Measuring Natural Risks in the Philippines

Transcription

1 Public Disclosure Autorized Policy Researc Working Paper 8723 Public Disclosure Autorized Public Disclosure Autorized Measuring Natural Risks in te Pilippines Socioeconomic Resilience and Wellbeing Losses Brian Wals Stepane Hallegatte Public Disclosure Autorized Climate Cange Group & Social, Urban, Rural and Resilience Global Practice January 2019

2 Policy Researc Working Paper 8723 Abstract Traditional risk assessments use asset losses as te main metric to measure te severity of a disaster. Tis paper proposes an expanded risk assessment based on a framework tat adds socioeconomic resilience and uses wellbeing losses as its main measure of disaster severity. Using a new, agent-based model tat represents explicitly te recovery and reconstruction process at te ouseold level, tis risk assessment provides new insigts into disaster risks in te Pilippines. First, tere is a close link between natural disasters and poverty. On average, te estimates suggest tat almost alf a million Filipinos per year face transient consumption poverty due to natural disasters. Nationally, te bottom income quintile suffers only 9 percent of te total asset losses, but 31 percent of te total wellbeing losses. Te average annual wellbeing losses due to disasters in te Pilippines is estimated at US$3.9 billion per year, more tan double te asset losses of US$1.4 billion. Second, te regions identified as priorities for risk-management interventions differ depending on wic risk metric is used. Cost-benefit analyses based on asset losses direct risk reduction investments toward te ricest regions and areas. A focus on poverty or wellbeing rebalances te analysis and generates a different set of regional priorities. Finally, measuring disaster impacts troug poverty and wellbeing impacts allows te quantification of te benefits from interventions like rapid post-disaster support and adaptive social protection. Altoug tese measures do not reduce asset losses, tey efficiently reduce teir consequences for wellbeing by making te population more resilient. Tis paper is a product of te Global Facility for Disaster Reduction and Recovery, Climate Cange Group wit te Social, Urban, Rural and Resilience Global Practice. It is part of a larger effort by te World Bank to provide open access to its researc and make a contribution to development policy discussions around te world. Policy Researc Working Papers are also posted on te Web at ttp:// Te autors may be contacted at bwals1@worldbank.org and sallegatte@worldbank.org. Te Policy Researc Working Paper Series disseminates te findings of work in progress to encourage te excange of ideas about development issues. An objective of te series is to get te findings out quickly, even if te presentations are less tan fully polised. Te papers carry te names of te autors and sould be cited accordingly. Te findings, interpretations, and conclusions expressed in tis paper are entirely tose of te autors. Tey do not necessarily represent te views of te International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or tose of te Executive Directors of te World Bank or te governments tey represent. Produced by te Researc Support Team

3 measuring natural risks in te pilippines: socioeconomic resilience and wellbeing losses brian wals & stépane allegatte Keywords: natural risks, resilience, risk assessment, welfare, Pilippines JEL: D15, D30, D63, D78, Q54, R11 Global Facility for Disaster Reduction and Recovery (GFDRR), World Bank Group

4 introduction 2 1 introduction Te Pilippines is by some measures among te most disaster-affected countries in te world. Te 100 million residents of te Pilippines, along wit teir omes and livelioods, are exposed to a wide variety of disasters, including typoons, eartquakes, floods, storm surges, and tsunamis. Wen major storms and eartquakes affect densely populated urban areas in te country, asset losses are regularly valued in te billions, wit additional untold uman costs. In addition, small events occur frequently across te 2,000 inabited islands in tis middle-income country. On November 8, 2013, Super Typoon Yolanda (internationally referred to as Typoon Haiyan) made multiple landfalls in te Eastern, Central, and Western Visayas regions of te Pilippines, claiming nearly 6,300 lives and directly affecting more tan 16 million individuals across nine regions. At least 1.1 million omes were damaged or destroyed, and te government estimated total losses at US$2.2 billion (P = 95 billion), making Yolanda te most costly urricane to affect te Pilippines to date. Since Yolanda, various regions of te Pilippines ave been affected by numerous wind, eartquake, and flood events, bot large and small. As tese statistics make clear, te consequences of disasters in te Pilippines extend beyond te replacement costs of destroyed assets or asset losses. Disasters ave well-documented consequences for social and economic development agendas and outcomes, including especially inequality, agriculture, education, and ealt [1, 2, 3, 4, 5, 6, 7, 8]. Despite tis growing body of researc, risk assessments still typically adopt asset losses as a singular metric of disaster impacts. Tis is unfortunate, because asset losses obscure te relationsip between disaster risk and poverty. By definition, wealty individuals ave more assets to lose, and terefore teir interests dominate in risk assessments tat are limited to asset losses. At te same time, asset losses do not measure many dimensions of disaster impacts tat accrue to te poor: wile tey by definition ave very little to lose, tey also lack te resources and instruments to smoot income socks wile maintaining teir consumption, and to recover and rebuild teir asset stock. Terefore, te poor are

