarily lessees, ten- ants, and small owner-cultivators. They account for 1.8 There are pronounced regional differences over 50 percent of the country's poor population. in poverty rates reflecting only in part differential Any serious effort aimed at addressing the poverty levels of urbanization and reliance on agriculture for problem in the Philippines must grapple with both income. The poverty headcount ranges from 3.5 the fundamental causes of underdevelopment in ag- percent in Metro Manila where 14.1 percent of the riculture and rural areas and the relatively slow pace country's population resides, to 87.5 percent in Sulu of structural transformation from an agriculture- province in ARMM (with less than 1 percent of the based-economy to one which is more reliant on the population). Among the regions, Bicol hosts the larg- higher productivityindustrial and service sectors. est number of poor although poverty incidence is somewhat higher in Eastern Visayas and ARMM. 1.11 It is interesting to note that while 12.7 per- cent of the population live in households where the Household characteristics head is unemployed or not working, povertyin this segment of the population is substantially belowthe 1.9 Household welfare vanres systemnaticallywith national average; the headcount index is 25 percent certain demographics, includg the household head's for the country as a whole and only 12.1 percent for Table 1.2: Who are the Poor? Table 1.3: Poverty by Class of Worker, 1997 (Data from 1997) Share among Share in total Contribution to the poor (%) population (%) Population Poverty total poverty' Households living in: Share Incidence Depth Severity Incidence Depth Severity Rural areas 77.4 52.5 Wage earners 52.7 17.6 4.2 1.5 37.2 34.6 34.4 Metro Manila 2.0 14.1 (0.36) (0.11) (0.05) CentralLuzon 5.4 10.3 Agriculture 7.8 43.8 11.7 4.4 13.7 14.3 14.9 Bicol 12.8 7.1 (1.25) (0.47) (0.24) EastrMMisya 60.0 3.1 Non-agriculture 44.9 13.1 2.9 1.0 23.5 20.3 19.5 (0.34) (0.10) (0.04) Houiseholds whose headis: Se/f-employed 46.7 33.5 8.9 3.3 62.6 64.6 66.8 Employed in Agriculture 67.8 40.1 (0.46) (0.16) (0.08) Self-employed 62.6 46.7 Agriculture 32.0 42.1 11.4 4.3 53.9 57.0 59.8 Elementarygraduateorless 75.6 51.1 (0.59) (0.21) (0.11) Male 92.4 87.8 Non-agriculture 14.7 14.8 3.3 1.1 8.7 7.6 7.0 Age 30-50 65.2 55.7 (0.62) (0.17) (0.08) Age 60+ 10.5 15.5 Note Figuresnpaentheses are robos standard errors correcledforsaple desa p effect) Source: Basedon 1997FlES data; Ba/isacan (1999a). Source: Ba/lsacan (1999a) basedon 1997F/ES 4 1: PROFIt F OF THE POOR flects the importance of private transfers for house- Table 1.4 Poverty by Sector of Employment, 1997IM holds with unemployed or elderlyheads. Regression Popultion Poverty Contribution to analysis of deterrnminants of transfers using 1997 data Share Incidence Depth Severity Incidence Depth Severit finds that having an unemployed or elderly head of EmployedHHhead household has a large and significant positive im- Agriculture 40.1 42.3 11.5 4.3 67.8 71.9 74.5 pact on transfers received bythe household, almost (0.53) (0.19) (0.10) all of which are from private sources. With regard Mining 0.6 30.0 10.0 4.5 0.7 0.9 1.1 (4.21) (2.02) (1.15) to unemployment, while the relativelyhigh rates may Manufacturing 7.0 13.5 2.7 0.9 3.8 2.9 2.6 signal certain inefficiencies in the labor market, they (0.90) (0.24) (0.10) do not appear to present an important povertycon- Utility 0.7 9.5 2.4 0.9 0.3 0.3 0.3 (2.17) (0.62) (0.28) cern, in part because informal/private transfers may Construction 7.7 23.1 5.0 1.6 7.1 6.1 5.4 be playing an important safetynet function.7 (1.03) (0.30) (0.13) Trade 8.8 13.5 2.9 0.9 4.7 4.0 3.5 (0.76) (0.20) (0.08) 1.12 Private transfers may be dampening pov- Transport 8.0 13.7 2.8 0.9 4.4 3.5 3.2 ertyfor two additional categories of households that (0.82) (0.23) (0.10) Finance 1.9 3.0 0.5 0.1 0.2 0.1 0.1 are commonly targeted through public programs, (0.70) (0.13) (0.04) especiallyin developed countries: households headed Services 12.5 9.9 2.2 0.7 4.9 4.4 4.0 byfemnalesandtheelderlyPovertyis higherfornmale- (0.59) (0.16) (0.07) HH head not 12.7 12.1 2.9 1.0 6.1 5.9 5.3 headed than for female-headed households in the employeot (0.61) (0.18) (0.08) Philippines, irrespective of the povertymeasure and Total Population 100.0 25.0 6.4 2.3 100.0 100.0 100.0 . . (0.29) (0.01) 0.05 age of the household head. Similarly the elderly Note: 'Includes unemployedoas well as those not working Figures in parentheses are robust (households where the head is aged 60+) have a standard errors (corrected for sample design effect). lower than average incidence of poverty (Table 1.2 Source:- Ba/sacan (1999a) based on f997 F/ES. and 1.5). This positive association between house- those not employed (Table 1.4). While available data hold living standards and female/elderly headship do not allow us to distinguish the unemployed from holds even after controlling for other characteristics those not in the labor force, it is possible to split the (Table 1.6). Both results may be reflecting the im- non-working heads categoryinto younger and older portance of private safety nets in reducing poverty (those above and below the age of 60) cohorts. The in these segments of the population. Regression non-working older cohorts are more likely to be analysis of the determinants of transfers suggests outside the labor force while the younger non-work- that female-headship has a large and significant posi- ing cohorts are more likely to be unemployed. tive impact on transfers received bythe household. However, the poverty incidence for both groups Similarly (private) transfers increase significantlyfor turns out to be very similar: 11.9 percent for the households headed by individuals aged 60+ after younger non-working cohort and 12.3 percent for controlling for pensions. This however should not the older non-working cohort. This is consistent with be construed to imply that poverty is not an issue two observations. First, it is in line with the oft-quoted for women and the elderlyin general, because both statement for low income countries that "the poor these groups often belong to poor households and cannot afford to be unemployed." It is also consis- also because intra-household allocations could be tent with the finding that the rate of unemployment biased against them even within non-poor house- in the Philippines increases with the level of school- holds. ing, and peaks among those who reach but fail to finish college. In January 1997, the unemployment 1.13 The educational attainment of the house- rate among those with elementaryeducation was 5.6 hold head is negatively correlated with poverty sta- percent while it was as high as 12.8 percent among tus, as expected. The poorest households are those those with some college education. Second, it re- whose heads did not receive any formal schooling, 5 PHIL IPPINES POVERTYASSESSMENT although these contribute less than 10 percent to national poverty. It is the households whose heads Pf p W had no more than elementary education that con- ' 50 tribute the bulk - about 75 percent - of total s1.0 0 =. poverty when poverty is decomposed by the edu- 3 x =0. , . . * ° 20 - As, S _ ,4_ t\ &\\ =0.6 cational attainment of the head. 20 * A0.4 1.14 Poverty appears positively correlated with C l2 345 6 7 80101112 household size, such that it is highest among house- Household size holds with seven or more members, while families Note., ; is the parameter for scale ec2nomies intmousehold size; a value of 1pmplies no with five or more members, account for over three- eeomdesandavalulnoempledspeteco7om Thegrhsasoaassurm, #atchildren fourths of total poverty (Table 1.5). But this result is Source: Staffcomputationbasedoon f997FESdafa driven by the use of per npzta consumption as the welfare measure, which is increasingly questioned in results for rneasures of the depth and severity of the lterature because it does not adjust for house- verty e similar.This pattern also holds at other hold ~ ~ . si. an.opsto LnowadBvln values of equivalent scales; for instance, it also holds 1995; Deaton and Zapision 19;Daonw andRa , if children's needs are set at half of those for adults. 1999). The use of per capita incomne or consump- tio9naste welofa mer makitain les wor a sumptions 1.16 What does a value of 0.6 for the scale tion as the welfare measure makes two assumptions: cnrisprmtrip?Sneiutaiecl (i) that all household members have the same needs, economies parameter imply? Some illustrative cal- regardless of whether they are adults or children (all culations may be helpful. For instance, if the mini- menbers are adult equivalent); and, (ii) that for any mum requirements for a family of 5 were fixed at, * , ~~~~~~~~~~~~~~~~~~~~~say, 100 units, then 0 = 0.6 wvould iml that the given composition, a household's needs imcrease pro- siply portionallywith household size (no scale economies requirements for a family of 10 would be 152 (and portionallywith household size (no) scale economies in household size)8 Anot 200 as implied by the per capita rule), and 58 for a family of 2 (not 40 as with the per capita rule). 1.15 Sensitivityanalysis us.ng aplausible range of The critical value of 0.6 at which the poverty-house- ,, . 1 . . . ~~~~~hold size relationshipbetween tends to disappear is parameters (for adult equivalence and scale econo- p pp rnies) shows that the link between household size similar to that obtained for Pakistan (Lanjouw and and poverty is tenuous. It turns out that while the Ravallion, 1995), though it is higher than the critical positive relationship between povertyand household value of 0.4 obtained for Mozambique (Datt, et. al., size is quite robust to alternative values for the equiva- 2000). For the Philippines itself, there is some evi- lent scale parameter, it is quite fragile with respect to dence that bears on this issue, based on surveys of the assumed magnitude of economies to household self-reported poverty thresholds conducted by the size. Figure 1.1 shows how the relationship changes for values of scale econonuies parameter ranging Fg~ 2 QI h~iIeadWueQdS~ from one (i.e., the per capita case), to 0.8, 0.6, and 0.4, while the adult-equivalence parameter is fixed at 200- No scale 0.8 (i.e., children's needs9 are fixed at 80% of adult 150 Scaleecon needs). The results show that as economies of scale E _parm.=0.6 are allowed to increase, the positive poverty-house- 1,0 8 Metrate a hold size relationship first flattens out, tending to 50 vanish for the scale economies parameter of 0.6. Then, for still greater economies of scale lower val- 2 ues of the scale economies parameter 0, in Figure 1.1), the relationship turns into a negative one. The Source Staff computations. Self-ratedpovertythresholdbasedonMangahas(t992) 6 1: PROFILE OF FtH POOR Table 1.5: Poverty Profile and Household Characteristics, 1997 Population Poverty Contribution to total poverty Share Incidence Depth Severity Incidence Depth Severity National 100.0 25.0 6.4 2.3 100.0 100.0 100.0 Sex and age group Male 87.8 26.4 6.7 2.4 92.4 92.5 92.7 (0.32) (011) (0.05) Below 20 0.0 17.5 5.8 2.5 0.0 0.0 0.0 (7.93) (3.14) (1.59) Between 20 to 30 6.8 27.0 6.3 2.1 7.3 6.7 6.2 (0.99) (0.30) (0.14) Between 30 and 40 24.7 32.4 8.8 3.3 31.9 33.8 35.3 (0.63) (0.23) (0.12) Between 40 and 50 26.5 28.8 7.4 2.7 30.4 30.9 30.9 (0.63) (0,21) (0.10) Between50and60 18.2 20.1 4.9 1.8 14.6 14.1 13.8 (0.66) (0.20) (0.09) Greaterthan60 11.6 17.5 3.8 1.3 8.1 7.0 6.3 (0.72) (0.20) (0.09) Female 12.2 17.0 3.8 1.3 7.6 7.5 7.3 (0.67) (0.21) (0. 10) Below 20 0.0 9.6 2.8 0.8 0.0 0.0 0.0 (8.98) (2.60) (0.75) Between 20 to 30 0.5 9.7 1.8 0.6 0.2 0.1 0.1 (2.45) (0,58) (0.24) Between 30 and 40 1.9 15.7 4.6 1.8 1.2 1.4 1.4 (1.76) (0.66) (0.33) Between 40 and 50 2.6 16.4 4.6 1.7 1.7 1.9 1.9 (1.53) (0.52) (0.24) Between 50 and 60 3,2 15.8 3.6 1.2 2.0 1.8 1.7 (1.35) (0.39) (0.17) Greater than 60 3.9 15.6 3.7 1.3 2.4 2.3 2.1 (1.12) (0.33) (0.15) Maritalstatus Single 2,0 9.6 2.2 0.8 0.8 0.7 0.7 (1.24) (0.34) (0.15) Married 86.7 26.1 6.7 2.4 90.3 90.6 91.0 (0.32) (0.11) (0.05) Widowed 10.1 20.4 5.1 1.8 8.2 8.0 7.7 (0.82) (0.25) (0.11) Divorced/Separated 1.1 14.6 3.8 1.3 0.6 0.7 0.6 (2.16) (0.71) (0.28) Unknown 0.1 27.3 5.5 1.9 0.1 0.0 0.0 (10.74) (2.81) (1.19) Educational attainment No education 3.8 46.2 13.8 5.5 7.0 8.1 9.0 (1.48) (0.59) (0.31) Elem. Undergraduate 22.8 41.6 11.4 4.4 37.8 40.8 42.9 (0.70) (0.26) (0.14) Elementary graduate 24.5 31.4 7.8 2.8 30.8 29.9 29.2 (0.66) (0.22) (0.10) HS undergraduate 11.5 24.4 5.7 1.9 11.2 10.3 9.5 (0.86) (0.26) (0.12) High school graduate 18.6 13.5 2.9 0.9 10.0 8.5 7.5 (0.53) (0.15) (0.06) College undergraduate 10.5 6.3 1.2 0.3 2.6 1.9 1.5 (0.47) (0.10) (0.04) Degree holder 8.2 1.7 0.30 0.1 0.6 0.4 0.3 (0.31) (0.07) (0.02) Not reported 0.2 3.1 0.0 0.0 0.0 0.0 0.0 (3.05) (0.04) (0.0) Family size 1-2 10.4 8.3 1.7 0.5 3.5 2.8 2.5 (0.39) (0.10) (0.04) 3-4 32.4 15.8 3.3 1.0 20.4 16.9 14.4 (0.37) (0,10) (0.04) 5-6 32.2 28.0 6.9 2.4 36.0 35.0 34.0 (0.52) (0.17) (0.08) 7-8 16.9 39.4 11.4 4.5 2&6 30.0 32.7 (0.90) (0.33) (0.17) 9&above 8.2 41.3 11.8 4.6 13.6 15.2 16.5 (h 15) (0.53) (0.27) Type of household Single family 73.2 27.7 7.2 2.6 80.9 82.2 82.8 (0.34) (0.12) (0.06) Extended family 26.4 18.0 4.3 1.5 18.9 17.6 16.9 (0.54) (0,16) (0.07) With unrelated members 0.4 7.1 (2.6) 1.2 0.1 0.2 0.2 (2.27) (1.02) (0.58) Note: Figures in parentheses are standard errors corrected for sample design effect. Source: Ealisacan (1999aA based on the 1997FJES. 7 PHILIPPINES POVERTYASSESSMENT Social Weather Stations (SWS). This evidence sug- Table 1.6: Spatial and Household Characteristics gests that the implicit scale economies may be even and Living Standards, 1997 higher, i.e., a value of the scale elasticity parameter Contribution below 0.6. This is shown in Figure 1.2 by the flatter Regression to variance Variable Coefficient t-stat, explained slope of the self-rated poverty threshold for Metro Manila based on the September 1989 SWS Survey HOUSEHOLDA 9.918 250.30 (Mangahas, 1992). There is no conclusive evidence AGE 0.011 7.87 4.2 however in favor of higher scale econornies because AGESQ 0.000 -6.31 -3.4 MALE -0.057 -4.39 0.8 the self-rated povertythresholds themselves tend to MARRIED 0.085 7.01 -1.0 increase with household living standards. Overall, ELEM 0.160 21.13 -3.5 HIGHSCH 0.427 47.32 10.1 therefore, whle we do not have a precise estimate COLLEGE 1.025 64.29 33.7 of scale economies for the Philippines, larger house- HOUSEHOLD holds may no longer be poorer once an allowance COMPOSITION 29.8 FSIZE -0.068 -39.83 12.7 is made for such economies within what appears to CHRATIO -0.538 -30.27 3.9 be a plausible range. EMPRATIO 0.248 14.00 7.2 ECONOMIC SECTOR 11.9 AGRI -0.196 -15.38 9.6 Determinants of living standards MINING -0.103 -2.99 0.0 EGW 0.230 6.05 0.3 1.17 The foregoing discussion focused on indi- TCOANDSET 0.297 -8.84 0.2 vidual correlates of poverty In many policy con- TRANSP -0.029 -2.02 -0.1 texts, it is also important to know if the correlation FINANCE 0.127 3.82 0.6 IMP ~~~~~~~~~~SERVICES 0.018 1.26 0.3 with a particular attribute holds up when controlling UNEMP 0.038 2.48 0.5 for other attributes. This can be explored through LOCATION 174 multivariate regression analysis. The results of such URBAN 0.162 23.20 7.2 REG1 -0.159 -9.84 -0.3 an analysis confirm a number of observations made REG2 -0.279 -16.29 0.4 above concerning spatial and household correlates REG3 -0.173 -13.13 -0.6 REG4 -0.203 -16.11 -0.9 of poverty (Table 1.6).Io Given other factors, for REG5 -0.409 -26.19 2.2 instance, the household head's educational attainmnent REG6 -0.155 -10.86 -0.2 REG7 -0.362 -22.47 1.0 and experience (proxied bythe household head's age) REG8 -0.467 -28.45 3.2 positivelyinfluence household welfare. Households REG9 -0.238 -13.79 0.5 headed bymnales have lower welfare levels than those REG10 -0.1205 -12.89 0.1 headed by females, holding other factors constant. REG12 -0.272 -15.91 0.4 Household size negativelyinfluences household wel- CAR -0.119 -7.15 -0.2 ARMM -0.408 -25.33 3.8 fare as does the proportion of children m the house- CARAGA -0.307 -17.51 0.7 hold, all other things remaining the same (subject to Sample size 39,520 the caveats discussed above). But given household R-square 0.534 size, the number of employed household members Note: Dependent vanable isnatura//oganthm of(cost-oi-iving-ad/usted)percapIta house- hold expenditure The mode/ls estimated uslng Statas's ~vyreg~prceourei wb,chtakes into positively affects welfare. Together, household com- accountsample design effects (/ e, stratiffcafionandwepftfs assignedioeacb obsewvation,) position and the houschold head's characten'stics, See Table At 1 forthe definition of variables used position and the household head's characteristics, Source: Ballsacan (t999a). especiallyeducational attainment, account for roughly three-fourths of the variance explained bythe model. factor in fact contributes the bulk - about 80 per- 1.18 Economic sector and location characteris- cent - of the variance explained by the employ- tics account for another one-fourth of variance ex- ment variables. Households located in urban areas plained by the model. Employment in agriculture is tend to have higher welfare levels than those in rural negatively associated with household welfare. This areas. Location appears to be an important deter- 8 7: PROFILE OF THE POOR munant of poverty even after controlling for sector of employment and education level (Table 1.6). V - 1 Households in regions other than Metro Manila have Poorer provinces in the Philippines also have lower levels of life lower welfare levels than those in the capital region, expectancy, literacy and school enrollment. all else remaining the same. The location variable cap- tures a variety of economic, political and social fac- * . *.. .* tors that may be influencing overall welfare levels, 65- E including the state of infrastructure, productivity of :2 *r land, proximity to markets, and quality of govern- 60 rnent. Regional dimension and social indicators nc 3 2 40 60 30 100 ncidence of Poverty, 1997 1.19 The regional differences in povertyrates are mirrored in equally profound differences in other 10° social indicators. Functional literacy ranges from a a * low of 48 percent in the province of Basilan in West- 80 * * emn Mindanao, to a high of almost 93 percent in Cavite in Southern Luzon. Enrollment rates in phi- mary and secondary school range from 43 percent in the province of Suiu in the ARMM to 99 percent 50X ---- T ---'- I in the Mountain Province of the Cordillera Admin- ° Incidence of Poverty, 1997 20 100 istrative Region. And, life expectancy ranges from a low of 52 years in the province of Tawi-Tawi in the ARMM to a high of 71 years in the province of ° Pampanga in Central Luzon (Annex B, Volume II). 80 * ; * Moreover, as Figure 1.3 shows, and as confirmed 0r by formal statistical analysis, functional literacy, 60 school enrollment rates, and life expectancy are all significantly correlated with poverty. poorer prov- 40 inces in the Philippines also have significantlylower 0 20 I40 so Poery 1 9o educational attainment and worse health outcomes. This correlation points to the instrumental value of Source: UNDP Human Development Report, 1999, 1997 F/ES improving social indicators in combating income/ consumption poverty but an improvement in these food. Not surprisingly, housing is a problem for indicators, especially in regions where they have the poor as 11 percent of the population in the bot- lagged behind, is irnportant in and of itself. tom three expenditure deciles are either squatters or live in poor quality housing (Table 1.8). What is sur- WHAT DOES IT MEAN TO BE POOR? prising is the extent to which housing is a problem for the population as a whole: the prevalence of 1.20 Poverty is a way of life. It affects what illegally occupied housing is highest for those in the people do with their time and money how theylive, middle of the distribution and remains high even their command over resources, their status in the among the wealthiest (2.5 percent for the 9th decile). comunity their relations with the government and It is quite likelythat housing, especially insecurity of their sense of self-worth (Table 1.7). Typically, the tenure, is a problem particularly in urban areas and poor save less of their income and spend more on more specifically in Manila. 9 PHfLIPPINES POVERTY ASSESSMENT 0 4/,\ Table 1.7: Pattern of Spending and Behavior for the Poor eas while the poor are less likely to use contracep- (Data from 1997 and 1998) tion but not so overwhelmingly (Table 1.7). Differ- The Total ences in the pattern of use among different segments Save()Poor Population of the Philippine population are dwarfed by the large Save (% of income) 1' -5.6 18.9 Rely on transfers (% of pre-transfer income)2'! 43.0 12.9 differences between the Phlippimes and Vietnam, for Spend on food (% of consumption) a 64.4 44.7 example, where almost 60 percent of married Live in sub-standard housing 3' 11.0 8,8 Obtain credit 4 22.5 24.3 women m rural areas use modern contraceptives. Belong to cooperatives 7.7 12.9 There are clearly deep cultural and religious forces Use modern contraceptives " Urban 21.4 29.5 at work, which overwhelm imcome effects. Rural 19.3 27.1 1 1.22 Another finding relates to the low level of Note: t/The poorare rankedbyincome. When ranked byconsumption the savings rate i fft.Spefcentf2 TransferincIdespublicandprivafetransfersfrombothdomestcsources association among the poor. The poor are less likely andabroarbutnotpensions, Tihepooraredefinedasthebottom30percentofthepopu- lao/onandrankedbypre-transferincome. 3/Thefigureforthepoorlsthatforthebottom30 to belong to a cooperative (Table 1.7) and fewer of percentolpopulation. 4Asshareofpop7lat/oninfam,//estamiowentrepreneunialact/vt//es them are also members of peoples organizations 5/Share ofmarriedpersons us/ngmodemcontraceptivemethods Poordefinedasbottom 20percentoft/epopulationand the averageforthetotalpopulationisassumedtobe or non-government organizations. It is not clear represen/edbrythe midd/e rguintt/e. Source: Basedon 1997FIES; t998APIS, 1993DHSlSBalisacan(t999a),;Alba(2000). whether this is a symptom of poverty or has some causal effect on welfare: low membership in coop- eratives may lead to less information and lower bar- Table 1.8: Housing Condition by Expenditure Decile, 1997 gaining power for the poor with negative impact Percentage of people in decile living in on economic activity while the findings on POs and Houses Houses with roofs with walls Illegally NGOs may suggest that the poor may be iad- made of made of occupied equately served bythese important intermediaries. Improvised makeshift makeshift houses Decile houses materials materials (squatters) Fecirpoest) 1.4ses 2.2e 2.8erials 3.2uatters) 1.23 The poor relyheavilyon transfers which play First (poorest) 1.4 2.2 2.8 3.2 .... Second 1.7 2.1 2.9 3.4 a significant role in reducing poverty in the Philip- Third 2.6 2.9 3.5 4.4 pines. In addition, they are equaling and benefit Fourth 2.1 2.5 2.9 3.8 Fifth 2.3 2.8 3.3 3.9 vulnerable groups more than proportionately Trans- Sixth 1.7 1.9 2.0 4.0 fers are large in the Philippines and are mostly pri- Seventh 1.3 1.6 1.9 3.7 Eighth 1.0 1.1 1.5 3.5 vate (Table 1.9). Theyhave also been mcreasing over Ninth 0.6 0.8 0.7 2.5 time. They accounted, on average, for 13 percent Tenth (richest) 0.2 0.2 0.3 1.0 of pre-transfer household incomes in 1997 with 57 PHILIPPINES 1.5 1.8 2.2 3.3 Source. Ba/isa can (I9a). based on the 1997 /ES data, percent coming from abroad. Data from the bal- ance of payments puts worker's remittances at 1.21 Fertility rates in the Philippines have come PhP134 billion (US$7.7 billion) for the same year, down but remain high by regional standards - 3.7 births per woman in 1995 down from 6.4 in 1970.11 Within the East Asia region, comparable rates are Table 1.9: Households and Transfers, 1997 3.4 in Malaysia, 2.7 in Indonesia and 1.8 in Thailand. Pesos per capita An important correlate of high fertility rates is the Total transfers received 3347 low use of modern contraceptive methods. Al- Transfers from abroad 1922 Domestic transfers 1424 though married women of reproductive ages have Private 1404 considerable awareness of contraceptive methods Government 20 (90 percent report awareness, 1998 APIS) and al- Transfers given 238 most all who are aware know where to avail of Net transfers received 3109 Transfers as percent of expenditure 14.1% services (97 percent), contraceptive use is low. There Transfers as percent of pre-transfer income 12.9% is negligible difference between rural and urban ar- Source: Based on t7dF/ES data. 10 7: PROFILE OF THE POOR lent~ ~ ~ ~lg toir 7.6, pecnrfproa osmto equivalent to 7.6 percent of personal consumption Table 1.11: Public Services and the Poor, 1998 reported in the national accounts and up from less than one percent in 1988. of the of total poor population 1.24 Transfers are highlyprogressive, benefiting Go to Public Schools "/ 94.8 74.5 more households with low levels of (pre-transfer) Receive Assistance forTertiary Education 2' 1.6 2.8 income per capita. For the poorest quintile (ranked Delivery by a Medically Trained Person c by pre-transfer incomes), transfers received consti- Urban 36.1 82.2 tute almost 60 percent of pre-transfer.incomnes (Table Rural 18.9 59.5 Receive extension services 2 3.4 3.0 1.10). For a large number of households, private Benefit from CARP2' 1.9 1.7 transfers are the tickets out of poverty Income pov- Benefit from housing program 2.0 4.4 erty would have been much higher in the Philippines fNofe: ,/Share of 6-24yearolds who were attendingschool in the absence of transfers: the headcount index 2/Share oftotalpopulation would have been 32 percent rather 25 percent.12 The 3/Share of those who isiteda health facility 4/Share o/deliveries in the live years prior to the survey. Poor- dolned as bottom 20 distributionally sensitive melasures would have been percent of the populationl and fhe average for the total population is assumed to be represented by tfhe middle quintile. even higher suggesting that transfers help some of Source.:Sasedon 1998APIS, 19980HS, Aba (2000) the poorest among the Philippine households. This is a crude estirate in that it assumes that household nets have been a powerful anti-poverty force in the behavior would be the same with and without trans- Philippines. fers - that is, work effort would remain unchanged - but it serves to underscore the importance of ACCESS TO PUBLIC SERVICES transfers in reducing poverty 1.26 How well do public services reach the poor, 1.25 Furthermore, analysis of the determinants in particular, sectors that are crucial for building hu- of transfers suggests that having an unemployed, eld- man capital or for programns that are specifically de- erly or female-head has a large and significant posi- signed to help the poor? This section gives an over- tive impact on transfers received by the household. view of the targeting of some basic public services. On the whole, private transfers or informal safety 1.27 Table 1.11 gives a summary of the target- Table 1.10. Private Transfers Benefit the Poor, 1997 ing record of some publicly-provided social ser- vices."3 Table 1.11 shows that the poor are more Income Quintile Income Gross Share of (ranked by per capita transfers transfers likely to go to public schools; public assistance for income before before gross per capita in pre-transfer tertiary education, however, benefits more the transfers) transfers received income (Php) (Php) (%) wealther segments of the population. The poor are also more likely to visit public health facilities, but Quintile 2 10877 2183 20.1 have less access to a medically-traimed person dur- Quintile 3 16203 2594 16.0 ing delivery Two programs designed to benefit the Quintile 4 25353 3224 12.7 poor have very different targeting profiles: the land Quintile 5 70935 5185 7.3 Average 25887 3347 12.9 reformprogram(CARP)appearssuccessfulireach Poverty and ing the poor whereas the housing program benefits inequality indices more the wealthier groups. Incidence 32.0 25.2 Depth 10.7 7.2 Severity 5.0 2.9 1.28 Table 1.12 gives a similar picture using the Theil 0.52 0.49 barangzy as the unit of observation, when bararg-s Log of variance 0.71 0.64 are ordered by median per capita consumption. Source: ased on 1997F/ES data. Placement of infrastructure is regressive if wealthier 11 PHILIPPINES POVERTY ASSESSMENT Table 1.12: Distribution of Public Infrastructure, Barangay areas are better endowed; it is progressive if more Level Information, 199811 infrastructure is placed in poorer areas. Progressive (+1 Significance Regressive (-) level2' P-value Street pattern - '' 1.29 Table 1.12 shows that road access, town Access to main roads halls, rarkets, high schools, colleges and libraries, Town haill Church + 0.17 hospitaLs, housing progmmns, communication services Park . and clean water are found more frequently in Cemetery + ... Market wealthier areas. Cemetenres and pnimaryschools can Primary school + be found more often in poorer areas and churches High school College - and clinics are distributed independently of wealth. Library - Note that one should not use the results in Table H-ospital- Clinic 0.18 1.12 to make inferences about causality Hence, ar- Phone program eas may be wealthier because they have access to Telegraph - main roads, or wealthier neighborhoods may be in Postal service - * a better position to influence decision maker to have Water Vote:' Deteirminedt'baPSI/evetOLSregesswon fd presence ofapub/!cfrfrastnctare greater access to main roads. dummyontiremedanpercapitaexpendiure(Manllapnces), usinrgrobouststandanderrors I 'significantat 1percentlevel, "at5percentlevel 'a 10opercentlevel Source. Basedon 1998APlS data. Table A1.11: Definition of Variable Used in the Log Consumption Model (Table 1.6) Notation Variable Description Household head attributes AGE Age of household head AGESO AGE squared MALE Dummy variable, household head is male MARRIED Dummy variable, household head is married COLLEGE Dummy variable, household head is at least a college graduate HIGHSCH Dummy variable, household head is at least high school graduate but did not complete college ELEM Dummy variable, household head is at least elementary graduate but did not complete high school Household composition FSIZE Family size CHRATIO Ratio of dependent (below 15 years old) to total number of children EMPRATIO Ratio of employed to total household members Economic sector of household head AGRI Agriculture, Fishery, and Forestry dummy variabe CONST Construction dummy variable FINANCE Finance and Banking dummy variable MINING Mining and Quarrying dummy variable TRADE Trade dummy variable TRANSP Transportation and Communication dummy variable SERVICES Services dummy variable EGW Electricity, Gas, and Water dummy variable UMEMP Unemployed Location REG1 llocos Region dummy variable REG2 Cagayan Valley dummy variable REG3 Central Luzon dummy variable REG4 Southern Tagalog dummy variable REG5 Bicol dummy variable REG6 Western Visayas dummy variable REG7 Central Visayas dummy variable REG8 Eastern Visayas dummy variable REG9 Western Mindanao dummy variable REG10 Northern Mindanao dummy variable REGi1 Southern Mindanao dummy variable REG12 Central Mindanao dummy variable ARMM ARMM dummy variable CAR CAR dummy variable CARAGA CARAGA dummy variable URBAN Dummy variable, household lives in an urban area 12 1: PROFILE OF THE POOR Endnotes 1 Details of the approach and its implementation are given in Annex A, Volume IL. 2 Another difference potentially contributing to the relatively higher official food poverty lines is that while the basic needs food poverty line is anchored to the norm of 2000 calories per person per day, the official line is calibrated to both that caloric norm as well as the recommended daily allowance (RDA) for proteins and 80% of the RDA for vitamins, minerals and other nutrients (NSCB, 1999, and Balisacan, 1999a). 3 There are other differences between the estimates reported here and the official estimates. An impor- tant one is the use of per capita income rather than consumption as the welfare measure in the official methodology. This however contributes in the direction of lower official poverty estimates reflecting some positive savings even among those below the poverty line. Using income per capita and the basic needs povertylines, povertyincidence is estimatedto be 22.1%. SeeAnnexA, Volume II, and Balisacan (1999a) for further discussion of the difference between the two approaches. 4 While there is more recent survey that is available, viz., the 1998 APIS, income and consumption modules in the APIS are not comparable with those in the FIES. 5 This estimate is consistent with the estimated imnpact of the crisis, as discussed further in Chapter 5. For instance, the estimates in Chapter 5 indicate a 9 percent increase in the incidence of poverty due to the crisis. 6 The Philippines is unique among developing countries in that Social Weather Stations (SWS), an NGO, has regularly collected survey data on self-rated poverty since 1985 (Mangahas, 1995). Between 1985 and 1991, surveys were fielded once to twice a year, and since 1992, surveys have been carried out at least once every quarter. 7 There may be measurement issues surrounding the unemployment rate in the Philippines. The unemployed are defined to include all persons 15 years old and above, who did not work even for one hour during the past week, and who reported actively looking for work. Also considered as unemployed are persons without a job or business who are reported not looking for work because of their belief that no work was available or because of temporary ilness/disability, bad weather, pending job application or waiting for job interview. This is a more inclusive definition of unemployment than used in some other countries. 8 One convenient way to pararneterize a mre general measure of welfare is as follows: x = x /(A + C)6 where x is total household expenditure, A and C are the numnber of adults and children, -y is the adult-equivalence parameter and 0 is the scale economies parameter. Both pararreters lie between zero and one. If both -y and B equal one, the familiar per capita measure is obtained. On the other hand, as 6 declines to zero, there are increasing scale economies. Similarly, a decline iny inplies a smaller allowance for children's needs relative to adults. 9 Defined to be below 15 years of age. 1 0 Strictly speaking, one can only interpret the estimates in Table 1.6 as explaining the variation in household welfare conditional on past decisions concerning employment and human capital development. They do not explain the process by which households have chosen employment or have accumulated human capital. To the extent that selectivity in employment and humnan asset accumulation takes place, the benefit to a typical household of finding employment or owning a certain asset could be overstated. _ 3 PHILIPPINES POVERTY ASSESSMENT 1 1 Fertility rates are high for poorer households (6.5 for the lowest quintile in rural areas) but the measure of welfare used in the DHS survey, which is based on per capita assets rather than income or consumption, suffers from the same problems discussed earlier regarding economies of scale and household composition. 12 This calculation estirnates pre-transfer consumption corresponding to pre-transfer income using a consumption function estimated from the survey data where pre-transfer income is the income net of transfers. The results for consumption poverty are less striking than those for income poverty because the consumption function predicts higher propensity to consume at low income levels. 1 3 For the first time, information on access to services is available from a national household survey. The discussion in this section exploits information available in the DHS and the APIS on use of public services. 14 CHAPTER GROWING OJT OF P "The progress of the past few years has not raised the standard of living of the majority of Filipinost decent levels. I campaigned on a platform of growth >' with equity. These twin objectives were based on the realization that growth is a precondition for reducing poverty. On the other hand, the fruits of such growth should be equitably shared." Joseph E. Estrada, . .' Forewordto the Medium-Term Philippine Development Plan, 1999-2004. 2.1 Economic growth is widely recognizedas the engie of povertyreduction (World Bankc, 1990 2000g. It has also been widelyrecognized that w1l \ growth generally reduces poverty, the extent of pov- ', erty reduction and welfare irnprovemnent associaed with a given rate of econornic growth can vary enop- . mously from one situation to another. There is noi6h4 ` ing axiomatic about the quality of growth, and A has to look at the actual country experience to asse how far has growth delivered the promised pov-. erty reduction. This would nonetheless be of lin \, ited use if it only served as an exercise in evaluation, of past perforrnance. The more irnportant reason for engaging in such a review is that it should offet insights into how the current development trajec- tory mlay have to be modified to yield greater re- turns in future poverty reduction. 2.2 The first chapter presented is the recent pro- file of the poor. This chapter reviews the track record of growth and poverty reduction in the Phlippimes , since the rmid-1980s. The mid-eighties is a useful benchmark The pernod maarked the return to for- mal democracy in the country. It also marked the beginning in 1997of the first Medium-Term Phiippine Development Plan (MT-PDP), mandated PHILIPPINES POVERTY ASSESSMENT bythe Constitution, as an effort in coordinated imple- Table 2.1: Average Living Standards, Poverty and Inequality, mentation of programs and policies for national de- 1981997 veloprnent.Theperiodsincethemid-1980s alsowit- 1985 1988 1991 1994 1997 Mean living standard 17,197 18.926 20,049 19,600 23.694 nessed important changes im Philippines economic (per person per year policy with a shift to greater outward onrentation as at 1997 prices) (9.9) (5. 5) (-2.4) (14.3) the country sought to embrace the still unfolding Poverty 'ncidence (percent) 40.9 34.4 34.3 32.1 25.0 East Asian Miracle. At a more practical level, this 7-112) (-aO. 1, -45) (-15 6) od also niarked the beginning of aseries of na- Depth (percent) 132 10.1 i0.6 87 S.4 peniod also marked the beginning ofa (-128 (2.5) I -99) (-14.6) tionallyrepresentative household surveys, conducted' Severity (percent) 5.8 4.2 4 5 _.4 2.3 everv three years by the National Statistics Office (12.1) i3.3) (-11.3) 1,2.8) Inequality (NSO), that enable a more systematic monitorng Gini 0.412 0.400 0.428 0.397 0.427 of the distribution of household living standards in Theil T 0.330 0.298 0.363 0.302 0.376 Theil L 0.282 0.264 0.306 0.260 0.303 the country NVotes: Living standards are deftnedas householdconsumption expenditures adjusted for family size and provincial cost-of-living differences. Povery estimates are based on spa- 2.3 While povertv is inherentlv multi-dimen- tially ixeopoverynormandonpercapitaconsumptionexpendituresad/ustedforprovinc/hi cost-of-living differences (see Bahisacan 1999, for details). Figures iniparentheses are t- sional in character, povertymeasures based on con- ratiosforreferenceyearagainstpreviousperiod Thef-testforthesignificanceofpovery r r , ~~~~~~differenceisbasedonihem,ethodologypropos,edloyKalcwdni!'RS93) sumption, or consumption poverty for short, offer Sourcn Balisacana(2000). a useful first assessment of how distribution of liv- ing conditions has evolved. Tis assessment is later d whole. Figure 2.1 shows this for the "two-dom ar a supplemented with both income and non-income day' poverty Line, although the same comparative based, including self-rated, measures, pattem also holds for the "dollar a day" povertv based, including self-rated, rneasures. , line. CONSUMPTION POVERTY 2.6 This begs the question: has siower poverty reduction been on account of slower growth in the 2.4 Consumption poverty appears to have de- .. . . d clined sipniia hemdMc8su o h ne fromne sir ifc lysince th.i-18supt h ne Philippines or rather due tO growth being less pro- o the economica crysi in lante1997. Based ondta poor? The evidence based on household surveydata FafilyeInconor crisis m late 1997. ands urveys dat suggests that probablyboth factors have been at work fromfive Family Income and Expenditure Surveys~(Figure 2.2).2 Growth was certainly slower in the over the period 1985-97, and using per capita con- . sumption as a measure of welfare, the incidence of , hippmes (with the exception of Malaysia), whether sumption as a measure of welfare, the incidence of measured by per capita private consumption from absolute poverty is estumated to have declined by the national accounts (NA) or by per capita house- about 39 percent, from about 41 percent of the Filipino population in 1985 to 25 percent in 1997 F A (Table 2l1).i The decline in poverty was not con- fined to those in the neighborhood of the poverty line, but was shared more wvidely among those be- low the povertyline. This is apparent from the evo- lution of the depth and severityof poverty which 1 [1 1 1 1 decmined even more rapidly than the proportion of the poor.:; * - 2.5 Impressive as it ,may appear, this record Note. Thegraphs/towsetheannual,rateof declineeintheopropofitionof peolarionliving belowS65.491mo,nth at1993 PFPPdollars Thlis sthe twvo-d~ollarsaday',po veltyline. Th7e needs to be put in perspective. The rate of poverty estrIMatesare baseOdon nationally,represen7tsaye ousetsold olsuiveyadata,om/he followinTg reduction since the mid- 1980s up to the Asian crisis years Chin (1987g 1996), indonesia (1984, 1996). Malaysia (1987 1995) Phildpplnes s1985, t99Z), Thailand(1988, 1996), andEastAsia (1987 t996) The esilmatesfor Ch1na has been slower in the Philippines than in several and Malaysia are incomebased; those for China are adjustedby the ratio oprivate con- sumption to GDP Forfurther petails on methodology see World'Bank I2000al. other countnres int the region and im the region as a Souice: Computedirom distribuion data usedin WcrldEank(2000d). 