1 Vulnerability to Tropical Cyclones: The Case of the Philippines Rio Yonson School of Economics and Finance Victoria Business School Victoria University of Wellington, New Zealand [email protected]As of May 2015 Abstract The study is apparently the first subnational empirical work using econometric approach to deduce the aspects of development that influence people’s vulnerability to disasters. The study focuses on tropical cyclones, and uses as case study the Philippines. A new provincial level panel dataset is constructed from separate sources, and we use panel data estimation methods to determine the effects of socioeconomic, as well as time-invariant geographic control variables. Results show evidence that the level of socioeconomic development provides protection and builds human capacities, thereby reducing vulnerability and disaster impacts. Rapid and unplanned urbanization increases people’s vulnerability and exposure to harm. The quality of local development governance can significantly alter the gravity of disaster impacts on people. Likewise, we find that geography is an important determinant of disaster impacts. ___________________ This paper comprises the first chapter of the author’s PhD dissertation, which is currently being prepared under the supervision of Professor Ilan Noy, Chair in the Economics of Disasters of Victoria University of Wellington, and of Associate Professor Jean-Christophe Gaillard of the School of Environment, The University of Auckland.
Transcript
1
Vulnerability to Tropical Cyclones: The Case of the Philippines
Rio Yonson School of Economics and Finance Victoria Business School Victoria University of Wellington, New Zealand [email protected]
As of May 2015
Abstract
The study is apparently the first subnational empirical work using econometric approach to deduce the aspects of development that influence people’s vulnerability to disasters. The study focuses on tropical cyclones, and uses as case study the Philippines. A new provincial level panel dataset is constructed from separate sources, and we use panel data estimation methods to determine the effects of socioeconomic, as well as time-invariant geographic control variables. Results show evidence that the level of socioeconomic development provides protection and builds human capacities, thereby reducing vulnerability and disaster impacts. Rapid and unplanned urbanization increases people’s vulnerability and exposure to harm. The quality of local development governance can significantly alter the gravity of disaster impacts on people. Likewise, we find that geography is an important determinant of disaster impacts.
___________________
This paper comprises the first chapter of the author’s PhD dissertation, which is currently being prepared under the supervision of Professor Ilan Noy, Chair in the Economics of Disasters of Victoria University of Wellington, and of Associate Professor Jean-Christophe Gaillard of the School of Environment, The University of Auckland.
1
I. Introduction
We make a new contribution to the strand of statistical empirical literature that establishes the
influence of development on disaster impacts. These studies seek to explain the direct disaster
impacts using socioeconomic variables that represent the vulnerability of the exposed population
and assets1. These existing works are all inter-country; one of their profound practical significance is
as inputs for evidence-based decisions, largely by the international disaster risk and reduction
community2.
We complement these inter-country work with a subnational level of study3. We use a deductive
approach, using econometric methods, to determine the factors that made people vulnerable to
disasters using a risk model.4 Several inductive subnational work have been undertaken but as
Pelling (2006) points out, a deductive approach “adds realism to the analysis” compared to an
inductive approach which are “not empirically verifiable against specific disaster-related outcomes”.
The importance of a subnational study lies in the production of “outputs that are meaningful for
development action”, particularly if the research outputs “are to contribute to development
planning” (Pelling, 2006) since specific key decisions on vulnerability and risk reduction are made
and implemented within a country (Birkmann, 2007). A subnational assessment will be of practical
usefulness for area-specific disaster mitigation planning (Peduzzi et al., 2009), given that risk and
vulnerability are very different across areas within countries. Moreover, subnational assessment
could potentially input towards translating internationally-agreed priorities for action, including
1 See Cavallo and Noy (2011) for a detailed account on the existing relevant research. 2 For instance, the work by Peduzzi, Dao, Herold, and Mouton (2009) on the Disaster Risk Index is undertaken for the UNDP to systematically analyse the linkage of vulnerability to development. The DRI is a global index for establishing the relative human vulnerability across countries that serve as guide to governments and aid organizations (Peduzzi et al., 2009) 3 Noy and Vu (2010) also undertook a subnational level study on disaster and development, using econometric approach, but with a different nature of inquiry. Specifically, the study seeks to estimate the short-run impacts of disasters on the macroeconomy. 4 In Section III, we present subnational measures of vulnerability. However, unlike our approach, they employ inductive approach.