5 introduction 3 more likely tan te wealty to forego consumption of food, ealt, or education in order to finance teir recovery, and to take longer to recover. To correct tis bias, te initial Unbreakable report introduced te concept of wellbeing losses. Wile proportional to traditional asset losses, wellbeing losses account for people s socio-economic resilience, including (1) teir ability to maintain teir consumption for te duration of teir recovery, (2) teir ability to save or borrow to rebuild teir asset stock, and (3) te decreasing returns in consumption tat is, te fact tat poorer people are more affected by a $1 reduction in consumption tan ricer individuals (see Figure 1). Te analysis presented in tis paper builds upon and expands te approac proposed in te initial report. It uses a new agent-based model along wit detailed natural-risk and ouseold survey data to examine te consequences of natural disasters on ouseolds, as measured by asset losses and alternative metrics. Tis analysis provides a multi-metric assessment of disaster risks at te regional level, using: (1) traditional asset losses; (2) poverty-related measures suc as poverty eadcount; (3) wellbeing losses, wic provide a balanced estimate of te impact on poor and ric ouseolds; and (4) socio-economic resilience, an indicator tat measures te ability of te population to cope wit and recover from asset losses. Tis broad perspective is intended to complement traditional, more spatially detailed risk assessments in te Pilippines. Te first conclusion of tis analysis is te close link between natural disasters and poverty in te Pilippines, and tis connection goes bot ways. First, natural disasters are a cause of poverty: on average, estimates suggest tat almost alf a million Filipinos per year face transient consumption poverty due to natural disasters. And in several regions trougout nortwestern Luzon, te number of individuals pused by disasters below subsistence level represents at least 20% of cronic subsistence incidence. In tese places, reducing risk is likely an efficient way of reducing poverty. But poverty also magnifies te impact of natural azards by making people more vulnerable and less resilient. Te bottom income quintile suffers only 9% of te national asset losses, but 31% of te total wellbeing losses. On average, te poorest quintile suffers from wellbeing losses tat are 1.5 times larger tan average individual loss in te country. As a result of te disproportionate impact on poor people, te

6 introduction 4 average annual wellbeing losses due to disasters in te Pilippines is estimated at US$3.9 billion per year (3.3% of ouseold expenditures), more tan double te asset losses of US$1.4 billion (1.3% of ouseold annual expenditures). A second conclusion is tat priority interventions bot in spatial terms (were to act?) and sectoral terms (ow to act?) are igly dependent on wic metric for disaster severity is used. Te most important interventions will focus around Manila if asset losses are used as te main measure of disaster impacts, wile regions like Bicol become priorities in terms of poverty incidence and wellbeing losses. And te least resilient region te one tat would struggle te most if it was affected by a disaster is te poorest, ARMM. Furter, one needs to use regional averages wit care: our results sow tat te poorest people in te ricest regions are almost as vulnerable as te poorest people in te poorest regions. An important consequence of tese findings is tat te coice of te metric used in risk assessments is not a tecnical question, but a political coice wit significant implications for wic interventions are desirable. Finally, te tird conclusion of tis work is tat new metrics of disaster impacts including poverty eadcount, poverty gap, and wellbeing losses can be used to quantify te value of interventions currently outside te traditional risk-management toolbox. Asset-informed risk-management strategies primarily focus on protection infrastructure, suc as dikes, and te position and condition of assets, for instance wit land-use plans or building norms. Wellbeing-informed strategies can utilize a wider set of available measures, suc as financial inclusion, private and public insurance, disaster-responsive social safety nets, macro-fiscal policies, and disaster preparedness and contingent planning. Even if tey do not reduce asset losses, tese measures can bolster communities socio-economic resilience, or teir capacity to cope wit and recover from asset losses wen tey occur, and reduce te wellbeing impact of natural disasters. Beyond tese policy conclusions, tis paper also describes for te first time an agent-based model developed to better understand te impacts of natural disasters on diverse ouseolds and teir pats to recovery. Its primary innovation is te use of te Family Income & Expenditure Survey (FIES) to disaggregate expected asset losses among representative ouseolds, resulting in a measurement of asset losses,

7 introduction 5 Figure 1: Traditional risk assessments evaluate asset exposure and vulnerability to azards to determine expected asset losses. Te Unbreakable model additionally incorporates te socio-economic resilience of te communities to predict wellbeing losses. poverty impacts, and wellbeing losses by income quintile and region in te country (or many oter possible categories, including gender, education level, and sector of employment). 1 Te analysis starts from azard- and asset-class-specific exceedance curves at te provincial level from te Government of te Pilippines Department of Finance Catastrope Risk Model (DFCRM)[9, 10]. Te first significant innovation of tis approac is to distribute tese losses among te representative ouseolds of te FIES. Tis step is based on teir asset vulnerability, estimated from available ouseold caracteristics (i.e., from ousing construction materials and condition). 2 Te second main innovation of te model is to explicitly represent disaster reconstruction dynamics at te ouseold level using an agent-based approac in wic (1) eac ouseold acts rationally to minimize its wellbeing losses, and (2) ouseolds interact troug firms activities and government budgets. Te model specifies eac ouseold s unique reconstruction and savings expenditure rate, assuming ouseolds optimize te fraction of income tey dedicate to repairing and replacing teir assets, at te expense of immediate consumption. For instance, people close to te 1 In geograpical terms, te current version of te FIES is representative only at te regional level. Te 2018 FIES will make it possible to perform te same analysis at te provincial level. 2 Unfortunately, te FIES does not currently include te geolocations of ouseolds.