16 2: GROWING OUT OF POVERTY -~~~~~~~~~~~~~~~~~~~~~~~~b s I th survey,S reason to suspect a systemati ba iny 17 T ~~~~~~~based growth rates, it may be more "correct"~ to G rSy- GS (NA) relate trends in poverty to changes in survey-based , t Elg t rn-th E at I CsY SL r@Y) G,Dtat E-a ,-y tND 1 mean consumption because the poverty rates are necessarily derived from the survey data. In either case, it is clear that growth is an undeniable part of the story It is equally clear that how that growth is -21 fX 2D distributed matters greatly for the poor. GROWTH, INEQUALITY AND TRICKLE Note: Thegmwnth rate is the annual compound growth ratein meanpercapita consumption DOWN? expenditure overthe surveyyear as repottedin Notes to Figure 2.1. The erpidcalgrowth elasticity is the ratio of percentage change in the headcount index to the percentage change in mean consumption from the survey NA denotes NationalAccounts. Source.:Basedonstaffestimafesandothosein WorldSank(2000a). 2.8 Over the period 1985-97 as a whole, how much did the poor gain from the growth that did hold consumption from the household surveys. But occur? The elasticity of the headcount index for the how the Philippines fared in terms of the respon- whole period to real mean consumption growth was siveness of its poverty to growth (the growth elastic- negative one. This is less in absolute terms than the ity of poverty) depends on whether the elasticities are elasticity that would be expected under distribution calculated with respect to NA or survey-based growth neutral growth, which would have been -1.2. Put iates. Based on the national accounts, the Philippines differently, had growth been shared equally by all appears to have the most pro-poor growth because consumption groups, the headcount index would relatively small rates of per capita consumption have declined from 41 percent in 1985 to 22.7 per- growth over the period (1.2 percent per annum) ap- cent by 1997, relative to the actual value of 25 per- pear to be generating substantial poverty reduction. cent. Thus, over the period as a whole, the beneficial But if elasticities are calculated using survev-based effects of growth on poverty reduction were miti- growth rates (3.2 percent per annum over the pe- gated by an adverse distributional change, but only riod), the Philippines ends up with the least pro-poor by a smnall magnitude. The effects of adverse redis- growth of all (the same deflator is used for calculat- tribution on the depth and severity of poverty were ing both the survey and NA growth rates). even more limnited. On the other hand, if there had been no growth in mean consumption, poverty in- 2.7 Which is correct? It is likelythat the discrep- cidence would have irwsed to 43 percent. Thus, ancy between consumption growth from the na- contrary to some popular claims, insofar as growth tional accounts and the surveys is the result of dif- occurred, it was poverty reducing. ferent measurement conventions. For instance, such a discrepancy could arise if, with a declining share 2.9 The importance of growth for poverty re- of the informal unincorporated enterprises in the duction is also confirmed by the evolution of economy, expenditures of this group, which were growth and poverty across regions and within a re- included in the residuallyestimlated private consump- gion (Figure 2.3). At both the inter- and intra-re- ton in the national accounts, are now increasingly gional levels, a strong negative relationship is exhib- excluded. However, there is no independent evi- ited between growth in mean consumption and dence for this, and the per capita GNP growth rates changes in poverty during 1988-97. are equally below the per capita survey consump- tion growth rates. This discrepancy is not easily re- 2.10 There has been some concern that inequali- solved, but it is likelyto be grounded more in mea- ties have been widening in the Philippines in recent surernent problemns rather than in changing economric years. The data on consumption and income distri- realities. Moreover, since there is no independent bution does not show a consistent pattern. Irrespec- 17 PHt!IPPINES POVERTY ASSESSMENT Economic growth is a powerful force for poverty reduction... ... across regions within the Philippines... and across provinces within Southern Luzon. \Mindoro 4% °1y =-1 5939x + 0.0004 10%t dentai 3%- AvuM R =0.702 8°% 2°h- \ CARAGA 6%°- 2%- p'~~~Fg on L a lrza- \ 40/o_ g ag< ~~~~~~~~~~~~~~~~~~~~~~~~~~~Ajrcra 1% Airc5~~~~~~~~~~~~~~~2 0%X - X Eastern Cavie v2% 0% 2% 4%; 6% -4% -2% 0% "Walawa, 4e% 5C eentral LuzoC IL a,ana, locos -2% Man Rn 30/ 0a75/inK Bcol -6% Mindoro O czcn 4% -her C nlra~ Vsayas -68% y =-2.4396x 0.0351 ,Bata.ga 4b/o - So.dherr Lit'5 4 \ C snlra: 9isat as -10t R' = 0.8942 .5% ttaWsJent - N MIindanao Visayas \ MMintanao 6%eo CAR MetroMaria lote: Tlhe graph plots annualpercentage change ki the hea6dcount lirdexn fa region (pro vilIce) against percentage change in mean consumptio 7In thatregion7 (province] durfrrg 1998-97 Source: Staff estimates basedon 1988and 1997RFES data, tive of the measure used, inequality seers to have 2.11 How do the levels of inequalityin the Philip- fluctuated without any significant trend. In particu- pines compare with other countries in the region? The lar, the rise in inequality during the most recent pe- results are shown in Figure 2.4. Thc levels of inequal- riod, 1994-1997, should not be confused with a trerS ity in the Philippines are comparable with those in increase. This holds for consumption as well as in- Thailand and Malaysia, but seem to be higher than come inequality those in Korea and China. As for changes since the mid- 1980s, with the exception of Korea and Malay- sia, inequality seems to be increasing in most coun- Figure 2.4: Income and Consumplon Inequaly tries. Although as alreadynoted above, increase inthe mid- 980 an ml-190s000000.40000t0:\0 case of the Philippines is misleading as it depends Thailand \ heavily on the end-years chosen for comparison. Philippines ,,4,W AHfi k mr SUSTAINED POVERTY REDUCTION? Malaysia ndonesa _ 2.12 It is apparent from Table 2.1 and Figure Korea 2.5 that poverty reduction in the Philippines was Cl,ina anything but sustained during 1985-1997. Most of _0, - ~ ~ -~ -r the decline in poverty was confined to the first and I Cons. 20 40md last three years of this period. The decline in the Thailand _ headcount index during 1985-1988 and 1994-1997, for imstance, accounted for more than 85 percent Philippines of the total decline during 1985-1997. Similarly the Malaysia ~ decline im the depth and severity of poverty during ndonesia the two sub-periods accounted for a little under 80 Korea percent of the total decline. Cnina i. 0 10 20 30 40 50 60 2.13 The decline in poverty during 1985-1988 rncome Gifts mid 805 n Incomre Ginis mid-gOs mnay however be somewhat misleading since 1984- Note The graph shows Gini indices (in percentages) fornominal per capita income or 1985 was a year of sharp econonic contraction. Per consumption. The estim7ates are based on nafional/y representativ/e household survey dataforthoeollowingyears:.China(1985 1995),Korea(t988l 1993), Indonesia(1984. capita GDP shrank by an average of 10 percent a 1996) Malaysia (1984, 1995), Philippines (1985. 1997) and Thailand (1986 1996). Source: Computed from distributional data used in World Bank (2000a). year during 1984-1985. Based on FIES data, real 18 2: GROWING OUTOF POVERTY Pgu .25: Evo~luionofGQRqMftyuiowvr:h ; -- 2.15 Some simple calculations are iUustrative of in th P .0 -: both the importance of sustained growth and the mean consumption, (suev) sustained quality of growth. Thus, for instance, if 1401 7 Meal csmption(NAosingCPI) . e the 1994- 1997 highest growth rate in per capita con- Poventy ncidence O 120- 7 G nf sumption of 6.3 percent were maintained over the 1< 4 + 'whole period, the incidence of poverty would have 0.oo- declined to 12.9 percent (compared with the actual A, . ^ value of 25.1 percent). But, if the highest empirical 80) * elasticity of povertyincidence with respect to growth A (of -1.58 during 1985-1988) were maintained over 1955 1588 1991 1994 1997 the entire period, then the existing growth rates would Source. Based on F/ES data. have impled a decline in the headcount index to 16.5 percent by 1997. mean consumption in 1988 was 10 percent higher than that in 1985, although arguablystill much lower INCOME VS. CONSUMPTION POVERTY than the level prevailing at the beginning of the 1980s. 2.16 An important issue in the measurement of 2.14 Underlying the uneven performance in pov- poverty concerns the choice of a metric of individual erty reduction was the uneven growth performance. welfare. Two key contenders here are consumption- For instance, survey-based mean consumption per and income-based measures. Either income or con- person grew by 3.2, 1.9, -0.8 and 6.3 percent per sumption is a defensible measure of an individual or annum over 1985-1988, 1988-1991, 1991-1994 and household's command over goods and services, 1994-1997, respectively (Figure 2.6). However, this though practical considerations in collecting reliable growth was also of varying quality for the poor. survey data in developing countries often lead to a Thus, over the same sub-periods, the elasticity of preference for consumption-based measures (see the headcount index with respect to growth in mean Ravallion, 1994; Deaton, 1997; and Box 1.1 in Chap- consumption also fluctuated from -1.6, -0.1, 2.9 to ter 1 for further discussion). Even if consumption -1.1. The growth elasticities of the poverty gap and maybe considered the preferred measure, it is useful the squared povertygap indices display a similar pat- to check if the evolution of incorne poverty has been tern. Thus, neither growth nor the pro-poor quality similar to that of consumption poverty An additional of growth was sustained over this period. reason for focusing on the income dimension is that the official poverty estimates for the Philippines have been based on per capita incomne (NSCB, 1999). 7.3 2.17 In order to compare income and consump- E 2 ~~~VG owMh INA, _nI Wp- g_ aV,,ncvj u-Ve tion poverty, we use an income poverty line that is C 3 2 calibrated to generate the same national headcount index of 25.1 percent in 1997 as obtained with the consumption povernvlines. The income povertylines .. j 3 r - 8 over tirne and space however raintain the same rela- - -85 6 tivities as for the consumption poverty lines. Thus, for instance, Li the 1985 Manila consumption pov- NVote: r.e empirncalgrowth elasticityis the ratio of petcentage change in the headcount hideexto thepe,centage changein(su.vvey-baseo2mean7consumption. Growth(SuveOy)' erty line was 3C.5 percent of The 1997 vanila con- is the a,7nuaico.mpoundgrowth rateinmean consumption estimated from the suiveys and s *umtio t v u ric, n thi I 5 Manila inc . me 'GroA7h(n(VA)Jistheannualcompound growthrate inoercapitapersonaiconsumptlon nst4na?eo fromt'reNationalAccounts/NA). Bothsunvey-basedandNAconsumptionare povertyline i' . i Ix; X it 1. 3. :rcernt of the 1997 'leflaledbythye Consumer Price Indexforallcommoditles. Source: SaseaGnF/ES pata, 3ndadorNatAccountsdata, NSCB. irncome pOvela rOc I X alM i. 19 PHIL IPPINES POVERTY ASSESSMENT 2.18 The choice of the welfare metric makes F 2: ncome.C inG tand Inequality: some difference to the assessment of how living o It Matter? standards have evolved in the Philippines. The fol- 50 l * Mean consumption (survey) Mean incoe (sunvey 1 ' ' 1 1 , ,,L,,, ~~~~~~~~~~~~~Poverty ncdence(Gj :; c-ty mncde-c(Y) lowing pomts are notable. 140 Pie oyn (a) For the period 1985-1997 as a whole, the de- 130- cline in income and consumption poverty rates ,1'0 * + is comparable (Figure 2.7). Thus, for instance, 10 . both consumnption and income headcount in- AL dices declined from about 41 percent in 1985 70 to about 25 percent in 1997. The rates of de- 60 --- 1 --- -L 1985198 199i 1994 1997 cline were also comparable for me'asures of depth of poverty, and were only marginally oe:(C)Refersto consumption eeinates, (Y) refers to incomeestmtes depth of poverty, and were nly niarginafly Source: Baedn FIES data. higher for income-based measures of the se- iod marked by errati growtls The pe- verity of poverty (igure 2.7). a.p .IC verity of povert (Figure 2.7).nod 1988-94 was a period of relative stagna- (b) Average income and consumption growth fol- ton Both G9P a person onsumptionpe lowed the same pattern over this peniod, withtinBohGPadprnlcnsminpe owdtsaeptroeth dwi capita were stagnant, and this is also reflected the exception of years 1985- 1988 when income in both income and consumption poverty growth appears to have outpaced consump- measures which changed little. On the other tion growth. This is consistent with some con- hand, both 1985-1988 and 1994-1997 were sumption smoothing by households during periods of growt, alhough the first spel was 1985 in response to the econormic contraction f more in the nature Of a recovery from the during the same and the preceding year. immediatelv orecedmn contraction of 1984. (c) The inequality measures for income were con- 1 t ed 1985. To the extent the poor tn'ed to protec-t sistently higher than consumption inequality I~~ ~ . their consumption in 1985, their consurmiptol. mreasures which 'LS unsurpn'smg, but the evolu-.. meas*es whicisunsrprsgbuthevu-growth over the subsequent recoverv during tion of income and consumption inequality 1985-88 could be expected to lag behind the over t1me was rernarkably simllar as shovn In Income growth, as thev reduce dissavimgs (or Figure 2.8. This is true for a range of inequality i . T w l t ... . tn~~~~~~creas e s avmgs'). This would lead to a smraller rmeasures, not just the Gin mdlces. measures,not justthe Giniindices.decline in consumption poverty relative to in- (d) However, there are some differences over sub- cme povrThpeiod 1994199 was how- pc >d.I atcAr 'cnepvrydcie come poverty. The pen'od 1994- 1997 was how- perods. In partilcular, income povertydeclined ever a more normal growth spell which was more rapidlyduring 1985-88, at aout the sae probably associated with some decline in pre- rate during 1988-1994, andmre slly dur- cautionary savigs by the poor, causing a rela- ing 1994-9 197. Thils probablvreflects the house- tvl agrdciei osmto oet < * ~~~~~~~~~~~~tively larger declme 'm consumption poverty. holds' attempts to smooth consumption over Figure 2.7: Income vs. Consumption Poverty: Does it Matter? SOCIAL INDICATORS AND SELF-RATED P.-tc dee (C S F-oney no daar~ POVERTY P.-tly deplh- i P-ty dropih (Y) y Pr crly ¢.erOo >C Poorrysereroy 9, 2.19 Has human and social development kept 80 ^ ^Dpace with the progress in reducing income ana con- sumption poverty? The evidence suggests an answer 60 in the affirmative: the Phlppines has achieved steady improvement in a number of social indicators over 4, t 046, r966 166r i594 1997 the 1980s and 199Cs (Table 2.2 Thus, life expect- Note: (C) Re/es to consumption estimnates (Y) refers to income estimates. ancy has improved, and illiteracy has declined fo Source Basedon F/ES data. both males and females. Similarly, immunization and 20 2: GROWING OUTOFYFOVERTY Table 2.2: Evolution of Select Social Indicators classify themselves as not poor or borderline are Social indicator 1987 1998 asked: "For a familyas large as yours but poor, how Life expectancy at birth, female (years) 65.9 70.5 much money do you think it would need to spend Life expectancy at birth, male (years) 62.2 66.8 each month for home expenses in order not to be Illiteracy rate, aduft total (% of people 15+) 8.4 5.2 considered poor anymore?" One way to under- Illiteracy rate, adult female (% of females 15+) 9.1 5.4 Immunization, measles (% of children under 12 months) 68.0 83.0 stand these two questions about a minimnum basic Low-birth weight babies (% of births) 18.0* 8.7 income" is that they provide information on a sub- Malnutrition prevalence (% of children under 5) 32.9 29.6 Mortality rate, infant (per 1,000 live births) 45.0 32.2 jective or self-rated povertyline, as identified bypoor Safe water, rural (% of rural population with access) 67.7* 81.0 and non-poor households themselves (Ravallion, Safe water, urban (% of urban population with access) 80.5* 91.0 " 1994). Note: *refers to 1988, " 1997 "' 1993, and ... 1996respectively Source: World Bank (2000f). 2.22 Four important conclusions emerge from access to safe drinking water in both rural and urban an analysis of these data on self-rated poverty areas has improved, while child malnutrition and in- (a) Unlike absolute poverty, there is no trend de- fant mortality rates have declined. These improve- cline in self-rated povertyoverthe period 1985- ments are also reflected m the Humnan Development 1999 (Table A2.1). But there is essentially no Index (HDI) for the Philippines, which rose from trend in real GDP per capita over the period. 0.683 in 1985, to 0.711 in 1990, to 0.740 in 1997 Hence, the fraction of people who consider (UNDP, 1999). This is all the more creditable as it themselves poor (essentially) remained un- occurred during a period of relative stagnation in changed during a period of (essentially) no per capita GDP. change in real GDP per capita.3 (b) But, self-rated poverty does respond to fluc- 2.20 Significant as these gains are, they may or tuations in real GDP per capita (see Figure 2.9 may not be reflected in people's own perception of and Table A2.2).4 In this sense, self-rated pov- their welfare. As mentioned in Chapter 1, Volume erty behaves like absolute poverty There is a II, the Philippines is unique among developing coun- clear negative relationship between self-rated tries in having a data base on self-rated povertybased poverty and real GDP per capita. Specifically, a on regular surveys conducted bythe Social Weather PhP100 (at 1995 prices) increase (decrease) in Stations (SWS) since 1985 (Mangahas, 1995). Between real GDP per capita is associated with a 0.67 1985 and 1991, surveys were fielded once to twice a to 0.74 percentage point decrease (increase) in year, and since 1992, surveys have been carried out poverty There is no relationship between self-rated at least once every quarter. How do these data on self-rated poverty supplement the picture on the evolution of poverty presented above? As a first step it is irnportant to understand the notion of self- 74 1 14515 rated poverty used in the SWS surveys. * 2.21 Three questions in the SWS surveys are of ' particular interest for an understanding of self-rated poverty First, households are asked about whether & 2L they consider themnselves "poor," Mborderline," or "not poor." Second, households who classifythem- 4 l - selves as poor are asked the following question: "In 5 10 t5 20 25 3e0 Semester your opinion, how much money would your fanily need for home expenses each month in order not to Note. The29semesersmreferto6-monthspellsfromthesecondhalfof 1985/othesecond be called poor anymnore?" Similarly, households who Source: Based on SWS and National Accounts data from NSCB. 21 PHILIPPINES POVERTY ASSESSMENT poverty and unemployment, and self-rated absence of important distributional changes poverty is only weakly affected by changes in which favor the poor, secular increases in the inflation. self-rated poverty line should lead to (signifi- (c) The threshold level which poor and non-poor cant) increases in the number of poor if there households regard as a "minimum acceptable are no (significant) increases in mean income. income" has risen dramatically during the pe- It is possible that these responses are consistent riod (Figure 2.10; Table A2.1). This suggests with household survey-based mean incomes that the self-rated poverty lines cannot be which show a significant rising trend over the thought of as absolute povertylines, which re- 1985-1997 period. An explanation could be main fixed in real terms over time. This is itn- that self-rated poverty lines track increasing portant in thinking about the potential use of mean incomes as captured in the surveys such self-ratedpovertyestirates forpolicypurposes. that these lines approximnate relative povertyin (d) Whydo the self-rated povertylines increase over the Philippines which has remained largelyun- time? The answer cannot be found in changes changed over the period, as reflected in the ab- in mean incorme, as approximated byreal GDP sence of a trend in self-rated povertyestimates. per capita from the national accounts, because Unfortunately, it is not possible to check this real GDP per capita does not show a signifi- hypothesis because while self-rated poverty cant trend over this period (Table A2.3).s The estimates are available on a quarterly basis, sur- unemployment rate does not appear to influ- vey-based income data are available only every ence self-reported poverty lines either and three years and therefore for five points in time changes in inflation are significant for only one during this period. of the two lines. A further puzzle is that it is difficult to reconcile the responses given to the 2.23 A better understanding of the relationship two questions - about poverty levels and the between self-rated and absolute poverty is imnportant poverty line - in a context of essentially stag- for the design of public policies in the Philippines. nating GDP per capita. Taken together,.these The recent March 2000 surveycarried out bySWS, in indicate that there has been no change in the collaboration with the World Bank, includes both the proportion of households that consider them- standard questions on self-rated povertyand poverty selves poor even while the povertyline has gone lines, and a consumption module identical to that in- up during a period of stagnant incomes. In the cluded in the 1998 APIS. Careful analysis of these data would provide useful insights that will comnple- ment the insights available from the analysis of abso- lute povertybased on income/consumption surveys. RURAL-URBAN TRENDS ra re d GOF pe!r eapfta p-,M lm-t p- h..t.W~ 2.24 One useful disaggregation of the consump- tion or income poverty data is by urban and rural Aw . . /\/' sectors. Poverty in the Philippines is, for example, often described as a largely rural phenomenon p.ttely hne-p., ht.h J dot (Balisacan, 1999a; also see Chapter 1, Volume II). --S- -----esr- - ___ Progress in reducing rural poverty will thus go a long 05 lo is 20 5 30 Semester wayin advancing the overall poverty reduction goal. Note: The 29 semestefs refer to 6-month spells from the second half of Y985to thes- Yet, tracking progress in rural and urban poverty ond half of R9R9 reduction is not as straightforward as it seems. As Source: Basedon data from Social Weather Stations, /ntemationa/ Monetary Fund nnd Intemalional Flnancial Ststics. noted in Amex A, rural/urban poverty idicators 22 2: GROWING OUTOF POVERTY constructed from the FIES for the 1980s are not Table 2.3: Living Standard, Poverty, and Inequality comparable with those for the 1990s owing to the by Locality, 1985-1997 urban-rural reclassification problem.6 Thus for in- 1985 1988 1991 1994 1997 stance, the share of urban households in all Filipino Urban Mean living standard 24,099 26,283 26,213 25,093 31,657 households estimated from successive rounds of the (per person per year FIES were 38 percent in 1985, 38 percent in 1988, at 1997 prices) (6.2) (-3.2) (11.5) 50 percent in 1991, 50 percent in 1994 and 48 per- Dimension of poverty ' 1997. Thesdden mpinthepoportoIncidence (percent) 21.7 16.0 20.1 18.6 11.9 cent m 1997. The suddenjump inthe proportion of (8.1) (-2.9) (-14.8) urban households in 1991 is on account of a large- Depth (percent) 5.9 3.8 5.7 4.4 2.6 (-9. 0) (-7.2) (-3.3) scale reclassification of erstwhile rural areas into ur- Severity (percent) 2.3 1.4 2.3 1.5 0.9 ban areas. Thus, inter-year comparisons for rural and (-8.2) (-8.7) ( 11.1) urban areas are only valid within the decade of the Inequality Gini 0.410 0.390 0.421 0.392 0.425 1980s and 1990s separately (since the satnping frame Theil T 0.327 0.286 0.355 0.295 0.379 and the rural-urban classification of geographic ar- Theil L 0.280 0.253 0.300 0.255 0.303 eas are common for individual decades) but not Rural across decades. Mean living standard 12,838 14,414 13,864 14,154 16,475 (per person per year at 1997 prices) (10.2) (1.8) (14.4) 2.25 Despite the problem of comparability Dimension of poverty across the two decades, certain patterns are clear. Incidence (percent) 53.1 45.7 48.6 45.4 36.9 (-94) (.40) (-11.7) There are large and persistent urban-rural dispari- Depth (percent) 17.8 14.0 15.6 13.0 9.8 ties. Average living standards are significantlyhigher (-11. 1) (.8.0) (-11.8) significantlyhighe Severity (percent) 8.0 5.9 6.8 5.2 3.6 (about twice) and poverty levels are significantly (-10.7) (-8.8) (-10.6) lower (about one-third) in urban areas. Thus, in 1997 Inequality a person in the ruraLL sector was three times more Gini 0.352 0.350 0.359 0.336 0.352 Theil T 0.226 0.217 0.238 0.205 0.230 likely to be poor than a person in the urban sector, Theil L 0.204 0.200 0.211 0.183 0.202 and the rural poor accounted for 77 percent of all Notes:Poveryandiinequa/ityestimatesarebasedonpercapita consumptionexpend. poor in the country The levels of urban poverty turesaiustedforprovincialcostfof/vingdifferenoes ThedenvaDonofprovcialpovertv linesisoesciffidinAnnexA, Volumell.Ftgur6eshnparenfthsesanet-raDosforoifferenes are lower despite the higher levels of urban imequal- beMeenpovedtymeasuresforthe referenceyearandoreprevioussear ity (Table 2.3 and Figure 2.11). Moreover, there are Source: BSaisacan (1999a. indications of widening urban-rural disparities im poverty rates over the most recent period 1994-97, although it is difficult to characterize it as a trend because of the area reclassification problem men- tioned above. Mean consumption Gini indices Incidence of poverty (Pesos/person/year at 1997 prices) 35,000 - _05- rbE 0 Urban 302000 - L 6u | 0.4- ural ERural 25,000 1104- 20.000 - 0.3- 40 - 15,000 - L i i i 02- 30 -_ 10,000 iI .- I 20 j 2,0000** 03 T: JJ- 1985 1988 1991 1994 1997 1985 1988 1991 1994 1997 1985 1988 1991 1994 1997 Note. Due to reclassiication of rural into utban areas, the figures forthe 1980s are not comparable &idth those forthe 1990s. Source: Ballsacan (2000), based on F/ES data. 23 PHILIPPINES POVERTY ASSESSMENT : t000f' N jiW_P-l_ 0,i950 100- 100 Z 80 > 80- a. _ Not employed a> X Not employed s 60- 1-Services 60- * Services Construction rG Construction a) 401 1 * Mining mfg, ub 40-l* Mining, mfg., util. 0CL. 20- , Agricuture 20 Agriculture 0 0 1985 1997 1985 1997 Source: Based on Balisacan (2000). 2.26 An apparent trend is the rising share of the tion of the population shifted in favor of sectors rural sector among the poor. This share has increased where poverty has declined relatively more rapidly both overthe 1980s and the 1990s, apparentlydriven thus leaving the sectoral shares in aggregate poverty by fact that during the two growth spells of 1985- relativelystable. For instance, while the incidence of 1988 and 1994-1997, urban poverty fell faster than povertydeclined least rapidlyin the agricultural, fish- rural poverty However, due to the reclassification ing and forestry sector, the population share of this of rural into urban areas, it is difficult to make de- sector also declined from 47 percent in 1985 to 40 finitive statemnents about trends in the rural-urban percent in 1997. composition of poverty 2.29 The overall decline in povertycan be decom- SECTORAL COMPOSITION OF POVERTY posed into a within component for the decline in poverty within sectors and a between component for 2.27 What was the sectoral composition of the shift of population across sectors (including an growth and povertyreduction? Table A2.4 summa- interaction element for how the change in poverty nizes the evolution of povertybyeconomic sectors.7 within sectors covanrcs with shift in population). Over Poverty declined significantlyin all sectors, though the 13-year period, 1985-1997, the overwhelning more in some sectors than others. The sector where proportion of the decline in poverty in the Philip- it declined the most (measured bypercentage change pines was due to poverty reduction within sectors, between 1985 and 1997) was finance, insurance, real accounting for 88 percent of the overall decline in estate and business services; this is true for all pov- povertyincidence and 91 and 93 percent of the total erty measures. The sector with the least decline was decline in the depth and severity of poverty agriculture, fishing and forestry for the headcount index, and mining and quarrying for measures of 2.30 In terms of the sectoral contribution to the the depth and severity of poverty overall povertydecline during 1985-97, it is the agri- culture sector that dominates, accounting for 46 per- 2.28 Despite the differential rates of poverty re- cent of the total decline in the incidence of poverty duction, the sectoral composition of the poor has This is despite the relatively slower decline in pov- remained remarkably stable since the mid- 1980s erty within this sector, and mnainly reflects the large (Table A2.4). Agriculture, fishing and forestry ac- share of this sector among the poor. The contribu- counted for two-thirds of the poor in 1985; it still tion of this sector to the decline in the depth and accounted for two-thirds of the poor in 1997 (Fig- severity of poverty is even higher at 59 and 67 per- ure 2.12). This is because the occupational distribu- cent, respectively 24 2. GROWING OUTOF POVERTY -~~~~~ ~~~~~~~~~~~~~ VM N& '-' Composition of employment Output per worker Capital per worker Agriculture E lndustry UServices Economy leftaxis) AGR(eftaxis} 100 ID(iNai) SR(ctai) -5 90 _ Economy AGR * IND * SER 45- NDrightaxis) * SER(lotaxis) 250 90- 40- 70 80 - 35- -200 70 30- * f , 150 10 10 a) I5 50 . I ,O- I20III - I I I I I I I O 20 -~~~~~~~~~~~~~~~~~~0 CO C 0 ro a a 0cO 0 a) cO N * a) CD - O CO 0 N m O Source: Staffcomputations tased on vanous data sources 2.31 The large share of the agricultural sector in spite of the service sector becoming more capital in- total poverty (two-thirds of the incidence and about tensive. In the case of the industrial sector itself most three-fourths for the severity of poverty) is the joint of the decline in capital intensity is on account of the product of the high levels of poverty in this sector rapid decline in the share of (capital-intensive) utilities and the high (though declining) share of population sector in total capital stock dependent on this sector for its livelihood. This has two policy implications for future poverty reduc- 2.33 The decline in capital intensityhas thus come tion strategy First, in the short to medium termn, at the expense of a decline in worker productivity there is no getting away from the task of agricul- which mitigated its potential contribution to poverty tural development if significant inroads are to be reduction. At the same time, there are large sectoral made into the problem of poverty reduction in the differentials in worker productivity which points to Philippines. And the reason is as simple as the fact the large untapped potential of structural shift in em- the 67-75 percent of the poverty problem rests in ployment and output scope for securing gains in liv- the agricultural sector. Second, poverty reduction ing standards for large segments of the population. over the medium to long-term will require struc- tural shift of employment and output from agricul- REGIONAL DISPARITIES ture to the more productive non-agricultural sectors of the economy 2.34 Regional disparities in living standards and poverty have been an important policy concem in 2.32 There has been only a limited shift in em- the Philippines. For instance, while the poverty inci- ploymnent from the agriculture to the service sector dence in 1997 was 3.5 percent in Metro Manila, it since the mid-1980s (Figure 2.13). The share of the was 71 percent for the Eastem Samar province in industrial sector has remained virtually constant at Eastern Visayas. In a similar vein, poverty incidence about 15 percent. Labor productivityin the economy in the poorest 10 provinces (which accounted for as a whole, has been stagnant since 1984, as has also 7.9 percent of the national population) was 59.5 been the case in agriculture and the services sector. percent while it was 5.5 percent in the 10 least poor Labor productivityin the industrial sector has actually provinces (accounting for 25.2 percent of the popu- declined since the mid- 1980s, which reflects the de- lation).' In other words, a person living in the poor- cline in capital intensityin that sector. There has been a est 10 provinces was ten times more likely to be decLineinoverallcapitalintensityinthe economylargely poorthan a person livingin one of the 10 least poor reflecting the decline in the industrial sector, and in provinces 25 PHILIPPINES POVERTYASSESSMENT 2.35 In 1988, poverty incidence in the poorest In general, inequality rose between 1988 and 1997 10 provinces was 73 percent while these provinces (Table 2.4). However, this was due to the increase in accounted for 8.8 percent of the population; and it inequalitywithin provinces. Inequalitybetween prov- was 6.4 percent inthe 10 least poor provinces where inces actuallydeclined. 21.3 percent of the population lived. This suggests there has been some reduction in regional disparities 2.38 One can further examine the evidence for in povertylevels. convergence by testing whether the rate of poverty reduction during 1988-97 has been greater in prov- 2.36 There was also significant re-ranking of inces that had higher poverty levels to begin with. provinces by their poverty levels. For instance, the The results show that percentage change in poverty poorest 10 provinces in 1988 and 1997 had only was indeed negatively related to the 1988 poverty four provinces that were common, and the poorest levels (Table 2.5) suggesting convergence over this 15 had only seven commnon ones. The rank correla- decade. A similar convergence result also holds for tion coefficient between 1997 and 1988 povertyrates the average standard of living and a measure of was about 0.6 for measures of incidence, depth and inequality. severity Re-ranking is also borne out by the graph in Figure 2.14 which plots povertyincidencein 1997 2.39 However, these broad convergence results against that in 1988. Most of the points are below hide a more complicated pattern of change in re- the 450 line indicates that povertyincidence fell in gional disparities. First, despite the substantial decline most (55 out of 76) provinces between 1988 and in national poverty poverty increased in 21 of the 1997. However, the graph is not monotonically in- 75 provinces. Of course, in manyof these cases the creasing which implies re-ranking. In particular, ev- increase is not statistically significant, but it is none- erydownward-sloping segment indicates a re-rank- theless true that poverty failed to decline in these ing of provinces in terms of their povertyincidence provinces. Second, it turns out that the poverty con- between 1998 and 1997. The graph shows that there vergence results do not hold if one were to restnrct was a lot of re-ranking. attention to the 55 provinces where povertydeclined. Rather, the overall convergence in poverty levels 2.37 There is some evidence that points to con- seems to be driven by the 21 provinces where pov- vergence of living standards across provinces over ertyincreased;in other words, within this set of prov- this period. For instance, this is suggested bythe de- inces, poverty increased less in provinces that had composition of inequality in per capita consump- higher levels of poverty to begin with (Table 2.5). tion into within and between provimce components. Figure 2.14: Change in Poverty Inc-dence Across Table 2.4: Decomposition of Inequality in Per Capita Provinces, 1988-97 ~~~~~Consumption, 1988-1997 '00 1 1988 1997 80* ,rr' Within Between Within Between Inequality Measure' Total Provinces Provinoes Total Provinces Provinces 70 ii / Generalized entropy (0) 0.2642 0.2034 0.0608 0.3026 0.2433 0.0593 aV ' zr' ,.'ft Generalized entropy (1) 0.2981 0.2368 0.0613 0.3762 0.3120 0.0642 A Sol?. 0gL ¢ Atkinson (1) 0.2322 0.1847 0.0583 0.2611 0.2256 0.0458 a1 A1 ,:/.)-:', | g|lt gAtkinson (2) 0.3816 0.3028 0.1131 0.4069 0.3543 0.0815 7: . ' t/'4,i$0 ' ' ^ \t Note * The parameter value for #te Generalized Entropy (GE) measure shovn in the parenfhe- five to tihe bottom o/fthe distibution than GE(t). TheD GE(t) measure is the same as te Theil T measure. Higher parameters for the Atkinson measure indicate greater aveislon to inequality I rT -v T-^r T Source: Computed from 1989 and 1997 FIES data. 0 tQ 20 30 40 50 60 70 80 90 Headcount 1988 Source. Based on F/ES data. 26 2: GROWING OUTOFPOVERTY Table 2.5: Convergence of Poverty and Living Standards Across Provinces, 1988-1997 All Provinces Experiencing Provinces Experiencing Provinces Decline in Poverty Increase in Poverty Percentage (Log) Change during 1988-97 Mean Poverty Theil Mean Poverty Theil Mean Poverty Theil Cons. Incidence Index Cons. Incidence Index Cons. Incidence Index Log initial -0.4872 -0.3678 -0.4647 -0.3119 0.0636 -0.6273 -0.1848 -0.3150 -0.1551 (1988) value (5.99) (4.78) (5.46) (4.78) (0.78) (6.57) (0.95) (2.88) (0.87) R-square 0.33 0.24 0.29 0.30 0.01 0.45 0.05 0.30 0.04 No. of provinces 76 76 76 55 55 55 21 21 21 Note: Loginiftal (1988) value refers to the naturallogarithm of the 1988 value ofeithermean (real) consumption per capita, orthepovernyinequalitymeasures forthe correspondlng equation. The data exclude the province of Rizal which was not surveyedin 1988 Absolute t-ratios are in parentheses. Source: Based on FIES data. 2.40 Thus, while there is quite a lot of re-ranking Eastern Visayas, Central Mindanao, and Central of provinces, there is no strong overall evidence for Luzon. In the case of four of them, this seems to convergence of poverty levels, and clearly a num- be driven bythe lack of significant growth in mean ber of provinces failed to participate in the signifi- consumption. For Eastern Visayas, there was sig- cant povertyreduction observed at the national level. nificant growth (though still low in comparison with Given the mixed picture at the provincial level, it is most other regions), but the poor failed to share also useful to look at how poverty evolved across adequately in the benefits of growth due to an in- the 16 regions. In particular, despite inter-provincial crease in inequalities. Together, these five regions ac- variation, did some regions as a whole lag behind? counted for 25 percent of the total population and Figure 2.15 summarizes the situation. 30 percent of the total number of poor in 1997. The lack of poverty reduction in these regions points 2.41 There is a large variation in both growth in to a major failure of the growth process to deliver mean consumption and rates of poverty reduction widelyshared benefits. It also points to the need for across the 16 Philippine regions. In terms of pov- a regional focus to the poverty alleviation strategy erty reduction, five regions have clearly lagged be- which accords high pnrority to the development of hind. These are the ARMM, the CARAGA Region, these poor and lagging regions. Figure 2.15: Evolution of Poverty and Lvin grs s, 198- 97 Annual percentage change in mean consumption Annual percentage change in poverty incidence CARAGA Region CAAGA Region _ 23 ARMM' -2.2 ARMM 3 3 CAR' _8 CAR' -6.3 Metro Man la' | 35 Melro Manla' * _ _ Gentra Mndanao 9 i Cnra inMidanao 9 Southerr M ndanao' l4Soitharn Mircanao' -62 Northern M ndanao' 4. Northern Mimanao' -6 Western M ndanao' _e52.3n Mianao' -3 Eastern Viaayas' j 1.7 l astern Visayas l 7 Ceetra Vrsayas' 31 Centrst V,sayas' 0 Western Vtsayas 2.C WesternVisayasy' -.1_ Bico | i _ _ " 3 2 icolr - .2 Southern Luzon' I l 2.2 Suterr Luzon' -4.6 Cental Luzon i -0.1 l l Ceital Luzo- -1.7 Cagayan _ 0 Cagayan' 2.9 I ocosne - I os -t22 3 -2 - 1 2 3 4 5 6 -12-10-8 -6-4 -2 0 2 4 6 Note: 'denotes significant change b-etween 1988and 1997at the 5percent level of significance. denotes significantchange at the 10 percent level. The standard errors are corfected for sample stratificatlon and c/usfeong Snurce: Based on .988 and 1997FIES data. 27 PHILIPPINES POVERTY ASSESSMENT Table A2.11: Trend and Seasonal Components of Variables in the Analysis Variable Trend component Seasonal component Self-rated poverty -0.0694 -1.506 (0.1358) (2.273) Self rated poverty line (poor households) 70.27*** 96.22 (18.12) (303.32) Self rated poverty line (not poor households) 73.32*** -40.82 (19.38) (324.47) Real GDP per capita 17.13 -913.30"** (11.34) (189.82) Inflation rate 0.0184 -0.2715 (0.1053) (1.762) Unemployment rate -0.0604*** 1.542 (0.0217) (0.3640) Note. ',nOcates significance at the tO percent level; " indicates significance at the 5 percent level; i'+ndicates significance at the f percent level Standard errors are reponledin parentheses A constant was calculated but is not reponed Sounce: Based on the data from ffte Social Weather Stations, Intemational Monetary Funo, International Financial Statistics, and Philippines National Statistics Office. Table A2.2: Correlates of Self-Rated Poverty Bivariate Bivariate Bivariate Multivariate model model model model Real GDP per capita (x100) -0.6683 -0.7416 (0.1912) (0.1969) -- Inflation rate 0.4128 0.3705* (0.2351) (0.2018) Unemployment rate 0.1257 -0.9307 (1.201) (1.063) Numberof observations 29 29 29 29 R-square 0.286 0.102 0.000 0.435 Dickey-Fullertest statistics -3.638 -3.918" -3.645** -4.263"' Note: 'indicates signilianceatthe lOporcentlevel '"indicatessignificanceatthe5percentlevel;, "'indicates significance atfhe 1percentlevel Standarderrors are repoftedinparenteeses. A constant was calculatedbut isnot reponted. Source: Based on data fron the Social Weather Stations, lnternationalMonetarlyFund International Financial Statistics and Philippines National Statistics Office. 