2
those in the recently adopted Sendai Framework for Disaster Risk Reduction 2015-2030, into
concrete country-specific actions.
We aim to answer the following broad question: What aspects of development at the subnational
level influence people’s vulnerability to disasters? For this paper, we focus on tropical cyclone
hazards and the level of analysis is provincial. We hypothesize that the level of socioeconomic
development, the characteristics of urbanization, and quality of local development governance
largely shape human vulnerability and the risk from disasters.
There have been subnational studies from various disciplines along this line of inquiry; they either
focus on a specific disaster event or undertake comparative analyses of few disasters events
(Florano, 2014; Gaillard, Liamzon, & Villanueva, 2007; Ginnetti et al., 2013). We adopt a more
general approach by looking at experiences across provinces in all tropical cyclones that occurred
during recent times. We construct a new provincial level panel dataset from separate sources on
disaster impacts on people, indicators of tropical cyclone strength, and indicators relevant to our
hypotheses.
We note that the theoretical literature offers numerous definitions of vulnerability in the context of
natural hazards, but despite the myriad of frameworks, a consensus has yet to be reached. For the
purpose of this study, we refer to factors influencing peoples’ vulnerability as those economic,
social, political, physical, and environmental factors that increase or reduce their capacity to
withstand the adverse impacts of natural hazards. This is a simplified adaptation of the selected
existing definitions of vulnerability (Blaikie, Cannon, Davis, & Wisner, 1994; Bohle, 2001; Cardona et
et al., 2009; Raschky, 2008; Toya & Skidmore, 2007) and those on the Philippines (Gaillard, 2008;
Ginnetti et al., 2013)
The paper is organized as follows: Section II provides a background on tropical cyclone-related
disasters and development in the Philippines. Among others, we initially explore the aspects of
development that influence vulnerability using descriptive statistics and existing accounts on
experiences with tropical cyclone-induced disasters. Section III briefly presents selected related
work across disciplines, and identifies the gap in research we aim to fill. Section IV presents the risk
model adopted and translated for use in our retrospective assessment, estimation method, and data
used. Section V presents our results and findings, while Section VI provides general conclusions and
next steps.
II. Background on Tropical Cyclones Disasters and Philippine Development
The Philippines is an archipelago located in Southeast Asia comprising of 7,107 islands that are
grouped into three major island groups, namely: Luzon, Visayas and Mindanao (Map 1). Natural
hazards are bound to occur in the country given its geographic, geologic and meteorological setting.
It is located within the Pacific Ring of Fire, as well as along the typhoon belt in the Pacific. As of
5
2013, the country has 81 provinces, a population of over 92 million as of the 2010 Census, and
population density of 308 per square kilometre (PSA, 2012).
In recent years, the Philippines has been given credit as one of the leaders in passing quality
legislations on disaster risk reduction and climate change adaptation, specifically the Climate Change
Act in 2009 and the Disaster Risk Reduction and Management Act of the Philippines in 2010.6 Even
before the corresponding institutional mechanisms were completely put in place, these laws were
put to test as the country was hit by highly lethal tropical cyclones. In 2013, Typhoon Haiyan left a
staggering trail of 6,092 deaths, while in 2012 and in 2011, Typhoon Bopha and Tropical Storm7
Washi claimed 1,248 and 1,258 lives, respectively (NDRRMC, 2014). These three tropical cyclones
were not only the most lethal in the Philippines, but also in the world during the years 2011-2013
(Guha-Sapir, Hoyois, & Below, 2012, 2013, 2014). Moreover, these tropical cyclones posted the
most costly disaster events in the Philippines in the said years (NDRRMC, 2014).