8 asset losses 6 subsistence level cannot set aside muc of teir income to rebuild teir assets witout experience large wellbeing losses, and may terefore take longer to recover. In extreme cases, tey may even be trapped in poverty, generating large wellbeing losses going well beyond te few years tat follow a disaster [11, 12]. Wit tis approac, we are able to develop detailed country risk profiles tat incorporate sub-regional variations in azard, exposure, asset vulnerability, and socioeconomic resilience and wellbeing losses, to be used as an input in a prioritization process at as fine a scale as data allow. We can also demonstrate te benefits of formal and informal risk saring mecanisms, including post-disaster support and private remittances, in terms of increased resilience and reduced impacts of natural disasters on lives and livelioods in te Pilippines. Te paper is organized as follows. Section 2 provides an overview of asset risk due to wind, precipitation floods, storm surge, and eartquake events in eac of te 17 regions. In Section 3, we overlay tis information wit ouseold-level income and asset vulnerability data from te 2015 Family Income and Expenditures Survey (FIES) to develop finely-grained estimates of income and consumption losses to disasters. Based on tese results, we estimate te number of ouseolds in te Pilippines facing transient income or consumption poverty eac year due to natural disasters. In Sections 4 to 6, we quantify wellbeing risk at te national and regional levels, respectively. In Section 7, we examine specifically te risk to assets and risk to wellbeing of te poorest quintile in eac region. Section 8 describes policy simulations, and Section 9 presents conclusions. A tecnical appendix provides details on te metodology and data, and te equations of te freely-available model. 2 asset losses Typically, risk assessments incorporate information on te azards (te natural occurrence of destructive events suc as severe winds, surges, floods, and eartquakes); exposure (te value of natural and built assets tat migt face a destructive event); and vulnerability (te expected consequences to exposed assets wen a destructive

9 asset losses 7 event occurs) of te targeted area. Togeter, tese tree dimensions describe average annual asset losses in te area of interest. Table 1 on te following page displays average annual asset losses, by azard type, for eac region in te Pilippines. 3 Already, we see tat disasters in NCR (metropolitan Manila) and in CALABARZON (region IVA) cost over US$300 million per year, on average. In NCR, tese losses are driven primarily by precipitation flooding, wile wind events are responsible for over 50% of annual losses in CALABARZON. Significant differences across regions and disaster type are explained by variations in te azard, exposure, and vulnerability of eac part of te country. For example, wile catastropic typoons can strike any part of te country, te norteastern coast of Luzon (regions II, IVA, and V) and te Eastern Visayas (region VIII) are te most frequently affected tat is, tese regions face elevated typoon azard. Furter, all te regions facing asset risk in excess of US$100 million are in mainland Luzon, te wealtiest and most developed part of te country (elevated exposure). By contrast, asset risk in te wole of Mindanao (regions IX, X, XI, XII, XIII, and ARMM) accounts for just 5% of total losses due to te relative infrequency of major events (low azard) and ig poverty in tese regions (low exposure, ig vulnerability). Figure 2 on page 9 maps multiazard asset risk in te Pilippines. On te left, we represent losses in millions of dollars (as in te rigtmost column in Table 1). On te rigt, te same results are presented as a percentage of eac region s AHI. Generally, te same regions (i.e., metropolitan Manila and surrounding parts of Luzon) are igligted in bot representations of asset risk. However, in te map on te rigt, Cagayan Valley and Bicol (regions II and V, respectively) ave replaced NCR and CALABARZON as te most eavily-impacted regions. Te simple comparison in Figure 2 illustrates an essential point: different metrics can lead to different disaster risk otspots" and, terefore, priorities. At a mini- 3 Compreensive asset risk is a direct output of te AIR catastrope model [9, 13], but tis analysis is particularly interested in te impacts of disasters on ouseold assets, consumption, and welfare. One issue we face is te difference between te Gross Domestic Income (GDI) derived from national accounts and te aggregated ouseold income (AHI) calculated from ouseold surveys (in tis case, FIES). As is well known, te latter tend to report lower incomes tan national accounts [14]. In tis analysis, we work based on te AHI. Terefore, te expected losses in Table 1 ave been scaled by AHI as a fraction of te nominal regional productivity (GRDP), a factor of 0.43 on average (cf. Tab. 6).

10 asset losses 8 Asset losses [mus$ per year] Region EQ HU PF SS All azards NCR IVA - CALABARZON III - Central Luzon V - Bicol I - Ilocos II - Cagayan Valley VIII - Eastern Visayas VII - Central Visayas VI - Western Visayas IVB - MIMAROPA CAR XI - Davao XIII - Caraga XII - SOCCSKSARGEN X - Nortern Mindanao ARMM IX - Zamboanga Peninsula National total ,443.3 Table 1: Expected annual asset losses in millions of US$ from eartquakes (EQ), urricanes (HU), precipitation floods (PF), and storm surges (SS) for eac region. mum, tis suggests tat asset risk does not give a complete picture of disaster impacts in te Pilippines. Bot maps in Figure 2 describe mostly wat appens to te wealtiest regions and wealtiest people in tese regions because tey are based on aggregate loss data, and wealty regions and people ave te most to lose. Altoug asset losses in ARMM total just US$5 million per year, disaster responses (and, terefore, disaster risk strategies) sould in some way account for te 54% poverty rate in ARMM (versus 4% in NCR). Clearly, a more spatially disaggregated approac will be required. But moreover, we need to understand and develop metrics for tose dimensions of disaster impacts tat accrue to te poor. To tese ends, te rest of tis analysis goes deeper in te distributional analysis, moving from te regional scale to te ouseold level. In te next section, we merge regional asset loss data wit ouseold-level socioeconomic caracteristics to examine disaster impacts on ouseold income and consumption. Tis novel approac will allow us to develop estimates of te number of individuals pused into transient poverty eac year by disasters.