28 2:GROWING OUTOFPOVERTY Table A2.3: Correlates of Self-Rated Poverty Lines Bivariate Bivariate Bivariate Multivariate model model model model POOR HOUSEHOLDS Real GDP percapita (x100) -8.922 -14.77 (30.70) (28.31) lnflaton rate 99.57*** 96.16** (27.02) (29.02) Unemployment rate -156.62 -66.77 (157.50) (152.80) Number of observations 29 29 29 29 R-square 0.003 0.335 0.035 0.344 Dickey-Fullertest statistics -4.701*** -5.901 *' -5.225*** 6.081** NOT POOR HOUSEHOLDS Real GDP per capita (xl 00) -2.79 -17.86 (32.89) (35.43) Inflaton rate 46.41 36.15 (34.29) (36.32) Unemployment rate -204.45 -195.88 (166.96) (191.20) Numberof observations 29 29 29 29 R-square 0.000 0.064 0.053 0.102 Dickey-Fullerteststatistics -3.401 *** 3.790*** -3.681 ** -3.860** Note: 'indicetessignillcanceatthe tOpercentlevel "indicatessrnlflcanceatteff5percent level, ''indicatessignilicanceatthe 1percent/evel. Standa-dfderrorsarerepoitedinparentheses A constant wascalculatedbutisnot repofted Source: Basedon the data from #he Social Weather Stations, Intemahonal MonetaryFund Intemational FinancialStatistics and PhilippinesNationalStatistics Office. 29 e~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 0 Table A2.4: Sectoral Poverty Profile, 1985-1997 Incidence Depth Severity 1985 1988 1991 1994 1997 1985 1988 1991 1994 1997 1985 1988 1991 1994 1997 Agnculture 57.7 51.2 51.9 49.9 42.3 20.0 15.9 16.9 14.5 11.5 9.2 6.7 7.4 5.8 4.3 (-73) ('08) (-25) (-9R7) (-10.0) (2.7) (-71) (-10.1) (-9.9) (3.1) (-82) (-9.2) Mining 46.4 34.4 44.7 37.1 30.0 13.8 8.9 12.5 7.4 10.0 6.0 3.5 4.9 2.1 4.5 (- 1.8) (1.7) (-12) (- 1. 1) (-1. 9) (1.6) (-2.4) (1. 1) (4.7) (1.3) (-2 9) (1. 9) Manufactunng 31.4 21.9 20.9 16.5 13.5 9.3 6.0 6.3 3.8 2.7 3.9 2.4 2.6 1.3 0.9 (4.8) (-0.6) (-3.0) (-22) (4.5) (0.5) (4.7) (-0.5) (-3.9) (0.8) (4.9) (-26) Utility 17.5 10.8 12.5 9.5 9.5 4.3 3.0 3.8 2.3 2.4 1.4 1.2 1.4 1.0 0.9 (-1.0) (0.3) (-0.7) (0.0) (-0.6) (0.5) (-1.0) (0 1) (-0.2) (02) (.05) (0.2) Construction 39.6 33.8 33.8 34.5 23.1 11.7 9.2 9.9 8.9 5.0 4.7 3.7 4.1 3.3 1.6 (-22) (0.0) (03) (-6.7) (-26) (0.9) (-1.4) (-6.8) (-2.0) (1.0) (-22) (-6.1) Trade 27.3 18.6 21.3 17.8 13.5 6.9 4.7 5.6 4.0 2.9 2.6 1.7 2.1 1.3 0.9 (4.9) (1.7) (-24) ('-35) (-3.9) (1.7) (-3.4) (-3.0) (-3.0) (1.6) (-3.7) (-25) Transportabon 27.8 24.1 22.5 21.2 13.7 7.0 5.4 6.0 4.7 2.8 2.6 1.8 2.4 1.6 0.9 (-1.8) (-0.8) (-0.8) (-53) (-24) (1.0) (-23) (4,6) (-2,5) (1.9) (-28) (-3.6) Finance 13.2 8.5 6.9 7.1 3.0 2.4 1.6 1.9 0.9 0.5 0.8 0.5 0.7 0.2 0.1 -1.7) (-0.7) (0. ) (-2.5) (- tl ) (0.4) (-1.6) (-1.5) (-0.7) (0.7) (-. d) (- 1. 1) Services 20.0 15.4 15.2 12.7 9.9 5.2 4.0 4.0 2.8 2.2 1.9 1.5 1.5 1.0 0.7 (-3.5) (-0.2) (-24) (-31) (-28) ('0. ) (-33) (-22) (-20) (02) (-3.1) (-20) Unemployed 21.5 18.3 16.8 17.1 12.1 6.1 4.9 4.5 4.1 2.9 2.6 1.9 1.8 1.4 1.0 (-2.3) (-1.2) (0.3) (-52) (-25) (-0.9) (-13) (4.1) (-2.6) (-04) (-23) (-32) Note: Figures in parentheses are I-ratios ofthe difference between thepovetyrmeasure forthe curren/yearandthatforthe previous year. Source sdalisacan (2000) 2: GROWING OUTOFPOVERTY Endnotes These estimates are based on a poverty line that is calibrated to a nutritional norm of 2.000 calories per person per day and allows for basic non-food expenditure (defined as the actual nonfood expenditure of households who can just afford the food poverty line, or whose total expenditure is in the neigh- borhood of the food poverty line). For further details, see Annex A, Volume II, and Balisacan (I 999a). 2 Figures 2.1 and 2.2 present estimates for the so-called "a dollar a day" poverty line, but a similar pattern also holds for the "two dollars a day" poverty line. Also note that for the Philippines, these poverty estimates differ from those in Table 2.1 both on account of the different poverty lines used and also due to the use of spatial (provincial) cost of living differentials in the construction of estimates in Table 2.1. 3 One limitation of the data on self-rated poverty and self-rated poverty lines is that, prior to 1992, information on self-rated poverty is not available for a number of quarters. This concern was ad- dressed by condensing the data into semesters, so that the first semester runs from January to June, and the second semester from July to December. Simple averages over two quarters are taken of inflation and unemployment, and of self-rated poverty when data at more than one point in time within a semester are available. GDP per capita in a given semester is simply given by the sum of GDP per capita in the two quarters in question. GDP per capita and the subjective poverty lines were normalized by dividing through by the consumer price index, so that all reported figures are in 1995 Pesos. Finally, to facilitate comparisons between the self-rated poverty-line and real GDP per capita, the self-rated poverty lines have also been transformed into per capita (rather than per-household), per semester (rather than per month) poverty lines. Results in Table A2.2 are based on regressions of the following form: Yt= t+ PTt+ 6Si t+ t where T, is a linear time-trend, SI, is a dummy variable for the first semester, and £, is an error term. 4 Figure 2.9 graphs de-trended, de-seasoned measures of self-rated poverty and de-trended, de-sea- soned measures of real GDP per capita. Results in Table A2. 1 are based on regression analysis to estimate the relationships between real GDP per capita, unemployment, inflation, and self-rated pov- erty, using regressions of the following form: Poverty, = 9 + yoGDP, + y,inflation, + y2unemployment, + ±t where all of the variables have had the trend and seasonal components removed. Two sets of regressions were considered, corresponding to those which include only one explanatory variable (real GDP per capita, inflation, or unemployment), and those which consider all three. Removing the trend and seasonal components from all variables makes sense from an econometric perspective to correct for possibly spurious relationships. Consider, for example, a sccnario in which self-rated poverty and real GDP per capita both have trend components. A regression of self-rated poverty on real GDP per capita might well find a significant relationship between the two. It may be, however, that no true underlying relationship exists, and the trend component in self-rated poverty is actually explained by trend components in other variables-say, an increase in education levels. From an economic point of view, de-trending and de-seasoning may also be sensible. During the 1985-1999 period real GDP per capita in the Philippines exhibits a small, insignificant positive trend, and a predictable "saw-tooth" pattern, generally being highest in the 4th quarter and lowest in the l" quarter of the year. Secular increases in real income may do little to explain self-rated poverty if the self-rated poverty line itself increases with income-something which has been observed elsewhere (Ravallion 1994; Kapteyn et. al., 1988). Also, if households do not associate predictable seasonal patterns in real GDP with changes in their underlying welfare, we would expect seasonally adjusted estimates of real GDP to be better predictors of self- rated poverty. 31 PHILIPPINES POVERTY ASSESSMENT 5 Table A2. 1 presents the results of regression analysis of the de-trended, de-seasoned poverty lines in terms of the de-trended, de-seasoned values of real GDP per capita, unemployment, and inflation. 6 Balisacan (1993) demonstrated that the failure to take account of the "shifting of physical areas" arising from reclassification of villages would distort the overall picture on the actual performance of rural areas from the late 1980s to the early 1990s. The sampling frame for the 1985 and 1988 FIES was based on the 1980 population census, while that for the 1991 FIES was based on the 1990 census. Both censuses applied the same set of criteria in classifying villages into "urban" and "rural" areas. 7 The sectors are identified by the primary occupation of the head of the household. 8 In general, larger provinces tended to be less poor. For instance, there was a significant negative correlation of -0.37 between provincial poverty incidence and population shares in 1997. 32 CHAPTER3 SOCIAL SERVICES AND THE POOR 3.1 The poor in most settings have few assets other than their own labor power. A key reason for, their poverty is that their labor has low productivity and earns low returns in the m-arket. Building their human capital through better education and health is thus a powerful instrument for their escape from povery It is valuable ini its own right too: it directly enhances the ability of the poor to lead better lives- and ensure better lives for their children. It also em - powers them to make better use of institutions to avail of the existing economic and social opportu- ; ninies. 3.2 An equitable provision of social services in education and health is an important means of in- vesting in the human capital of the poor. How well do these social sector services reach the poor in the Philippines? This chapter focuses on the education and health sectors. It examimes the degree of imequal- ityin education and health endowments, assesses the extent to which government spending is progres- sive in these sectors, and in the case of education, looks at how differential endowments play out in the labor market. EDUCATION, LABOR MARKETS, AND POVERTY 3.3 This section reviews the evidence on edu- cation, wages, and poverty n the Phiippines. The five basic conclusions are: first, the broad pattern of public education spending in the Philippines iLs mnildly j j progressive. Poor households receive a Iiigher frac- ~\\ tion of benefits than the non-poor. However, the extent to which spending is progressive varies a great deal bylevel, so that spending on prirnaryeducation .,<,j.~~~~~~~~~~ DIES PHILIPPINES POVERTYASSESSMENT is progressive, spending on secondary education is Table 3.1: Government Spending in Education; 1997 neutral, and spending on tertiary education is Million % of Total regressive.Second, given its per capita income level, Php government the Philippines has achieved impressively high en- expenditure roilment rates in primary and secondary education; Elementary 43,804 8.9 Secondary 13,268 2.7 as a result, the average number of years of school- Tertiary D 13,800 2.8 ing completed has increased over time, and is high Vocational C 1,300 0.3 Other d2,152 0.4 compared to other countries in the region. Third, Total - 74,324 15.1 however the high averages mask large differences As percentage of GNP (%) 2.9 by region and by income quintile in educational at- MOte I represents allocation at regional level, including GATSPE and school building program; tamrnent; providing high quality education to some OSUCsandCHED; tradit.ona yd. advantaged groups sho*-ild therefore vocational education is underTESQA since 1968 traditionally disadvantaged groups should therefore pre-school andnon-fonmal programs after t994. be an important priorityfor the future. Fourth, there Source: WorldBank (1998c), Philippines: SocialExpendiure Porinties, basecon DBM data. appear to be large differences in the rate of return ementary school - and therefore any deviation from to education across levels. Specifically, the private rate that line suggests that the distribution of benefits of return is much higher for college than for pri- varies by income; public spending is progressive if mary school or secondary school. This suggests ei- the curve is concave (above the 450 line) and re- ther that the basic education which is provided is gressive, if it is convex (below the 450 line). Because not of sufficientlyhigh qualityand raises productiv- the analysis uses a standard unit cost per level of ity by only a moderate amount, or that the structure school, the incidence of spending for a given level of the economy is such that it does not reward lower of schooling simply reflects the pattern of utiliza- levels of education. Fifth, there is no evidence of tion of that level of schooling byincome category. gender discrimination in access to education; but the Incidence analysis (that is, unit cost information) be- same cannot be said for the labor market, where comes important when assessing the distribution of women appear to eam less than would be expected aggregate education spending because costs vary given their education and experience. widely by level of schooling. PATTERNS OF EXPENDITURE ON 3.6 The fact that overall public spending on EDUCATION education is rnildly progressive hides important dif- ferences in the incidence of expenditures at differ- 3.4 The Philippines government spends 2.9 per- ent levels: spending on primary education is pro- cent of GDP on education (Table 3.1); this is below poor; the distribution of expenditures on second- spending levels in Thailand (4.8) and Malaysia (0.2) aryeducation is largelyneutral; and public spending but substantially above those in Indonesia (1.4) and on tertiary education is highly regressive. Approxi- China (2.3).' The overall pattern of spending is mildly rnately 20 percent of the education budget is spent progressive.2 on tertiary education which caters to less than one million students from relatively wealthier back- 3.5 Figure 3.1 plots the cumulative share of ben- grounds. The scholarship programs of the govern- efits from public education (y-axis) against the cu- ment do not attenuate the pro-rich bias of tertiary mulative share of the population ranked by con- education; they appear instead to be equally skewed sumption (x-axis). This requires information on fre- (Figure 3.2). This may be driven, in part, by the fact quencyof service use by sub-category and unit cost that government scholarships are handed out on the information for the relevant sub-categories. The 450 basis of merit, rather than need, and children from line represents neutralityin the distribution - if, for poorer households are less likely to have access to example, the poorest 20 percent of the population the high-quality secondary education which would accounts for 20 percent of public spending on el- allow them to perform best on standardized tests.3 34 3: SOCIA L SER VICES A ND THE POOR Figure 3.1: Public Spending on Education Figure 32: Tertiary Scholarships Overall public spending on education is m Idly progressive but tertiary Scholarships for tertiary education overwhelmingly benefit the education benefits the well-off, 1998. better off in urban areas, 1998. All public education Total beneficiaries 7 1 7- 03- 2 0 T -r F ~1- -_T_ T _ o r 2 3 4 5 6 .7 .0 .9 1 1 2 3 0 0 0 7 8 9 Cum proportion of populat on (rank by per cap ta cons)i Cum proportion of popu ation (rank bv per capita cons) Primary education Rural beneticiaries 9~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ O t 3 .4 D5 7 a . 1o-- 005 4- ~ ~ ~ ~ ~ ~ ~ ~ ~ o 03- 2- 3 - a Second ary ed ucation umrI proportion ot pnp0iabir (rank oy per capira c000' X_ ___________L________ Urban beneficiaries Cun proportion of popuiation trank by per capita Consl o --___T--__ Tertaryducto 1 2 3 4 5 6 7 .8 .9 1 Curn proportion Of prpultiotn irirk by per oi'ota curio, .2 9- I-X | Source:Staffcacationsbased ono998yAPIStdata. 00 7- ,D 6 / / i ~~~ ~ ~~~~~3.7 Incidence analysis is useful in gauging the ex- rD 5 / / [ ~~~~~~~tent to wh,ich programs are able to target the poor 2o 3 but they cannot tell the full storybecause they are nec-. E 2 / essarilysilent on the qualityof services provided. It is . : f~~~~~~~~ possible to have highlyprogressive government spend- O Corn 3 .4 5 .6 .7 3 1 ing which delivers low qualityeducation or low qual- Cmproportion of popuiation leonk by per capita coons) ityhealthcarc; in fact, if the services are of low qual- Note. The vertrcaflinein thesegraphsrepresenfsthepoveflyheadcountwhich was estimated ity, public facilities are more likely to be successful in at 31.S8%in 1998SusingthreAPlS dataset; note, ho weverthat this Is not cormpa rabletothe targeting the poor who will have fewer options than 25% tr eadcount estimateri for t997 using FIES data. Source: Staff calculationa based on 1998 APIS data. the wealthy in seeking alternative service providers. 35 PHILIPPINES POVER TYA5ESSMENIN Educational attainment Figure 3.4: Educational Attainment, Ages 15-19 3.8 Are Filipino children and adults well-edu- cated byinternational standards? The answer to this D (93 question clearly depends on both the access which children have to education, and the quality of the t education which theyreceive. Figure 3.3 graphs gross r° 3 secondary enrollment as a function of per capita 05 incorne for a sample of developed and developing j countries in 1980 and in 1995. Figure 3.3 suggests that access to secondaryeducation in the Philippines, , Grade atained as measured by secondary enrollnent rates, has been Source: Fllmer(1999b), basedonDemographicandHealthSurveys(DHS)forbothcoun- and remains high.4 tnes. 3.9 The high enrollment rates achieved in the 3.10 Figure 3.5 graphs the mean years of school- ing completed by different age cohorts i the Phil- Philppines are confirmed in Figure 3.4, which shows n the educational attainment of 15 to 19-year olds for ippines. Figure 3.5 clearlyshows that older individu- the Philippines and .ndonesia, two countries in te .als are much less likelyto have high levels of educa- tion than their younger counterparts: for example, region with similar per.capita icomes. The x-axis of the mean educational attainment of the 1920 birth the graph con-esponds to a given gradLe 'm pnrlmaryv or thegraphc oeondsryschoo toe vaxgive grad res i e p aryor cohort is about 5 years of schooling, compared to secondary school, while the y, ams measures the pro- almost 9 years of schooling for the 1965 cohort. portion of 15 to 19-year olds who have attaimed at Still, even today, fewer than two-thirds of Filipino least that grade. Figure 3.4 suggests that the fraction children wo en er tma c inrad 1 Fin- of chlUdren who continue after elementary school is ch Grade 6(iin- higher inthePhppiesthan I a. ' e ae ish Grade 6 (Asian Development Bank and World higher in the Phlippines than in Indonesia. There are Ba,19,p.2) no important changes im educational attainment ithe Philippines between 1993 and 1998.5 3.11 Average educational attainn-lent can hide im- portant differences byregion, income group, or gen- Figure 3.3: Gross Secondary Enrollment and Income, e.Fgr . hw h vrg ubro er Various Countries der. Figure 3.6 Shows the average number of yean of school completed by 15 tol9-year olds, broken 1 ina Secondar enrollment rate 1995 down by gender, and place of residence. It shows 14 ; Phffippines K Korea 120e that the educational attainment of women is higher 80-a a than that of men, in both urban and rural areas. It iiO - s - M~".abys a 240- \:dn 0 ielum 5000 I 0oo0 150,00 zoooo 25000 Figure 3.5:- Educational Attainment by Birth Cohort Per capita GDP (US$ 1995 prices) Chfna Secondary enrollfnent rate 1980 .ne a r \' u I II 5 rI 40 3' 20 a am0 4030 6000 6000 I=o 12"0 140t Per capita GDP (US$1995 pnces) ,,, T 1 Birlh Cohorl Source: World Development Indicators, various issues. Source: Staffcalculations based on 1998 APIS data. .36 3: SOCIAL SFRVICF.5AND IHF POOR also shows that there are very large differences be- 13 (roughly corresponding to graduation from el- tween the educational attainment of urban and rural ementary school). By age 16, children in the richest residents. Young meniin rural areas, in particular, drop income quintile have completed two more years of out of school early, presumably to take up employ- schooling than their counterparts in the poorest in- ment in agriculture.6 come quintile. Note that these differences could be driven both by higher repetition rates and higher 3.12 Figure 3.7 graphs the proportion of chil- drop-out rates among the poor. dren aged 6 to 14 who are currently enrolled in school. Three different lines are presented - corre- 3.14 A recent study states that the "foremost is- sponding to the wealthiest 20 percent, the middle sue in education is the lack of equity between those 40 percent and the poorest 40 percent of the popu- in the Philippines "who have" and those who "have lation.7 Figure 3.7 shows that at any given age, the not" (Asian Development Bank and World Bank, proportion of poor children who are enrolled is 1999, p. 24). Given the clear relationship between substantially below the corresponding fractions for education and living standards, it is important that children from the better-off households. access to education in the Philippines be extended to traditionally disadvantaged groups - in particu- 3.13 Fimally, Figure 3.8, graphs the average years lar, to poor households in rural areas. Access to ba- of schooling attained by age and income quintile sic education may be determined, in part, by physi- for children aged 6 to 16.8 It shows large differ- cal access - that is, whether there is a primary or ences m the educational attainment of children in secondary school nearby As discussed earlier, sec- different income quintiles, especially after ages 12- ondaryschools are less likelyto be located in poorer barangays.9 However, there are clearly other factors Figure 3.6: Educational Attainment, Ages 15-19, by Gender which determine educational attainmcnt, including and Place of Residence, 1998 the language of instruction, which may make it more difficult for children from households where Fili- pino is not the mother tongue to make satisfactory D8 '- - ''-., "' progress; the direct private costs of education, which tE 07_ ', '.,, may be important for poorer households; the qual- ity of education; the difficulties which poor house- holds face in making the transition from secondary to tertiary levels; and the higher discount rates of poor households, which would lead them to value the income they can earn in the labor market now Grade attained Figure 3.7 Proportion Currentty Enrolled, Ages 6-14, Figu" 3i8: Average Years of Sihooling Attained, by Wealth Group, I998 by-Age and Income Quintile, 1998 01 _ 7 -. = = = 8 ' /~~~~~~~~~~~~~~~ <2 /; ~ ~ ~~ ~~~~~~ ~~~~~~~~~~ = nO ~~~~~~ 0 o F3 0 &~~~~~~~~ 0 6 7 1 1 1 Os 11 01 2 1S 11 10 1' 17 202Z Age n Years Age Source: Filmer (1999b), based on 1998 DHS. Source:Alba (2000), basedon 1998APIS data. PHIL IPPINES POVERTY ASSESSMENT more than the higher income they could eam in the child im primary secondary and tertiary school." future if they received more education. Table 3.2 shows that there are very large differences in the private urit costs of education by consump- 3.15 Children in better-off households receive tion quintile. The richest households spend more than more education in the Philippines. Poor households 20 times as much as the poorest households for ev- are doubly disadvantaged; however, if the quality ery child enrolled in primnary school, and more than of education they receive is lower than the quality 10 times as much for every child enrolled in tertiary of education received by the non-poor. Differences education. These extremely large differences across in the quality of education received by poor and quintiles are driven, in part, by the fact that poorer non-poor households may result in eventual differ- households are much less likely to send their chil- ences in productivity If so, the poor may not re- dren to private schools, where the unit costs faced ceive as high a labor-mnarket premium to an addi- by households are much higher than in the public tional year of schooling as the non-poor.'0 system.'2 However, there are also large, significant differences byquintile in the amount spent byhouse- 3.16 There is good reason to suspect that chil- holds per student within the public (or private) sys- dren from poor households receive lower quality tems: households in the richest quintile, for example, education than their better- off counterparts in the spend 12 times as much per student enrolled in public Philippines. A larger fraction of non-poor students prinaryschool as households in the poorest quintile, are enrolled in private institutions, where quality ap- 6 times as much per student in public secondary pears to be higher: almost one-half (46%) of 6 to school, and 9 times as much per student in public 24-year olds who are enrolled in school in the rich- university This suggests that better-off households est income quintile attend a private school, compared spend more per student on vital education inputs to less than 5 percent of 6 to 24 year olds in the such as textbooks and learning materials - all of poorest income quintile. The fact that a large frac- which have an important effect on the amount of tion of those who can afford to send their children learning which takes place in the classroom and, ul- to private schools likely reflects underlying differ- tirnately, on the productivity of workers. ences in quality between the public and private sys- tems. Indeed, cohort-survival rates, and the perfor- mance of students on standardized tests are both Table 3.2: Direct Private Costs of Education, by Education higher in private than in public schools. Qualitym inLevel and Consumption Quintile All Quintile the public sector is also compromised by the fact HH 1 2 3 4 5 that a very high - and growing - fraction of total (poorest) (richest) expenditures in the sector are devoted to teacher Primary salaries, at the expense of maintenance and other (per student) operating expenditures (MOOE). Inper-pupilteris, %oftotalHH 0.7 0.6 0.7 0.9 1.3 2.0 maintenance and other operating expenditures was consumption only PhP135 in 1997 (less than US$4), which was Secondary Unit costs 2745 661 1263 1696 3052 7295 only about a quarter of what it had been in 1990 (per student) (Asian Development Bank and World Bank, 1999). % oftotal HH 2.0 1.5 1.9 1.9 2.4 2.6 consumption Tertiary 3.17 The problem of qualty differentials in the Unit costs 13334 1879 3487 5690 8512 20355 schooling received by poor and non-poor children (per student) is compounded by differences in the amount spent consumption 9.7 4.1 5.3 6.5 6.8 7.2 per student. Table 3.2 sumtnarizes the amnsumptien per student. Table 3.2 summarizes the amount spent Note; Quintiles are 1997per capita consumption quintules. byhouseholds in different consumption quintles per Source: Staff calculations based on 1997FIES (FIES) and 1998APIS data. 38 3: SOCIAt SERVICES AND THE POOR Labor market outcomes Table 3.3: Rates of Return to Education and Experience Specification 3.18 Education is a key determinant of wages, Household Within- Adults and wage income is an important source of income Basic heads family without mincerian only estimates children for poor households. This section examines the re- lationship between education and wages in the Phil. Years of education 0.133l -.124k" 0.114- 0.124** (0.002) (0.002) (0.006) (0.003) ippmes and tnres to provide answers to three ques- Experience 0.027*" 0.025*** 0.054 .0.022** tions: first, what are the private rates of return to (0.002) (0.003) (0.007) (0.008) education? Second, is there evidence of discrmina- Experience squared -0.033* -0.034* 0.004 0.129*** tion bygender in the Philippine labor market? Third, (xlOO) (0.004) (0.005) (0.019) (0.026) are certain kinds of education associated with par- Constant 1.157* 1.374 1.331* 1.559 ticularly high or low prernia im the labor market?13 (0.034) (0.048) (0.104) (0.077) R.square 0.291 0.278 0.265 0.243 3.19 The basic approach to estinating the pri- N 25,231 13,719 3023 9061 vate retums to education taken in this report is to use a regression framework to decompose the rela- Note: The dependent vaable in every specification is the log of the houry wage. Standard tionship between (the log of) hourly wages on the aged26-65whoreportearninrgwagesareincludedinspecification l;specification2islimited to household heads aged 25-65 only; specification 3 is limited to siblings in households in one hand, and years of education and experience which there are at least two siblings who report wage income, and includes an additional ternn 14 ~~~~~~~~~~~~~~~~~~for the average years of education of all siblings who report wage income, and includes an on the other. Results from this estimation are sum- additionalterm forthe average years of education offall sib ings in a household who report wage marized in Table 3.3." Column 1 presents the basic income; specification 4is limited to adults aged25-35 who are household heads or spouses in households which havenonmembers whose relation to the householdheadis "son ordaugh- regression results for all wage- earners aged 25 to 65 ter;andhouseholdmembers whoserelationship to thehouseholdheadis reportedas 'son, daughter,son-in-law,ordaughter-in-lawinhouseholdswheretherearenomemberswhose mti the sample; column 2 umlrnts the sample to house- relationshiptothehouseholdheadisreportedas grandsonorgrarddaughter.' hold heads as a partial correction for selection into ... significant at the 0.1% level; tsignificant at the 1% level; 'significant atthe 5% level. Source: Staffcalculations based on the 1998 APIS data. the labor rarket. This is an issue because not all people who receive education enter the labor mar- ket: many work on their own farms, or do house- within a given household.17 Column 4 finally, ad- work at home. We therefore observe wages only dresses an important concern, namely, the fact that for a self-selected group of potential wage earners. the measure of experience in the regression is a If, as appears likely, these wage earners are funda- measure of "potential experience" rather than "ob- mentallydifferentfromnon-wageearners, estirated served experience," where potential expenrence is results may be inaccurate;'6 column 3 limits the defined as age minus years of schooling minus six sample to siblings in households in which there are Actual experience is rarely measured in household at least two siblings who report wage income, and surveys in the Philippines or elsewhere, so including includes an additional term for the average years of "potential experience" in a wage regression is a com- education of all siblings in a household who report mon compromise. Unfortunately, "potential expe- wage income. If people who earn higher wages do rience" may seriously over-estimnate the number of so, at least in part, because they are more able, and years of experience of women if women who are ability is not properly measured, a "simple" regres- currently in the labor market have taken time out at sion framework would (incorrectly) attribute all of an earlier point in their career to raise children. This the differences in earnings to differences in school- is a particular source of concern for estimates of ing. The approach we take to partly correct for gender discrimination. We correct this problem by unobserved ability is to assume that this ability is lirniting the sample to men and women who are shared by some members in a household - say, all extremelyunlikelyto have had anychildren. Specifi- siblings. This means, in effect, that we estimate the cally, we run regressions which include only house- rates of return to education by comparing siblings hold heads and their spouses aged 25 to 35 years 39 PFYR UiMNES- POVfKTf A5SE5SA4hW old in households where the household head reports women. This gives the "explained" difference in having no children, as well as children and children- earnings - that is, the predicted wages which women in-law of the household head aged 25 to 35 years would earn if their characteristics were rewarded in old in households where the household head reports the labor market in the same way that they are for having no grandchildren."8 men. The remaining difference in mean wages is un- explained, and may be a result of labor market dis- Crininat' 21 3.20 Three points are worth noting on the re- cnmination. sults in Table 3.3. First, the estimnated coefficients show that education has a large, significant effect on wage 3.22 In the Philippines, the mean hourly wage earnings in the Philippines. On average, an additional of women is above the corresponding wage for year of education increases hourly wage earnings by men. For example, in the full sample of wage-earn- 12 to 14 percent.9 These estimates are high byinter- ers in the APIS survey, women earn, on average, national standards. A global review by about 6 percent more than men per hour. At face Psacharopoulos (1994) reports a mean rate of re- value, this would seem to suggest that there is no turn to education of 10.1 percent for the world, discrimnination against women in the labor market. and of 9.6 percent for a sample of Asian countries. However, women in this sample also have an aver- Education increases wagc income bya large amount age of almost two more years of education than in the Philippines, and should therefore be a keyin- men. The question, therefore, is whether the differ- gredient in any strategy designed to reduce poverty ential in mean hourly wages between men and Second, the estimated coefficients are quite similar women would be larger even if men and women across specifications. It therefore appears that selec- were rewarded in the labor market in the same way tion, unmeasured ability, and the imperfect nature for education and experience. of our measure of experience do not result in large biases in the estimates. Third, the coefficients on ex- 3.23 Table 3.4 presents estimates of the explained perience and experience squared show a familiar and unexplained components of wage differentials hump-shaped pattem, whereby additional experi- in the Philippines. The upper panel gives the esti- ence is rewarded for the first 40 to 45 years, and is mated rates of return to education and experience associated with lower mean earnings thereafter. for men and womnen, for the full sample (columns 1 and 2), for the sample of household heads only (col- 3.21 We next turn to estimates of the presence umns 3 and 4), and for the sample of young adults (or absence) of labor market discrimination against without children (columns 5 and 6). A number of women in the Philippines. Specifically, is there evi- things should be noted about these results. First, the dence that women earn lower wages than we would rate of return to education is substantiallyhigher for expect them to, given their characteristics? The basic women than for men: an additional year of educa- approach we take decomposes observed differences tion leads to an increase in mean hourlywages earned in mean hourly wages between men and women by women of 16 percent to 18 percent, compared into "explained" and "unexplained" components.20 to an increase of 12 percent to 13 percent for men. The observed difference is simply the difference in Elsewhere, it has also frequently been observed that mean hourly wages between men and women in the rates of return to education for women are the sample. To break this observed difference down higher than the corresponding rates for men. In the into "explained" and "unexplained" components, review by Psacharopoulos (1994), for example, the we run regressions similar to those for Table 3.3, average rates of return for all countries in the sample separately for men and women. The estimated co- are 11.1 percent for men's education, and 12.4 per- efficients on the regressions for one group - say, cent for women's education. Philippine data show, men - are then applied to the mean years of edu- however, that the rates of return for both men and cation and experience of the other group - say, women are large byinternational standards, as is the 4t _ 3: SOCIAL SERVICES AND THE POOR Table 3.4: Rates of Return to Education and Experience, by Gender Specification Basic household Adults without mincerian heads only children M W M W M W Years of education 0.123' * 0.166''* 0.123--- 0.153- 0.111-- 0.167--- (0.002) (0.003) (0.002) (0.006) 0.004) (0.007) Experience 0.031--- 0.017--- 0.025"' 0.011 -0.019 -0.030- (0.002) (0.003) (0.003) (0.009) (0.010) (0.013) Experience squared (x100) -0.039--- -0.016 -0.031 --- 0.009 0.1 10-- 0.177--- (0.004) (0.006) (0.005) (0.015) (0.033) (0.054) Constant 1.248--- 0.758--- 1.391"-- 1.019--' 1.720 0.965--- (0.040) (0.051) (0.50) (0.143) (0.099) (0.128) R-square 0.253 0.392 0.269 0.398 0.207 0.323 N 16,738 8,493 12,472 1,247 6,163 2,898 Observed mean hourly wage 17.83 18.02 18.54 17.50 15.93 18.14 Observed % difference (W,0 W°)/W ° 1.1% -5.6% 13.9% Predicted mean hourly wage 17.83 21.73 18.54 21.39 15.93 20.19 Predicled % difference (W,p - W,P)/WNP 21.9% 15.4% 26.7% Residual % difference (W,P - )/W0N 20.8% 21.0% 15.5% Note: The dependent variable in everyspecification is the log of the houry wage. Standarderrors conrectedforheteroskedasticityandclusterng are reportedin parentheses. Alladultsaged 26-65years old who reporteaming wages are includedin specification I; specification 2is limited to householdheads aged25-65years oldconly; specification 3is limded to adults aged25- 35years old who are householdheads orspouses in households which have nomembers whose relabon to the householdheadis 'son or daughter,"andhouseholdmembers whose relationship to thehouseholdheadis reportedas 'son, daughter, son-in-law, ordaughter-in-lawin households where there are nomembers whose relationship to the householdheadis reportedas 'grandson orgranddaughter.' "Observed % difference"is the observeddifference in mean hourly wages between women andmen, asa percentage of men's observedmean hourly wages. *Predicted % difference'is the predicted difference in mean hourly wages if women were paidaccording to men's salary structure, as apercentage of men's predictedmean hourly wages. "Residual % diflerence"is simply the differencebetween the observedandpredicteddifferences in mean hourly wages. *-t significant at the 0.1% level; **significant at the 1% level; 'significant at the 5% level. Source: Staffcalculations basedon the 1998APIS data. difference between themL High rates of return to similar for men and women. The third point worth female education likely encourage women to get noting about the top panel of Table 3.4 is the con- more schooling. Education is therefore a particu- stant in the regression. This constant gives the mean larly important vehicle for women to earn income expected log hourly wage for men and women with and escape poverty Moreover, higher levels of fe- zero years of experience and no education. This es- male education and higher wages, which are associ- tirated "starting" wage is consistentlyhigher for men ated with them, are likely to have other beneficial than it is for women. For examnple, in the full sample, effects. International evidence suggests that both the expected hourly wage for men with no educa- women's education and the contribution of women's tion and no experience is PhP3.48, while the com- income to total household income improve child parable wage for women is PhP2.13. Additional health outcomes, and increase the proportion of the years of experience add to the wages of men and household budget that is devoted to expenditures in women in much the same way and additional years education and health. of education add more to the wages earned by women than they do to those earned by men. How- 3.24 A second point in Table 3.4 is that there do ever, women enter the labor market at a substantial not appear to be large differences in the wayin which disadvantage.22 labor market experience is rewarded for men and women. Separate calculations show that the returns 3.25 The lower panel in Table 3.4 decomposes to experience for any given level of education are the differences in mean hourly wages between men 41 PHILIPPINES POVERTYASSESSMENT and women into the explained (or predicted) per- first column of Table 3.3 as a reference point, our cent difference and the unexplained (or residual) results would suggest that (controlling for experi- percent difference. Table 3.4 clearly shows that al- ence) workers with one year of schooling would though the mean hourly wages of women in the eam, on average, an hourly wage which is 14 per- Philippines are higher than those of men, women cent higher than their counterparts with no educa- would make even more in relation to men if their tion, just like those who have completed prirnary characteristics were rewarded in the labor market in school would have an hourly wage which is 14 per- the same way Specifically we would "expect" to cent higher than that eamed by workers who have see a wage gap of between 15.4 percent and 26.7 only completed fifth grade. It is not clear whether percent, rather than the observed wage gap of nega- this is a reasonable assurnption to make in the Phil- tive 5.6 percent to 13.9 percent. Note finally, that ippine context. This is a critical issue because differ- the difference between the observed and predicted ences in the marginal rate of return to education differences im wages is smallest for the sample of may have an important effect on the amount of childless men and women. In columns 5 and 6, the schooling chosen by households of different char- observed difference in wages is about one-half the acteristics. predicted difference. This suggests that much, but by no means all, of the very, high gap between ob- 3.28 In Figure 3.9, we graph the relationship be- served and predicted wage differences for women tween wages and education, for different levels of in the full sample can be attributed to the fact that experience; years of experience have been lumped our measure of "potential experience" is unlikelyto together into five-year groups, such that the first coincide perfectly with actual experience for women group (experience = 1) corresponds to workers with who have bome children. 