6These laws are “often in advance of so many European countries” (Shepherd et al., 2013). The Special Representative of the UN Secretary-General on DRR has been quoted as saying that these laws are the “best in the world” and indicate a shift from a reactive to a proactive approach in addressing disasters (Ginnetti et al., 2013). 7 In the Philippines, a typhoon is a tropical cyclone with a maximum wind speed of above 118 kilometres per hour (kph), while a tropical storm has a maximum wind speed of 64 kph to 118 kph. A third classification is tropical depression, which has a maximum wind speed of 63 kph (PAGASA, undated). It will be noted, however, that typhoon is used in a general sense to refer to tropical cyclones over the West Pacific Ocean, and which is equivalent to cyclone over the Indian Ocean, and hurricane over the East Pacific and Atlantic Ocean (PIDS, 2005).
6
Map 1. Administrative Map of the Philippines
7
As can be seen from Table 1, a total of 652 tropical cyclones entered the Philippines for the period
1980 – 2013 (PAGASA, 2014). About half of these are reported by the National Disaster Risk
Reduction and Coordinating Council (NDRRMC) as destructive having had adverse impacts on the
population (in terms of fatalities, injuries, and disruption in typical daily activities) and on properties.
Tropical cyclones are the most lethal and destructive as it also usually triggers widespread flooding,
rain-induced landslides, and storm surges. Column 3 of Table 1 shows the annual number of
fatalities from tropical cyclones and associated hazards. The cumulative death toll from 1980 to
2013 reached over 30,000, while average annual fatalities is 885. On a per destructive tropical
cyclone basis, an average of 102 persons die.
Column 4 of Table 1 indicate that about 5 million persons are affected in a year, and over 570,000
are affected per destructive tropical cyclone. Column 5 indicates that costs of damage from tropical
cyclones are likewise massive. Annual average cost is USD355 million. Damage costs are highest in
2012 and 2013, which are mainly due to Typhoons Bopha and Haiyan, respectively. Average damage
per destructive event is USD41 million.
8
Table 1. Number of Tropical Cyclones and Impacts on Population and Properties, 1980-2013
Sources: Number of Tropical Cyclones that Passed the Philippines - Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA) Number of Destructive Tropical cyclones, Impacts of Tropical cyclones - National Disaster Risk Reduction and Management Council (NDRRMCC) Annual average exchange rates used to convert cost in PhP to USD taken from Bangko Sentral ng Pilipinas (Central Bank of the Philippines) website (BSP, 2014) Disaster impacts include those resulting from tropical cyclone-induced flooding, landslide, and storm surge.
Meanwhile, we find indications and accounts that several aspects of Philippine development
influence vulnerability and aggravate disaster impacts. These include persisting poverty, unplanned
urbanization, environmental degradation, and poor governance, among others (Gaillard et al., 2007;
9
Ginnetti et al., 2013; Shepherd et al., 2013). Despite the sustained high economic growth rate
shown by the Philippines in recent years, poverty reduction has not been as encouraging. The
country has recently been manifesting high levels of economic growth. In 2013, its 7.2% real GDP
growth rate was higher than most of its neighbouring countries and almost at par with that of China
(WB, 2014). However, as of 2012, poverty incidence among population in the Philippines stood at
25.2%, only 1.4 percentage points lower than that in 2006 (WB, 2014). While incidence declined
over the 2006-2012 period, the number of poor people increased by 1.1 million. There is great
variation across provinces, with poverty incidence in 2012 ranging from a low of only 3.4% to a high
73.8%. Over a third of the provinces have incidences of above 40%. A recent study estimates that
among poor households, at least half are classified as chronic poor (Bayudan-Dacuycuy & Lim, 2013).
Without adequate levels of income, the poor face the unfortunate circumstance of prioritising food
expenditures and underinvesting in other basic needs, such as safe shelter, education and health.
This is an unfortunate circumstance given that human capital (health and education) are important
in reducing vulnerability and increasing resilience (UNDP, 2004). Without access to land, poor
people crowd and build makeshift houses in informal settlements located in areas that are hazard
prone, and have to deal everyday with the communities’ poor living conditions (Ginnetti et al.,
2013).