11 income, consumption, and poverty Figure 2: Total risk to assets (expected annual losses) from eartquakes, urricanes, precipitation floods, and storm surges for eac of te 17 regions of te Pilippines. At left, annual expected asset losses are expressed in US$. At rigt, losses are sown as a percentage of regional AHI. 3 income, consumption, and poverty Te tecnical details of asset loss disaggregation to te ouseold level are discussed in te tecnical appendix to tis report. In tis section, we present te main insigts generated by te union of te DFCRM wit te FIES, focusing on ow individual ouseolds socioeconomic caracteristics can mitigate or magnify te impact of disasters. 9

12 income, consumption, and poverty 10 Income losses Wen disasters damage or destroy te assets on wic individuals rely for teir liveliood including not only teir own sop or field, but also somebody else s factory affected ouseolds face income losses. Some may receive extraordinary public assistance or additional remittances, wic supplement teir regular income wile tey rebuild. Insofar as tey incorporate some of te socioeconomic caracteristics tat influence ouseolds recovery patways and describe better tan asset losses te real impact on wellbeing, net income losses are a useful metric of disaster impacts. To provide a istorical example: wen Super Typoon Yolanda made landfall in te Eastern Visayas (region VIII), it caused an estimated US$1.4 billion in damages to tat region alone. In terms of asset losses, ten, Yolanda was a rougly 100-year typoon (including damage from wind, storm surge, and precipitation flooding) in te region. Te top panel of Figure 3 on te following page illustrates te expected impact of a Yolanda-like wind event on individual incomes in te Eastern Visayas region. Te black outline indicates te regional income distribution as reported in FIES, wile te red istogram illustrates te expected income distribution immediately following a Yolanda-like event in te region. Tis distribution sows a mode around US$350 per person, per year, wit nearly 40% of te population living below te poverty line. Te large bar on te rigt sows te number of people wit incomes iger tan US$2,500 per year; te poverty and subsistence lines are indicated by te dotted lines around US$350 and US$450 per year, respectively. For tis Yolanda-like event, te wind destruction alone is expected to pus over 160,000 individuals into income poverty in Eastern Visayas, and over 170,000 below te subsistence income level (4% of te regional population).

13 income, consumption, and poverty year urricane in VIII - Eastern Visayas Subsistence line Increase of 172,700 (3.8% of regional pop.) in income subsistence Poverty line Increase of 160,300 (3.5% of regional pop.) in income poverty Population (,000) Pre-disaster income (FIES data) Post-disaster income (modeled) Income [USD per person, per year] Subsistence line Increase of 231,300 (5.1% of regional pop.) in consumption subsistence Poverty line Increase of 176,800 (3.9% of regional pop.) in consumption poverty Population (,000) Pre-disaster consumption (FIES data) Post-disaster consumption (modeled) Consumption [USD per person, per year] Figure 3: Expected impact of Yolanda-like (100-year) urricane on per capita income (top) and consumption (bottom) in te Eastern Visayas (region VIII). Income losses take into account te lost productivity of disaster-affected assets, wile consumption losses additionally include reconstruction costs, post-disaster support, and savings.

14 income, consumption, and poverty 12 Consumption losses Income losses provide new insigt into disaster impacts and are closer tan asset losses to te actual impact on people. However, tey still do not take into account a range of caracteristics and coping mecanisms tat can mitigate or exacerbate te effects of disasters on individual ouseolds. For example, we ave not yet accounted for te reconstruction costs tat directly-affected (and even tose indirectly affected) ouseolds must pay to rebuild teir assets after a disaster. Furter, many ouseolds ave some amount of savings to be used in case of a disaster, and wealtier ouseolds may benefit from formal and informal post-disaster transfers [15, 16]. Tese costs and resources all impact ouseolds consumption losses, or teir (in)ability to maintain consumption wen teir income drops. Te red istogram in te bottom panel of Figure 3 represents te ouseold consumption distribution in te Eastern Visayas region immediately after a Yolanda-like urricane in te region. After accounting for reconstruction costs and precautionary savings, te event is expected to generate a net increase of 177,000 individuals wit consumption at or below te poverty line (4% of te regional population), and over 230,000 wit consumption at or below te subsistence line (cf. Fig. 3). At te oter end of te distribution, we note tat tere is no longer a discernible difference between pre- and post-disaster consumption for ouseolds wose pre-disaster income is at least US$2,500 (compare to income losses). Tis suggests tat te poor struggle to cope wit lost income, wile te wealty are able to use savings and oter instruments to maintain teir consumption, even wen tey are affected by large disasters. Multiazard risk and cronic poverty On average, inclusive of all azards and regions, we estimate tat almost alf a million Filipinos per year face transient consumption poverty due to natural disasters. Tis is equivalent to 2.2% of national poverty incidence, toug te sub-national results indicate significant regional variation (cf. Table 2 on te next page). For example, in several regions trougout nortwestern Luzon, including NCR, Cagayan Valley,