0 to 5 years of experience, the second group (expe- nience = 2) corresponds to workers with 6 to 10 3.26 In conclusion, there are large, unexplained years of experience, and so on. Each point on a differences in the earnmings of men and women in graph corresponds to the observed mean wages for the Philippines. There are numerous possible expla- a given combination of education and experience.23 nations for these differences. First, women may be The vertical lines in each graph correspond to gradu- penalized by employers for the fact that they could ation from primary and secondary school. drop out of the labor market in the future to raise families. Second, the wage differentials at very low 3.29 Figure 3.9 shows that the log-linear regres- education levels could be explained by the fact that sion model which was used as the basis of the esti- the productivity of men in sectors which require mations in Tables 3.3 and 3.4 is, in general, a reason- physical labor - such as construction, or agricul- able approximnation to observed earnings patterns ture - is higher than that of women. This proposi- in the Philippines. However, there are some patterns tion cannot be verified in the absence of gender- in the data that would not be captured by a simple disaggregated data on productivity for these sec- regression such as that which is the basis for the es- tors. Third, it is possible that employers simply dis- tirnations in Table 3.3. First, in many of the graphs, criminate against women because of their gender. especially those which correspond to the earlier ex- perience cohorts, the slope of the earnings function 3.27 All the results in Table 3.3 assume a con- appears to be less steep for primaryschool and (less stant marginal return to education - that is, they clearly) for secondaryschool than for universityedu- assume that the percentage increase in wages is the cation. Second, the step-size for the last year of same from an additional year of education, regard- schooling within a given level (primnary school, sec- less of whether this year of education is, say the ondary school, and university) often appears to be first, the sixth, or the ninth. For example, taking the much larger than for other years. This suggests that 42 exper == exper ==2 exper= =3 exper= 4 3.879611 I ~X 0) .909694 (0 exper =5= exper =6 =xe= 10p15 15 .90969421 1-1 I _ l I I I 15 1S Years of Completed Schooling Source: Staffcalculations basedon the 1998APIS survey. Years of experience havebeen lumpedtogetherintofive-yeargroups, such that the firstgroup (expenence= 1) corresponds to workers with O to5years of experience, thesecondgroup (expenence =2) corresponds to workers with 6 to 10years of experience, andso on. Eachpointon a graph corresponds to the observedmean wages foragiven combination of education andexperience. 0) PHILIPPINES POVERTYASSESSMENT there are large wage preniia to graduation from a (primarily) that there are very high rates of return to given level of education - wage premia which far universityeducationin the Philippines. exceed completion of any other year. 3.33 One conccm with the graphs in Figure 3.10 3.30 Separate estimation of a model which al- is that theymaynot accuratelyreflect the relative costs lows the rate of return to education to vary by level and benefits of additional education if the prob- shows that the mean rate of return to education in ability of employment differs for primary school, the Philippines is lowest for primaryschool (7.2%), secondary school, and university graduates. For ex- higher for secondaryschool (10.4%), and highest for ample, if university graduates are much more likely college (19.3%). This estimation also shows that there to be unemployed than secondary school graduates, are signifieant increases in wages from completing young people will factor this in when they make the last year within any given level: on average, wages decisions about whether to seek additional school- of a primary school graduate are 9.9 percent higher ing. Specifically, economic theory predicts that the than those of a comparable worker with only five "expected" payoff to additional schooling would years of education. Likewise, a secondary school be given by the product of the rate of return and graduate eams 13.4 percent more than someone with the probability of employment. The APIS survey three years of secondary education, while a univer- suggests that there are, indeed, differences in unem- sity graduate earns 26.6 percent more than a com- ployment byeducation level: the unemployment rates parable worker with four years of uiu'versity educa- are 2.2 percent, 15.3 percent, and 17.2 percent for tion.24 These results suggest that credentials, in par- workers with primary education, secondary educa- ticular, a university degree, are themnselves rewarded tion, and universityeducation, respectively The large m the Filipino labor market. difference in unemployment rates between workers 3.31 The graphs in Figure 3.9 can be combined with inforrnation on the costs of education to esti- DSWUMr Rate: 0 025 mate the net present value of total lifetime earnings for individuals with different levels of education. In Figure 3.10, we present estimates of these earning streams for individuals who have completed pni- mary, secondary, and tertiaryeducation, respectively. Ursb /r We show graphs corresponding to two different '/ discount rates of 0.025 and 0.05. (The discount rate X is a measure of the extent to which income now is - valued relative to income in the future).25 1 0 5 ,T T 1 40 1 4' Years Since Graduaion f-ron Primary Schoo 3.32 Figure 3.10 shows that most Filipino chil- Dicoant Rate: 0.05 dren would be well-served by seeking additional schooling beyond primary school: only those who expect to be in the labor market for less than nine ,ec i years should stop their schooling after graduation from prirary school. Indeed, university education is the dominant strategy for all those who expect to - // Pores spend at least 15 years after graduation from pri- u- T - T r mary school seeking additional education or in the Years SnceGraduatio fr Pr SOo labor market. The explanation for this finding is Source:Staffcalculationsbasedonthe1998APISdata. 44 3: SOCIAL SERVICES AND THE POOR with prinmary school and secondary school - 13 does not fully capture actual and differentials in the percentile points - would tend to make secondary quality of care received by the poor and non-poor. education less attractive to primnary school gradu- To the extent that the quality of services is higher in ates. On the other hand, the small difference in un- private facilities which the non-poor tend to prefer, employment rates between workers with secondary the poor may suffer from lower health status de- school and university- 2 percentile points - should spite having reasonably good access to public facili- have a much smaller effect on the decisions made ties. In fact, the inferior quality of public facilities, by secondary school graduates. Incorporating dif- especiallyof non-hospital facilities, probablyaccounts ferences in the probabilities of unemployment into for their ability to target the poor effectively the analy,sis confirms that secondary education is poorly rewarded in the Filipino labor market. 3.36 A recent draft report by the Department of Health (DO") concludes that "the health sector 3.34 A number of conclusions can be drawn in the Philippines is not doing well."27 This report from Figures 3.9 and 3.10. First, further education, identifies three main problem areas: (i) the inappro- in particular, university education, appears to be a priate service deliverysystem; (ii) the inadequate regu- worthwhile investment in the Philippines. Second, latory mechanism; and, (iii) the poor structure of the direct private costs of secondary education, and financing. While mnany of the problemns in the health even university education, are more than offset by sector have a negative effect on the entire Philippine the higher expected income stream of workers with population, some have consequences, which are par- more education. Even so, the direct private costs of ticularly devastating for the poor. The DOH report tertiary education mnay be too high for the poor to points out, for example, that while the overall physi- afford - especiallysince theyare unlikelyto be able cian-to-population ratio in the Philippines is com- to borrow against future income. A program of parable to that in Taiwan, and exceeds the ratios in targeted scholarships fortertiaryeducation forprom- Thailand and Indonesia, these physicians are not dis- ising high school graduates from poor farnilies may tributed evenly across the country only 10 percent therefore be appropriate. This is a recommendation of the doctors, dentists and pharmacists, 20 percent that has been made in earlier reports (Asian Devel- opment Bank and World Bank, 1999). Table 3.5: Total Government Spending on Health HEALTH SERVICES AND THE POOR Personal Public health health care care Others Total % Share 3.35 The Philippine government spends about 1991 (billion PhP) 1.6 percent of GNP onprovidinghealth care (Table National Government 7.19 2.09 2.95 12.23 71.9 3.5), close to levels in Thailand (1.7) and Malaysia Local Government Units 0.28 0.73 0.38 1.39 8.2 3.5), close to levels in Thailand (1.7) and Ivialaysia Social insurance 1.82 1.57 3.39 19.9 (1.3) but substantially above Indonesia (0.6) (World Total Government 9.29 2.82 4.9 17.01 100.0 Bank, 2000f). Access to health services shows a dis- % share 54.6 16.6 28.8 100.0 tribution profile similar to education (Figure 3.1 1). Total Govt. as % of GNP 0.7 0.2 0.4 1.4 tribution profile similar to education (Figure 3.11). 1997 (billion PhP) Overall spending on public health facilities is pro- National Government 9.89 4.07 4.67 18.63 46.0 gressive. However, not all types of public health Local Government Units 3.92 7.88 3.68 15.48 38.3 Social insurance 3.87 2.49 6.36 15.7 spending is progressive."6 The overall incidence re- Total Government 17.68 11.95 10.84 40.47 100.0 flects a combination of a pro-poor distribution of % share 43.7 29.5 26.8 100.0 . , . .. ~~~~~~~Total Govt. as % of GNP 0.7 0.5 0.4 1.6 barangayhealth facilties and rural health centers but Note: Nationat government includes the Department(Health (DOH) and olhernational agen- regressive distribution of government hospital ser- cies; forexampletheDefenseDepartmentPersonalhealthcareincludeshospital-based vices. As in education, the information on public services. Socialinsuranceincludesthenationalhealthinsuranceprogram. 'Others in- cludes general administration and other support services. facilityuse and mcidence of spending presented here Source:Solon, et at. (1999). 45 PHILIPPINES POVERTYASSESSMENT All public health facilities Barangay health station .7- 7-1 o 1 2 .3 4 .5 6 .7 .8 .9 1 o 1 2 .3 .4 5 .6 .7 8 9 1 Cum proport on of population (rank by per cap ta cons) Cum. proportion of populabon (rank by per cap la cws) Rural health unit/center Government hospital 3.1 2.3 475 6.7.8 .8 1 o .r .23 4.5 6.7.6.8 1 Cum proporn on of populalon (rank by par capita cons) Cum. propornior of pJopulation (runk oy. par capita coos) Note: Jhe vertica/l le un these graphs represents the poverty headcount which was estimated at 31.8 % in 1998 using the APIS dataset; note, however that this is not comparable to the 25 % headcosunt estimated for 1997 using FIES data. Source Staff calculations based on the 1998 APIS data. Of nurses and medical technicians, and 35 percent Table 3.6: Health Status and the Poor of nurses practice in rural areas in the Philippines, Coumtn where the vast mnajority of the poor live. Moreover, auintiles Lowest Second Middle Fourth Highest even when the poor can visit a doctor or nurse, the Infant mnortality rate 2' cost of drugs and mnedicine is often prohibitive. Urban 49.7 40.1 37.6 24.8 17.7 Rural 48.7 38.7 28.4 25.1 (35.5) Under 5 Mortality rate a~ 3.37 Poor health status both contributes to and Urban 70.5 62.9 57.9 33.2 26.9 Is an outcomne of poverty Inequallty in access toRu ral1 5. 38.8c33.7a(39.8) quality care is reflected in large variances in impor- 1/qtuintiees are based on the entireApopudation;2adeathsounderageth 2 monthscper thousand births; 3/deaths under 5 years per thousand births; Figures in parentheses indicate large tant indicators of health status (Table 3.6). The poor sampling errs rs due to smallnumber ofcases. Sourepetence muc hig r rates of ia n cLd11 Source:Fimer(1999b,based on1998PhilippinesAHS. of nurses n d m dc al t echnicirate s , ormand 35u pecen ale36 eat tau ntePo mortalit,vthan theirnwealthier counterparts. There are negligible rural-urban differences in infant mortality of the national wealth distribution; however, this among the poor but the poorest in rural areas expe- requIres more analysis. nence higher child mortalit,v, note that the lowest quintile is nationally def ined so that rural-urban dif- 3.38 The incidence analysis also suggests that the ferences that exist are not driven by consumnption devolution of health spending mnay have contrib- differentials. The data suggest that the health status uted to making the overall pattern of health spend- of the middle quintile is lower in urban than in rural ing more pro-poor. This is the result of a change in areas; this mnay reflect poor sanitation conditions the composition of public health spending, with a among the urban poor who maywell be in thebmiddle near doubling of the share of public health care in 46 3: SOCIAL SERVICES AND THE POOR total (public) health spending since the devolution, brought about by the devolution of health services from 17 percent in 1991 to about 30 percent in 1997 (DOH p. 5). Other assessments of the devolution (Table 3.5). This tends to make the overall spending experience in health care delivery note several defi- more pro-poor because public health care spending ciencies (Solon, et. al., 1999). These include substan- refers largely to spending on facilities other than tial increases in the burden on local health systems; hospitals, which were found to benefit particularly continued problems with procurement and logistics the poor. The shift in the composition of overall (documentation requirements for the procurement spending mainly reflects the dramatic increase in of drugs, medicines and hospital supplies can re- public health care spending bylocal government units quire from 30 to 60 signatures); and problems of (LGIJ, whose share in the total health budget in- health financing which have also been compounded creased from 8 to 38 percent over the same period. by the passage of several laws since the 1993 devo- With the devolution, public heakh care spending by lution mandating increases in salaries and benefits LGUs was expected to rise faster than that by the of health workers. In many respects, these prob- national government, but the rise in LGU spending lems are not unique to the Philippines. International more than compensated for the slower rise in na- experience suggests that devolution is a complex tional government spending. policyintervention where success depends on a large number of factors (see Box 3.1). 3.39 While devolution mayhave helped make the pattern of public health spending more pro-poor, 3.41 There remain important challenges for the the current health system still has a pronounced per- health sector in the Philippines. As mentioned above, sonal health care bias, especially at the level of na- the Department of Health has already identified tional government. As noted above, national some keyareas for future reforn. In operationalizing government's relatively higher spending on personal the sector's future reform agenda, perhaps the most health is partlythe intended consequence of devolu- important challenge would be realizing the consid- tion. But how the personal health care budget is erable unexploited potential for better targeting of allocated both functionally and geographically also the benefits of heath services to the poor in a de- makes an imnportant difference to the distribution centralized setting. There also remainmimportant gaps of benefits from that spending. For instance, the in our knowledge, which points to a large agenda Department of Health spends more than half of its for future analytical work including: analysis of the budget on about 50 hospitals, which concentrates determinants of health outcomes (disaggregated for the benefits of this large component of the national rural and urban areas), the most important of which health budget to the limnited catchment areas of these may lie outside the sector (e.g., water, education); hospitals. review of the experience with decentralization to determine imnpact on the quality of services pro- 3.40 Overall, devolution appears to have been a vided, and success in reaching the poor; review of mixed blessing. The DOH report concludes that the the geographic and functional distribution of spend- "ineffective mechanism for providing public health ing by the Department of Health; and, review of programs" in the Philippines is, in part, due to "the the program of health insurance. fragmentation of the primary health care system 47 PHILIPPINES POVERTYASSESSMENT Box 3.1: Redistribution and the Provision of Social Services Efforts to decentralize various government functions, including the provision of social services, are underway in scores of developing countries. Perhaps the most important economic appeal of decentralization is that it can result in increases in allocative efficiency if lower tiers of government have better information about household preferences. Because decen- tralization can make more apparent the connection between taxes collected and services provided, it may increase consumers' willingness to pay for these services. Decentralization can also provide better opportunities for local residents to participate in decision-making, resulting in greater accountability of public officials, and strengthening democratic processes. On the other hand, decentralization may lead to increases in regional disparities and greater inequity. Moreover, if local governments do not have adequate capacity, some of the expected increases in efficiency could fail to materialize. There is no cookbook recipe for successful decentralization, but the following issues require consideration. The function or service to be decentralized: Decentralization is not an "all-or-nothing" proposition. In many instances, efficiency gains may be possible without increases in inequality if the central government keeps primary responsibility for financing while local governments take over responsibility for spending decisions, inputs, and implementation. Some social services may inherently be more difficult to decentralize than others. Economic theory suggests that redistribution may best be carried out by higher levels of government, because labor mobility will make attempts by lower jurisdictions to change the distribution of income self-defeating as the poor gravitate to areas of high redistribution, while the nch cluster in areas of low redistribution. Still, even if central governments take primary responsibility for financing of safety nets, and establish the criteria which determine eligibility for transfers, local govemments may have an informational advantage in screening applicants. Decentralization of the health sector is also complicated because of the need for effective referral across levels-from health posts which provide basic services, to high-technology hospitals. Unless these inter-linkages are considered carefully, decentralization can result in a deterioration of some aspects of the services provided, as may have happened in the Philippines, Bolivia, and Zambia. The level of the sub-national government to which responsibilities are decentralized: Economists often argue that decentralization should follow the principle of "subsidiarity," whereby decisions are made at the lowest level of government consistent with allo^ative efficiency. This often involves a careful parsing out of responsibilities. In education, for example, national governments are often responsible for setting standards, curriculum development, and textbook production and distribution; and local governments, communities, and parent-teacher associations are responsible for construction and maintenance of school facilities, and the day-to-day running of schools, as what happens in a multitude of countries, from the United States to Bhutan. International experience suggests that efficiency gains in the provision of social services frequently materialize when the central government devolves responsibilities to the community or facility level, but rarely when they are devolved to provinces or regions. In every instance, it is important that revenues for social services follow responsibilities for their delivery. The extent of community mobilization and oversight: Increases in allocative efficiency in the delivery of social services can only take place if more accurate local information can reach decision-makers, and if there are mechanisms whereby these decision-makers are held accountable for their performance. In Colombia, accountability to constituents pushed local mayors to concentrate more on training and hiring effective civil servants. In Northeast Brazil, community oversight and fear of job loss helped motivate civil servants. Initial conditions: The initial distribution of income is important. If income is distributed more unevenly within jurisdictions than across them, decentralization could be equalizing if local authorities have the capacity to transfer income to the poor and share the equity objectives of the center. On the other hand, if there are large initial differences in capacity or jurisdictions do not share the same equity objectives, some sub-national governments may not effectively target the poor. Central governments may therefore have to target poverty funds themselves or create stronger incentives for sub-national governments to do so. 48 3: SOCIAL SERVICESAND THE POOR Endnotes 1 See World Bank (2000f). 2 Unit cost estimates are from the 1998 Philippine Education Sector Study (1999), Table B12, and are for 1997. The enrollment information used in this analysis is from 1998 APIS. While survey data and administrative data on primary enrollment in public schools are quite close, there is a large gap for tertiary education. Survey data show 60 percent more students in public colleges and universities in 1998 than adrministrative data for 1997. The overall incidence of public spending on education would likely be more progressive using administrative enrollment data given that they show smaller number of students in the most regressive and costly level of education. 3 There is a huge discrepancy between survey data and administrative data on scholarship recipients. 339,000 families responded positively to the question in the survey on whether any members of the farmily had received tertiary education assistance. This suggests that a minimum of 339,000 had received some assistance compared with 36,551 from administrative data. There are some difficulties in interpreting the survey question: it could be interpreted to include private assistance and it is not limited to currently enrolled students. 4 These graphs are based on administrative data kept by the countries themselves. See World Devel- opment Indicators, 1998. 5 Figures 3.4, 3.6 and 3.7 are based on data from Demographic and Health Surveys conducted in the Philippines and Indonesia (Filmer, 1999b); by contrast, Figures 3.1 and 3.2 are based on the APIS survey. 6 Similar graphs for Indonesia show an even larger gap between the educational attainment for resi- dents of urban and rural areas, and no difference by gender. 7 The DHS do not collect information on income or consumption. They do, however, collect informa- tion on a number of measures of household assets - for example, whether a household owns a car, bicycle, clock, and so on - as well as information on the characteristics of living quarters - for example, the number of rooms in a home, and the material used for construction of the floor, walls, and roof. Principal components can be used to aggregate these data into an "assets composite" which provides a measure of the wealth of the household, and households can be sorted into wealth quintiles accordingly. See Filmer and Pritchett (1998). 8 This graph is based on the 1998 APIS survey, and households can therefore be sorted into income quintiles. 9 Note that one cannot simply regress enrollment rates or measures of educational attainment on the presence (or absence) of a school if program placement is responsive to local conditions, and if these local conditions are imperfectly measured. For example, if schools are placed in areas where there is a high demand for education, simple cross-sectional regressions may over-estimate the impact of educational infrastructure because these areas would have had better education outcomes even without a school. Conversely, if schools are placed in poorer areas with poor education outcomes, simple cross- sectional regressions may under-estimate the impact of educational infrastructure because education outcomes in these areas would have been even worse in the absence of a school. A discussion of these issues can be found in Pitt, Rosenzweig, and Gibbons (1993), and Paxson and Schady (1999). 49 PHILIPPINES POVERTYAS5ESSMENT 10 The low quality of education in the Philippines is an oft-voiced concern. A recent report states that "education quality in the Philippines is not nearly as good on average as it could be, and there is evidence that average quality has fallen in recent years" (Asian Development Bank and World Bank, 1999, p.24). In the 1996 Third International Mathematics and Science Study, which measured the mathematics and science achievement of basic education students in 41 countries, the Philippines ranked 39' in mnathematics, and 40' in science. This comparison may not be altogether fair because the Philippines was one of only four lower income and lower middle income countries in the sample. But even in domestic tests, students perform well below the expectations of the examination section of the Department of Education, Culture and Sports (DECS) (Asian Development Bank and World Bank, 1999, p.25). 11 Disaggregated data on the private costs of education per household are available in the 1997 Family Income and Expenditure Survey (FIES), but not in the 1998 APIS, while data on household composition, including the number of children attending different education levels are available in the APIS, but not in the FIES. Fortunately, 60 percent of the households in the APIS were also surveyed in the FIES, so a panel can be constructed. This feature of the data is used to estimate the private unit cost of primnary, secondary, and tertiary education by running an auxiliary regression of the total amount spent by a household on education in 1997 on the number of children attending each education level in 1998: Exh = (Ph + YSh + bTh + Eh where Ex is the total amount spent by household h on education in 1997, and P, S, and T are the total number of children who are attending primary, secondary, and tertiary education in 1998, and the constant in the regression has been suppressed. The parameters 3, -y, and a will provide a reasonable approximation to the private unit costs, by level, if household composition, the fraction of children who are attending school at every level, and actual (rather than estimated) unit costs remained constant between 1997 and 1998. There is no reason to suppose that household composition or unit costs changed in the Philippines between 1997 and 1998. Attendance rates may have dropped, however, because of the effect of the East Asian crisis on incomes in 1998, even though the Philippines was much less affected by the crisis than other countries in the region. Administrative data show no significant impact of the crisis on enrollment rates in primary or secondary school; but attendance rates may still have fallen if households were more likely to keep their children at home or doing odd jobs. A weaker assumption for the calculation of unit costs is that attendance rates in primary, secondary, and tertiary school in 1998 are a constantfraction of attendance rates in 1997, such that Ep98 = OEP97, Es98 = OES97 and E,98 = \Et97, where the subscripts stand for education levels, superscripts for years, and the parameter ; is less than one. Under this assumption, our estimation procedure will over-estimate unit costs by a factor 1/*, but will produce accurate estimates of the ratio of unit costs across levels. 12 Note that these are the unit costs faced by households, rather than the total unit costs, which would (obviously) include the public costs as well (see, for example, Asian Development Bank and World Bank, 1999, p.117). 13 The analysis in this section is based on the 1998 APIS survey, and is a summnary of a more detailed discussion found in Schady (2001). 50 ENONOTES mmN 1 4 Specifically, we estimate the basic Mincerian wage regression logWi = oc + 3Si + XXi + 6X2, + Ei, where S is the years of completed schooling, X is the number of years an individual has worked since completing schooling, E is an error term, and the subscript i refers to individuals in the labor market who earn non-zero wages. This is the standard approach to estimating the private rates of return to education (see, for example, World Bank, 2001). 15 The APIS gives seven categories for class of worker: (i) worked for private household; (ii) worked for private establishment; (iii) worked for government / government corporation; (iv) self employed without any employee; (v) employer in own family operated farm or business; (vi) worked with pay on own family-operated farm or business; and, (vii) worked without pay on own family-operated farm or business. The analysis is restricted to workers whose main job is in categories (i), (ii), (iii), and (vi) above, and to the wages earned from this main job only. 16 An oft-used solution to this selection problem, first proposed by Heckman (1976), is to specify the nature of the selection process itself, and use this to correct the estimates in the wage equation. Unfortunately, these "corrected" estimates can be as (or more) inaccurate as the simple OLS estimates if a number of stringent econometric assumptions are not met. Specifically, the Heckian correction involves finding a variable which affects the probability of participation in the labor market but does not have an independent effect on wages. One possible candidate for such a variable is the amount of non-wage income available to members of a household, including public transfers, pensions, interest and dividends. Unfortunately, in the APIS sample, these variables are not significant predictors of participation in the labor market conditional on age and education. Moreover, even when such an identifying variable can be found, the corrected Heckman estimates will only be unbiased if there is homoskedasticity and joint normality in the residuals from the switching and wage regressions, which may not be the case. For this reason, we do not attempt to implement the Heckmnan correction. Rather, we present a series of estimnates in which the sample is limited to household heads only. Household heads may be less able to "choose" not to work for wage income if they have to provide for other household members, so that these estimates of the rate of return to education may be more accurate than OLS estimates for the full sample. 17 A general discussion of these issues can be found in World Bank (2001). An application of such "within" estimators to the evaluation of child health and family planning programs in the Philippines can be found in Rosenzweig and Wolpin (1986). Behrman and Deolalikar (1995) use such within estimators to estimate rates of return to education for men and women in Indonesia. Another option to purge the OLS estimates of the biases induced by both unobserved ability and measurement error is to use instrumental variables (IV) estimators. IV estimates are based on a variable, the instrument, which is correlated with schooling but uncorrelated with the error term in the wage regression. In a recent paper which uses data on the Bicol region in the Philippines, Maluccio (1998) instruments years of completed education with measures of school proximity and household wealth at the time schooling took place, and concludes that the estimates of the rates of return to schooling derived from simple OLS regressions are not biased up. Rather, they may be biased down by as much as 80 percent because measurement error biases all coefficients, including the estimated coefficient on education, towards zero. 51 PHILIPPINES POVERTYASSESSMENT 18 The confusion between "actual" and "potential" experience may also introduce biases in the estimation of the effects of experience on earnings -- for example, if children start school late, or if repetition rates are anything other than zero. "Potential experience" will then tend to overstate actual experience, and the estirates from the regression will therefore under-estimate the increases in wages which are associated with increases in actual experience. In general, we do not think that this is a big source of concern as we are prinarily interested in estirating the rates of return to education, not experience, and these rates of return to education should be unaffected by the confusion of actual and measured experience. 19 To convert the coefficients on years of schooling to percentages, calculate 100*(exp(pl)-l). 20 This approach was developed by Oaxaca (1971), and has since been applied to many different contexts and countries (for example, Malkiel and Malkiel,1973; Altonji and Blank, 1999). 21 There is some debate in the literature as to whether one should control for job categories or occupation in these regressions (see for example, Altonji and Blank, 1999). Introducing these controls will lead to an under-estimate of discrimination if the choice of job is itself dictated, in part, by societal attitudes about gender. Failing to control for job category or occupation, on the other hand, will lead to an over-estimate of discrimination if differences in the choice of jobs by men and women are driven, in part, by differences in "preferences" or "abilities." Ideally, one would present both sets of estimates -- with and without controls. However, detailed breakdowns of occupation categories for individuals sampled in the APIS was not available for this report. Estimnating the extent of discrimnination within job categories in the Philippines should be an important priority for the future. 22 These figures are, to some extent, a construct: Given that experience is defined as X-- A - S - 6, workers with zero years of education and zero years of experience would have to be seven years old, and are therefore not even within the sample, which is limited to those aged 25 and older. Given that experience is rewarded in much the same way for men and for women, however, one way to understand the differences between the intercepts in the regressions is as a measure of the premium paid to men with no education over those paid to women with no education at any level of experience. 23 This is equivalent to estimating a very flexible regression model in which the log of the individual wage rate is regressed on a set of 15 schooling dummy variables, 10 experience dummy variables, and 150 interactions between experience and schooling: logW1., = n/ Si, + 0, Xu -+- k., (Sie * X,) + P1i,, where the subscripts i, s, and x stand for individuals, years of schooling, and experience cohort. This model therefore estimates 160 coefficients, rather than the three in the model which is the basis for Tables 3.2 and 3.3 (fifteen dummy variables have to be dropped to avoid perfect co-linearity). We group experience into five-year brackets to avoid the very small sample sizes which would have resulted if we had created dummy variable for every year of experience. Similarly, to avoid having some cells with very small sample sizes, we do not include individuals who report having postgraduate education, or post-secondary education other than college. 52 ENDNOTES 24 The following model, based on a linear spline function, is used to estimnate rates of return to schooling by education level: logWi = a + XXi + 5X2i + PSi + O[(Si-6)*D61] + X[(Sj-lO)*DlOJ] + vi, where S is the years of completed schooling, X is years of "potential experience", [(Si-6)*D6,] is an interaction term between a dunmmy variable for those who have completed at least six years of schooling and S-6, [(Si-10)*DlO] is an interaction term between a dummy variable for those who have completed at least ten years of schooling and S-10, and vi is the error term in the regression. The rate of return to primary education is then given by the coefficient P; the rate of return to secondary education is given by the sum of the coefficients ( and 0; and the rate of return to tertiary education is given by the sum of the coefficients 3, 0, and X. To test for so-called "sheepskin effects" -- particularly high rates of return for the comnpletion of the last year of primary, secondary, or tertiary education -- dunmmy variables for those who have achieved at least 6, 10, and 15 years of education, respectively, are also included in the model: logWi = a + XXi + SX2 i + (S, + nD6i + 8[(Si-6*)D6i] + yDlO, + X[(S1-l0)*DlO] + MDl51 + vs. In this discontinuous spline function, the coefficients 1, Ey, and ¢ give the increases in log wages which are associated with the last year of primary, secondary, and tertiary school, respectively. These effects are often interpreted as the rewards to a credential or degree above and beyond the inherent value of the education. See Hungerford and Solon (1987); Park (1994); and World Bank (2001). 25 Non-parametric regressions (specifically, Fan regressions) of log wages on experience for (separately), primnary school graduates, secondary school graduates, and college graduates are the basis for these calculations. The expected earnings stream is given by the area below these curves, appropriately discounted. These calculations assume that: (i) the earnings profiles of current labor market participants can be applied to future labor market entrants. This may not hold if there have been systematic changes in the structure of the economy with corresponding changes in the relative premia paid to workers with different levels of education, or changes in the relative qualty of some forms of education over others; and (ii) that there is no correlation between schooling and other characteristics which can affect earnings, such as unmeasured ability. For example, if people with more schooling are also more able, the earnings stream of these more able people may not provide an accurate representation of the earnings streams which less able people would have had even if they had received more schooling. If there is self-selection on schooling of this sort, the calculations presented will likely over-estimate the expected earnings of college graduates compared to secondary school graduates, and of secondary school graduates compared to primary school graduates. 26 For this incidence analysis, the unit costs of government hospitals, and barangay health facilities and rural health centers are based on the 1997 public spending estimates and the number of visits to different facilities estimated from the 1998 APIS survey. Unit cost of government hospitals is based on spending on personal health care which is mainly hospital-based, while the unit costs for barangay health facilities and rural health centers are based on public health care spending. Using these data, a ratio of 3.3: 1, of unit costs of personal to public health care is derived. This ratio is used in deriving the incidence of overall health spending. This ratio is likely to be an underestimate because some hospital-based spending is probably also included under public health care. To that extent, results on the overall incidence may overstate the progressivity of public health spending. 27 Department of Health, "Health Sector Reform Agenda," p. 1. 53 TARGETED POLICJES AND 4.1 Social sector spending tends to be broadly targeted in that universal coverage is often a guiding _ objective in the social sectors, as frequentlye lated in phrases lik~e "education and health Such spending can still be targeted to the is concentrated on services that matter most poor (for example, basic education anLd bas"ic Chapter 3 looked into the incidence of sp, such broadly targeted public spending. 