15 income, consumption, and poverty 13 CALABARZON, and Central Luzon, te number of individuals pused into subsistence represents at least 20% of subsistence incidence, as measured by FIES. Altoug we do not model economic growt or oter patways out of poverty, tese results indicate tat natural disasters are major drivers of extreme poverty in certain regions of te Pilippines. In tese areas, disaster risk management strategies can be a igly efficient tool to decrease poverty and subsistence incidence, eiter by reducing asset losses or by enancing ouseolds ability recover from disasters. Consumption poverty impacts of natural disasters Poverty Subsistence Increase % of regional Increase % of regional Region [tousands] incidence [tousands] incidence IVA - CALABARZON III - Central Luzon NCR V - Bicol II - Cagayan Valley I - Ilocos VII - Central Visayas VIII - Eastern Visayas VI - Western Visayas XII - SOCCSKSARGEN XIII - Caraga IVB - MIMAROPA CAR XI - Davao X - Nortern Mindanao ARMM IX - Zamboanga Peninsula Total Table 2: Annual, regional impacts of natural disasters on consumption poverty and subsistence incidence. Results are expressed in tousands of individuals, and as percentages of te FIES 2015 regional poverty and subsistence rates. Sifts are defined as te number of individuals in consumption poverty or subsistence immediately after a disaster occurs, less te pre-disaster eadcount. Poverty incidence is inclusive of subsistence. Figure 4 on te following page maps te annual consumption poverty impacts of multiazard exposure in eac region. Te map on te left indicates disaster-related poverty incidence as a percentage of te total population in eac region, wile te map on te rigt describes disaster-related poverty incidence relative to te FIES

16 income, consumption, and poverty Figure 4: Tese maps indicate natural disaster-related increases in te number of Filipinos expected to face poverty or subsistence for any duration eac year. At left, map colors indicate new poverty incidence as a percentage of regional population. At rigt, map colors indicate new poverty incidence as a percentage of regional poverty rate (inclusive of subsistence). estimate of cronic poverty in eac region. Tese relationsips provide additional metrics by wic to measure te uman impacts of natural disasters, and can elp policy makers to understand te link between natural disasters and poverty. 14

17 income, consumption, and poverty 15 Socioeconomic caracteristics and disaster recovery dynamics Asset losses (and te foregoing income poverty analysis) describe individual ouseolds status in te instant after a azard occurred, and provide useful insigts into ow governments and oter first responders sould target umanitarian relief. However, anoter important question is weter disaster-affected ouseolds become mired in poverty or acieve a speedy recovery. In contrast to asset losses, income and consumption losses can inform on te temporal dimensions of disaster impacts. Returning to te Yolanda-like event, Figure 5 on te next page maps te expected time to recover 90% of te assets destroyed wen a 100-year wind event strikes eac region. Unsurprisingly, te map indicates tat metropolitan Manila and surrounding parts of Luzon are te quickest to recover. In oter words, altoug disasters are more frequent and costlier in tese regions tan elsewere, te ig density of ouseolds and productive assets in nortern Luzon may elp disaster-affected individuals to recover on teir own. Wen major disasters occur outside tese areas, recovery is expected to proceed slowly if at all, and disaster strategies seeking to minimize recovery times sould focus on providing post-disaster support to tese areas. Notably, tis modeling result is largely independent of initial asset losses, expressing instead te overall socioeconomic resilience of eac region. Of te estimated alf million Filipinos pused into transient consumption poverty by disasters eac year, some 25,000 will still be in poverty 10 years later. In total, tis complex picture suggest tat te costs of natural disasters are far-reacing for certain ouseolds, communities, and regions. Accounting for loss eterogeneity and ouseolds different abilities to reconstruct suggests terefore te existence of a long-term impact of disasters on income, wic may be difficult to detect in aggregate economic data because of te small income of te people suffering from tese long-term effects, but ave been documented after some large-scale disasters in oter countries [12, 11, 17, 18]. Tis eterogeneity could elp explain wy some studies ave found long-term impacts of disasters on growt [19, 20, 21], wile oters ave not [22, 23, 24, 25]. As discussed, te asset loss map in Figure 2 igligted te impacts of disasters on te ricest people. In contrast, metrics tat focus on poverty and reconstruction (e.g.,

18 income, consumption, and poverty 16 Figure 5: Map of time to recover 90% of assets destroyed in 100-year urricane. Figure 4 on page 14 and Figure 5) reveal ow poor Filipinos experience and cope wit socks, and are terefore essential inputs to anti-poverty development policies. Still, poverty eadcounts do not provide insigt into ow ouseolds already in poverty nor tose wose income does not drop below te poverty line are impacted by disasters. Tis is an important limitation, as te foregoing discussion is meant to be as inclusive and integrative as possible, witout discounting disaster-related costs to te wellbeing of any ouseolds. Tis is wy we now introduce te concept of

19 risk to wellbeing at te national level 17 wellbeing losses, wic can be used to measure disaster impacts on all ouseolds, witout creating a bias toward te ricer ones. 4 risk to wellbeing at te national level Te foregoing as sown tat, even if disaster-affected ouseolds suffer identical asset losses, teir consumption losses and recovery time will vary according to teir socioeconomic status and te resources available to tem (for instance because insurance, savings, remittances, and public support elp some ouseolds to smoot te consumption socks). Going furter, $1 in consumption losses can ave very different consequences for individual ouseolds, depending on teir income. In particular, wile ric ouseolds will be able to spend down teir savings or cut on luxury consumption, poorer ouseolds will often ave to cut on basic needs and essential consumption like food, ealt, or education, treatening teir ealt, uman capital, and long-term prospects. Disaster risk strategies and budgets sould account for tese differences, and wellbeing losses do precisely tis. Wile $1 in asset or consumption losses affects a poor individual more tan a ric one, wellbeing losses are defined suc tat a $1 wellbeing loss affects a ric and a poor individual equally. Wellbeing losses are calculated from consumption losses using a classical welfare function. Tis operation translates into wellbeing te value of a ouseold s consumption at eac point in its unique recovery, wit decreasing returns to represent te fact tat increasing consumption by $1 increases more te wellbeing of a poor individual (compared wit a ric person). 4 Te difference in te wellbeing generated by $1 of consumption is a simple proxy for te continuum from survival consumption (te very first units of consumption tat ave te largest impact on wellbeing) to luxury consumption (wic enances wellbeing less and less). 4 Here, te unit of analysis is te ouseolds. It would be useful to do it at te individual level, to uncover intra-ouseold distributional effects, linked to gender or age. In te absence of witinouseold data, we make te strong assumption tat pre-disaster consumption and disaster losses are distributed equally per capita witin eac ouseold.

20 socioeconomic resilience to disasters 18 Wellbeing losses integrate eac ouseold s consumption losses over te duration of its recovery, and give more weigt to te consumption losses experienced by poor people tan to te losses experienced by ricer people. In tis way, wellbeing losses account for socioeconomic differences among ouseolds, and correct for te prowealty bias inerent in asset losses witout relying on binary tresolds like te poverty line. As a metric, tey capture more fully te costs of disasters and te benefits of prospective DRM investments tan do asset losses. Terefore, wellbeinginformed strategies are not merely more equitable, but more cost-effective tan assetinformed strategies. As discussed in Section 2 on page 6, we estimate annual asset losses to ouseolds from natural azards (inclusive of urricanes, precipitation floods, storm surge, and eartquakes) in te Pilippines at over US$1.4 billion (P = 72 billion, cf. Table 3 on page 21), equivalent to 1.3% of ouseold annual expenditures. Wellbeing losses are muc iger, at US$3.9 billion (P = 193 billion; 3.3% of ouseold expenditures) per year. In oter words, te real impact of disasters in te Pilippines is equivalent to a decrease in national consumption by almost US$4 billion. 5 5 socioeconomic resilience to disasters Te analysis as moved from asset to income, consumption, and wellbeing losses, incorporating additional relevant ouseold-level socioeconomic caracteristics at eac stage. To summarize tese developments, we return to te traditional risk-assessment framework (cf. 1 on page 5). Te tree traditional components (i.e., azard, exposure, and vulnerability) predict asset losses, but not te capacity of affected populations to cope wit and recover from tese losses. In response to tis teoretical and practical sortcoming, we ave effectively added a fourt component, called socioeconomic resilience, wic we now define as te ratio of expected asset losses to wellbeing losses. 5 More precisely, te impact on wellbeing is equivalent to a = P 193 billion decrease in consumption tat would be optimally sared across ouseolds in te country and across time. Or it is equivalent to a P= 193 billion decrease in consumption tat would be equally sared across ouseolds in te country, if all ouseolds ad te same income (i.e. if all inequality ad disappeared).

21 socioeconomic resilience to disasters 19 At te national level, expected asset and wellbeing losses total US$1.4 billion and US$3.9 billion, respectively. Terefore, te wellbeing impact of disaster losses in te Pilippines is 170% ( = 2.7) iger tan asset losses. On average, every $1 in asset losses is equivalent to a $2.70 consumption loss, as experienced by a ouseold earning te national average income. Alternatively, we can say tat te socioeconomic resilience of te Pilippines is 37%. 6 Altoug tis is an informative result, te national average of socioeconomic resilience belies significant regional variation in te capacity of Filipino ouseolds to cope wit and recover from disasters. Figure 6 on te following page maps socioeconomic resilience to disasters at te regional level. Among all regions, metropolitan Manila enjoys te igest resilience (61%). Tis is due not just to its overall wealt and poverty rate of just 4%, but also to its ig degree of financial inclusion and social protection coverage (30% iger tan te national average). On te oter and, despite tese advantages, wellbeing losses to disasters in metropolitan Manila are still 64% ( = 1.64) iger tan asset losses, and most of tese losses still accrue to te poorest residents of te capital region. On te oter end of te resilience ranking, ARMM as te lowest resilience among all regions (15%). Low annual asset losses in ARMM indicate tat major events are rare (low azard) and tat te total value of assets in te region is relatively small (low exposure), but resilience is low, and additional policies and measures sould be implemented to elp tis fragile region cope wen disasters inevitably occur. In addition to tese diagnostics, te extension of te traditional azard-exposurevulnerability framework to include socioeconomic resilience as te benefit of expanding te disaster risk management "toolbox." Traditional risk-management strategies seek to mitigate azards as well as te exposure and vulnerability of assets. Oter tools (e.g., post-disaster support, financial inclusion, private and public insurance, and sovereign contingent credit) are not easily included in risk assessments 6 Tis estimate is significantly lower tan tat from te Unbreakable report, and cannot be directly compared, considering te difference between te models. Te difference is explained by te improved consideration of distributional impacts (using te full survey instead of only two categories of ouseolds) and te explicit representation of te reconstruction patway. Te difference confirms te need to model te reconstruction patway in a dynamic manner and to include te impact of sort-term consumption drops.

22 socioeconomic resilience to disasters 20 Figure 6: Regional map of multiazard socioeconomic resilience to disasters. because tey do not affect asset losses. In a wellbeing-informed framework, owever, te benefits of tese interventions become obvious and quantifiable: to te degree tat tey allow ouseolds to maintain a ealty degree of consumption wile tey rebuild, tese interventions increase socioeconomic resilience and reduce wellbeing losses to disasters.