4.2 Thais chapter directs atteninta~ cies that are more expressly redsrutv There is, of course, a large numb ventions in the Philippines, inclui4 small programs aimed at specific t comprehensive review of these progis beyond the scope of this study Instead, this chapter tively takes a fresh look at a few key policies in the areas of rice trade and distribution, land refo housing, and inter-governmnental allocations. While these policies have a redistributive core, to view them as just that would however be taking a very narrow view of these interventions. For instance, the limited pace of structural transformation of the Philippine economy towards the non-agricultural sectors is partly on account of lirmited economic growth ove the last two decades, but is possibly also delayed b policies supporting high levels of protectionobf agricultural sector. Similarly, efficiencygains andhigh growth through greater investment in agriculture 4: TARGETED POLICIESAND THEPOOR Table 4.1: Nominal Protection Rates of Major Agricultural Commodities 1980-84 1985-89 1990-94 1995 1996 1997 1998 Rice -13 16 19 63 91 82 34 Corn 26 67 76 104 54 96 72 Sugar 42 154 81 91 93 66 99 Coconut Oil -4 7 18 10 5 0 0 Pork -9 43 31 44 n.a. n.a. n.a. Chicken 46 39 74 84 n.a. n.a. n.a. Source: David (1999). important among the expected outcornes of land re- by rice farmers, and it sells rice to consumers at a form programs. More equitable inter-governmental subsidized price. These two elements of the price transfers can similarlyalso serve as an important means policy have different irnplications for the poor. of fostering growth in poor and lagging areas. While they maywork through multiple channels, this chap- Agricultural protection ter highlights the keydistributive irnplications of these policies on the welfare of the poor. 4.4 Nominal protection rates (NPR) for rice, and agricultural comrnodities, in general, have been RICE POLICY high and increasing over the 1990s (see Table 4.1).' For instance, the average NPR for rice over the 1990s 4.3 Food staples have often been deemed "po- up to the 1997 crisis was about 41percent, and for litical conrmodities" where the interests of produc- the period 1995-97, it was nearly 80 percent. The ers and consumers are pitted against each other. Pro- devaluation of the Peso following the financial crisis ducers want higher prices, while consumers want brought this down to 34 percent in 1998. low prices. In the Philippines, rice has been a key political commodity There are two main dimen- 4.5 What are the distributional implications of sions to the rice price policy in the Philippines. The this protection? Standard economic theory tells us first has to do with prevailing high levels of protec- that such high levels of protection favor the pro- tion, mainly by way of quantitative restrictions; the ducers at the expense of the consumers. On the con- National Food Authority (NFA) has statutory mo- sumption side, it is clear that rice is an important nopoly on the import of rice. The second has to do component of poor people's consumption; for in- with NFA operations in the dornestic rice market, stance, it accounts for about 22 percent of total con- where it procures rice domestically with a view to sumption of the poorest two deciles (Figure 4.1). setting a protective floor for producer prices faced Rice budget share declines slowly for higher deciles, 25- o 20:~~~~~~~~~~~~~~~~~~~~0 0~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~~~~0 E~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~E 01 p10-~~~~~~~~~~~~~~~~~~0 0 C 8 ? ----2---- -6 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Deciles (ranked by per capita consumption) Deciles (ranked by per capita consumption) Note: The decaies are deciles of population rankedbypercapita consumption, adjustedforspatal costoflivingdifferentials. The 30,40and50% referto assumptions on the mark-ups on farm-gateprices used toadjust the value of rice produced (valued atfarmgate prces) to make it comparable with the value of rice consumed(valuedatlocalretailprices). Source: Balisacan (2000), and calculations based on 1997FIES data. 55 PHILIPPINES POVERTYASSESSMENT and is about 14 percent even for the sixth decile. eration, NFA procurement is unlikely to have any This has two implications: while it is clear that the discernible effect on the producer price for rice. The poor would derive prqortiantely larger consump- main channel of protection to the producers is by tion benefits from a lower price of rice, a large share way of restrictions on the import of rice. of total consumption benefits would still accrue to the non-poor on account of their large share in total 4.8 On the consumption side, the storyis more rice consumption. complex. First, the NFA rice distribution operations are more substantial, in recent years, accounting for 4.6 The distributive impact of protection criti- about 10 percent of total consumption on average cally depends on who the mt consumers of rice are, (Table 4.2). However, it is not clear if this level of and where they are located in the overall income operation has had any effect on the market price of distribution. Estimates based on the 1997 FIES data rice. There are indications that the cheaper rice sold suggest that the poor in the Philippines are net con- in NFA stores is actually inferior quality rice, even sumers of rice. Figure 4.1 shows how the net con- more inferior than that imported by the NFA sumption of rice as a proportion of total per capita (Roumasset, 1999). It is likely then that the NFA consumption varies across deciles (ranked by per operation has little influence on the price of this in- capita consumption), under three altemative assump- ferior quality rice since the traders are likely to dilute tions on the differential between retail and farm gate the quality of rice till its effective market price ap- prices.2 The figure shows that under each of the proximates the announced NFA rice price. In this three assumptions, it is reasonable to conclude that case, the NFA subsidy is not really a subsidy to the the poor (bottom one-fourth of the population) are consumers either, because theyare either consuming net consumers of rice, while the net producers are inferior quality rice or, in fact, paying the cost of spread over the mniddle and higher deciles. Overall, propping up prices for rice producers. therefore, a reduction of implicit tariffs on rice is likelyto have progressive welfare effects, and in par- 4.9 NFA tends to import whenever there is ticular, is likely to be beneficial to the poor.3 production shortfall, and in deficit years, these im- ports can be a significant fraction of the total sup- NFA domestic market operations ply, as for instance, recently during 1998 when rice production was hit by the El Nifmo drought. During 4.7 The effects of a protective rice regime could 1998, NFA rice distribution amounted to 22 per- be potentiallyrmitigated bythe other elemnent of NFA cent of total consumption. While this level of distni- operations, viz., its domnestic mnarket interventions, bution is clearly imnportant and has potential price which involve both domnestic procurement of rice stabilization benefits for the consumers, it is unclear at support prices aining to set a floor price for the why the same function would not have been per- rice farmers, as well as distribution of rice to con- formed by private traders if quantitative restrictions surners at below market prices. On the procuremnent on rice imports were relaxed. side, the size of NFA operations is miniscule, ac- counting for less than one percent of total produc- 4.10 The potential benefits of NFA rice distri- tion during 1995-98 (Table 4.2). At this level of op- bution are also linited by its largely untargeted na- ture. The distribution of NFA rice across regions Table 4.2: NFA Rice Procurement and Distribution has little relation to the regional povertylevels (Table 1995 1996 1997 1998 Average 4.3). For instance, during 1995-98, 36 percent of Quantity procured ('000 ml) 8.19 124.31 100.47 97.41 82.60 NFA rice went to Southern Luzon, which only ac- As % of total production 0.08 1.10 0.95 1.14 0.81 Quantity distributed ('000 mt) 256.70 731.40 622.80 1628.20 809.80 counted for 12.5 percent of the poor in the Philip- As % of total consumption 3.60 9.30 7.90 22.20 10.70 pines. Moreover, the Southern Luzon includes the Source: BureauofAgnculturalStatistics. National Capital Region, which appears to account 56 4: TARGETED POLICIESAND THE POOR 11 liilii lliili ii fm" _ R - Table 4.3: NFA Rice Distribution and Poverty groups in the Philippines. A number of such pro- Across Region grams were implemented nationallyin the 1970s and % Share in NFA % Share in the 1980s,including the SlumImprovement andResettle- rice distribution number of poor (1995.98) (1997) ment (SIR) program, and the CommunityMortgage llocos 3.6 4.6 Program (CMP). Since the 1990s, manyof the pub- Cagayan 2.0 4.7 lic resources for housing have been channeledthrough Central Luzon 12.4 5.4 local governments. In 1998, for example, local gov- Southern Luzon 36.0 12.5 (including NCR) emments spent PhP3.8 billion on housing prograrms, Bicol 9.2 12.8 and the central government spent a further PhP2.1 Central Visayas 52.6 101 billion. Based on the 1998 APIS data, the housing Eastern Visayas 6.1 10.3 program benefited 3.2 million persons, of whom Western Mindanao 3.8 5.6 2 . . Northern Mindanao 5.0 7.0 2.1 mion were urban and 1.1 mllion were rural. Southern Mindanao 5.4 7.8 These resources have apparently been directed Central Mindanao 3.4 4.3 through a multitude of progras- including the CAR 2.3 1.6 porm h ARMM 2.5 5.8 A bot-Ka)a Pahahry Fund, the Group Land Acquisi- Source:.ureauotAgricuIturaJStabistcs,and1997FIESdata. tion Program, the Local Housing Fund, the LGU Resettlement Assistance Program, the LGUPahVay for the bulk of NFA rice in that region (almost 70% Fund of Pagibig, and others. Data from the APIS during 1991-93). There could be an element of self- are not disaggregated enough to allow for a com- targeting if the subsidized rice were of an inferior parison of these (and other) housing programs. The quality, as typically consumed by the poor. question asked in the survey is: "Did you avail your house and its lot through the assistance of the gov- LAND REFORM AND HOUSING emment housing program or financing program?" Analysis of responses suggests however, that in the 4.11 The government's land reform and hous- aggregate, these housing programs have not been ing programs are among the most important public pro-poor at all. A serious evaluation to assess which, efforts to remedy constraints faced by the poor in if any, publicly-funded housing programs actually acquiring assets and irproving their living condi- benefit the poor in the Philippines should be an im- tions; a third set of policies and programs around portant priorityfor the future. rice production and consumption were discussed earlier. Surveyresults showthat the land reformpro- 4.13 Assessing the effectiveness of programs in gram had reached 1.2 million, mainly rural benefi- reducing poverty or improving well-being requires ciaries (77%), while 3.14 million, mainlyurban people going beyond examining targeting efficiency It re- (65%) benefited from the housing program. The quires evaluating the net impact of a program on distributional profile of these two programs de- beneficiaries over time as against the counterfactual signed to target the poor is very different: the land without the intervention. It also requires weighing reform program is progressive, overall, whereas the the program benefits against the costs of the pro- housing program fails starkly in reaching the poor, gram which should include not just the financial but especiallythe urban poor (Figure 4.2). Since the analy- also the economic costs of program implementa- sis here does not impute value to the housing assis- tion. For example, a significant cost of a wage sub- tance received, it is quite likely that the distribution sidy program which is intended to increase overall of benefits is even more skewed than simple access employment levels is the deadweight loss (i.e., com- to the program. panies would have hired the employees even in the absence of the wage subsidy) and the displacement 4.12 There is a long history of efforts to pro- cost (companies do not increase the overall number vide housing and slum upgrading to low income of employees hired but choose those who come 57 PHILIPPINES POVERTYASSESSMENT Housing program: total Land reform program: total O 17 - r5- 4O .t .2 3 4 5 . . 4 - . 3 . . 7 . za 3 - 4 c3 0 .1 .2 .3 4 .5 .6 .7 .8 9 1 0 .1 .2 3 4 .56 7.8.9 1 Cumr. proportion of population (ranked by percapia cons.) Curn. proportion of population (ranked by per capita cons) Housing program: rural areas Land reform program: rural areas 8 .7- ~~~~~~~~~~~~~~~~~7- .6- 6~~~~~~~~~~~~~~~ 1 2 . 4 . 6 . 8 . with the progrCa m. pOoeralnl pop efitio (ranke by p ner-cpt cn. Cebat , poolticyssue, pplthorned b erginni cngs.) hchdt consideation. Horusngtl progam urba arlea olo no cladscald reforme program:nt u orband areas r,thr t5 .8 .2 .6V,2 0- 5 0 s1 2 .3 .4 5 .6 .7 .8 9 i i a t2 t i l4 an 6 g body Cum. proporion of popuat ion (ranked by pet capita cons.) Cornn prc rtiof of popilason (ranked by per capia cons.) with asubsidy Hover ngthose rwho dobanot aressoitd.4 Land reform ingrm thea Phlipreas i ol wthe thren program.4O eeiswl eudr debaedt plcyrissue,unthe beeginningis pofwichngt dthe some detail a ated costs~4 4s-oee,ls aoal n ugss o mprclwr htsget hr a ego Nte:he vtcurr ieinthpro graphrpeenshpve.haonwihwssiat Rtl8ercenti cro8sighASdts-coun ryl ,hoevelw rkhtt isnpoitingarblto the 258prethacutetmtdfr19 srgFE aa 4: TARGETED POLICIESAND THE POOR existence of a negative relationship between the in- quantitative redistribution targets, effectiveness of equality in the distribution of assets and subsequent targeting the poor, attainment of higher productivity, growth.6 At the micro level, low initial wealth and and longer term gains through increased investment. the ensuing borrowing constraints have been shown Available inforination suggests: (i) that the Philippines to limit households' ability to make productive in- land reform program has fallen short of attaining its vestments, acquire human capital, and start-up en- quantitative targets; (ii) that it has helped increase se- terprises.7 There is also growing evidence for the curityof tenure among those who have access to land presence of "povertytraps," i.e., situations in which but has not succeeded in reducing landlessness; (iii) poverty is perpetuated not because of lack of abil- that it is reasonably well-targeted towards the poorer ity but lack of endowments! Finally globalization landed households; (iv) that increased securityof ten- may have increased the premium on asset owner- ure has a significant positive impact on household in- ship as this is seen as a critical determinant of house- come; and, (v) that land reform beneficiaries have holds' ability to utilize the economic opportunities invested, more than non-beneficiaries,in physical capi- opened up through macro liberalization. As asset talandthe education of theirchildren. These are pow- transfersgo,landreformhastraditionallybeenviewed erful results in favor of continued implementation as an ideal redistributive policy in view of the im- of land reformnEvidence on some of the associated movability and indestructibility of land, land reform costs is, however, less favorable and suggests that there may actuallyprovide a basis for a non-distortionary is a need to rethink certain paramneters of the current lump-sum transfer that could have a major impact program. on the welfare of the poor. 4.16 Modern land reform programs in the Phil- 4.15 What has been the Philippines expenence ippines have two distinct phases, relating to the Presi- with land reform? This question can be answered at dential Decree 27 of 1972 and the Comprehensive several different levels, including success at reaching Agrarian Reform Law (CARL-) of 1988. (Box 4.1) Box 4.1: Land Reform in the Philippines The foundation for "modern" land reform in the Philippines was laid by the 1972 Presidential Decree (PD) 27 which provided the basis for "Operation Land Transfer" (OLT) and "Operation Leasehold", respectively, and the 1988 Republic Act (RA) 6657, also known as the Comprehensive Agrarian Reform Law (CARL). PD 27, the applicability of which is limited to rice and corn lands, consists of two key provisions. First, it calls for a land ownership ceiling of 7 hectares and the mandatory sale of all land that is owned in excess of this limit to tenants. Second, it outlaws share tenancy and mandates the conversion of tenants on landholdings below 7 hectares to leaseholders with a rent ceiling of 25% of crop revenue after appropriate deductions for inputs. The 1988 CARL expands the scope of land reform in two respects. First, land reform was no longer limited to rice and com land but expected to cover the whole country, with implementation expected to occur in three main phases over the subsequent 10 years and to be funded by a special "Agrarian Reform Fund" that would be constituted from confiscation of ill-gotten wealth in the Marcos era. Second, the range of beneficiaries was to be increased to include not only cultivators but, at least in principle, landless households. Finally, full land ownership was to be granted automatically to all the beneficiaries who had benefited from the earlier land reform under PD 27. Implementation of these measures was expected to proceed in three distinct steps over a total period of 10 years. A first phase, expected to last 4 years, aimed to complete coverage of tenanted rce and corn lands, transfer ownership to PD 27 beneficiaries, and include lands voluntarily offered for sale, alienable public lands, and holdings with a size above 50 hectares. A second phase of three years' duration, was to cover agricultural holdings between 24 and 50 hectares. Finally, land reform was to be completed during the last 3-4 years through redistribution of lands between 5 and 24 has. Agri- business operations and plantations were exempted from land reform for a period of 10 years to allow recovery of fixed investments and supposed to be treated once the remainder of CARP had been implemented. 59 PHILIPPINES POVERTYASSESSMENT Table 4.4: Land Distribution Status by Land Type and Mode of Coverage, 1972-1997 Land Type Target Accomplishment % Accomplishment Balance Department of Agrarian Reform Tenanted Rice/Corn 579,520 500,643 86.39 78,877 Voluntary Offer-to-Sell 396,684 265,744 66.99 130,940 Voluntary Land Transfer 284,742 276,307 97.04 8,435 Govt. Financing Institution-owned 229,796 148,900 64.80 80,896 Compulsory Acquisition: >50 has. 456,588 74,687 16.36 381,901 Compulsory Acquisition: >24-50 has. 312,355 6,251 2.00 306,104 Compulsory Acquisition: >5-24 has. 736,420 20,483 2.78 715,937 Kilusang Kabuhayan at Kaunlaran lands 657,843 606,347 92.17 51,496 Settlements 662,727 662,727 100.00 0 Total DAR 4,316,675 2,562,089 59.35 1,754,586 Department of Environment and Natural Resources Public Alienable & Disposable Lands 2,502,000 927,734 37.08 1,574,266 Integrated Social Forestry areas 1,269,411 832,651 65.59 436,760 Total DENR 3,771,411 1,760,385 46.68 2,011,026 Total CARP 8,088,086 4,322,474 53.44 3,765,612 Source: Deininger, Lara Jr., Maertens and Quisumbing (2000). After 25 years of land reform and ten years after average household expenditure in five villages is above the promulgation of CARL, i.e., at a point where the poverty line (US$266) but close to or significantly land reform should already be a thing of the past, below this line in the marginal villages - Rizal with achievements on quantitative targets remain well US$266 and Signe with US$212 (Table A4.1). below expectations (Table 4.4). Even if the histori- cal pace of imnplementation were to be maintained, 4.18 Comparison of changes in land tenure be- completing the government's targets would take tween 1985 and 1998 suggests that, in the aggre- more than 17 years. Financial constraints may lead gate, there is a significant increase in the number of to even more protracted implementation because owners from 25 to 37 percent, and a reduction of unlike much of the land redistributed earlier, almost the proportion of share tenants from 12 to 8 per- all of the remaining land is privately-owned and will cent (Table A4.2).11 But there are large differences require compensation. Also, large parts of land that in tenure structure and the level of land reform irnple- could be redistributed remain outside the scope of mentation across villages and legislative intervention CARL, either, because they were artificially subdi- does not appear to have been successful in eliminat- vided or because they were held under special ar- ing share tenancy in marginal environments. Land rangements to which the land reform legislation is reform was implemented most quickly and effec- not applicable.9 This suggests a need to review ex- tively in favorable environments - share tenancy perience, to date, and to evaluate alternative options. was already virtually non-existent in Gabaldon and Maragol in 1985 and was completely eliminated 4.17 Analysis of the impact of land reform pre- thereafter. By contrast, land reform was much less sented in this report relies on data from five villages in successful in eliminating share tenancy in marginal Central Luzon and Iloilio - typical of the rice and areas where between one-sixth and one-third of land- com growing areas that were the focus of the 1972 owners continued to remain share tenants. More land refornLl' Between 300 and 400 households in importantly, land reform implementation following these studyvillages were surveyed three times, in 1985, the 1988 Comprehensive Agrarian Reform Program 1989, and 1998. With an equivalent of US$311 in 1998, (CARP) appears to have consolidated the gains from 60 4: TARGETED POLICIESAND THE POOR Table 4.5: Original Tenure Status of Households Who Benefited from Land Reform, 1972-1988 Highest tenure Non- % of % of In 1972 Total Beneficiaries Beneficiaries Group Beneficiaries Owner 98 83 15 15 16 Amortizing owner 12 6 6 50 6 Leaseholder 72 47 25 35 26 Share-tenant 86 40 46 53 48 Landless 69 66 3 4 3 Total 337 242 95 Source: Deininger, Lara Jr., Maertens and Quisumbing (2000). the first phase of land reform but was much less nificantly lower incomes, holding asset ownership successful in reducing landlessness. and education levels constant. Incomes are 60 per- cent lower and statistically significant for share ten- 4.19 Who benefited from land reform? Evi- ants and the landless in 1998, compared to land- dence on the original status of land reform benefi- owners (Table 4.6). In addition, if greater security cianres indicates that in line with program design, the of tenure is associated with higher levels of invest- 1972 land reform has not benefited the landless but ment and therefore asset accumulation, land reform rather share tenants and leaseholders (Table 4.5). Al- would have an even greater impact on welfare in most 50 percent of beneficiaries from this land re- the long-run because returns to assets appear to have form effort came from the ranks of share tenants, increased over time, from 10.6 to 18 percent be- one quarter from leaseholders, and another quarter tween 1985 and 1998 (Table 4.6). from the class of owners. In the absence of infor- mation on income in 1972, it is difficult to assess 4.21 Indeed, land redistribution appears to have how well the program targeted poor households. had a strong impact on investment in physical as But based on available information, analysis suggests well as in human capital and, in the longer term on that the earlyphase of land reform benefited house- growth of income, productivity, and investment. holds with low levels of education and low land There are two channels through which a one-off assets - not an outstanding targeting performance asset transfer could induce higher levels of accumu- since those with no land were largely excluded but a lation of physical and humnan capital. First, the in- reasonably good one, given program objectives. creased security of property rights associated with 4.20 What were the benefits of land reform? Between 1985 and 1998, profits before payment of Table 4.6: Determinants of Income, 1985 and 1998 rent and family labor increased by about 80 percent 1985 1998 due largely to diversification and real price move- Coefficient t-slatistics Coefficient t-statistics ments. A larger share of these profits actually ben- Log of assets 0.11 2.85 0.18 3.55 efited producers - profits after payment of rent Education 0.04 2.24 -0.02 -0.90 increased faster than profits before,rent payment - CLT holder -0.16 -0.71 -0.23 -1.04 increased faster than profits before rent payment - Leaseholder -0.18 -1.12 -0.24 -1.46 suggesting that rent payments declined, thanks to land Share tenant -0.63 -2.75 -0.60 -2.35 reforrrL In a,chtion, because tenure status appean Landless -0.06 -0.27 -0.65 -3.68 reform. In addition, because tenure status appears Mago 0.3.6-.4 -25 ... .. . rr ~~~~~~~~~~~~~Maragol 0.13 0.69 -0.46 -2.52 to have a significant irnpact on incomes in the Phil- Pandon -0.06 -0.29 -0.76 -3.45 ippines, measures aimed at increasing ownership and Rizal 0.14 0.69 -0.71 -3.50 Signe -0.79 -2.95 -0.92 -3.20 improvng security of tenure would be expected to Intercept 3.82 12.01 4.75 10.90 benefit the poor: compared to land ownership, less Adjusted RI 0.17 0.17 secured forms of tenancy are associated with sig- Source: Deininger, Lara Jr., Maertens and Quisumbing (2000). 61 PHILIPPINES POVERTYASSESSMENT the transition from tenancyto ownership is likelyto faster rate in 1985-98 than that of non-beneficiaries, increase the payoff to investments, especially for by some US$73-92 or more than half of the origi- those relying on family labor. Second, by increasing nal income level. A long-term impact on rice yields beneficiaries' level of asset ownership and thereby is also notable: households who benefited from the eliminating credit constraints, land reform may ex- 1972 land reform programns increased their rice yields pand their ability to make investments that require by between 565 and 637 kilograms more than non- credit access. beneficiaries. These results support the hypothesis that a transfer of productive assets to the poor acted as 4.22 The analysis relies on assessing the net im- a catalyst to facilitate a permanent change in house- pact of land reform by comparing the behavior and holds' pattern of asset accumulation as well as in achievements of households who benefited from their welfare and productivity land reform with those of a control group who did not. The results suggest that land reform made a 4.24 The benefits of land redistribution under significant and quantitatively important contribution CARP could have been enhanced further if the re- to non-land investment, the magnitude of which distribution effort was coordinated with the provi- varied between US$1,486 and US$860 (Table A4.3). sion of agricultural extension services. The evidence Land reform participation is, in effect, equivalent to (based on the 1998 APIS) suggests that these two an increase in the household head's initial levels of programs have operated with virtually no overlap education of between 5.8 and 11.5 years, implying a (Figure 4.3). Only 16 percent of the CARP benefi- doubling or more of initial levels of education in ciaries also benefited from agricultural extension ser- the sample (6.2 years). This suggests that land redis- vices, while only 9 percent of the beneficiaries of tribution which increases households' endowments extension services were also CARP beneficiaries. could make a significant contribution to poverty re- Better provision of agricultural extension to land duction in countries where households' levels of reform beneficiaries should be viewed as part of education are high but were not able to make pro- the general revamp of the entire agricultural exten- ductive use of their abilities due to prevailing socio- sion program that has been lagging in performance economic conditions and lack of access to produc- since its decentralization under the Local Govern- tive assets. ment Code (LGq. 4.23 In addition to contributing to higher accu- 4.25 While the land reform program generated mulation of physical capital, land reform appears significant benefits for those who participated, a to have also enabled households to increase their positive social return requires that the costs, both level of human capital accumulation. The educa- direct and indirect, be lower than the benefits. The tional advance of children "affected by land reform" was between 0.60 and 0.83 years higher than that of non-beneficiaries (Table A4.3). This is over and E E I E E E above a very strong overall convergence effect 0.26% whereby overall educational expansion is estimated _ 5 to have helped children from lower educated famif- lies to make up for between 70 and 76 percent of 1.68% - Both CARP and Extension their shortfall.11 Thus, over and above the impact (1,215_00 Beneficianes - Onjy CARP Beneiicianes on accumulation of physical capital, the one-time Only Extension Benelciaries asset transfer implicit in land reform appears to have induced beneficiary households to significantlyin- 2.% crease investment in their children's education. There 2,146,tO6) is also evidence that beneficiaries' income grew at a Source: Based on 1998 APIS data. 62 4: TARGETEDPOtICIESAND rHEPOOR 3] Non-parametric regression of cultivated on owned land, 1985 31 Non-parametnc regression of cultivated on owned land, 1998 2.5 2.5 ] - 15 | O 15-4 50 0 0 . --- ------ ----- L ------ - -- - --- --------- ------- 1- -T-- ---------- .5 1 1.5 2 2.5 3 .5 1 1.5 2 2.5 3 Owned land (ha) Owned land (ha) Source: Deininger, Lara Jr., Maertens and Quisumbing (2000). survey data used in this analysis suggest that there less likelyto rent out while small ones were less likely was reduced land rental narket activitybetween 1985 to rent in."3 At the same timne, land access for the and 1998 and that as a result, the landless experi- landless has decreased significantly and is estimated enced a decline in their access to land. Figure 4.4 to be less than half of what it was in 1985. Although shows that there is a strong relationship between the data at hand do not allow us to distinguish ownership and cultivated land in 1998 where only a whether this was due to CARP or to somne other weak one existed in 1985, suggesting reduced land factors, they provide a first indication that land ac- rental market activityin 1998. In 1985, land markets cess mnight indeed have decreased during the period appear to have been functioning relatively well. Al- under consideration. most everybody seems to have converged to an "optimal" operational farm size of slightly more 4.26 The notion of a structural shift in land ac- than one hectare, and even landless households (i.e., cess related to CARP legislation is corroborated by those with an owned land endowment of zero) do transition matrices for the 1971-89 and the 1989-98 not seem to have faced significant barriers in land periods (Tables 4.7A and 4.7B). In the first phase, access, cultivating about the same amount of land land markets'were still relatively active and there was as those who owned much larger plots. Things are considerable upward movement not only among different in 1998. The slope of the overall regres- leaseholders and share tenants, but also among the sion line is much steeper, implying that households' landless: 50 percent of the landless were able to move cultivation decision was more closelylinked to their up the agricultural ladder, about 20 percent making land endowment, and that larger land holders were it to leaseholder or beyond, about 20 percent mov- Table 4.7A: Transition Matrixfor Movement in Land Tenure Status Between 1971 and 1988 Tenure Status in 1971 Tenure Status Amortizing Lease Share As % of Total in 1988 Owner owner holder tenant Landless Total in 1988 Owner 90 - 4 5 4 103 30 Amortizing owner 1 12 20 10 1 44 13 Leaseholder 6 - 37 31 8 82 24 Share tenant 1 - 3 30 13 47 14 Pawning out - - 6 2 5 13 4 Pawning in - - 1 - 3 4 1 Landless 1 4 8 34 47 14 Total 99 12 75 86 68 340 As % of Total in 1971 29 3 22 25 20 Includes sub-tenants and sub-leaseholders 63 PHILIPPINES POVERTYASSESSMENT Table 4.7B: Transition Matrix for Movement in Land Tenure Status Between 1988 and 1998 Tenure status in 1988 Tenure Status Amortizing Lease- Share Pawning Pawning As % of Total in 1998 Owner owner holder tenant out in Landless Total in 1998 Owner 53 13 16 8 1 2 1 94 38 Amortizing owner - 12 12 1 - - 25 10 Leaseholder 8 6 25 8 1 1 3 52 21 Share tenant 5 - 1 13 - - - 19 8 Landless 4 6 8 4 10 1 23 56 23 Total 70 37 62 34 12 4 27 246 As % of Total in 1988 28 15 25 14 5 2 11 Source: Deininger, Lara Jr, MaeRtens and Quisumbing (2000). ing to share contracts, and the remainder engaging INTERGOVERNMENTAL TRANSFERS in pawning transactions."4 Thus, even though the 1972 Land Reform legislation did not make provi- 4.28 As noted earlier, there are important and sion for landless non-cultivators (e.g., agricultural persistent geographic differences in social and eco- wage workers) to become direct beneficiaries, this nornic outcomes in the Philippines, including the dis- group was able to imnprove its tenure status by mov- tribution of income-based poverty, malnutrition, ing up the ladder through the regular land market. health indicators such as infant mortality, enrollment and educational attainment. Public policies can af- 4.27 The transition matrix for the 1989-98 pe- fect the extent to which these geographic differences nod illustrates that in the wake of CARP legislation, are corrected. For example, social expenditures can the ability of the poor and landless to gain access to be targeted geographically, and transfers from the land through established markets has worsened. central government to local governments can be Even though landless non-cultivators were now ex- designed in such a way as to favor poorer provinces plicitly included as potential beneficiaries, their op- and municipalities. portunities have declined significantly15 It is impor- tant to note that this phenomenon was not due to a 4.29 The main block grant from the central gov- general lack of movements up and down the "agri- ernment to local governments in the Philippines is cultural ladder," in fact, large portions of leasehold- the Internal Revenue Allotment (IRA). IRA transfers ers and amortizing owners moved towards full to provinces, cities, and municipalities are based on ownership and a significant population of share ten- a formula which considers the population, area, and ants managed to improve their tenure status. There an "equal share" component, while IRA transfers to is, however, hardly any upward movement for the baranga,s only take into account population and the landless.16 These results suggest that even though land equal share component. There is, therefore, no re- reform has helped improve the lives of its direct distributive intent in the IRA formula. To see beneficiaries, its unintended consequences mayhave whether IRA transfers are redistributive in practice, been costly, in particular as they were borne by the we aggregate transfers to provinces, cities, munici- landless who are among the most vulnerable in the palities, and barangas for every province, and graph rural population. This is a disturbing finding. There the per capita transfers to provinces as a function of is clearlya need to undertake additional analysis to check the incidence and depth of poverty Figure 4.5 shows if the findings from this snall surveyin five villages is that there is essentiallyno relationship between pov- robust. erty and IRA allocations, suggesting that the main 64 4: TARGETED POLICIESAND THE POOR 0 ~~~~~~~~~~~~~~0 4000- 4000- 0)0 _0) C) s 1 o o9 ci 3000- o 0 300 °0 22 ~ ~~~ ~~~~ 0 44 00 °e goeme2000- othep o0edc regonl n cde 2000 LG scneae0it oet iec equa s 100eeilr0 geO 0° ° ° a given0 00 o los ° a l,.q. u 0 0 0, 0 I 0 20 40 60 80 g0 0 1 0 2 0 3 4.30 Inerg nvementalo Povraty, 1997 Dept imof Povesy thtaeincas3 997bvewcaels tatlokurant ofromtevenuetfra subnatmenal goven local than manytheaveragesprviniaclasses fican2thionun- eqaitis but 0000e islrevrainwti ie ls Fg a~~~~~~~~~~~~ ~r 006) Hece thr0salre ubrorv 4.3 0 40egoenmna 60nt 80e 100 0mor ine 10a 1re 20 25as 30 3n5bveaels tant source of nrevenes ofPovrty 1997tina Depthn porfa Povpovn erty 1997se1ad.Thsi ments in most developing countsies. The design not altogether surprising because the economic clas- of these transfers affects the efficiencyand equity sification of Local Govemment Units (LGUs) is of local service provision. Box 4.2 susmarizes based on total revenues at the disposal of LGUs, some considerations which should be taken into of wliich IRA is by far the largest component. As account for the design of an optimal scheme of discussed above, the allocation of IRA itself was intergovernmental grants. One important objec- not found to be correlated with poverty levels. Rel- tive of intergovevmental transfers in sanygcoun- ance on the economic classification under the LGC tries is to reduce regional fiscal disparities with- for targeting of anti-poveryprogranm is therefore out providing disincentives to local resource mo- not advisable; there information for more bilization. One reconmmendation of this report is effective target'n at the provincial level. that the Philippine Government consider incor- po rating an explicitly redistributive component A into the IRA scheme to enable aopmal schm govef- ments to provide adequate social services to their On average, provinces are poorer as the economic classification moves cntrisistoueduergonlfsaldsaite.ih fror to u argetn ofaniti-onihnagvent progams iugss thsierefore bcoinstioent. One recommendation of ths report is effectroom for improved targetinga 100 * Incidence . 4.31 At location of ceral govemnment resources c0 o Averager is sornetimes also guided by the economic classifi- moves 10>0 * niec N cation of local government units under the Local t Government Code from I to 6, with 1 considered 40 U the most well-off and 6 the least well-off. How well 20 does this classification reflect poverty incidence at s 0 1 2 3 4 5 6 7 the provincial level? Figure 4.6 shows the results. Economic class of provinces Source: Staffcalculations. 65 PHILIPPINES POVERTYASSESSMENT Box 4.2: The Design of a System of Intergovernmental Grants The design of the most appropriate formula for intergovernmental transfers is a difficult issue, and often reflects some combination of the desire to equalize fiscal capacity or to reduce disparities in the levels of public service and to encourage local governments to mobilize resources. First, the formula may be designed to reflect differences in expenditure needs. Different indicators can be used to measure expenditure needs: * Population. * Indicators of physical factors that may lead to greater costs of service provision, e.g., land area, population density, urbanization. a Measures to reflect the concentration of high cost population in the local government area, for example, the percent of families living below the poverty line, the percent of the population on pensions, the percent of school aged children. * Indicators of infrastructure needs, such as the miles of paved highways, percent of households with access to adequate water supply, etc. Second, the intent is income or fiscal capacity equalization, which means an equalization in the capacity of local govemments to finance a given level of services. The grant formula attempts to provide more money to those local governments with a lower capacity to raise taxes. Such a formula may allocate funds according to the level of average income in the local area. Since this provides no incentive for the recipient government to increase its tax effort, the formula often includes a measure of tax effort, or requires the maintenance of some level of revenue mobilization as a condition of receiving the grant. Third, the grant formula could reflect a balance between revenue raising capacity and expenditure needs. But once local governments get additional resources, there is no guarantee that these resources are used to finance services that the poor value. Different types of intergovernmental grants can be used to influence the allocation of local government resources. For example, if national governments want to ensure minimum standards of service provision, grants could be provided conditional on the funds being used for a particular purpose, such as basic health services, and with conditions on standards of service and access. Such transfers ensure that the recipient government spending on a particular category will at least be equal to the amount of the grant. There are alternative grant types, such as matching grants or cost-sharing programs, which are conditional transfers that require funds to be spent for specific purposes and the recipients match the funds to some degree. Such transfers have an income effect, as the subsidy gives the community more resources, and a price or substitution effect, since the subsidy reduces the relative price of the subsidized service. Matching transfers may change local priorities, which may be the desired outcome if the objective is to achieve national policy objectives. This type of grant has the potential to be inequitable, since richer communities can raise matching funds more easily. But this effect can be offset by adjusting matching rates to local government wealth. Source: Bahl (1999). 66 4: TARGETED POL ICIESAND THEPOOR Table A4.1: Socio-Economic Characteristics and Land Tenure Status in Sample Villages Across Three Generations Central Luzon Panay Island Total sample Gabaldon Maragol Pandon Rlzal Signe GENERATION I Year of birth 1909 1906 1912 1909 1907 1907 Males (number) 678 127 198 110 147 96 Born in the municipio 37.6% 45% 19% 44% 38% 70% Mean educational level (years) 3.78 4.24 371 3.75 3.88 3.23 Main occupation, farmer 79.3% 81% 88% 59% 73% 92% Main occupation, agricultural laborer 2.7% 4% 3% 5% 2% 0% Main occupation, permanent non-ag. 6.3% 4% 4% 11% 10% 3% Main occupation. temporary non-ag. 11.4% 11% 6% 25% 13% 4% Females (number) 674 127 194 110 147 96 Born in the municipio 32.5% 25% 24% 51% 32% 50% Mean educational level (years) 3.26 3.99 2.74 3.36 3.39 3.04 Main occupation, housekeeper 91.3% 87% 89% 92% 93% 99% Main occupation, agricultural laborer 1.6% 2% 2% 4% 0% 0% Main occupation, permanent non-ag. 0.3% 0% 1% 0% 1% 0% Main occupation, temporary non-ag. 4.3% 8% 6% 5% 1% 1% GENERATION II Year of birth 1940 1944 1942 1938 1937 1938 Males (number) 344 64 100 57 74 49 Born in the municipio 49.4% 55% 47% 37% 66% 37% Mean educational level (years) 6.22 6.47 6.21 6.67 6.42 5.1 Main occupation, farmer 79.1% 78% 84% 61% 76% 96% Main occupation, agricultural laborer 7.6% 9% 12% 11% 3% 0% Main occupation, permanent non-ag. 2.6% 0% 1% 5% 5% 2% Main occupation, temporary non-ag. 8.7% 12% 3% 14% 14% 2% Females (number) 299 54 97 45 62 41 Born in the municipio 45.8% 37% 49% 40% 61% 32% Mean educational level (years) 6.32 6.31 5.89 7.09 7.05 5.44 Main occupation, housekeeper 79.3% 76% 80% 67% 79% 95% Main occupation, agricultural laborer 6.0% 11% 8% 7% 2% 0% Main occupation, permanent non-ag. 3.0% 2% 3% 4% 5% 0% Main occupation, temporary non-ag. 7.0% 7% 4% 18% 8% 0% GENERATION IlIl Year of birth 1966 1968 1967 1966 1965 1966 Males (over 14 years old) (number) 645 109 198 105 122 111 Mean educational level (years) 8.32 7.86 8.46 9.08 8.44 7.65 Main occupation, farmer 30.6% 26% 37% 9% 42% 32% Main occupation, agricultural laborer 19.8% 20% 28% 10% 7% 27% Main occupation, permanent non-ag. 6.6% 6% 7% 6% 11% 4% Main occupation, temporary non-ag. 20.9% 25% 13% 36% 18% 21% Still in school 20.0% 23% 13% 36% 19% 14% Residing in the municipio 71.6% 65% 83% 63% 67% 51% Residing in other region: 11.3% 8% 2% 13% 20% 27% Mindanao 3.4% 1% 1% 4% 12% 5% Manila 6.0% 6% 2% 4% 7% 15% Residing in other country 1.2% 0% 1% 4% 0% 2% Females (over 14 years old) (number) 600 101 183 92 136 88 Mean educational level (years) 9.37 8.75 9.40 10.24 9.92 8.27 Main occupation, farmer 4.7% 5% 10% 0% 2% 1% Main occupation, housekeeper 36.1% 38% 30% 40% 36% 43% Main occupation, agricultural laborer 7.7% 9% 14% 2% 6% 2% Main occupation, permanent non-ag. 9.0% 4% 8% 17% 10% 6% Main occupation, temporary non-ag 21.5% 25% 16% 21% 21% 30% Stil in school 8.4% 20% 19% 16% 21% 12% Residing in the municipio 60.0% 48% 65% 38% 56% 44% Residing in other region: 13.7% 14% 7% 19% 22% 25% Mindanao 1.7% 1% 1% 5% 2% 1% Manila 9.5% 12% 6% 9% 16% 16% Residing in other country 3.0% 2% 3% 6% 3% 4% 67 PHILIPPINES POVERTYASSESSMENT Table A4.2: Household Income, Assets, and Production Structure, in Sample Villages, 1985 and 1998 1998 Central Luzon Panay Island Total sample Gabaldon Maragol Pandon Rizal Signe Income and expenditure Total expenditure 311.6 344.7 321.4 393.7 266.4 211.6 Total income, wet season 281.7 348.6 329.0 203.8 264.8 161.8 Farm income (wet season) 194.3 180.2 262.0 127.6 171.3 127.9 Rice farm income 110.6 101.6 201.0 54.1 49.9 33.2 Other crop farm income 24.3 31.8 17.4 14.2 32.8 32.4 Livestock income 59.4 46.8 43.6 59.3 88.5 62.3 Non-farm income (wet season) 87.4 168.3 66.9 76.2 93.5 33.8 Off-farm income 24.1 92.3 17.2 9.3 9.1 2.4 Non-farm income 23.1 19.9 25.8 10.4 32.1 14.7 Unearned income 40.2 56.2 23.9 56.5 52.3 16.7 Assets Total assets 7021.9 2721.9 8467.9 7845.8 7680.1 5576.8 Land 2674.5 878.7 4234.8 1922.8 2198.1 2164.8 Housing 2500.4 827.2 1960.1 3705.6 3876.5 1739.7 Productive assets and savings 848.9 467.8 1279.9 1200.6 463.7 295.3 Animals 513.7 369.1 420.3 307.8 605.3 1215.9 Consumer durables 484.5 179.2 572.8 709.0 536.5 161.1 Production Farm size (ha.) 1.26 1.24 1.46 1.05 1.07 1.37 Rice yields (kg/ha.) 3357.4 4040.5 4527.7 3344.7 1934.9 1847.7 Profits per ha. before rent 422.4 605.1 644.8 381.4 141.7 154.8 Profits per ha. after rant 354.5 526.8 587.4 272.5 85.6 92.2 Land Tenure Structure Owners 38% 24% 39% 32% 45% 48% CLT holders 10% 18% 17% 8% 0% 0% Leaseholders 21% 21% 20% 27% 24% 4% Share tenants 8% 0% 0% 5% 11% 43% Landless 23% 37% 23% 27% 19% 4% 1985 Income Total income, wet season 136.1 140.5 151.9 110.7 153.6 62.2 Farm income (wet season) 87.4 89.9 108.6 53.5 95.7 34.4 Rice farm income 68.8 74.8 88.6 34.9 76.9 15.5 Other crop farm income 3.7 3.2 4.7 1.9 1.2 9.7 Livestock income 15.0 12.0 15.2 16.7 17.6 9.3 Non-farm income (wet season) 48.7 50.5 43.4 57.2 57.9 27.7 Off-farm income 11.9 20.8 10.9 6.6 13.4 5.7 Non-farm income 20.7 11.8 18.4 34.9 21.1 20.3 Unearned income 16.1 17.9 14.1 15.6 23.4 1.8 Assets Total assets 4547.7 2814.2 5960.9 4027.4 4783.1 2360.8 Land 2839.8 1240.9 3264.2 2873.2 3752.3 1703.1 Housing 667.1 431.2 967.5 630.9 447.5 338.3 Productive assets and savings 581.2 727.8 1024.1 360.2 188.4 69.8 Animals 210.9 274.6 199.9 108.9 249.9 210.4 Consumer durables 248.6 139.7 505.2 54.3 145.0 39.1 Production Farm size (ha.) 1.65 1.83 2.14 1.02 1.29 1.13 Rice yields (kg/ha.) 3395.4 3274.6 3612.0 4189.4 3306.9 2063.8 Profits per ha. before rent 233.1 232.9 194.8 297.5 297.3 158.8 Profits per ha. after rent 167.5 175.9 132.1 205.1 239.5 81.7 Land Tenure Structure Owners 25% 21% 13% 22% 37% 48% CLT holders 13% 24% 24% 3% 0% 0% Leaseholders 34% 39% 48% 30% 24% 9% Share tenants 12% 0% 2% 16% 13% 39% Landless 17% 13% 11% 30% 23% 4% Note; Expenditures are in adult equivalents. Incomes, expenditures, assets andprofits are expressed in 1998US$. 68 4: TARGETED POLICIESAND THE POOR Table A4.3: Impact of the Land Reform on Human Capital Formation, Asset Accumulation, and Long-Term Productivity and Income Dependent Independent OLS Robust Regression Median Regression variable variables Coefficient t-statistics Coefficient t-statistics Coefficient t-statistics Asset accumulation: Beneficiary 1486.50 3.21 860.99 3.94 1138.84 3.51 Inherited to 1988 Education 258.18 3.51 43.71 1.26 96.19 1.59 (1998 US$) Initial assets 0.81 0.48 1.49 1.85 0.82 0.48 Asset accumulation: Beneficiary 1525.74 3.24 996.95 4.38 1330.26 4.22 Inherited to 1988 Education 258.52 3.51 51.43 1.49 116.24 1.67 (1998 US$) Initial assets 1.37 0.68 1.71 1.81 2.15 0.89 Initial assets' Beneficiary -1.91 -0.52 -39.54 -1.75 -1.51 -0.07 Increase in education Beneficiary 0.60 1.83 0.74 2.37 0.83 1.80 (years) Initial education -0.76 -14.19 -0.70 -13.66 -0.75 -9.42 Increase in income Beneficiary 86.12 2.08 73.84 2.76 92.24 2.82 (1998 US$) Education -0.57 -0.09 -2.32 -0.55 -1.72 -0.24 1985-98 Income in 1985 -0.68 -6.76 -0.48 -7.40 -0.57 -2.63 Asset accumulation Beneficiary 605.18 0.78 525.57 1.77 612.67 2.28 (1998 US$) Education 263.77 2.01 124.41 2.48 165.93 2.25 1985-98 Increase in rice yields Beneficiary 637.91 2.26 580.16 2.28 565.45 1.67 (kglha.) Education 8.18 0.17 -11.31 -0.26 14.46 0.29 1985-98 Increase in profits Beneficiary 102.40 1.88 79.56 1.65 82.07 1.89 (1998 US$) Education -0.84 -0.09 -3.63 -0.44 -5.81 -0.60 1985-98 Source: Deininger, Lara Jr., Maertens and Quisumnbing (2000). 69 PHILIPPINES POVERTYASSESSMENT Endnotes 1 The effective protection rates have been increasing even faster given the declining protection rates for agricultural inputs (David, 1999). 2 The FIES survey values the household production of rice at farn gate prices while household con- sumption of rice is valued at local retail prices. An (upward) adjustment therefore needs to be made to the value of rice produced in order to make it comparable with the values of consumption. Using data on farm gate and retail prices from the Bureau of Agricultural Statistics and Roumasset (1999) and using a 65% recovery rate for rice (from palay), the effective differential between retail and farm gate rice price is estimated to range between 30-50% (of farm gate price). Thus, three altemative adjustment factors of 30, 40 and 50% are used to derive estimates of the value of rice production, and hence of the value of net consumption of rice. 3 For a similar analysis for Thailand, see Deaton (1989). 4 This section draws on a background paper by Deininger, et. al., on "Redistribution, investment and human capital accumulation: the case of Agrarian Reform in the Philippines." 5 The first systematic attempt at land redistribution, undertaken by the US colonial government in 1903, involved the purchase of about 166,000 hectares from the Catholic Church. It was followed by the 1933 Rice Share TenancyAct which limited share rent to 50% and imposed a ceiling of 10% per annum on credit extended by the landlord and two major pieces of Agrarian Reform legislation under the Magsaysay administration, namely, the 1954 Agricultural Tenancy Act and the 1955 Land Reform Act. The former aimed to improve the situation of tenants by limiting share rent to 30%, further reducing the interest ceiling to 8-10%, and improving the enforcement of existing legislation. The latter was to provide for expropriation of large estates, but in practice remained rather ineffective (Riedinger, 1995). 6 Birdsall and Londono (1997); Deininger and Squire (1998). 7 Blanchflower and Oswald (1998). 8 Fafchamps and Pender (1997); Jalan and Ravallion (1998). 9 An additional issue is that subdivision and renting out of land that had been received under the land reform program is strictly prohibited. Many of the beneficiaries who received land in 1972 are by now too old to farm themselves and there is considerable anecdotal evidence suggesting that their chil- dren have acquired education and taken on non-agriculturaljobs (e.g., Hayami, Marciano and Kikuchi, 1998). If this is true, elimination of these restrictions would be vital to ensure not only efficiency of land use but also the ability of a "second generation" to benefit (at least indirectly) from the earlier land reform. 10 One of the villages in each province (Maragol in Central Luzon and Pandon in Iloilo) represents favorable agro-climatic environments with irrigated rice production. Two other villages in less favor- able environments combine reliance on rain-fed production with supplemental irrigation during the dry season. The fifth village (Signe) is located in a mountainous and marginal enviromnent. 11 The analysis is based only on panel households interviewed both in 1998 and 1985. 70 ENDNOTES 12 In addition, for 68 beneficiary households who had both children who completed their schooling before receiving land and children who did so after the household benefited from land reforn, the same test was conducted "within" the same household. A simple t-test points towards an even higher and highly significant (t = 5.01) difference of 3.32 years, although other factors may of course come into play such as unobserved ability and life-cycle phenomena unrelated to land reform (house- holds taking their first child out of school early to have a secured successor and send the others to school later). 13 Note that the ability to make inferences about large landholders is constrained by the fact that in a survey people are unlikely to admit to owning more than the legal maximum of land. 14 24 of the 7 5 leaseholders (3 0%) moved up to become either amortizing or full owners, and 36% of the share tenants in the sample moved up to become leaseholders, with another 18% becoming either full owners or CIT holders. 1 5 Descriptive evidence indicates that the majority of cultivators moved "down" the tenancy ladder - mostly to give up farming -did so voluntarily, mainly because they receive remittances from kin who had moved out of the village. Neither age nor intergenerational transfers appear to be able to explain the phenomena observed -only about 10% of households had bestowed part of their landholdings to their offspring and excluding them does not change the substantive conclusions reported here. 16 Note that it is not possible to discern which of the upward movements were due to land reform and which ones were due to other events. 71 fTH E ( 'N ANS IS A NN 1997~~ make thee beinn of the Asa finanm _ S r 0 N CHAPTERg,; :\ E THE CRISIS ANTD THNN the Philippines was enjoyingnfvorableaeonomico;o 5.1th Whnfdealuaion ofd rthmed tha BahnaJulyb levels after the double-digit ates of 1988-91A thefal levels, and the budget was in surplus. Pove of pover.ty faing from 32 percn i1994 to go2 odnhae.ntheles thre years prioratoith crisis,N thephidlyippnsittes was tenjoying ifaorabe economi as large capital outflows instantly cre downward pressure on the Philippine Pes N N \ NN leesatrthe doubluar e-dgtrae of 19 88-91,d the Philip~4 econorny stalled 'm 198 Real GDP sha compounded by the worst drought inl 30 years 0S This was reflected in the 1998 sectral groth levels, and te bnudgtnwas prdcinosurplus Poeby 1.7>4v ANQN N ~NNNNN ratesn. bendclining as w w tinienc N \ 7 0 - :: '-:K :~~~~~~~~~~~~~>4N N N 5: THE CRISIS AND THE POOR Table 5.1: Indicators of Economic Activity and Household Survey (APIS) conducted by the National Statistics Impact During the Crisis Office (NS0).3 The 1998 APIS surveywas designed Pre-crisis 1998 to be a longitudinal survey forming a panel with the Per capita GDP Growth 0.4 (1990-96) -2.6 1997 FIES. Twenty-three thousand one hundred fifty Per capita private consumption grcwth 1.0 (1990-96) 1.3 percent of the APIS were Inflation (CPI) 9.8 (1990-96) 97 households (59.8 sample) Unemployment Rate 8.6 (1996) 10.1 thus common to both surveys. While these data could Poverty Incidence 25.0 (1997) in iple provide a direct measure Government Spending (as % of 1997) in pric of welfare Education 102.6 change, the potential usefulness of the longitudinal Health 97.6 nature of these data was seriouslyimpaired byprob- Source: World Bank (2000e). lems of comparability of income and consumption across the two surveys.4 As a result, it is virtually 5.3 With the slowdown in output growth impossible to separate how much of any observed came the slowdown in employment. decline in consumption or income would be attrib- Unemployment rates increased to double-digit utable to a real crisis-related welfare shock versus levels during 1998 (averaging 10.1 percent in 1998 how much is simply on account of measurement. against 8.7 percent in 1997); underemployment also rose. Inflation accelerated to double-digit 5.5 The analysis here is based on exploiting a levels. With the plummeting of agricultural separate section of the APIS survey, where output, food prices increased even faster than the households were asked if they were adversely general level of prices. The crisis also reduced affected by the crisis in different ways.5 The results government revenues which constrained public suggest that the impact of the crisis was modest spending despite an overall counter-cyclical fiscal relative to what has been estimated for other crisis- policy adopted by the government. However, the affected countries in the region, leading to a 5 government was successful in protecting percent reduction in average living standards and expenditures in the social sectors. a 9 percent increase in the incidence of poverty, with higher increases indicated for the depth and WELFARE IMPACT OF THE CRISIS severity of poverty. The impact on measures of overall inequality was minimal. However, these 5.4 What was the impact of these results could underestimate the full impact of the macroeconomic developments on household crisis to the extent they do not factor in the effects welfare? With smaller output falls than any of of the price shock reported by nearly 90 percent the other crisis-affected countries in the region, of the population. On the other hand, there may the Philippines escaped the worst of the regional also be an element of overestimation of the impact financial crisis.' But relatively little is known on account of potential measurement error in the about the distribution impact of the crisis, which self-reported shocks. for the Philippines turned out to be a combination of financial and weather-related shocks. For other 5.6 The largest share of the overall impact on countries in the region, the distribution impact poverty is attributable to the El Nifmo shock as op- of the crisis has been analyzed based on posed to shocks mediated through the labor mar- household survey data before and after (or ket. Not all households were equally vulnerable to during) the crisis, in some cases involving the the crisis-induced shocks and the distribution im- construction of a counterfactual based on a pact of the two shocks was different: while the la- predicted value from past trends and in another, borrnarket shockwas progressive (inequalityreduc- also exploiting the existence of panel data.2 For ing), the El Nifio shock was regressive (inequality the Philippines, the latest available household increasing). Ownership of land made households survey is the 1998 Annual Poverty Indicators more susceptible to the El-Nifno shock (which is 73 PHIL IPPINES POVERTYASSESSMENT unsurprising) but higher levels of education made incorporate a measure of the intensity of the effect them more vulnerable to wage and employment of anyone of these shocks. shocks. The impact of the crisis increased with the level of commercial development of the 5.10 Virtually everyone - nine out of every ten community but the crisis dampened the positive persons- reported being affected by the price effects on the living standards of the households' shock (Table 5.2). However, a large share of the social network (such as membership in co- population was also hit by other shocks. For operatives and NGOs) and community social instance, about two-thirds of the population capital (such as a town hall, a church, a park or reported being hit by at least one of the other library in the community). Finally, occupational four shocks. In addition, most households diversity within the household helped mnitigate reported being affected by more than one shock the adverse impact of crisis-related shocks. For instance, less than 30 percent of the population was reportedly affected by only a 5.7 Despite the relatively small magnitude of single shock. If one were to disregard the price the overall impact of the crisis, households did shock, which was experienced by virtually try to protect their consumption. For three- everyone, less than 3 percent of the population fourths of the affected households, consumption reported being affected by only a single shock. impacts were smaller in magnitude than the The multiplicity of shocks reported by the house- income impacts; the median consumption impact holds reflects both the multidimensional nature of was about one-third lower, while the mean consumption impact was about four-fifths of the income impact. But the ability of the poor to prcotct thirpac consumption wasility morel the pTable 5.2: The Incidence of Crisis-Related Economic Shocks protect their consumption was more limited; the mean consumption to income shock ratio for the Crisis-related Percent of sample Percent of non-poor was 78 percent whereas it was 94 percent shocks households population affected affected Price shock 89.9 91.4 Domestic employment shock 18.4 20.3 5.8 Five potential shocks were identified in the Overseas,migranf APIS questionnaire. Households were asked: employment shock 4.3 4.9 Wage shock 15.3 16.7 "During the past six months, did the following Drought/EI-Nino shock 56.6 59.8 problems affect you and your family?" Price shock only 28.2 26.0 (i) increasing prices of food and other basic Domestic employment shock only 0.16 0.16 commodities; Overseasimigrant (i) loss of job within the country, employment shock only 0.02 0.02 Wage shock only 0.23 0.19 (i loss of job due to retrenchment of mngrant/ Drought/EI Nino overseas workers of the family, shock only 2.36 2.28 (iv) reduced wages; and, Hit by at least one of the five shocks 93.0 944 (v) drought or "El Nino. Hit by at least one of the four shocks other than the rc hc 64.8 68.4 5.9 It is not entirely clear what is meant by e prce shock being "affected by a problem." The allusion to Labor market shock (regardless of the price shock) 8.1 8.6 being affected by a "problem" rather than just an El-Niho shock (regardless event" is indicative of the intent to elicit responses of the price shock) 39.0 40.4 Both El-Nino and labor on potentially adverse impact. But the responses from market shocks the households are in yes/no format and do not (regardless of price shock) 17.6 19.4 Source Cacuilated from 1998 APIS data. 74 5: THE CRISISAND THE POOR the crisis as well as the multiple sources of income The commnunity social capital index is constructed within the household. as an average of binary variables pertaining to the presence of a town hall, a communityhall, a church, 5.11 Because the universal nature of the price or a park in the barangay The comrnmercial capital shock makes the identification of its impact virtually index is based on barangay-level variables on the impossible (the size of the sub-sample not affected number of financial institutions, industrial establish- by the price shock is far too small to construct a ments and stores8 (see Table A5.1 for sumrmarysta- credible control group), the analysis of impact fo- tistics on the model variables). In addition, the analy- cuses on the remaining two main categories of sis allows for variation inl the severity of th-e shock shocks: the labor market shock applying to house- and the ability of the household to cope based on holds who experienced either reduced wages or a certain household and communitycharacteristics as loss of job within the countryor overseas (this com- there is no direct measure of the severityof the shock bines shocks ii, iii and iv above); and the drought or (Model 2 in Table A5.2). These include households' El-Nifio shock (shock v above). Further, given that endowments of labor, land and human capital, some households experienced both types of shocks, households' social network and emplovment diver- three mutually exclusive categories of shocks were sity as measures of their risk-management abilities, constructed, viz., (i) labor market shock alone, (ii) community-level indices of infrastructure, and so- El-Nifio shock alone; and, (iii) joint labor market cial and commercial capital. and El-Nifio shocks. Using these definitions, about 9 percent of the population (8 percent of sample) 5.13 The results (Model 1 in Table A5.2) suggest was affected by the labor market shock, about 40 that the labor market shock had a 12 percent negative percent (39 percent of sample) by the El-Nifio impact on per capita consumption of households shock, and about 19 percent (18 percent of sample) affected bythat shock alone, and the El-Nifio shock was affected by both. Altogether, the three shocks reduced the consumption of those affected by 5 per- account for more than two-thirds of the Filipino cent, while households who were affected by both population (Table 5.2). shocks suffered a negative imnpact on their consump- tion of the order of 9 percent. The snaller effect of 5.12 The analysis tries to identify the impact of the joint shock relative to the single labor market shock being affected by one of these three categories of may appear strange, but it is entirely consistent with shocks on household welfare after controlling a large the possibilitythat those hit bythe single labor market number of household attributes, the attributes of shock were hit harder than those affected by both the communities they live in and the geographic fixed shocks. This could reflect the fact that households who effects (Model 1 in Table A5.2).6 Household char- reported being hit by more than one shock had more acteristics include household demographics, charac- diversified income sources to begin with, which helped teristics of household head, education level of adults, them to cope better with the effects of the shocks. occupational characteristics, including a measure of Allowing for variation of the household level impact diversity of employment,7 and access to land, elec- of the various shocks based on household and com- tricity social networks and public assistance. Com- munity attributes yields some interesting patterns munity or barangay characteristics include three in- (Model 2 in Table A5.2). dices based on barangay or community-level data: one for infrastructure capital, one for community 5.14 Labor Market Shock. The labor market social capital, and one for commercial capital. The shock significantly reduces the returis to education. infrastructure capital index is constructed as an aver- The impact on consumption is greater at higher lev- age of binary variables indicating the presence in the els of education (see Model 2 in Table A5.2 for in- bramrzy of a phone, a telegraph, postal services, a teractive variables for the labor shock). The coeffi- laid-out street pattern and access to national roads. cient on employment diversityis not significant, but 75 PHIL IPPfNES POVER TYASSESSMENT its positive sign points to a mitigating effect on the increases with the level of education, the ownership impact of the shock. The adverse impact of the of land, and the level of the community's commer- labor market shock increases significantly with the cial development. It decreases with the diversity of conununity's level of commercial developrnent. This employment within the household. suggests that the more commercially-developed communities, by virtue of their superior integration 5.17 Since there is no reliable information on with the rest of the economy, are also likely to be the pre-crisis- consumption or income level of more exposed to shocks associated with macroeco- households affected by the shocks, estimating the nomic and financial crises. Understandably, owner- overall impact of the shocks on poverty and ship of land is not a significant factor influencing the inequality requires estimating the counterfactual impact of the labor market shock consumption of households in the absence of the crisis. For households who 'were not affected by 5.15 El-Nifio Shock. Not surprisingly, the any of the three shocks, the impact of the shocks adverse impact of this shock increases significantly is by definition zero and their counterfactual with the ownership of land. The household's consumption is the same as their actual educational endowment on the other hand does consumption. For those affected, the impact of not seem to have a bearing on the impact of this the shock is derived by comparing the level of weather-related shock. The significant negative (predicted) consumption in the absence of the coefficient on the household social network shock with the level of (predicted) consumption indicates that the El-Nifio shock significantly having been affected by the shock.'° eroded the beneficial effects of such networks, rather than these networks being able to protect 5.18 The results indicate a modest impact of the household living standards, possibly reflecting crisis (Table 5.3). There is a negative effect on the covanrate nature of the shock A similar effect mean consumption of about 5 percent. This may is also observed for community social capital. be compared with the 2.6 percent decline in real per However, a greater degree of employment diver- capita GDP but a 1.3 percent increase in per capita sitywithin the household does protect its living stan- personal consumption between 1997 and 1998 esti- dards against the El-Nimo shock mated from the national accounts (Table 5.1). The impact on the headcount index is of the order of 5.16 The results for the joint occurrence of both about 9 percent (an increase from 29.1 to 31.7 per- shocks are somewhere in-between those of the in- cent),1' while the povertygap index is about 11 per- dividual shocks. The adverse impact of the joint shock cent higher due to the crisis, and similarly the squared Table 5.3: Impact of the Crisis on Consumption Poverty and Inequality Povertylinequality Actual Counterfactual Impact Countertactual Measure (all shocks zero) (%) (without L-shock) (without E-shock) (without LE-shock) Mean consumption* 26,482 27,859 -4.9 26,837 27,004 26,982 (per capita per year) (25.8] [37.91 [36.3] Headcount index (%) 31.7 29.1 8.7 31.2 30.5 30.7 [17.0] [46.5] [36.5] Poverty gap index (%) 9.43 8.47 11.3 9.32 8.91 9.10 [11.5] [54.1] [34.4] Squared poverty gap index (%) 3.93 3.48 13.0 3.89 3.67 3.78 [9.5] [57.0] [33.5] Theil's T-index 0.513 0.512 0.1 0.513 0.513 0.511 Variance of logs 0.628 0.633 -0.9 0.635 0.624 0.631 Generalized entropy measure (e=2) 2.02 1.97 2.4 1.99 2.05 1.98 Note: Figures in square brackets give the relative contribution of the three shocks to the total impact. *At 1998 Metro Manila prices (nominal values adjustedforinter-province cost of izvingdifferences, basedon provincialpoverty lines developedby Balisacan, 1999a). Source Calculated from 1998 APIS data. 76 5: THE CRISIS AND THE POOR poverty gap index is indicated to be about 13 per- cent higher. The effects on living standards and pov- 7 erty are, of course, larger for the crisis-affected population. \ 5.19 The measured magnitude of the impact of the cnrsis on poverty depends on the poverty line. To examine this, Figure 5.1 shows how the entire t cumulative distribution functions (GDF) shifted as a result of the crisis. The figure plots the | difference between the actual ("post-crisis") and counterfactual ("pre-crisis") CDFs against log per 1 71 - 17 capita consumption. Thus, for instance, at the ta Pt P capita,consumption.Thus, ,Source: Staff calculations based on the 1998 APIS data. poverty line, we can read off a value of 2.6 whlich confirms the result in Table 5.3 that the crisis induced an increase in the headcount index by this masks the contrasting effects of the labor mar- 2.6 percent points. But the key point here is that ket and El-Nmo shocks. The labor market shock is the impact of the crisis could have been higher progressive (proportionate impact increases with pre- or lower dependimg upon where the poverty line cnsis consumption), whe the El-Nio shock is re- is drawn. This comes out even more strongly in gressive (proportionate impact declnes with pre-crisis Figure 5.2, which shows the ratio of the actual- consumption).12 This is also evident in Table 5.4 which to-counterfactual distribution functions. Again, shows the distribution of the different shocks across while at the poverty line used in this report the the deciles (based on the 1997 FIES).'3 This is con- crisis appears to induce about a 9 percent increase sistent with the notion that the labor market shock in the headcount index (Table 5.3), the percentage affected the relatively better-off wage eamers nmore impcth .could b em highe for, oer poer t severely, while the impact of the drought was heavier mTpact could be much higher for lower povertyim lines (up to 16 percent if the poverty lines were on the relatively poorer agriculture-based house- 1 . .\ . , , . . ~~~~~~~~~holds. Theimpact of the joint labor rarket-El Nifo halved), and much lower for higher poverty lines IMP j (falling down to 3 percent if the poverty lines were shock is found to be neither progressive nor regres- doubled). sive. 5.20 The crisis appears to have had little effect onmasrsofoeal nqult(al 5.3. Bu Table 5.4: Impact of Economic Crisis and El Nin~o on rineasures of overall mequaLty (Table 5.3). But Per Capita Percent of Households Affected by Expenditure Loss of Loss of F~~~ure 54: C t.<~~~~~~~~~~~ N' ~~ Decile Price domestic overseas Reduced '4t*Xfr*,4ti*fr , (1997 FIES) increases job job earnings El Niflo ,, . . . 1 (Poorest) 93.5 17.0 3.8 15.4 78.6 2 91.5 16.6 3.2 13.9 72.7 3 90.9 18.3 2.9 15.5 68.3 ; 9 4 91.7 18.5 4.1 17.1 64.5 t '1 12' 5 5 90.0 21.5 4.5 17.1 61.7 6 90.2 20.5 3.8 16.8 55.0 E41 t t t7 89.7 20.7 4.7 17.1 51.4 - , q ,1 * \ 8 89.6 19.4 4.8 15.2 45.2 9 88.3 18.3 5.1 14.2 43.5 10 (Richest) 84.7 14.7 4.8 11.2 37.8 Overall 90.0 18.5 4.2 15.3 57.9 LPg per capta tp --pspt., Note: Calculation are basedon paneldata (23,150 households) constructedfrom the 1997 FIES and the 1998 APIS. Source: Staff calculatiors based on the 1998 APIS data. Source: Balisacan (1999a), based on 1998 APIS data. 77 PHILIPPINES POVERTYASSESSMEN T 5.21 The relative contributions of the three the contribution of the El-Nimio shock to the depth shocks to the total impact on mean consumption and severity of poverty would be upwards of 80 and poverty point to the importance of the El percent. Nino shock (Table 5.3). The labor market shock alone accounted for about a quarter of the total 5.23 An analysis of the crisis impact with respect impact on mean consumption, the El-Nifio shock to income (rather than consumption) shows alonle contributed 38 percent and the remaining somewhat larger effects. The impact on incomes 36 percent was attributable to the joint labor for affected households are larger (in absolute market-El Nino shock. It is not possible to terms) than the corresponding consumption decompose the last component for the joint shock impacts: 17, 8 and 15 percent as against 12, 5 and any further; thus the shares for the individual 9 percent for the labor market, El Nimo, and the shocks can be interpreted as lower bounds for joint labor markct-El-NiAo shock respcctively the labor market and El Nifio shocks. Hence by (Model 1 in Table A5.2). Similarly, the effect on these estimates, the contribution of the labor mean income is of the order of -7 percent market shock was somewhere between 26-62 compared with -5 percent for mean consumption percent while that of the El Nino shock was (Table 5.5). The impact on income poverty is somewhere between 38-74 percent. correspondingly larger: an increase of 12, 16 and 18 percent in income-based H, PG and SPG 5.22 The contribution of the El-Ninio shock to measures, as against a 9, 11 and 13 percent increase the total impact is higher for the poverty in corresponding consumption-based measuires.14 measures. For instance, this shock alone Incidentally, the El Nimio shock still accounts for contributed about 46 percent to the increase in the the bulk of the total impact on average incomes as headcount index. For the poverty gap and the well as income poverty squared poverty gap measures, more than half the impact is on account of the El-Nimio shock alone, 5.24 The relatively modest difference in the im- while the labor market shock alone contributed only pacts on income and consumption-based measures about 10-11 percent. About one-third of the total of living standards and poverty, which are averages impact is attributable to the joint El Nino-labor for the population as a whole (including those not market shock. If one were to split this joint shock in affected bythe shocks), maynonetheless hide some proportion with the shares of the individual shocks, consumption smoothing by those affected by the Table 5.5: Impact of the Crisis on Income Poverty and Inequality Poverty/lnequality Actual Counterfactual Impact Counterfactual Measure (all shocks zero) (%) (without L-shock) (wfthout E-shock) (without LE-shock) Mean consumptior,i 26,547 28,437 -6.6 27,010 27,254 27,268 (per capita per year) [24.5] [37.4] [38.1] Headcount index (%) 31.7 28.3 12.0 31.2 29.9 30.6 [14.63 [536.5 [3t 9] Poverty gap index (%) 11.1 9.5 16.0 10.9 10.2 10.6 [10.4] [57.0] [32.6] Squared poverty gap index (%) 5 48 4.67 17.5 5.41 5.00 5.22 [8.5] [59.6j [31.93 Theil's T index 0.636 0.627 1.5 0.635 0.634 0.631 Variance of logs 0.785 0.779 0.8 0.795 0.770 0.789 Generalized entropy measure (e=2) 2.87 2.76 4.2 2.81 2.92 2.77 Note: Figures in square brackets give the relsave contribtition of the three shocks to the total impact (not calculatedfor the ineouality measures for wnich these contnobutons are no! additve). 'At 1998 Metro Manila prices (nominal values adjustedcfor inter-province cost of living differences, basedon provincialDoverty lines develoDedby Balisacan, 1999a). Source Calculated from 19.98 APIS data. 78 5: THfE CRISIS AND THE POOR consumption. But results also point to the more ) lirIiited ability of the poor to maintain their con- 0.4 . surmptionin the face of crisis-induced income shocks: the mean consumption to income shock ratio for the income-poorandnon-poorwas94and78per- proportmno ' ,,,- countaa a ;. cent, respectively ,noorne GOVERNMENT AND o_ A HOUSEHOLD RESPONSE 0 03 Income shock as a proporhon of cownteiactual ncome Source: Staffcalculabons basedon 71998 APIS data. 5.25 As with other countries in the region, once it became clear that the crisis would have a shock Figure 5.3 plots the consumption impact as significant impact on domestic output, the a proportion of (counterfactual) income against the govemment relaxed its fiscal stance (see Table 5.6 income impact also expressed as a proportion of for a timeline of key events and government (counterfactual) income. For nearlythree-fourths of policies during the crisis). The result was a national the affected households,the estimated consumption govemment deficit of 1.8 percent of GNP in 1998 impact is smaller than the estimated income impact. from a surplus of 0.1 percent in 1997 (World If households that experienced an income shock were Bank, 2000b). Much of the increase in the deficit to curtail their consumption by the proportion of was driven by a sharp fall in revenue collection but that shock, we would see them clustered around the the partial lifting of the initial budget cuts, imposed 450Line. In fact, the (non-parametric regression) line in 1998 through a 25 percent forced savings scheme, shown in Figure 5.3 lies belowthe 450 line, implying helped sustain the level of expenditures. smaller consumption than income impacts. For all hiouseholds, the mean consumption to income shock 5.26 The government was, in general, successful ratio was 83 percent (73 percent for all affected in protecting social expenditures during the crisis. households), suggesting that despite the relativclylim- As Table 5.7 shows, expenditures on education ited magnitude of the shock, the affected house- were essentiallyunchanged. Expenditures on health holds did resort to actions aimed at smoothing their dropped marginally, but the drop came mairny in Tabie 5.6: Timeline of the Crisis and Response Pre-July 1997 BSP tries to maintain stable exchange rate by raising interest rates Jui-97 Currency depreciation starts Dec-98 BSP reverses its high interest rate policy but market rates still keep increasing Jan-98 Peso at a low of PhP45 per dollar, treasury rates at a peak of 20 percent Feb-98 National Economic Summit organized by the government to articulate response to the crisis Feb-98 Social Accord for Industrial Harmony and Stability signed between government and labor to discourage lay-offs and strikes Mar-98 BLES reports 43,000 workers being laid off temporarily or permanently during the first quarter of 1998 Mar-98 Interest rates start going down Mar-98 Philippines enters into a new stand-by arrangement with the IMF of $1.37 billion Apr-98 Income tax collection falls PhP8 billion short of target Apr-98 Rice subsidy program in selected provinces and municipalities; also distribution of iron-fortified rice May-98 Reflecting the revenue shortfall, national government deficit increases to PhP15 billion Jul-98 Mandatory savings of 25 percent lifted for spending on health and social services Jul-98 ERAP San-saristores established to sell basic food commodities at below market prices Jul-98 Additional funds for rural works program for displaced workers in Mindanao region Aug-98 Emergency loan package and distribution of rice to displaced sugar workers Nov-98 Phil Jobnet launched to facilitate job placement and applicant-matching through a computerized system Dec-98 During 1998, the NFA imported 2 million tons of imported rice to moderate domestic prices Apr-99 Lins Bayan program launcned to provide casual jobs as well as improve local hygiene Apr-99 Signs of recovery: drop in year-on-year unemployment rates and a drop in inflation during the first quarter Source: Compiled by staff from several sources. 79 PHIL IPPINES POVERTYASSESSMENT central government expenditures on curative health, duction fell by 25 percent between 1997 and 1998). which benefits the poor less than preventive care. As a result, it was possible to contain the increase in The 1999 National Government budget incorpo- the price of rice on the domestic market despite the rated a fiscal stimulus equivalent to approximately depreciation of the Peso"5 because domestic nrce 0.5 percent of GNP. The bulk of this stimulus prices were higher than their world market paclkage was for infrastructure but also included equivalents. In addition to stabilizing the irrigation, basic education, public health and consumer price level for rice, the government hospital services and housing and community started piloting a targeted rice subsidy program development. It is, however, not clear as to how in selected provinces and municipalities, much has been actually spent in these sectors, distributed iron fortified rice and established given the continued difficulty in mobilizing special stores (so-called ERAP sari-sari stores) in revenues and the need to contain the budget poor communities to sell basic food commodities deficit. at below market prices. There is no information available on the scale of these operations or their 5.27 In addition to attempting to protect social effectiveness.1" spending, government policy focused primarily on ensuring an adequate supply of cereals during 5.28 There is no unemployment insurance this period. The National Food Authority (NFA) system in the Philippines, except for workers in increased imports of rice in 1998 to 2 million met- the government sector. The output contraction nc tons, or equivalent to 25 percent of total con- in the agriculture sector was absorbed largely sumption (up from 9 percent in 1997) to make up through higher levels of underemployment (see for the shortfall in domestic production (palay pro- below). In other sectors of the economy, as im many of the other crisis-affected countries in the Table .7: Social Sector Expenditures, 1997-98 region, much of the adjustment is likely to have (in 1995 Million PhP) taken the form of real wage declines, even though 1997 1998 unemployment levels also increased during this Consolidated Government 1384 141.1 Education 89.8 92.1 period. Minimum wages eroded with inflation Health22.5 22.0 during the course of the crisis; the annual round Labor and Employment 1.5 1.9 of wage adjustment implemented in early 1998 Housing and Community Deveopment 5.7 5.9 Social Welfare 18.0 18.3 was moderate; and hardly any petitions for wage Othern.6 116.2 116.1 hikes were received during negotiations in 1999. Education 83.1 84.5 Labor was more interested in job security than Basic 68.1 67.6 Tertiary 13.0 13.6 pay increases as reflected also in the two pacts be- Other 2.0 3.2 tween trade unions and employers in February and Health12.4 10.7 November 1998, which emphasized labor retention Preventive 1.3 1.1 Curative 6.7 5.9 as a pnonty Other 4.2 3.7 Labor and Employment 1.4 1.4 Housing and Community Development 2.1 2.1 5.29 One of the few targeted actions m response Social Welfare 16.3 16.4 to the crisis was the setting up of an Emergency Other 0.6 0.6 Loan Facility for Displaced Workers in 1998. Local Governmrent Units 22.3 25.1 La aiiyfrDslcd\Jresi 98 Education 6.7 7.6 Workers who had lost jobs as a result of the cn'sis Health10.2 11.3 (not earlier than July 1997) were elble for loans Labor and Employment 0.1 0.4 Housing and Community Development 3.6 3.8 from the facility so long as theywere current in their Social Welfare 1.8 1.9 . social security contnrbutions. Loans totaling some Note: Itemsmaynotequaltotalsduetorounding. 