23 regional risk summary 21 6 regional risk summary Te annual risk to assets, risk to wellbeing, and socioeconomic resilience for eac region of te Pilippines are summarized in Table 3. In absolute terms, risk to ouseold assets is igest in metropolitan Manila (NCR) and in CALABARZON. Asset losses in tese two regions total over US$610M (P = 30 billion) per year, over 40% of te expected losses in te entire country. Te asset and population density in tese areas make risk management interventions effective, but also expensive. Socio-economic Asset losses resilience Wellbeing losses Region [mus$/year] [%] [mus$/year] IVA - CALABARZON V - Bicol III - Central Luzon NCR I - Ilocos II - Cagayan Valley VIII - Eastern Visayas VII - Central Visayas VI - Western Visayas IVB - MIMAROPA XIII - Caraga XI - Davao CAR XII - SOCCSKSARGEN ARMM X - Nortern Mindanao IX - Zamboanga Peninsula Total 1, ,879.2 Table 3: Asset losses, socio-economic resilience, and wellbeing losses for te entire population of eac region of te Pilippines. Asset and wellbeing losses are denominated in USD, and socio-economic resilience is te ratio of asset losses to wellbeing losses in eac region. Wen disaster impacts are measured in wellbeing losses, an alternative set of regional priorities emerges. In particular, Table 3 indicates tat Bicol and CALABAR- ZON suffer te greatest losses eac year. In absolute terms, CALABARZON suffers te igest wellbeing losses, surpassing US$665 million (P = 31 billion) per year, equivalent to 3.4% of AHI in te region. Tese losses are due to elevated exposure to

24 poorest quintile risk summary 22 urricanes and storm surges (51% and 22% of annual regional asset losses, respectively), wic result in major wellbeing losses despite an above average value for socioeconomic resilience (45%) and a low regional poverty rate (just 9%). In Bicol, were asset losses are muc lower, wellbeing losses are driven by a very ig cronic poverty rate (36%), low ouseold savings (60% of te national average, across all deciles), and low social transfer receipts (75% of national average). Overall, wellbeing losses are less geograpically concentrated tan asset losses, and tis result is indicative of te magnitude of te callenge facing disaster managers in te Pilippines. At te same time, it signifies new opportunities to mitigate disaster risk and cronic poverty by investing in socioeconomic resilience outside te most developed parts of Luzon. 7 poorest quintile risk summary Because our results are based on information at te ouseold level, we are able to assess asset and wellbeing losses not just at te regional level, but also for different income groups. 7 Table 4 on page 24 lists annual asset and wellbeing losses for te poorest quintile in eac region. Across all regions, te asset losses of te poorest 20% are US$125 million (P = 6.3 billion per year), or just 9% of total asset losses. On te oter and, te wellbeing losses of te poorest quintile give a better sense of teir experience of disasters: te wellbeing losses of te poorest quintile are valued at US$1.2 billion (P = 61 billion) per year, or 31% of total wellbeing losses. Tis means tat, on average, individuals in te poorest quintile suffer wellbeing losses tat are 50% larger tan te average individual loss in te country. Critically, Table 4 sows tat te regional variations in socioeconomic resilience decreases significantly wen we narrow our focus to te poorest Filipinos. Wile NCR as te igest overall resilience (61%), te poorest ouseolds in te capital ave a resilience of only 17%. To wit: on average, te poorest residents of NCR lose 7 It is also possible to look at different subgroups in te country or at te regional scale (e.g., per occupation, ead of ouseold gender, ouseold size, etnic background or religion, social transfer enrollees), witin te limits of te representativeness of te ouseold survey.

25 poorest quintile risk summary Figure 7: Expected (average annual) wellbeing losses from eartquakes, urricanes, precipitation floods, and storm surges for eac region of te Pilippines. Losses are sown in US$ (left) and as a fraction of regional aggregate ouseold income (AHI, at rigt). US$9.50 per person, per year to natural disasters, but tis equates to nearly US$60 per person and per year in wellbeing losses (cf. Table 5 on page 25). In terms of disaster impacts and recovery prospects, tis result suggests tat te poorest Manileños ave more in common wit te poor in oter regions tan wit teir wealtier neigbors. Tis is an important caveat to regional-scale assessments: te aggregate wealt of te capital region does not imply tat its poorest residents are well-protected from or resilient to natural disasters. To te contrary, te low resilience of te poor to asset losses, combined wit te large contribution of disasters to poverty incidence in 23

26 policy simulations 24 Socio-economic Asset losses Resilience Wellbeing losses Region [mus$/year] [% of total] [%] [mus$/year] [% total] IVA - CALABARZON V - Bicol III - Central Luzon NCR I - Ilocos II - Cagayan Valley VIII - Eastern Visayas VII - Central Visayas VI - Western Visayas IVB - MIMAROPA XIII - Caraga XI - Davao XII - SOCCSKSARGEN CAR X - Nortern Mindanao ARMM IX - Zamboanga Peninsula Total Table 4: Natural disaster impact on te poor: risk to assets, socio-economic resilience, and risk to wellbeing for te poorest quintile (20%) in eac region. Manila (cf. Tab. 2 on page 13), suggests tat interventions to reduce asset losses and build resilience of te poor could be very effective for reducing disaster exposure and poverty incidence, particularly in Manila and oter wealty regions. 8 policy simulations Te teoretical framework and model presented ere are not limited to disaster risk diagnostics. Tey also allow for detailed analyses of te costs and benefits of diverse instruments and investments to reduce risks, or make te population better able to deal wit tem. Wile disaster response tools suc as post-disaster cas transfers do not affect asset losses, teir benefits can be measured by multiple poverty metrics (e.g., poverty eadcount, poverty gap), reconstruction time for various ouseolds, and wellbeing losses. And since te model estimates losses at te ouseold level, te