'netoftransferstoLGUs PhP433 rnillion (out of PhP500 million, or 0.1 per- Source: Budget of expenditures and sources of financing, 1999 and2000, Diokno (1999); Medium-Term Philippine Development Plan, 1999-2004. cent of the budget) were approved in 1998-99 with 80 5: THE CRISIS AND THE POOR the bulk of the funds going to workers in Metro that the poor were cushioned from the full impact Manila; given the eligibilityrequirements, it is almost of the crisis, this appears to have happened through certain that the bulk of the beneficiaries were regu- informal safetynets and reliance on household level lar formal sector workers. coping strategies. 5.30 Over the past two decades, the Philippines 5.31 The 1998 APIS includes a second question has made notable attempts to increase labor absorp- pertaining to the crisis which inquires about the ex- tion through public works especially in rural areas. perience of households affected by the crisis. This The last major effort focused on rural infrastructure informnation is helpful in assessing household cop- in 1994-96 under the Ramos Administration and ing responses and the extent of assistance received targeted the poorest 19 provinces and the fifth and from the government and other households. Again, sixth class towns with pooled resources from dif- because responses to these questions are in the yes/ ferent govemment agencies estimated at PhP23 bil- no format, they do not provide insight into the in- lion."7 The Estrada Administration's main initiative tensity of the response: for examnple, it is not pos- in this area was to set up an inter-agency task force sible to distinguish between a one-hour and a five- charged with formulating recommendations on in- hour increase in working hours or gauge the amount creasing the labor content of the government's in- of assistance received. The crude nature of the ques- frastructure programs. Beyond the general lip ser- tion makes it difficult to interpret responses and the vice paid to encouraging labor-based production large number of households that responded in the technologies, little use appears to have been made affirmative on some questions raises doubts about of the potential of public works programs as a cri- the accuracy of the informnation. Nonetheless, it ap- sis response to provide a safety net for the poor. pears that most households responded to the crisis The government's main contribution towards miti- by changing their eating patterns (Table 5.8). Data gating the irnpact of the shocks was a reasonably from Indonesia confirm that it is not unusual for effective management of macroeconomic policies, households to resort to this coping mechanism: in and the abilityto contain rice price increases through Indonesia, urban (rural) households reduced real massive inports and naintenance of a flexible wage spending on food by 28 (8) percent during the crisis economy Targeted interventions were limnited in scale (between 1997 and 1998) but this reflected a large and unlikelyto have reached the poor. To the extent shift from high cost items such as meat, towards Table 5.8: Household Responses to Crisis Percent of HH Responding to Crisis by: Changing Taking children Migrating Receiving Receiving Increasing Income Decile Total HHs eating out of to city or other assistance from assistance from working (1997 FIES) Responding pattern school countries otherhouseholds government hours 1 2,256 56.7 12.4 7.8 16.5 10.7 37.5 2 2,223 52.3 9.3 5.4 17.1 8.8 36.8 3 2,211 50.7 7.3 5.4 16.3 8.4 33.6 4 2,206 51.0 8.7 5.2 17.0 6.8 33.1 5 2,180 47.8 7.1 4.5 17.2 5.9 29.4 6 2,155 48.3 5.6 3.8 16.4 5.7 27.0 7 2,138 47.0 5.0 3.7 15.0 4.5 26.1 8 2,125 44.1 3.5 3.4 12.5 2.9 22.3 9 2,097 41.4 3.2 3.1 13.8 3.9 23.1 10 2,011 33.3 1.2 3.5 12.0 2.6 18.2 Total 21,602 47.5 6.4 4.6 15.4 6.1 28.9 Note: Calculations are based on panel data (23,150 households) constructed from the 1997 FIES and the 1998 APIS. Source: Balisacan (1999a), based on 1998 APIS data. 81 PHILIPPINES POVERTYASSESSMENT staples, such that the share of food spending on 1996 to 23.7 percent in 1998, with a particularlylarge staples increased from 22 (40) percent to 32 (49) increase in rural areas in 1997, presumably reflecting percent. Data fromthe Philippines also suggest that the impact of the El Nimo shock This may or may the poor were more likely to change their eating not be consistent with the APIS responses: higher pattern than the non-poor. participation rates could induce an increase in un- deremployment as existing work is shared among 5.32 Increasing work hours also seem to be a larger number of employed persons and still be major response, again especially, for households in consistent with more hours of work at the house- the lower deciles. It is difficult, however, to recon- hold level; but this seems unlikely cile this response with information from the labor force surveys (Table 5.9). Labor force survey data 5.33 The proportion of households who re- suggests an increase in labor force participation in sponded by taking their children out of school is 1998. The increase is particularly large when com- too high to be credible."8 Administrative data sug- pared with the increase in the working age popula- gest that enrollment rates continued to increase for tion; at the margin, 87 percent of the increased work- both elementary (from 95.09 in 1997 to 95.73% in ing age population in 1998 entered the labor force 1998) and secondary schools (from 64.04 in 1997 (compared to an average labor force participation to 65.22% in 1998) with no perceptible change even rate of 65 percent) but only 39 percent of them in the rate of increase compared to previous years.'9 found employment. This is consistent with an in- Data from Indonesia where the crisis and its impact crease in hours worked for the household as a unit. on poverty were much more severe (poverty in- But the data also show that underemployment in- creased by about 70% between 1996 and 1999), creased during 1997 and 1998 from 19.4 percent in show only a srnall drop in junior secondary school Table 5.9: Employment Impact of the Crisis In thousands of persons Working Age Population Labor Force Employed Unemployed Underemployed Visibly Under-employed 1/ Total Rural Urban 1994 42670 27488 25171 2317 5261 2618 1170 1448 1995 42770 28042 25700 2342 5089 2519 1111 1408 1996 45034 29635 27440 2195 5323 2771 1156 1616 1997 46214 30256 27879 2377 6356 2927 1853 1074 1998 47415 31306 28290 3016 6705 3565 2071 1494 1996-1995 2264 1592 1739 -147 235 253 45 208 1997-1996 1180 621 439 182 1033 156 697 -541 1998-1997 1201 1050 411 639 348 637 218 419 In percent Labor force Employment Unemploy- Underemploy- Visibly under- Rural visibly Urban visibly participation rate rate ment rate ment rate employed as % unemployed unemployed of employed as % oftotal as % oftotal 1994 64.4 91.6 8.4 20.9 10.4 44.7 55.3 1995 65.6 91.6 8.3 19.8 9.8 44.1 55.9 1996 65.8 92.6 7.4 19.4 10.1 41.7 58.3 1997 65.5 92.1 7.9 22.8 10.5 63.3 36.7 1998 66.0 90.4 9.6 23.7 12.6 58.1 41.9 1996-1995 70.3 109.2 -9.2 13.5 14.5 17.8 82.2 1997-1996 52.6 70.7 29.3 235.2 35.5 447.3 -347.3 1998-1997 87.6 39.1 60.9 84.8 155.1 34.2 65.8 1/ Visible underemplomentrefers to the population tnat is working less than 40 hours but wouldlike to work longer. Those who are already workirig 40 hours but would like to work longermake up the rest of the underemployed. Source; 1999 Philippine Statistic-al Yearbook. 82 5: THE CRISIS AND THE POOR enrollment. It is possible that households' responses 5.35 The above results suggest a possible link to this question are reflecting an increase in the inci- between a household's pre-crisis living standard and dence of absenteeism which would not be captured its response to a racroeconomic shock. Regres- in the administrative data on enrollments. However, sion analysis confirms that the probability of house- this interpretation would be consistent with the de- holds changing their eating patterns, taking children cline in achievement results in 1998, as measured by out of school, and increasing working hours is in- the national tests (NEAT and NSAI) especially for verselyrelated with pre-crisis living standard. On the secondaryschools.20 other hand, the probability of receiving assistance/ relief from the public sector, as well as other house- 5.34 Households in the Philippines also had ac- holds, is not significantly related with pre-crisis liv- cess to private and public transfers in response to ing standard. This suggests that during the crisis, ex- the crisis. The proportion of households that re- isting social safety nets in the Philippines, whether ceived assistance from relatives and friends was more from formal or informal sources, did not have a than the proportion that received assistance from pro-poor bias. It is, of course, possible that the the government. Only 6 percent of households re- amount of income transfers received by the poor is ported receiving anyassistance from the govemment higher (in absolute terms or as a proportion of their but 15 percent did receive from other households. pre-transfer incomes) than that received bythe non- Interestingly, for pnvate income transfers, responses poor. Unfortunately the data do not contain infor- across expenditure decdiles exhibit little variation, sug- mation on the type and amount of income transfers gesting that recipients of such transfers do not have received from either the public or the private sector. to be the poorest groups in society 83 PHIL IPPtNES POVERTYASSESSMENT Table A5.1: Descriptive Statistics of Model Variables (1998 APIS) Model Variables Mean Std. Dev. Min. Max. At least one adult household member - in agriculture, fishery or forestry 0.441 0.497 0 1 - in mining or quarrying 0.007 0.084 0 1 - in manufacturing 0.152 0.359 0 1 - in electricity, gas or water 0.010 0.098 0 1 - in construction 0.107 0.309 0 1 - in wholesale or retail 0.243 0.429 0 1 - in transport, storage, communication 0.130 0.337 0 1 - in finance, real estate, business services 0.043 0.202 0 1 - in communal, social, personal services 0.303 0.460 0 1 Produces food for own consumption 0.455 0.498 0 1 Urban household 0.593 0.491 0 1 Member of a cooperative or NGO 0.147 0.281 0 1 Beneficiary of government assistance 0.028 0.086 0 1 (extension servicesischolarship/housing/land reform) Owns land 0.178 0.383 0 1 Family size 5.058 2.260 1 24 Family size squared 30.690 27.898 1 576 Head of household is female 0.162 0.369 0 1 Age of head of household 47.246 14.220 6 99 Age of head of household squared*100 2434.4 1460.4 36 9801 Avg. years of education of adult household members 8.063 3.172 0 17 Avg. years of education of adult household members squared 75.064 50.014 0 289 No. of children between 1-6 years 0.762 0.973 0 7 No. of children between 7-14 years 1.051 1.197 0 7 No. of male adults (at least 15 years) 1.571 1.027 0 10 No. of female adults (at least 15 years) 1.572 0.935 0 10 Head of household is single 0.038 0.192 0 1 Head of household is widow(er) 0.128 0.335 0 1 Head of household is divorced 0.015 0.123 0 1 Household has electricity 0.743 0.437 0 1 Social capital index 0.474 0.229 0 1 Infrastructure capital index 0.458 0.285 0 1 Commercial capital index 0.435 0.281 0 1 Diversity of employment 1.476 0.669 1 6 Shock variables: Labor market shock: S(L) 0.081 0.273 0 1 El Nino shock: S(E) 0.391 0.488 0 1 Joint labor market-El Nino shock: S(LE) 0.176 0.381 0 1 Note: Numberofobservations= 38,710. Source: Calcuated from 1998 APIS data. 84 5: THE CRISIS AND THE POOR Table A5.2: The Estimated Consumption and Income Models (1998 APIS) Dep. Variable: Log Dep. Variable: Log Income Consumption per Person per Person Model (1) Model (2) Model (1) Model (2) OLS restricted OLS restricted param. t-stat param. t-stat param. t-stat param. t-stat At least one adult household member - in agriculture, fishery or forestry -0.1113 -6.73 -0.1111 -6.71 -0.0041 -0.20 -0.0035 -0.18 - in mining or quarrying 0.0607 1.45 0.0558 1.33 0.2480 5.01 0.2469 4.96 - in manufacturing -0.0170 -1.00 -0.0150 -0.88 0.2195 10.76 0.2226 10.92 - in electricity, gas or water 0.1769 5.59 0.1765 5.59 0.4380 13.13 0.4406 13.27 - in construction -0.0991 -5.85 -0.0994 -5.87 0.1737 8.39 0.1739 8.41 - in wholesale or retail 0.0144 0.87 0.0155 0.94 0.2214 11.14 0.2234 11.23 - in transport, slorage, communication -0.0348 -2.06 -0.0336 -1.99 0.1889 9.20 0.1915 9.33 - in finance, real estate, business services 0.1301 5.80 0.1306 5.82 0.3607 14.08 0.3626 14.15 - in communal, social, personal services 0.0384 2.35 0.0375 2.29 0.3214 16.14 0.3212 16.12 Produces food for own consumption -0.1181 -10.97 -0.1177 -10.98 -0.1203 -10.48 -0.1202 -10.49 Urban household 0.0686 1.05 0.0702 1.08 0.0587 0.92 0.0634 1.00 Member of a cooperative or NGO 0.2584 18.97 0.2945 17.30 0.2203 14.86 0.2030 12.39 Beneficiary of government assistance 0.2162 4.86 0.2249 5.06 0.1230 2.54 0.1251 2.59 (extension services/scholarship/housing/land reform) Owns land 0.1428 13.24 0.1738 9.06 0.0870 7.12 0.1486 6.11 Family size -0.2939 -28.65 -0.2952 -28.84 -0.2954 -25.57 -0.2970 -25.76 Family size squared 0.0104 23.48 0.0105 23.82 0.0101 20.29 0.0102 20.71 Head of household is female 0.0922 7.17 0.0920 7.15 0.1160 7.59 0.1155 7.57 Age of head of household 0.0122 8.00 0.0121 7.97 0.0109 6.56 0.0109 6.56 Age of head of household squared*100 -0.0001 -6.76 -0.0001 -6.71 -0.0001 -5.85 -0.0001 -5.86 Avg. years of education of adult -0.0627 -13.09 -0.0605 -12.65 -0.0656 -11.71 -0.0625 -11.13 household members Avg. years of education of adult 0.0100 32.20 0.0100 32.29 0.0105 29.63 0.0105 29.83 household members squared No. of children between 1-6 years -0.0099 -1.09 -0.0097 -1.06 -0.0022 -0.20 -0.0016 -0.15 No. of children between 7-14 years 0.0384 4.33 0.0381 4.31 0.0311 3.05 0.0309 3.03 No. of male adults (at least 15 years) 0.1005 11.00 0.1020 11.08 0.1326 12.39 0.1335 12.47 No. of female adults (at least 15 years) 0.1047 10.86 0.1090 11.09 0.1183 10.75 0.1215 10.97 Head of household is single 0.0217 1.19 0.0209 1.15 -0.0046 -0.22 -0.0057 -0.28 Head of household is widow(er) -0.1153 -8.86 -0.1143 -8.80 -0.1218 -7.74 -0.1209 -7.71 Head of household is divorced -0.1556 -6.43 -0.1563 -6.47 -0.1366 -4.74 -0.1376 -4.78 Household has electricity 0.2578 27.01 0.2581 27.10 0.2161 19.63 0.2145 19.51 Social capital index -0.0043 -0.12 0.0190 0.48 0.0041 0.14 0.0573 1.63 Infrastructure capital index 0.1384 4.70 0.1360 4.61 0.1413 4.91 0.1371 4.77 Commercial capital index 0.1041 3.33 0.1231 3.76 0.1159 4.14 0.0997 3.17 Diversity of employment 0.0597 3.69 0.0406 2.31 -0.0618 -3.17 -0.0853 -4.20 Shock variables: Labor market shock: S(L) -0.1190 -9.29 0.1238 2.49 -0.1657 -12.19 0.1072 2.24 El Nino shock: S(E) -0.0502 -5.11 -0.0351 -1.32 -0.0756 -6.90 -0.0930 -3.18 Joint labor - market-El Nino shock: S(LE) -0.0924 -7.54 -0.1468 -11.96 85 PHILIPPINES POVERTYASSESSMENT Table A5.2: The Estimated Consumption and Income Models (1998 APIS) Dep. Variable: Log Dep. Variable: Log Income Consumption per Person per Person Model (1) Model (2) Model (1) Model (2) OLS restricted OLS restricted param. t-stat param. t-stat param. t-stat param. t-stat S(L)'Avg. years of education of adult household members -0.0215 -4.93 -0.0217 -4.70 S(L)*No. of male adults (at least 15 years) -0.0113 -1.19 S(L)*No. of female adults (at least 15 years) -0.0147 -1.45 -0.0226 -2.01 S(L)*Diversity of employment 0.0263 1.63 S(L)'Member of a cooperative or NGO S(L)'Owns land -0.0957 -1.80 S(L)*Social capital index S(L)'lnfrastructure capital index S(L)tCommercial capital index -0.1092 -2.54 -0.0767 -1.82 S(E)'Avg. years of education of adult household members S(E)'No. of male adults (at least 15 years) S(E)*No. of female adults (at least 15 years) S(E)*Diversity of employment 0.0287 2.96 0.0408 3.78 S(E)'Member of a cooperative or NGO -0.0877 -3.75 S(E)'Owns land -0.0395 -1.77 -0.0782 -2.93 S(E)'Social capital index -0.0691 -1.58 -0.1034 -2.18 S(E)lInfrastructure capital index S(E)*Commercial capital index 0.0620 1.62 S(LE)*Avg. years of education of adult household members -0.0095 -3.53 -0.0184 -6.67 S(LE)*No. of male adults (at least 15 years) S(LE)*No. of female adults (at least 15 years) -0.0139 -1.67 S(LE)'Diversity of employment 0.0259 2.28 0.0280 2.41 S(LE)*Member of a cooperative or NGO 0.0818 2.56 S(LE)'Owns land -0.0473 -1.75 -0.0798 -2.43 S(LE)'Social capital index -0.0890 -2.18 S(LE)'lnfrastructure capital index S(LE)'Commercial capital index -0.0669 -1.73 Number of observations 38710 38710 38585 38585 Number of estimated parameters 233 249 233 263 R - square 0.5890 0.5902 0.5890 0.5904 F test F(14,3917) = 0.65 Ptob > F 0.83 Note: The t-statistics allow fordesign effects due to thestratification andclustering of theAPISsample. The sample has 168 strata and3378primarysampling urits. Each of the regrrssions alsoallows forstrata fixedeffects and32provincialdummy variables to control formissingbarangaydata. The estimatedincome models exclude 125 observations with negative reoorted incomes. Source: Calculated from 1998 APIS data. 86 5: THE CRISIS AND THE POOR Endnotes 1 See, for instance, World Bank (1999). 2 See, for instance, estimates in World Bank (2000e). Some of this literature is also reviewed in Booth (1999). For Thailand, see Kakwani and Pothong (1998); for Korea, Kakwani and Prescott (1999) and for Indonesia, Suryahadi, Sudamo, Suharso, and Pritchett (1999). 3 A second round of the APIS for 1999 was also recently fielded by the NSO, though data from this survey are not yet available. 4 The comparability of income was impaired because the reference period used in the APIS is limited to a six-month period (from April to September 1998), while the FIES incomes relate to the full calendar year (January to December 1997). A partial-year recall of incomes introduces unknown seasonal biases in the estimates of incomes, and has particularly serious implications for estimates of agricultural incomes and incomes from other self-employment which are best defined in annual terms. There is some limited abbreviation of the income module in the APIS, but for the most part, the income modules of FIES and APIS are comparable. The comparability of consumption, on the other hand, was compromised by the use of a much shorter consumption module in the APIS. The APIS uses a two-page module identifying only major categories of consumption while a detailed forty-page consumption module is used in the FIES with a detailed coverage of items within categories. A shorter consumption module generally introduces a downward bias in measured consumption levels (see, Jolliffe, 1999, for instance). 5 The analysis is here is based on a background paper by Datt and Hoogeven (2000). 6 Formally, the specification is as follows: =,, f XjX +y Ls +yES + yLESjE+J where Cj is the average consumption per person in household j, Xj is a set of household characteristics and other determinants of household j's per capita consumption, S(' are binary variables indicating if the household experienced crisis-related shocks, and c, is a random disturbance term. As discussed, three measures of shock are distinguished: the labor market shock (SL), the El-Nilio shock (SE), and both (S'E). Consumption per capita is adjusted for spatial cost of living differentials and is expressed in 1998 Manila prices using spatial price indices estimated by Balisacan (1999a). 7 Occupational background of household members is represented by a set of binary variables for different occupational sectors they were employed in. The binary variables take a value of unity if there is at least one household member working in a given sector, zero otherwise. This specification is preferred to one with the number of members in each occupational category because the latter is more likely to be responsive to a shock than the affiliation of at least one member to an occupational category. 87 PHILIPPINES POVERTYASSESSMENT 8 For each of these three indicators, the APIS questionnaire records the actual number if less than 10, and records 10 if there are 10 or more of them in the barangay. 9 The augmented version of model interacts shock variables with a set of household and community characteristics (Zj): lnC=,BX +X$S +rEE +)1ESLE+'6,Z SL +5E,Z SE +61E,Z SE+U The interaction terms could also be interpreted as indicative of how the returns to specific characteristics are altered by the shocks related to the crisis. The vector of interacted characteristics, Zj, could be the same as the set of determinants of consumption, Xj. However, in the interest of maintaining a more parsimonious specification, these are limited Zj to a subset of Xj. 10 The pre-crisis or counterfactual consumption (CjG) is thus derived as actual consumption (C,) minus the impact of the shocks. Thus, C= C - [exp(lntj Sy = O)- exp(lnt0)] where the ternms in the square bracket measures the impact of the crisis-related shocks. Strictly speaking, the impact is measured as the maximum of the estimate in the square brackets and zero. There is no guarantee that the estimated parameters of the model would yield a negative impact on consumption for all households reportedly affected by the crisis. In general, a larger number of interaction terms in the model tended to generate a larger number of cases of positive impact. Using the pruned model estimates, the impact tumed out to be positive for 1,115 (or 4.4%) of the 25,079 households affected by the three shocks. The impact for these households was set to zero. This makes negligible differences to the results. 1 1 Based on provincial poverty lines developed by Balisacan (1 999a). In effect, these poverty lines are used to express all nominal consumption values at 1998 Manila prices, and the Manila poverty line of PhPl1,677 per person per year is used. This corresponds to a nutritional norm of 2000 calories per person per day and allows for basic nonfood expenditure. See Balisacan (1999a) for more details. The counterfactual poverty level reported here is different (higher) than the actual for 1997 in Chapter 1 because the counterfactual estimate is a prediction based on 1998 APIS data while the 1997 FIES data, and the consumption modules in the two surveys are not comparable. 12 This is done by running a simple (tobit) regression of the absolute shock as proportion of pre-crisis consumption on log of pre-crisis consumption. 83 ENDNOTES 13 Table 5.4 is based on panel households only. As a result figures on proportion of households affected by a given shock are slightly different than those in Table 5.2 which is based on the entire sample. 1 4 Note that for the income-based poverty measures, the poverty line is calibrated to yield the same headcount index as obtained with per capita consumption as the measure of welfare. 1 5 The depreciation of the peso had the effect of reducing the nominal protection rate for rice by reducing the differential in peso terms between the domestic price and the world price. 16 The survey on users' perspectives on govemrnment services includes questions on NFA rice and the Erap sari-sari stores and should yield some useful information on access and degree of satisfaction. 1 7 Esguerra et. al. (1999). 18 Given the formulation of the question, households would have responded with a yes even if they took their children out of school for a day as a result of the crisis. 1 9 Early administrative data had suggested a decline in secondary enrollment (Reyes, De Guzman, Manasan and Orbeta, 1999). 2 0 National Anti-Poverty Action Agenda, NAPC (2000). 89 PHILIPPINES POVERrYASSESSMENT AnnexA CONSTRUC77NG ABSOL UTE POVER7Y LINES: A DISCUSSION OF ME THODOL OGY HOUSEHOLD SURVEYS recent census, the 1995 POPCEN, applies to the 1997 FIES. 1. The mnain data sets for this study are drawn from the nationwide household surveys conducted 4. The classification into urban or rural areas is bythe National Statistics Office (NSO). In particu- based on population density and the presence and lar, the study draws upon five rounds of the Family number of public infrastructure, facilities, and es- Income and Expendituref Survey (FIES) fromthe mid- tablishments, i.e., public buildings, plazas, streets, 1980s to the late- 1990s, and the recent 1998 Annual hospitals, etc. Because these indicators can change Poverty Indicator Survey (APIS). from census to census, temporal welfare compari- son based on this classification may be misleading. 2. Conducted every three years since 1985, the The physical area of the "urban sector" is, almost FIES is the main survey data employed in the gen- bydefinition, shifting overtime. As population grows eration of poverty and income distribution statistics and/or economic activity expands, an initially rural on the Philippines. While earlier surveys covering area will be classified as urban, sooner or later. While 1961, 1965, and 1971 are also available, these are this may not be problematic for purposes of mea- not included in this studysince these are either beset suring, say, urbanization trends, it tends to create a bytechnical problems or available only in published systematic upward (downward) bias on urban (ru- forms'. Unit record data are available for the 1985 ral) performance indicators. to 1997 surveys. 5. Thus, when disaggregating byurbanity, the only 3. The 1985 FIES covers a sample of 16,971 strictly comparable years are 1985 with 1988, and households; the 1988 FIES 18,922 households; the 1991 with 1994 since, foreach pair, the classification 1991 FIES 24,789 households; the 1994 FIES 24, is based on the same census. Substantial reclassifica- 797 households; and the 1997 FIES 39,520 house- tion of villages occurred between the 1980 and the holds. The urban and rural areas of each province 1990 POPCEN.2 In addition, since not much re- are the principal domains of the survey In addi- classification appears to have occurred from the 1990 tion, areas with 150,000 or more population based to the 1995 POPCEN, comparison between 1994 on the latest Census of Population (POPCEN) are andt 1997 appears to be valid. also domains of the survey with urban and rural dimensions. For the 1985 and 1988 FIES, the rel- 6. Until 1994, the sampling design allows only evant census is the 1980 POPCEN. For the 1991 regional disaggregation. Significantlylarge standard and 1994 FIES, it is the 1990 POPCEN. The most errors are observed when data are disaggregated l See Balisacan (1994, 1999a) foran extensive discussion on thecomparability of the FIES data. 2A large numberofinitiallyruralareas in 1980became urban areas in, 1990 wen they were foundtosatisfythecritena forurban areas. This reclassifiraton, inadditionto netmigration fronm nral touran areas, reduced the population share of FIES rural areas from 62percent in 1968 to 50 percent in 1991. In contrast, the estimated ruralpopulation share based on fixedphysicalareas was virtually the same - 64 percent- dunng the same period. 90 ANNEXA M " into smaller geographic subgroups. The design since areas, especiallyin agriculture and agriculture-depen- then permits provimcial disaggregation. dent economic activities, comparability of even the income data from the two surveys is a major prob- 7. The FIES captures a wide range of implicit lem. Thus the two surveys could not be used for expenditures, such as consumption of home-pro- welfare comparison between 1997 and 1998. This is duced goods and services, gifts, assistance or relief indeed unfortunate considering that the APIS is in- in goods and services received by the household tended partly to inform changes in poverty for the from various sources, the value of durable goods intervening years when there are no FIES data. (including owner-occupied dwelling units), This makes these data valid for economic welfare com- 10. The 1998 APIS covered 38,710 sample house- parisons among provinces, between urban and rural holds; the sample households came from the same areas, and among socioeconomic groups. sampling frame as that of the FIES. Both surveys (1997 FIES and 1998 APIS) have a sample overlap 8. The APIS, on the other hand, covers variables of about 58 percent, i.e., over one half of the sample other than incomes and expenditures, thereby pro- households interviewed for both surveys can be viding more comprehensive indicators of poverty formed into panel or longitudinal data. We exploit status than income- or expenditure-based poverty this feature of the two data sets to inform the influ- indicators that could be generated from FIES. It ence of certain household characteristics (socioeco- contains information about the demographic and nomic and initial income conditions prevailing in economic characteristics of individual household 1997, the year irmrnediately preceding the onset of members, as well as items related to health, educa- the Asian economic flu in the Philippines) on the tion, family planning, and family access to housing, relative welfare impact of, and household responses water and sanitation, and credit. The APIS also in- to, the crisis. cludes two questions pertaining to the Asian eco- nomic crisis. The first question inquires whether or POVERTY LINES not the household was affected by price increases, loss of jobs, reduced wages and the El Nino. The 11. The rest of this Annex outlines the approach second question inquires about the response of used for constructing poverty lines. The approach households and pertains only to those households respects the principle of consistencyfor spatial com- affected bythe crisis. parison of absolute poverty, i.e., poverty lines con- structed for various areas or population subgroups 9. Intendedto be runeveryyearbeginning in 1998, are fixed in termns of a given standard of living. The the survey does not, however, provide consumption intent is not to derive an altemative estirmate of the and expenditure data as detailed and robust as the level of national poverty, but rather to come up with FIES. This is even more so for household expendi- a practical approach to constructing poverty lines tures, in which the expenditure items in the survey that can be used for consistentlyranking povertystatus instrument were reduced to just two pages (27 ex- across provinces, regions, or socio- economic groups, penditure lines), compared to over 20 pages (over as well as for monitoring performance in absolute 400 expenditure lines) in the FIES. Moreover, the APIS poverty reduction over time. The underlying as- reference periods were for the second and third quar- sumption is that the main objective of poverty mea- ters of the year, while those of the FIES were for the surement is to inform policy choices for reducing first and second semesters. Since there is significant absolute poverty across space and over time.3 seasonality of economic activities across geographic 3 The approach closely resembles thatsuggested by Ravallion (1984, Annex 1; 1998). 91 PHILIPPINES P0VERTYASSESSMENT 12. The approach involves (i) setting a bundle of 14. The main source of data for fixing the refer- food in each province which is the average con- ence group is the 1997 FIES of the National Statis- sumption of a reference group fixed nationally in tics Office. This survey captures a wide range of terms of their expenditure, (ii) adjusting this bundle market-purchased and implicit expenditures, such as to satisfy the minimum nutritional requirement of use value of durable goods (including owner-occu- 2,000 calories per person per day (iii) valuing the pied dwelling units), consumption of home-pro- adjusted bundle at consumer prices prevailingin each duced goods and services, gifts and assistance or province, and (iv) estimnating the non-food spend- relief goods and services received bythe household ing of the reference households in the neighborhood from various sources. The urban and rural areas of of the point where total spending equals the food each province were the principal domains for the threshold. The approach does not require that the survey This makes these data valid even for welfare same bundle of goods be used in each province; comparisons among provinces, between urban and rather it requires that the bundle is typical of those rural areas, and among socioeconomic groups. within a pre- determined interval of total consump- tion expenditure nationally Put differently the ap- 15. The FIES data file does not, however, contain proach fixes the standard of living used for provin- information on either average unit values or quanti- cial comparison but not the composition of goods ties of goods consumed by the household, which used in each province. Differences in composition are required to transform the food bundle into calo- mnay anse as a result of spatial differences in relative ries. In this annex, average provincial prices of com- prices faced by households. monly purchased commodities, together with calo- ne conversion ratios obtained from the Food and Food thresholds Nutrition Research Institute (FNRI), were used to "recover" the calorie content of the bundle. The 13. As in the official approach, the estimation of price data, covering 73 provinces and 11 main cities povertylines proposed in this studystarts with speci- (including Metro Manila), were obtained from the fication of a food bundle for each province which Prices Division of NSO.5 would attain the nutritional norm for good health. TIhe differences in food bundles should reflect sub- 16. However, not all food items in the FIES have stitution effects arising from differences in relative corresponding price data. Also, for some provinces, prices, not differences in real incomes.4 To this end, the price information on some commodities is miss- the bundle for each province is set as the average ing or appears to have been erroneously recorded. consumption of a reference group fixed nationally In the first case, these items were dropped in the in terms of their expenditure (adjusted for family bundle. In the second case, the prices of those com- size). In this study, the reference group pertains to modities were imputed from the average prices of the bottom 30 percent of the population fixed na- nearby provinces, i.e., provincial price arbitrage was tionally, the average consumption bundle is obtained assumed to hold. After these adjustments, the for that reference group in each province. Each matched data still have 54 food items. For the refer- bundle is then transformed into calories and adjusted ence group, these itemns account for an average to satisfy the food energy requirement of 2,000 of about 93 percent of the total food calories per person per day. expenditures. 4 This implies thatthe foodbundles allie on thesame indifference curve. If oneknows the demandmodel, one can easilyset the trndle foreachprice regime (representing aprovince, say). However, in practice, the demand model is notalways known. The approach employedhere does not require knowledge of such a model. 5 These are the same prices as used in the computation of the current CPI series. 92 ANNEXA 17. To calculate the food expenditures for each Indeed, of all the data required in measuring pov- province that will just yield the calorie requirement, erty, the setting of the non-food line is probablythe the cost of the bundle with price information is most contentious. The main issue is whether the pro- multiplied by the ratio of the recommended to the cedure to construct poverty lines allows for consis- computed calories. This assumes that the average cost tent spatial or subgroup poverty comparisons. The per calorie of the items without price information procedure used for constructing nonfood poverty is equal to that of the matched items. Furthermore, lines closelyfollows Ravallion (1998). it is supposed that, within the relevant income range, the composition of the food basket (in terms of 20. This procedure appeals to the notion that "ba- expenditure shares) is fixed. The resulting provincial sic needs" come in hierarchy, beginning with sur- food thresholds are shown in column 1 of Annex vival food needs, basic non-food needs, and then Table 1. basic food needs for economic and social activity This assumes that once survival food needs are sat- Non-food component isfied, as total income rises, basic non-food needs have to be first satisfied before basic food needs. 18. The official approach to estimnating the non- Furthermore, once survival food and non-food food component of the poverty line utilizes the con- needs are mnet, both food and non-food become sumption patterns of households within the ten per- nornal goods. Thus, when a person's total income centile of the food threshold in the income distribu- is just enough to reach the food threshold, anything tion. The average food share for these households is that this person spends on non-food items can be derived and used to divide the food threshold to considered a minimum allowance for "basic non- arive at the povertyline. This procedure carries over food needs," since she/he is sacrificing basic food the inconsistencyproblem inherent in the estimation intakes to purchase such non-food items. It follows of the food threshold. Since the food thresholds that adding this minimum allowance to the food reflect the consumption patterns (and hence overall threshold is a reasonable procedure to setting the living standards) prevailing in each region, as well as poverty line. in rural/urban areas within each region, the average food share is expected to be lower in progressive 21. In practice, the consumption pattern of those areas or regions of the country than in backward sample households whose expenditures are at or near areas or regions. It is well known that food share the food line is used in order to estimnate this mini- correlates well, albeit not perfectly, with standard of mum allowance. The estimation takes the weighted living. That is, for two households with different average of the households whose per capita expen- food shares, the one with the higher food share tends ditures fall within a ten percent band around the food to have lower standard of living, regardless of their line. The weights are selected so as to decline lin- demographic differences (Deaton and Muellbauer, early, the farther the per capita expenditure is from 1980). Thus, by construction, the nonfood compo- the food line. The resulting poverty lines for each nent of the poverty lines in economically progres- province and region of the country are summarized sive regions also implies higher level of living stan- in Annex Table 1. dard than that for the economically backward re- gions. 22. The above procedure of estimating poverty lines gives what has been referred to as lower- 19. Admnittedly, it is unlikely that there exists a bound line (Ravallion, 1998). One may also set - procedure to setting the non-food component of though not pursued here - an upper bound by the poverty line that does not invite disagreement. also appealing- to the same notion of needs hierar- 93 PHI/LIPPINES POVERTYASSESSMENT chy and noting that the assumptions imply that the REAL EXPENDITURES AND povcrty linc cannot exceed the total spending of COST-OF-LIVING INDICES those whose actual food spending achieves basic food needs. A person with this level of spending 24. Povertymeasurement requires combiningpov- must have reached the normative activity level erty lines with information on consumption expen- underlying the food energy requirement (i.e., the ditures. If individual data on money incomes are food threshold), as well as achieved basic non- given, a straightforward way to do this is to com- food needs considered necessary prerequisite to pare these money incomes with poverty lines con- that activity level in a given society. However, at structed for each region, province, or area. Thus, a this level of spending, and since total food household located in province j is deemed to be spending usually does not rise at the same rate as poor if its per capita money income m is less than total spending, it is likely that: (i) spending on the poverty line z for province j. food exceeds survival needs, and (ii) the amount spent on non-food goods exceed the amount re- 25. Another wayto accomplish the same thing is quired to achieve basic non-food needs. For this to deflate each money income m by the "true cost reason, povertyline generated from the total spend- of living index" P, defined for fixed reference prices ing of households whose per capita food expendi- and reference household characteristics. P is just the ture achieves the food threshold is deemed a "high" ratio of each person's poverty line to the reference estimate of the poverty line. povertyline,the latter defining a household with given demographics at a given location and time. The nor- 23. This manner of establishing the povertyline is malized value rn/P gives what is often termed "real in essence similar to the official approach, except expenditure" or "real income" (also referredto else- that the food threshold for each province is set as where in this Report as "living standard"). Thus, a the average consumption of a reference group fixed person is deemed poor if that person's real expen- nationally in terms of their expenditure, not by the diture is less than the base (reference) poverty line. FNRI-determined food consumption bundle con- structed for each province or region. Note that in 26. The cost-of-living indices (with Metro Manila the approach suggested here, both the food and non- as the base), as well as per capita nominal expendi- food components of the poverty line rmake use of tures and per capita living standard averaged for information generated from the same household provinces and regions,are presented in Annex Table survey, i.e., FIES. In contrast, in the official approach, 2. For the analysis of temporal welfare compan'son the "food menu" is prepared by FNRI using in- overthe period 1985-97,we haveincorporated price formation from its food consumption survey, increases overtime by applying the official regional while the non-food component of the poverty CPI to the provincial cost-of-living index.' Annex line is generated from the FIES. Consistency is thus Table 3 summarizes the resulting regional indices for not ensured in the official approach. 