27 policy simulations 25 Region Per capita losses Q1 population Asset Well-being [tousands] [US$ per cap., per year] V - Bicol 1, II - Cagayan Valley I - Ilocos 1, VIII - Eastern Visayas IVA - CALABARZON 2, III - Central Luzon 2, Total 20, NCR 2, VII - Central Visayas 1, XIII - Caraga IVB - MIMAROPA VI - Western Visayas 1, CAR XI - Davao XII - SOCCSKSARGEN ARMM X - Nortern Mindanao IX - Zamboanga Peninsula Table 5: Per capita risk to assets and wellbeing for te poorest quintile (20%) of eac region. model also makes it possible to examine te distribution of tese costs and benefits trougout te population. Post-disaster support in te Pilippines In te Pilippines, te Department of Social Welfare and Development (DSWD) is te lead agency for disaster response witin te government s National Disaster Risk Reduction and Management Plan (NDRRMP). In response to Yolanda, DSWD implemented a variety of SP and social welfare programs: distribution of in-kind relief items, cas transfers (unconditional and conditional), selter, and community-driven development. 8 Initially, te empasis was on food and nonfood items (like mats, blankets, tarpaulins, ygiene kits, and cloting) to meet te immediate and urgent survival needs, plus temporary selter assistance for displaced ouseolds. 8 A more complete description of te response to Yolanda is provided in te Sock Waves report [26].

28 policy simulations 26 After immediate survival needs were addressed, DSWD delivered a number of cas-based response programs, suc as Cas for Work, Cas for Building Liveliood Assets, and cas for selter (Emergency Selter Assistance) ten transformed into te Core Selter Assistance Program to rebuild permanent ousing. DSWD also temporarily removed all conditionality of te Pantawid Pamiliya Pilipino Program (4Ps), a usually conditional cas transfer program. In addition, at least 45 international umanitarian agencies implemented cas transfers (unconditional and conditional), partly delivered troug te 4Ps infrastructure. Four agencies alone distributed around US$34 million, benefiting 1.4 million disaster-affected people. Modeling post-disaster support Here, we do not try to reproduce te very complex response to typoon Yolanda, but instead we assess te benefit from an idealized post-disaster support (PDS) provided to te population after a disaster. As an illustrative exercise, we consider a simple post-disaster support system. In tis system, all disaster-affected ouseolds receive a uniform cas payout, equal to 80% of te average asset losses suffered by te poorest quintile. Tis system ensures tat poor people are compensated for a large fraction of teir losses and assumes tat all affected ouseolds are supported, wile total costs remain acceptable. Te cost of te program is distributed among all ouseolds in all regions via a flat tax on income. Returning to te Yolanda-like urricane event: expected wind damage to ouseold assets in te Eastern Visayas region is valued at US$633 million. Wellbeing losses from te same event are valued at US$2,176 million, for a socioeconomic resilience of 29%. 9 In Figure 8 on page 28, we plot per capita asset and wellbeing losses, grouped by income quintile. Te figure sows tat te ricest ouseolds lose te most assets, wile te poorest ouseolds suffer te greatest wellbeing losses. 9 Here, we report expected asset losses from te 100-year wind event in te Eastern Visayas (a product of te DFCRM, scaled to matc AHI), and tis represents 45% of te reported US$1.4 billion in losses to Yolanda in te Eastern Visayas. Te total value includes damage from precipitation flooding and storm surge (not included in tis example), as well as te assets tat account for te difference between AHI and nominal GRDP.

29 policy simulations 27 In response to tis disaster in tis location, te post-disaster program disburses a total of US$187 million (P = 9.4 billion), distributed uniformly among all affected ouseolds. Note tat te flat tax payment mecanism used ere effects a net transfer from te top quintile to te bottom four; progressive taxation, social insurance-like systems, and more complex alternatives can also be modeled. Te two clusters on te rigt in Figure 8 sow ow post-disaster support would reduce wellbeing losses (witout aving any effect on asset losses), especially for te poorest ouseolds. Te first quintile sees its wellbeing losses alved, wile te impact on te ricest quintile is small. In total, post-disaster support reduces wellbeing losses to US$1,265 million, a 42% decrease relative to te nominal simulation. Because post-disaster support does not impact asset losses, suc programs cannot be subjected to traditional cost-benefit analyses. However, cas transfers do increase te socioeconomic resilience of te region to 50%, and te wellbeing-informed approac projects a benefit-to-cost ratio of 4.9 for tis intervention. More detailed analysis would allow te costs and benefits of post-disaster support to be optimized, including realistic limitations on budgeting, targeting, and delivery. 10 On an annual basis, te cost of te post-disaster program is US$472 million, equivalent to 32% of expected annual losses to all azards. Te program is expected to reduce wellbeing losses from all disasters by 17%, or US$598 million (P = 30 billion), acieving a benefit-cost ratio of 1.3 on average. Tis benefit-cost ratio is lower tan in te previous example because, for any individual event, te benefit-cost ratio depends on wo is affected. Wen a very ric area is affected, te system may redistribute resources from poor (but non-disaster-affected) people to ricer (but disasteraffected) people, wic reduces te benefit-cost ratio. Tis system is provided just as an example te cost of a post-disaster support package could be significantly re- 10 It is important to note tat even if te amount of post-disaster support is equal to asset losses, it does not fully cancel wellbeing losses: indeed, post-disaster support maintains consumption, but consumption losses are larger tan asset losses. Tis result is consistent wit intuition: even if people are immediately given in cas te cost of rebuilding teir ouses and replacing teir assets, tey would still experience wellbeing losses during te reconstruction period, since assets and ouses cannot be replaced instantaneously.