1985-1997. Clearly these indices suggest substantial regional price variation in any given year, as well as marked regional differences in rates of price increases during the period. 6 CPI data are not available at the provincfal level 94 ANNEX A Table Annex A.1: Estimates of Food Thresholds and Poverty Lines: Absolute Cost-of-Basic-Needs Approach (1997, Pesos per capita per year) Province Food threshold Poverty line Melro Manila 7,669 10,577 iiocos 7,561 llocos Norte 4,912 7,084 llocos Sur 5,829 7,906 La Union 5,702 7,669 Pangasinan 5,645 7,542 Cagayan Valley 8,318 Batanes 7,512 10,492 Cagayan 6,573 8,717 Isabela 6,337 8,546 Nueva Viscaya 5,360 7,091 Quirino 4,871 6,649 Central Luzon 9,442 Bataan 6,819 9,117 Bulacan 7,204 9,935 Nueva Ecija 7,968 10,805 Pampanga 7,109 9,073 Tarlac 5,950 7,834 Zambales 6,116 7,789 Olongapo City 7,280 10,184 Southern Luzon 9,239 Aurora 6,382 8,657 Batangas 6,982 9,928 Cavite 7,426 10,510 Laguna 7,057 9,443 Marinduque 6,404 8,544 Mindoro Occidental 5,426 7,020 Mindoro Oriental 5,994 8,123 Palawan 5,516 7,311 Quezon 6,077 8,372 Rizal 7,717 10,804 Romblon 6,155 8,047 Bicol Region 8,256 Albay 6,717 9,043 Camarines Norte 5,422 7,495 Camarines Sur 5,818 7,654 Catanduanes 5,676 7,426 Masbate 6,113 8,117 Sorsogon 7,046 9,274 Western Visayas 7,403 Aklan 6,000 7,988 Antique 5,093 6,803 Capiz 5,407 7,350 lioilo 5,325 7,436 Negros Occidental 5,316 7,131 Bacolod City 5,884 7,607 Iloilo City 6,559 9,018 Central Visayas 7,392 Bohol 4,921 6,433 Cebu 5,887 7,803 Negros Orient. 4,949 6,158 Siquijor 5,188 6,930 Cebu City 6,711 9,387 95 PHILIPPINES POVERTYASSESSMENT Table Annex A.1: Estimates of Food Thresholds and Poverty Lines: Absolute Cost-of-Basic-Needs Approach (1997, Pesos per capita per year) Province Food threshold Poverty line Eastern Visayas 7,570 Eastern Samar 6,036 8,240 Leyte 5,896 7,746 Northern Samar 4,920 6,584 Western Samar 5,758 7,538 Southern Leyte 5,679 7,595 Western Mindanao 7,264 Basilan 6,072 8,558 Zamboanga del Norte 5,138 7,093 Zamboanga del Sur 4,998 6,738 Zamboanga C ty 5,542 8,061 Central Mindanao 6,294 Bukidnon 4,314 5,699 Camiguin 5,358 7.300 Misamis Occidental 4,946 6,593 Misamis Oriental 4,961 6,659 Southern Mindanac 7,079 Davao del Norte 4,934 6,605 Davao del Sur 5,065 6,515 Davao Oriental 4,627 6,406 South Cotobato 5,190 7,301 Davao City 5,942 8,002 General Santos City 5,712 7,548 Eastern Mindanao 7,042 Lanao del Norte 5,264 6,906 North Cc'obato 5,10B 7,077 Sultan Kudarat 5,119 7,024 Cotabato City 5,366 6,979 Marawi City 6,374 8,371 CAR 7,646 Abra 5,053 6,474 Benguet 6,057 8,708 Ifugao 4,667 6,447 Mt. Province 4,827 6.558 Baguio City 7,680 10,759 ARMM 8,990 Lanao del Sur 5,452 7,618 Maguindanao 4,900 6,357 Sulu 9,274 12,700 Tawi-Tawi 7,379 10,423 CARAGA 8,990 Agusan del Norte 5,304 7,048 Agusan del Sur 4,593 6,077 Surigao del Norte 5,610 7,348 Surigao del Sur 5,154 6,931 Source: Balisacan (1999a). 96 ANNFX A Table Annex A.2: Mean Expenditure, Cost-of-Living Index and Living Standard, by Province (1997, Pesos per capita per year) Province Average Cost-of-Living Average Living Expenditure Index Standard (Metro Manila = 100) Metro Manila 42,367 100.0 42,367 Ilocos llocos Norte 18,435 67.0 27,514 Ilocos Sur 18,321 74.7 24,526 La Union 15,847 72.5 21,858 Pangasinan 15,180 71.3 21,291 Cagayan Valley Batanes 23,003 99.2 23,189 Cagayan 13,411 82.4 16,276 Isabela 13,978 80.8 17,299 Nueva Viscaya 17,027 67.0 25,414 Quirino 13,970 62.9 22,210 Central Luzon Bataan 26,122 86.2 30,304 Bulacan 21,874 93.9 23,295 Nueva Ecija 16,579 102.2 16,222 Pampanga 21,123 85.8 24,619 Tarlac 17,069 74.1 23,035 Zambales 17,998 73.6 24,454 Olongapo City 25,723 96.3 26,711 Southern Luzon Aurora 16,675 81.8 20,385 Batangas 22,063 93.9 23,496 Cavite 25,887 99.4 26,043 Laguna 25,554 89.3 28,616 Marinduque 14,610 80.8 18,081 Mindoro Occidental 14,180 66.4 21,356 Mindoro Oriental 13,783 76.8 17,947 Palawan 13,674 69.1 19,789 Quezon 16,662 79.1 21,065 Rizal 26,759 102.1 26,209 Romblon 10,122 76.1 13,301 Bicol Albay 14,432 85.5 16,880 Camarines Norte 12,912 70.9 18,212 Camarines Sur 12,776 72.4 17,646 Catanduanes 13,387 70.2 19,070 Masbale 9,665 76.7 12,601 Sorsogon 12,615 87.7 14,384 Western Visayas Aklan 15,616 75.5 20,684 Antique 14,922 64.3 23,206 Capiz 15,793 69.5 22,723 Iloilo 14,554 70.3 20,702 Negros Occidental 13,356 67.4 19,816 Bacolod City 26,353 71.9 36.652 Iloilo City 26,321 85.3 30,857 Central Visayas Bohol 10,204 60.8 16,784 Cebu 13,683 73.8 18,540 97 PHIL IPPINES POVERTY A SFSSMANNT I- Table Annex A.2: Mean Expenditure, Cost-of-Living Index and Living Standard, by Province (1997, Pesos per capita per year) Province Average Cost-of-Living Average Living Expenditure Index Standard (Metro Manila= 100) Negros Oriental 12,519 58.2 21,510 Siquijor 9,411 65.5 14,368 Cebu City 22,606 88.8 25,457 Eastern Visayas Eastern Samar 9,294 77.9 11,931 Leyte 12,224 73.2 16,700 Northern Samar 8,752 62.3 14,048 Western Samar 10,117 71,3 14.190 Southern Leyte 10,830 71.8 15.083 Western Mindanao Basilan 12,713 80.9 15,714 Zamboanga del Norte 12,351 67.1 18,408 Zamboanga del Sur 12,081 63.7 18 965 Zamboanga City 16,810 76.2 22,060 Norlhern Mindanao Bukidnon 12,330 53.9 22,876 Camiguin 12,178 69.0 17,650 Misamis Occidental 11,576 62.3 1 8,582 Misamis Oriental 18,501 63.0 29,367 Southern Mindanao Davao del Norte 12,467 62.4 19,978 Davao del Sur 11,263 61.6 18,285 Davao Oriental 10,143 60.6 16,738 South Cotobato 12,086 69.0 17,516 Davao City 24,048 75.7 31,767 General Santos City 18,936 71.4 26,521 Cenlral Mindanao Lanao del Norte 14,592 65.3 22,346 North Cotobato 11,460 66.9 17,130 Sultan Kudarat 12,817 66.4 19.302 Cotabato City 17,119 66.0 25,938 Marawi City 11,622 79.1 14,692 CAR Abra 14,361 61.2 23,465 Benguet 15,979 82.3 19,416 Ifugao 12,487 61.0 20,470 Kalinga Apayao 13,120 58.7 22,351 Mt. Province 11,120 62.0 17,935 Baguio City 32,880 101.7 32,330 ARMM Lanao del Sur 8,813 72.0 12.241 Maguindanac 9,421 60.1 15.676 Sulu 9,313 120.1 7,755 Tawi-Tawi 12,924 98.5 13,121 CARAGA Agusan del Norte 13,367 66.6 20,070 Agusan del Sur 11,251 57.5 19,567 Surigao del Norte 11,165 69.5 16,065 Surigao del Sur 12,560 65.5 19,176 Source. Balisacan (1999a). 98 ANNEX A Table Annex A.3: Regional Cost-of-Living Indices (NCR 1997 = 100) 1997 Classification 1985 Classification of Provinces _of Provinces Region 1985 1988 1991 1994 1997 1998 1997 1998 NCR 30.5 38.1 58.5 79.9 100.0 110.2 100.0 110.2 1 llocos 27.2 30.5 45.5 58.8 72.8 80.3 71.5 78.9 2 Cagayan Valley 30.4 32.7 48.3 61.0 76.0 83.1 78.6 86.0 3 Central Luzon 32.6 38.3 57.5 71.7 89.3 98.4 89.3 98.4 4 Southern Luzon 33.4 36.8 56.4 70.2 87.4 96.0 87.4 96.0 5 Bicol 27.7 31.3 48.4 60.3 78.1 85.1 78.1 85.1 6 Western Visayas 26.5 29.9 46.9 57.8 70.0 75.4 70.0 75.4 7 Central Visayas 24.4 27.3 44.8 55.6 69.9 77.3 69.9 77.3 8 Eastern Visayas 26.8 29.6 44.1 56.2 71.6 77.5 71.6 77.5 9 Western Mindanao 29.6 32.9 50.3 62.7 79.0 86.8 68.7 75.4 10 Northern Mindanao 24.8 26.8 39.0 49.2 61.5 67.8 59.5 65.7 11 Southern Mindanao 28.8 31.3 43.2 53.7 66.8 73.0 66.9 73.2 12 Central Mindanao 25.1 28.3 43.4 54.1 66.0 72.1 66.6 72.7 CAR 72.3 77.8 ARMM 85.0 93.7 CARAGA 65.2 71.0 Source: Salisacan (1999a). 99 PHIL IPPINES POVERTYASSESSMENT WS f;~~ R MdE A,7 °X__ Annex B POVER7YAND SOCIAL INDICA TORS o _0 Table Annex B.1: Consumption Distribution in the Philippines, 1985-1997 Decile 1985 1988 1991 1994 1997 Ranked by Share in Consump. Share in Consump. Share in Consump. Share in Consump. Share in Consump. per Capita Total /Person/Year Total /Person/Year Total /Person/Year Total /Person/Year Total /Person/Year Consumption Consumption (PhP) Consumption (PhP) Consumption (PhP) Consumption (PhP) Consumption (PhP) 1 2.53 4342 2.63 4984 2.39 4790 2.78 5447 2.57 6087 2 3.75 6444 3184 7263 3.49 7004 3.93 7707 3.62 8567 3 4.65 8011 4.75 8976 4.42 8860 4.79 9383 4.46 10570 4 5.57 9580 5.65 10687 5.34 10697 5.68 11118 5.35 12682 5 6.57 11308 6.69 12666 6.37 12767 6.70 13129 6.35 15044 7 7.80 13419 7.98 15085 7.63 15299 7.89 15471 7.53 17859 8 9.33 16044 9.57 18118 9.29 18626 9.45 18528 9.11 21581 8 11.63 19982 11.91 22529 11.85 23512 11.70 22935 11.44 27102 9 15.64 26905 15.87 30042 15.74 31898 15.72 30809 15.47 36670 10 32.52 55951 31.12 58929 33.48 67140 31.37 61478 34.10 80787 Gini/Mean 0.412 17197 0.400 18926 0.428 20049 0.397 19600 0.427 23694 Note: Percapita consumption expenditures are expressed at 1997Metro Manila prices and have also been adjustedforprovincial cost of living differentials (Balisacan, 1999). Thelast rowshows Gini indices and mean per capita consumption levels. Source: Based on 1998 FIES data. 2 .o~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - l~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~- Table Annex B.2: Income Distribution in the Philippines, 1985-1997 1 i 1985 1988 1991 1994 1997 Decile Ranked Share in Income Share in Income Share in Income Share in Income Share in Income by per Capita Total /PersonNYear Total /PersonNear Total /PersonNYear Total /PersonNear Total /PersonNYear Income Income (PhP) Income (PhP) Income (PhP) Income (PhP) Income (PhP) 1 2.18 4330 2.18 5105 2.00 4993 2.33 5580 2.04 5952 2 3.31 6566 3.32 7761 3.08 7706 3.46 8323 3.03 8873 3 4.22 8368 4.21 9833 3.96 9907 4.33 10410 3.86 11287 4 5.11 10152 5.09 11911 4.86 12151 5.26 12625 4.73 13826 5 6.09 12089 6.14 14351 5.89 14747 6.30 15139 5.72 16714 7 7.30 14492 7.40 17325 7.19 17981 7.54 18106 6.99 20411 8 8.90 17665 9.15 21416 8.90 22266 9.19 22061 8.68 25367 8 11.18 22197 11.55 27015 11.46 28676 11.63 27921 11.21 32754 9 15.54 30829 15.80 36971 15.85 39645 15.89 38173 15.74 45980 10 36.17 71813 35.17 82292 36.82 92116 34.07 81827 37.99 110990 Gini/Mean 0.453 19848 0.447 23397 0.468 25018 0.434 24016 0.478 29214 Note: Per capita incomes are expressedat 1997 Metro Manila prices and have also been adjusted forprovincial cost of living differentials (Balisacan, 1999a). The last row shows Gini indices and mean per capita income levels. Source: Based on 1998 FIES data. Table Annex B.3: Provincial Living Standard and Poverty, 1997 Province Average Living Stds.* Poverty Life Expectancy at Functional Literacy Primary and High School (PhP/person/year) Incidence Depth Birth(years) 1994 Rate (%) 1994 Enrollment Rate (%) 1997 METRO MANILA 42,367 3.5 0.6 (0.08) 69.0 92.41 91.72 1. ILOCOS llocos Norte 27,514 8.3 (2.42) 1.2 (0.61) 68.9 84.69 85.19 Ilocos Sur 24,526 13.3 (2.19) 2.0 (0.42) 66.6 83.29 90.32 La Union 21,858 22.6 (2.67) 5.8 (0.84) 68.9 87.43 87.51 Pangasinan 21,291 25.2 (1.98) 4.7 (0.49) 68.5 87.38 90.58 2. CAGAYAN Batanes 23,189 21.7 (5.13) 3.3 (0.99) 64.1 92.68 84.7 Cagayan 16, 276 31.7 (2.70) 6.5 (0.73) 65.1 86.72 85.27 Isabela 17,299 36.1 (2.32) 10.3 (0.84) 67.0 89.45 87.14 Nueva Viscaya 25,414 10.8 (2.59) 2.5 (0.71) 65.2 78.2 85.83 Quirino 22,210 18.5 (3.24) 3.4 (0.79) 63.5 80.14 86 3.CENTRAL LUZON Bataan 30,304 7.0 (1.78) 1.2 (0.37) 67.7 88.74 90.71 Bulacan 23,295 10.1 (1.36) 1.8 (0.32) 69.8 90.59 92.22 Nueva Ecija 16,222 26.7 (2.29) 6.1 (0.68) 69.0 92.42 84.26 Pampanga 24,619 5.8 (1.28) 0.6 (0.16) 71.0 79.23 85.82 Tarlac 23,035 15.4 (2.22) 3.0 (0.57) 68.1 82.22 86.28 Zambales 25,399 13.8 (2.49) 2.5 (0.58) 67.0 81.71 89.97 4. SOUTHERN LUZON Aurora 20,385 19.2 (2.85) 3.5 (0.65) 64.6 84.16 87.77 Batangas 23,496 17.4 (1.84) 4.2 (0.55) 70.1 90.4 93.83 Cavite 26,043 9.1 (1.44) 1.7 (0.35) 69.1 92.8 87.81 Laguna 28,616 8.2 (1.21) 1.4 (0.25) 67.9 86.09 93.03 Marinduque 18,081 38.2 (3.17) 10.8 (1.19) 64.9 91.25 88.96 Mindoro Occidental 21,356 17.3 (2.42) 3.3 (0.63) 63.9 83.12 80.25 Mindoro Oriental 17,947 32.8 (3.89) 7.7 (1.17) 64.9 91.54 86.16 Palawan 19,789 26.1 (2.46) 5.6 (0.71) 62.9 77.35 83.45 Quezon 21,065 30.3 (2.66) 7.4 (0.79) 66.6 87.25 82.89 Rizal 26,209 12.3 (1.61) 2.2 (0.35) 69.3 89.17 88.55 Romblon 13,301 61.5 (3.51) 17.5 (1.41) 64.3 85.92 86.15 5. BICOL Albay 16,880 49.8 (3.12) 13.8(1.24) 67.6 82.31 84.31 Camarines Norte 18,212 39.5 (3.61) 9.7 (1.21) 63.6 90.01 77.75 Camarines Sur 17,646 35.1 (2.62) 8.5(0.91) 68.7 85.97 86.34 Catanduanes 19,070 29.6 (4.35) 6.7 (1.35) 64.7 87.01 83.02 Masbate 12,601 64.9 (2.68) 20.6 (1.3) 64.0 75.21 80.69 Sorsogon 14,384 50.3 (2.86) 14.6 (1.16) 66.6 79.38 86.87 Table Annex B.3: Provincial Living Standard and Poverty, 1997 Province Average Living Stds.- Poverty Life Expectancy at Functional Literacy Primary and High School (PhP/person/year) Incidence Depth Birth(years) 1994 Rate (%) 1994 Enrollment Rate (%) 1997 0 6. WESTERN VISAYAS Aklan 20,684 32.8 (3.20) 7.0 (0.99) 63,8 83.01 74.25 Antique 23,206 23.5 (2.99) 5.0 (0.84) 63.3 78.45 83.97 Capiz 22,723 26.0 (3.05) 4.7 (0.74) 64.6 76.45 85.82 Guimaras 19,002 17.5 (3.47) 3.7 (0.93) 67.7 83.59 73.27 Iloilo 22,749 22.0 (1.71) 4.8 (0.53) 68.6 83.59 85.47 Negros Occidental 22,271 18.8 (1.67) 4.2 (0.50) 67.6 78.30 85.02 7. CENTRAL VISAYAS Bohol 16,784 43.0 (2.96) 11.9 (1.15) 68.3 84.86 80.80 Cebu 20,317 31.8 (1.58) 9.8 (0.66) 70.5 80.18 83.35 Negros Orient. 21,510 35.1 (3.16) 9.3 (1. 10) 65.5 73.82 73.55 Siquijor 14,368 57.5 (3.96) 18.1 (1.81) 64.4 86.27 77.36 8. EASTERN VISAYAS Biliran 13,345 57.0 (4.50) 15.4 (1.83) 65.0 79.45 82.60 Eastern Samar 11,931 70.9 (3.15) 25.1 (1.77) 61.6 86.25 90.68 Leyte 16,700 41.9 (2.46) 13.2 (1.04) 65.9 79.45 75.86 Northern Samar 14,048 55.0 (2.98) 19.5 (1.51) 62.1 73.63 80.29 Southern Leyte 15,083 45.9 (3.50) 12.2 (1.18) 65.0 86.35 80.08 Western Sarnar 14,407 55.1 (3.05) 15.6 (1.22) 60.9 76.41 76.80 9. WESTERN MINDANAO Basilan 15,714 30.2 (2.71) 5.9 (0.68) 61.4 48.08 69.45 Zamboanga del Norte 18,408 44.2 (3.09) 12.0 (1.21) 63.5 74.49 75.95 Zamboanga del Sur 19,871 31.9 (1.96) 6.9 (0.59) 65.4 77.23 76.44 10. NORTHERN MINDANAO Bukidnon 22,876 23.1 (2.49) 4.9 (0.72) 65.2 83.15 67.31 Camiguin 17,650 33.6 (4.01) 9.1 (1.48) 63.7 85.90 73.76 Misamis Occidental 18.582 37.1 (2.38) 10.9(0.94) 64.7 84.83 73.84 Misamis Oriental 29,367 22.9 (1.89) 5.8 (0.65) 66.9 84.54 76.02 11. SOUTHERN MINDANAO Davao del Norte 19,978 26.2 (2.70) 6.4 (0.87) 64.3 85.49 73.78 Davao del Oriental 16,738 40.2 (4.36) 12.4 (1.64) 65.9 74.61 70.40 Davao del Sur 26,013 21.6 (1.94) 4.6 (0.555) 68.0 68.78 77.27 Sarangani 16,223 38.1 (4.64) 10.4 (1.64) 66.3 73.63 63.85 South Cotobato 20,520 25.4 (3.03) 6.9(1.02) 66.6 73.63 72.02 12. CENTRAL MINDANAO Lanao del Norte 22,346 32.9 (2.12) 9.4 (0.79) 63.3 73.39 69.57 North Cotobato 17,130 42.7 (2.77) 13.4 (1.14) 660 72.76 81.54 Sultan Kudarat 19,302 21.6 (3.04) 3.2 (0.63) 63.7 78.63 83.68 Table Annex B.3: Provincial Living Standard and Poverty, 1997 Province Average Living Stds.- Poverty Life Expectancy at Functional Literacy Primary and High School (PhP/person/year) Incidence Depth Birth(years) 1994 Rate (%) 1994 Enrollment Rate (%) 1997 13. CAR Abra 23,465 22.0 (3.44) 4.7 (0.87) 63.9 90.11 93.39 Apayao 19,781 19.7 (3.69) 4.7 (1.04) 61.9 70.35 87.64 Benguet 23,808 19.7 (2.20) 4.6 (0.70) 67.0 83.89 90.90 lfugao 20,470 31.3 (4.17) 4.4 (0.74) 60.9 51.07 76.18 Kalinga 24,066 16.3 (3.06) 2.2 (0.51) 61.7 70.35 87.00 Mt. Province 17,935 31.4 (4.09) 5.9 (1.02) 61.7 81.08 99.10 14. ARMM Lanao del Sur 12,520 40.8 (2.42) 10.4(0.89) 57.1 59.31 78.81 Maguindanao 17,043 24.0 (2.16) 4.0 (0.53) 55.8 68.71 51.70 Sulu 7,755 87.5 (1.50) 33.1 (0.90) 52.9 57.73 43.48 Tawi-Tawi 13,121 52.1 (3.62) 13.4 (1.37) 51.9 52.67 67.53 15. CARAGA Agusan del Norte 20,070 32.3 (2.38) 9.2 (0.90) 62.9 88.16 77.35 Agusan del Sur 19,567 36.3 (3.30) 8.8 (1.10) 61.1 71.84 73.27 Surigao del Norte 16,065 43.0 (3.30) 10.8 (1.23) 65.8 81.64 76.31 Surigao del Sur 19,176 36.4 (3.33) 10.0 (1.29) 62.6 82.43 77.45 Note: 'Mean per capita household expendture at 1997 Metro Manila prices adjusted forprovincial cost-of-living differences. Cities are incorpo rated in provinces in which they are located. Figures in parentheses are robust standard errors (corrected for sample design effect). Source. Balisacan (1999) and HDN and UNDP (2000). 2b P1t11UPPINES POVERTYASSESSMENT References Alba, Michael M. (2000). "Exploring the APIS Data on the Accessibility of Public Services." July. Alba, Michael M., and E. F. Esguerra. (1999). "Estimating the Modes of Labor Force Participation in the Philippines." Working Paper Series No. 2. January. Altonji, Joseph G.,and Rebecca M. Blank. (1999). "Race and Gender in the Labor Market." In Orley Ashenfelter and David Card, eds., Handbook of Labor Economics, Vol.3. Elsevier Science Pub.Co., NewYork. Angrist, Joshua D., and Alan B. Krueger. (1999). "Empirical Strategies in Labor Economics." In Orley Ashenfelter, and David Card, eds., Handbook of Labor Economics, Vol. 3. Elsevier Science Pub.Co., New York. Asian Development Bank. (1999a). "Assessing the Social Impact of the Financial Crisis in Asia." Integrated Report. Manila. __ . (1 999b). "Report and Recommendation of the President of the Board of Directors on Proposed Loans and Technical Assistance Grant to the Republic of the Philip- pines for the Grains Sector Development Program." Manila. Asian Development Bank and World Bank. (1999). "Philippine Education for the 215'Century." The 1998 Philippines Education Sector Study. Manila. Bahl, Roy. (1999). "IntergovernmentalTransfers in Developing and Transition Countries: Principles and Practice." Prepared for the World Bank's Municipal Finance Thematic Group, April 1999. Balisacan, Arsenio M. (I 993). "Agriculture Growth and Rural Performance: A Philippine Perspec- tive." Journal of Philippine Development 20 (Second Semester): 289-317. . (1994a). "Philippines." In M.G. Quibria, ed., Rur-al Poverty in Developing Asia, Vol. 2. Asian Development Bank, Manila. . (1 994b). "Poverty, Urbanization and Development Policy: A Philippine Perspective." University of the Philippines Press, Quezon City. _ (1995). "Anatomy of Poverty During Adjustment: The Case of the Philip- pines." Economic Development and Cultural Change 44 (October): 33-62. . (1997). "In Search of Proxy Indicators for Poverty Targeting: Toward a Framework for a Poverty Indicator and Monitoring System." Report prepared for the UNDP- assisted Project: Strengthening Institutional Mechanismns for the Convergence of Poverty Alleviation Efforts. . (1999a). "Poverty Profile in the Philippines: An Updated and Re-examina- tion of Evidence in the Wake of the Asian Crisis." University of the Philippines, Manila. . (1999b). "What Do We Really Know -or Don't Know - about Economic Inequality and Poverty in the Philippines?" In A.M. Balisacan and S. Fujisaki, eds., Causes of Poverty. Myths, Facts and Policies - A Philippinze Study. Quezon City: University of the Philippines Press in cooperation with the Institute of Developing Economies, Tokyo. 106 REFERfNCES (1 999c). "Growth, Redistribution and Poverty: Is the Philippines an Excep- tion to the Standard Asian Story?" Journal ofAsia Pacific Economy. In press. . (2000). "Growth, Inequality, and Poverty Reduction in the Philippines; A Re-examination of Evidence." University of the Philippines, Quezon City. January 21. Balisacan, Arsenio M., Ruperto P. Alonzo, Toby C. Monsod, Geoffrey M. Ducanes, and P. Jude Esguerra. (1998). "Conceptual Framework for the Development of an Integrated Poverty Monitoring and Indicator System." Report prepared for the UNDP-assisted Project: Strength- ening Institutional Mechanisms for the Convergence of Poverty Alleviation Efforts. National Economic and Development Authority, Pasig City. Balisacan, Arsenio M., and Rosemarie G. Edillon. (1 999). "Human Face of the Asian Crisis: What Do Nationwide Panel Data on Philippine Households Show?" Paper Presented at the National Conference of the Philippine Political Science Association, 23 July, Quezon City. Bautista, Victoria A. (1999). "Combating Poverty through the Comprehensive and Integrated Deliv- ery of Social Service (CIDSS)." University of the Philippines, Quezon City. Beegle, Kathleen, Elizabeth Frankerberg, and Duncan Thomas. (1999). "Measuring Change in Indo- nesia." RAND. Behrman, Jere R., and Anil B. Deolalikar. (1995). "Are There Differential Returns to Schooling by Gender? The Case of Indonesian Labour Markets." Oxford Bulletin of Economics and Statistics 57: 97-118. Birdsall, N., and J.L. Londono. (1997). "Asset Inequality Matters: An Assessment of the World Bank's Approach to Poverty Reduction." American Economic Review (Papers and Pro- ceedings) 87(2): 32-37. World Bank, Washington, D.C. Blanchflower, DavidG., andAndrew J. Oswald. (1998). "WhatMakes an Entrepreneur?" Journal of LaborEcono7nics 16(1): 26-60. Booth, Anne. (1999). "The Social Impact ofthe Asian Crisis: What Do We Know Two Years On?" Asian-Pacific Economic Literalure 13(2): 16-29. Bouis, H.E., and L.J. Haddad. (1992). "Are Estimates of Calorie-Income Elasticities Too High? A Recalibration ofthe Plausible Range." Journal ofDevelopmentEconom7ics 39 (October): 333- 362. Buhmann, Brigitte, Lee Rainwater, Guenter Schmauss, and Timothy Smeeding. (1988). "Equivalence Scales, Well-Being, Inequality, and Poverty: Sensitivity Estimates Across Ten Countries Using the Luxembourg Income Study Database." Review of Income and Wealth 34: 115-142. Card, David. (1999). "The Causal Effect of Education on Eamings." InOrleyAshenfelter, and David Card, eds., Handbook ofLabor Economics, Vol.3. Elsevier Science Pub.Co., New York. Conning, Jonathan, and Michael Kevane. (1999). "Community Based Targeting Mechanisms for Social Safety Nets." Draft. 107 PHILIPPINES POVERTYASSESSMENT Cowell, Frank. (1995). Measuring [nequality. Prentice Hall Haevester Wheatsheaf, London. Cox, Donald, and Emmanuel Jimenez. (1995). "Private Transfers andEffectiveness of Public Income Redistribution in the Philippines." In Dominique van de Walle and Kimberly Nead, eds., Public Spending and the Poor. Tlheory and Evidence. Johns Hopkins University Press, Baltimore. Cutler, David M., and LawTence F. Katz. (1992). "Rising Inequality? Changes in the Distribution of Income and Consumption in the 1980s." American Economic Review 82 (May): 546-551. Datt, Gaurav, and Hans Hoogeveen. (2000). "El Niiio or El Peso?: Crisis, Poverty and Income Distribu- tion in the Philippines." Policy Research Working Paper No. 2466. World Bank, Washington, D.C. Datt, Gaurav, Kenneth Simler, Sanjukta Mukherjee, and Gabriel Dava. (2000). "Determinants ofPoverty in Mozambique: 1996-97." Food Consumption and Nutrition Division; International Food Policy Research Institute, Washington, D.C. January. Datt, Gaurav, and Martin Ravallion. (1992). "Growth and Redistribution Components of Changes in Poverty Measures: A Decomposition with Applications to Brazil and India in the 1980s." Journal of Development Economics 38: 275-295. . (1998). 'Why Have Some Indian States Done Better Than Others at Reducing Rural Poverty?" Economica 65: 17-38. David, Cristina C. (1999). "Constraints to Food Security: The Philippine Case." Journal of Philippine Development 26(2): 1-46. Deaton, Angus. (1989). "Rice Prices and Income Distribution in Thailand: A Non-Parametric Analv- sis." The Journal of 'the Royal Economic Society (U.K.) 99: Supplement, p.1 -37. (1997). The Analysis of Household Survevs: A Microeconomet7ic Approach to Development Policy. Johns Hopkins University Press. Baltimore. Deaton, Angus S., and Christina H. Paxson. (1998). "Economies of Scale, Household Size, and the Demand for Food." Journal of Political Economy 106: 897-930. Deaton, Angus S., and John Muellbauer. (1980). Economics and Consumer Behavior. Cambridge University Press, New York. Deaton. Angus S., and S. Akbar Zaidi. (1999). "Guidelines for Constructing Consumption Aggregates for Welfare Analysis." Forthcoming. World Bank, Washington, D.C. Deininger, Klaus. (1999). "Making Negotiated Land Reform Work: Initial Evidence from Colombia, Brazil, and SouthAfrica." World Development Report 27(4): 651 -672. Deininger, Klaus, Francisco Lara, Jr., Miet Maertens, and Agnes Quisumbing. (2000). "Agrarian Reform in the Philippines: Past Impact and Future Challenges." Mimeo. Washington, D.C. Deininger, Klaus, Fransisco Lara Jr., Pedro Olinto, and Miet Maertens. (2000). "Redistribution, Invest- ment, and Human Capital Accumulation: The Case of Agrarian Reform in the Philippines." Mimeo. Washington, D.C. 108_ REFERENCES Deininger, Klaus, and Hans P. Binswanger. (1999). "The Evolution ofthe World Bank's Land Policy: Principles, Experience, andFuture Challenges." World BankResearch Observer 14(2): 247-76. Deininger, Klaus, and Lyn Squire. (1998). "New Ways of Looking at Old Issues: Inequality and Growth." Journal ofDevelopment Economics 57 (2): 257-285. Department of Health. (1999). "Health Sector Reform Agenda, Philippines." Mimeo. Manila. Diokno, Benjamin E. (1999). "Local Government Expenditures for Social Services Delivery." Report presented at the Manila Social Forum, 11 November, Manila. Esguerra, Jude, et. al. (1999). "Some Thoughts on Reform of the Budget and Performance Information System in the Philippines." World BankAide Memoire. October. Estrada, Joseph E. (1999). "Foreword to the Medium-Term Philippine Development Plan 1999-2004." National Economic and DevelopmentAuthority. September. Fafchamps, Marcel, and John Pender. (1997). "Precautionary Saving, Credit Constraints, and Irre- versible Investment: Theory and Evidence from Semiarid India." Journal of Business and Economic Statistics (U.S.) 15:180-94. April. Filmer, Deon. (1999a). "The Effect of Household Wealth on Educational Attainment: Evidence from 35 Countries." Population anid Development Review (U.S.) 25:85-120. March. -___. (1999b). "Educational Attainment and Enrollment Profiles." A Resource Book Based on an Analysis of Demographic and Health Survey Data. Development Research Group, World Bank. July. Filmer, Deon, and Lant Pritchett. (1998). "Estimating Wealth Effects without Expenditure Data, or Tears: An Application to Educational Enrollments in States of India." World Bank Policy Research Working Paper No.1994. Washington, D.C. Foster, James, and Amartya Sen. (1997). "On Economic Inequality after a Quarter Century." In Amartya Sen, On Economic Inequality (expanded edition) . Clarendon Press, Oxford. Foster, James, J. Greer, and Erik Thorbecke. (1984). "A Class of Decomposable Poverty Measures." Econometrica 52: 761-765. Hayami, Yujiro, Agnes Quisumbing, and Lourdes Adriano. (1990). "Toward an Alternative Land Reform Paradigm: A Philippine Perspective." Ateneo de Manila University Press, Quezon City. Hayami, Yujiro, E.B. Marciano, and M. Kikuchi. (1998). "DidPhilippine Land Reform Reduce Inequal- ity? A Perspective from a Laguna Village." Prepared at the Social Sciences Division of the International Rice Research Institute as part of the IRRI-Japan Shuttle Project. Hecknian, James. (1976). "The Common Structure of Statistical Models of Truncation, Sample Selec- tion and Limited Dependent Variables and a Simple Estimator for Such Models." Annials of Economic and Social Measuirement 5, No.4: 475-92. 109 PHIL IPPINES POVERTYASSESSMENT Heckman, J., H. Ichimura, J. Smith, and P. Todd. (1998). "Matching as an Econometric Evaluation Estimator." Review of Economic Studies 65: 261-294. Human DevelopmentNetwork and the United Nations Development Programme. (2000). Philippine Human Development Report 2000. Oxford University Press, New York. Hungerford, Thomas, and Gary Solon. (1987). "Sheepskin Effects in the Returns to Education." Review of Economics and Statistics 69 (1): 175-77. Intal, P.S., and M.C.S. Bantilan. (1994). "Understanding Poverty and Inequality in the Philippines: A Compendium of Policy and Methodological Researches." National Economic and Develop- ment Authority, and United Nations Development Programme. Jalan, Jyotsa, and Martin Ravallion. (1998). "Determinants of Transient and Chronic Poverty: Evi- dence from Rural China." Policy Research Working Papers No. 1936. World Bank, Washing- ton, D.C. . (1999). "Income Gains to the Poor from Workfare: Estimates for Argentina's Trabajar Program." Policy research working paper No. 2149. World Bank, Washington, D.C. July. Jolliffe, Dean. (1999). "Survey Design and Household Welfare: The Sensitivity of Estimated House- hold Consumption to Survey Design." Mimeo. Kahneman, Daniel, and Carol Varey. (1991). "Notes on the Psychology of Utility." In Jon Elster, and John E. Roemer, eds., Interpersonal Comparisons of Wlell-Being. Cambridge University Press, New York. Kakwani, Nanak. (1986). Analyzing Redistribution Policies: A Study Using Australian Data. Cambridge University Press, New York. . (1993). "Statistical Inference in the Measurement of Poverty." Review of Eco- nomics and Statistics (U.S.) 75: 632-39. November. Kakwani, Nanak, and J. Pothong. (1998). "Impact of Economic Crisis on the Standard of Living in Thailand." Asian Development Bank and the Development Evaluation Division, National Economic and Social Development Board. Pasig City Kakwani, Nanak, and Nicholas Prescott. (1999). "Impact of Economic Crisis on Poverty and Inequal- ity in Korea." Mimeo. Kapteyn, Arie, Peter Kooreman, and Rob Willemse. (1988). "Some Methodological Issues in the Implementation of Subjective Poverty Definitions." The Journal of Human Resources 23: 22242. Knowles, James C., Emesto M. Pernia, and Mary Racelis. (1999). "Assessing the Social Impact of the Financial Crisis in Asia: Integrative Report." Asian Development Bank, Manila. June. Krugman, Paul R., James Alm, Susan M. Collins, and Eli M. Remolona. (1992). "Transforming the Philippine Economy." National Economic and Development Authority, and United Nations Development Programme, Manila. 110 REFERENCES Lanjouw, Peter, andMartinRavallion. (1995). "Poverty and Household Size" Economic Journal 105 (November): 1415-1434. Lanjouw, Peter, Branko Milanovic, and Stefano Patemostro. (1998). "Poverty and the Economic Transition: How Do Changes in Economies of Scale Affect Poverty Rates for Different House- holds?" Policy Research Working Paper No. 2009. World Bank, Washington, D.C. Lirn, JosephY. (1998). "The Social Impact and Responses to the Current EastAsian Economic and Financial Crisis: The Philippine Case." Country paper prepared for the United Nations Devel- opment Programme/ Regional Bureau for Asia and the Pacific. Llanto, G., A. Orbeta, M. T. Sanchez, and M. Tang. (1997). "A Study of Housing Subsidies in the Philippines." Report prepared for the World Bank and the Housing and Urban Development Coordinating Council. Manila. MacIsaac, Donna J. (1995). "Labour Market Discrimination Against Indigenous People in Peru." Journal ofDevelopment Studies (U.K.) 32:218-33. December. Malkiel, Burton G., and Judith A. Malkiel. (1973). "Male-Female Pay Differentials in Professional Employment." The American Economic Review 63(4): 693-705. September. Maluccio, John. (1998). "Endogeneity of Schooling in the Wage Function: Evidence from the Rural Philippines." FCND Discussion Paper No. 54. Intemational Food Policy Research Institute, Washington, D.C. November. Manasan, Rosario G. (1988). "Tax Evasion in the Philippines, 1981-85." Journal of Philippine Development 15: 167-189. Mangahas, Mahar. (1992). "Self-Rated Poverty in the Philippines, 1981-1992." SWS Occasional Paper. Quezon City. . (1995). "Self-Rated Poverty in the Philippines, 1981-1992." International Jour- nal ofPublic Opinion Research 7: 40-5 5. . (1999). "MonitoringPhilippinesPoverty." OperationalSocialIndicators,Social WeatherStationsOccasionalPaper. Manila. July. Marquez, Nelia R., and Romulo A. Virola. (1997). "Monitoring Changes in the Characteristics of the Philippine Poor: 1985 to 1994." Papers and Proceedings of the Sixth National Convention on Statistics. Monsod, Solita C., and Toby C. Monsod. (1999). "Intemational and Intranational Comparisons of Philippine Poverty." InA. M. Balisacan, and S. Fujisaki, eds., Causes of Poverty. Myths, Facts and Policies - A Philippine Study. University of the Philippines Press, Quezon City. National Anti-Poverty Commission. (1998b). "National Anti-Poverty Action Agenda." Manila. National Economic and Social Development Board. (1999). "Poverty and Inequality During the Economic Crisis in Thailand." Indicators of Well-Being and Policy Analysis, Nevsletter. of the NESDB, Vol. 3, No. 1. January. _117 PHIL IPPINES POVERTYASSESSMENT National Statistical Coordination Board. (1999a). "1997 Philippines: Final Poverty Statistics." Manila. _ (1999b). "1999 Philippine Statistical Yearbook." The Na- tional Statistical Information Center, Makati City. October. Oaxaca, R.L. (1971). "Male-Female Wage Differentials inUrban Labor Markets." PrincetonUniversity Industrial Relations Section Working Paper No. 23, Park, Jin Heum. (1994). "Retums to Schooling: A Peculiar Deviation from Linearity." Working Paper No. 335. Princeton University. October. Paxson, Christina H., and Norbert Schady. (1999). "Do School Facilities Matter?: The Case of the Peruvian Social Fund (FONCOD)." World Bank Policy Research Working Paper No. 2229. Washington, D.C. Pitt, Mark M., Mark R. Rosenzweig, and Donna M. Gibbons. (1993). "Deterninants and Consequences of the Placement of Government Programs in Indonesia." World Bank Economic Review 7: 319- 48. September. Pollak, Robert. A., and Terence J. WVales. (1979). "Welfare Comparisons and Equivalence Scales." American Economic Review (Papers and Proceedings) 69: 216-221. Ponciano, Intal Jr., Melanie Milo, Celia Reyes, and Leilanie Basilio. (1998). "The Philippines." In Ross H. McLeod and Ross Garnaut, East Asia in Crisis: From Being a Miracle to Needing One. Routledge, London and New York. Psacharopoulos, George. (1994). "Returns to Investment in Education: A Global Update." World Development Vol. 22, No. 9: 1325-1343. Putnam, R. (1993). Making Democracy Work: Civic Traditions in Modern Italy. Princeton t 'niversity Press, Princeton. Ravallion, Martin. (1994). Poverty Comparisons. Chur, Switzerland: HarwoodAcademric Publishers. . (1996). "Issues in Measuring and Modeling Poverty." Economic Journal 106 (September): 1328-1343. . (1998). "Poverty Lines in Theory and Practice." LSMS Working Paper 133. World Bank, Washington, D.C. Ravallion, Martin, and Shaohua Chen. (1997). "What Can New Survey Data Tell Us about Recent Changes in Distribution and Poverty?" World Bank Economic Review 11 (May): 357-282. Republic of the Philippines. (1999). Budget of Expenditures and Source of Financing. Manila. . (2000). Budget of Expenditures and Source of Financing. Manila. Reyes, Celia M., and A. B. E. Mandap. (1999). "The Social Impact of the Regional Financial Crisis in the Philippines." MIMAP Research Paper Series No.41. February. 112_ REFERENCES Reyes, Celia M., Generoso G. De Guzman, Rosario G. Manasan, and Aniceto C. Orbeta. (1999). "Social Impact of the Regional Financial Crisis in the Philippines: Preliminary Report." Asian Develop- ment Bank, Manila. June. Riedinger, J.M. (1995). "Agrarian Reform in the Philippines: Democratic Transitions and Redistribu- tive Reform." Stanford University Press, Califomia. Rosenzweig, Mark R. and Kenneth I. Wolpin. (1986). "Evaluating the Effects of Optimally Distributed Public Programs: Child Health and Family Planning Interventions." American Economic Re- view 76: 470-82. Roumasset, James. (1999). "Market-friendly Food Security: Alternatives for Restructuring NFA." Report prepared for the USAID-supported AGILE Project. Schady, Norbert R. (2001). "Convexity and Sheepskin Effects in the Human Capital Earnings Func- tion: Recent Evidence for Filipino Men." World Bank Policy Research Working Paper No. 2566. Washington, D.C. Sen, Amartya. (1985). Commodities and Capabilities. Amsterdam, North-Holland. Solon, Orville, A.N. Herrin, M.H. Racelis, M.G Manalo, V.N. Ganac, and GV. Amoranto. (1999). "Health Care Expenditure Patterns in the Philippines: Analysis of National Health Accounts, 1991- 1997. " UPSE Discussion Paper No. 9908. School of Economics, Univcrsity of the Philippines, Quezon City. SouthAsia Partnership Canada. (1999). "The Good, the Bad, the Well, and the Ill." Popular Perspec- tives on Poverty from four Asian Countries. August 31. Subbarao, K.,AkhterAhmed, and Tesfaye Teklu. (1995). "Selected Social SafetyNet Programs in the Philippines: Targeting, Cost-Effectiveness and Options for Reform." World Bank Discussion Paper No.317. Washington, D.C. Subramanian, Shankar, andAngus Deaton. (1996). "The Demand for Food and Calories." Journal of Political Economy 104: 133-162. Suryahadi, Asep, Sudarno Sumarto, Yusuf Suharso, and Lant Pritchett. (1999). "The Evolution of Poverty during the Crisis in Indonesia, 1996 to 1999." Social Monitoring and Early Response Unit (SMERU), Jakarta. United Nations Development Programme. (1999). HumanDevelopmentReport 1999. OxfordUniver- sity Press, New York. Villegas, Bemardo M. (1999). "Philippine Economic Prospects Under the EstradaAdministration." In Linda Y. C. Lim, Frank Ching, and Bernardo M. Villegas, eds., The Asian Economic Crisis: Policy Choices, Social Consequences, and the Philippines Case. Asian Update, Asia Soci- ety, New York. Wilkinson, Barry. (2000). "Employment Agencies and Labour Market Flexibility in the UK." Final Report on Employment Agencies, World Bank/ASEM-EU Study. March. 113 PHILIPPINES POVERTYASSESSMENT World Bank. (1990). Poverty, WorldDevelopment Report. OxfordUniversityPress. Oxford. June. . (1995). A Strategy to Fight Poverty: Philippines. East Asia and Pacific Region, Washington, D.C. . (1998a). "Philippines: Promoting Equitable Rural Growth." Report No. 17979-PH. Washington, D.C. . (1998b). "The Socioeconornic Impact of the Financial Crisis in the Philippines." Social Development Sector Unit, World Bank, Washington, D.C. July 29. . (1 998c). "Philippines: Social Expenditure Priorities." ReportNo. 18562-PH. Philippines Country Management Unit, World Bank, Washington, D.C. November. . (1999). "Philippines: The Challenge of Economic Recovery." Report No. 18895-PH. Washington D.C. . (2000a). "Global Economic Prospects and the Developing Countries 2000." Washing- ton, D.C. . (2000b). "Philippines Growth with Equity: The Remaining Agenda." Social and Struc- tural Review. Poverty Reduction and Economic Management Unit, Manila, May 3. _ (2000c). "Combating Corruption in the Philippines." Philippine Country Management Unit, World Bank, Manila. May 3. . (2000d). "Judicial Sector Study for the Philippines." Report No. 20446-PH. Washington D.C. . (2000e). EastAsia: RecoveryandBeyond. Washington,D.C. . (2000f). WorldDevelopmentIndicators. Washington,D.C. . (2000g). "AttackingPoverty." WorldDevelopmentReport2000/01. OxfordUniversity Press, Oxford. . (2001). "Philippines: Filipino Report Card on Pro-Poor Services." Report No. 22181-PH. Environmental and Social Development Sector Unit, Washington, D.C. 114 Acknowledgements Tis report was prepared by a team of Bank staff comprising Gaurav Datt, Tamar Manuelyan Atinc and Norbert Schady. The team drew upon several background studies, including A. by Arsenio Balisacan (poverty profile), Michael Alba (access to public services), Gaurav Datt and Hans Hoogeveen (social impact of the crisis), Klaus Deininger (land reform), Norbert Schady (education), and Norbert Schady and Gaurav Datt (self-rated poverty). Martha Aiinsworth, Benu Bidani, Joven Balbosa, Shaohua Chen, Hans Hoogeveen and William Rex made valuable contributions to specific sections of the report while Deon Filmer and Emanuela Galasso were generous in sharing their work and responding to queries. David Bisbee provided able assistance with data and Taranaki Mailei and Araceli Tria with the production of the report. The team gratefully acknowledges the many useful comments received, especially from Emmanuel Jimenez and Jesko Hentschel (peer reviewers). Others who were generous with their comments and time included Vinay Bhargava, Masahiro Kawai, Sanjay Dhar, Bernard Funck, Heidi Hennrich-Hanson, Gurushri Swamy, Vijay Jagannathan, Aniruddha Dasgupta, Rahul Raturi, Syed Husain, Richard Anson, Nor Gonzales, Bhuvan Bhatnagar, Jayshree Balachander and Teresa Ho. The team valued enormously the overall strategic guidance and intellectual support received from Homi Kharas. The study team would like to express its appreciation for the collaboration received from the National Economic Development Authority in carrying out its work. Ms. Ofelia Templo ably led the Philippine counterpart team, providing valuable input on content and fit with the Government's agenda. The team benefited from many useful insights offered during a video conference with members of the Philippines NGO community to discuss the concept of the study (December 1999) and during several workshops held in Manila (February, June and September 2000) and in Cebu (September 2000) to discuss findings with government officials, academe and civil society. Representatives of the donor community, in particular, of the Asian Development Bank and the United Nations Development Programme, were also generous with their time. The report could not have been prepared without the good quality data collected by the Philippines National Statistics Office (NSO). The team would like to note the important contribution made to the quality of poverty monitoring in the Philippines by the government's policy of putting survey data in the public domain. The team is grateful to Mr. Tomas Africa and Ms. Josie Perez of NSO for their valuable support and would like to thank Ms. Sotera de Guzman and Mr. Gene Lorica who responded to many data queries. The team would also like to acknowledge the excellent collaboration received from Mr. Mahar Mangahas who has been a leader in the region in the collection and analysis of data on self-reported poverty. Finally, the study tearn would be remiss if it did not acknowledge the major contribution made by donors to the Asia Europe Meeting (ASEM) trust fund. By supporting the collection of very useful informa- tion at the time of the crisis in the Philippines (the first Annual Poverty Indicators Survey, ASEM donors contributed to a better appreciation of the social impact of the crisis and helped shape policies. The World Bank 1818 H Street N.W Washington D.C 20433 USA www.worldbank.org World Bank Office Manila 23rd Floor, The Taipan Place Emerald Avenue, Ortigas Center Pasig Gty, Philippines Telephone: (632) 637-5855 ww-w.worldbank.org.ph