In terms of urbanization, the rapid influx of people into the urban areas has reportedly outpaced the
creation of employment and provision of adequate services (WB-EASPR, 2003).8 These resulted to
high levels of urban poverty that transfigure to greater vulnerabilities, as well as greater hazard
exposure as poor communities expanded further in hazard prone areas (Gaillard, 2008; Gaillard et
8 The Philippine population grew at an average of 2.69% during the period 1950-2010, higher than the averages for South East Asia, the
whole of Asia and the World (UN, 2014). Urban population grew much faster, driven mainly by migration of people from rural areas. During the period 1950 – 1990, urban population grew at an annual average of 4.47%, also higher than the averages for South East Asia, the whole of Asia and the World (UN, 2014). Thereafter, urban annual population growth rate slowed down, ranging from 1.12% to 2.21% from 1990 to 2010.
10
al., 2007; Ginnetti et al., 2013; WB-EASPR, 2003), and without regard for existing policies (such as
land use plans, zoning ordinances, as well as water, forestry and building codes). The encroachment
of built-up areas in hazard prone areas has persistently been one of the prevalent land use conflicts
across provinces in the Philippines (Corpuz, 2013). Areas demarcated as hazard-prone are ironically
among those with densest human settlements. It will be noted, however, that in some cases, the
growing settlements in these areas are not always without legal permission. The consequences of
unplanned urbanization, along with the poor enforcement of land-use plans, zoning ordinances and
other pertinent policies and laws (such as water, forestry and building codes) combine together in
building up exposure and exacerbating vulnerability to disasters.
The country’s experiences with disasters reveal that governance can largely alter the impacts. In
2011, Tropical Storm Washi entered the Philippines with a maximum 24-hour rainfall volume of 180
mm and maximum wind speed of 75 kilometres per hour (PAGASA, 2014). The tropical storm made
several landfalls. It first hit one of the eastern coastal province in the Caraga Region, Surigao del Sur
(NEDA, 2012b). There was only one death recorded in the entire region (NDRRMC dataset, 2014).
Reportedly, relocation of residents started days before the typhoon.
With similar magnitude in terms of rainfall volume and wind speed, Tropical Storm Washi crossed
Northern Mindanao Region, which has always been known for its comparative advantage of being
tropical cyclone-free (NEDA, 2005). Historically, tropical cyclones pass the region once in every
twelve years and generally were not strong enough to create destruction (NEDA, 2005). While there
were advisories on its coming and expected strength of the tropical cyclone, there was no adequate
evacuation initiated by the local government units in areas expected to be exposed. Death toll in the
region was very high at 1,259 (NDDRMC dataset, 2014).
In Cagayan de Oro City, the regional centre of Northern Mindanao that is located within the
provincial boundary of Misamis Oriental, the aftermath of the disaster mirrored the kind of
governance that existed in the city for several years (Ginnetti et al., 2013). Majority of the recorded
11
provincial total of fatalities and affected persons of 698 and about 400,000, respectively, are from
Cagayan de Oro City (NDRRMC dataset, 2014). The rampaging waters carrying debris and mud
wiped out the burgeoning settlements along the banks of the Cagayan de Oro River, and particularly
on former riverbeds and sandbars that formed from silt (Ginnetti et al., 2013; NEDA, 2012b) .
A study on internal displacement due to Tropical Storm Washi accurately depicted the failure of
governance, particularly the grave negligence of local officials that lead to the high death toll during
the tropical storm in one of the hardest hit areas, Cagayan de Oro City. (Ginnetti et al., 2013). As
reported, settlements along the riverbank and sandbars grew as a result of the mayor’s housing
program that offered a token price of just a Philippine peso (about USD 0.02) to poor families for
them to have the right to build houses in said areas. With this outright infringement of existing land
use policies, coupled with the provision of inappropriate incentives by the city leadership, the
settlements continued to be densely populated. This is despite several recommendations from the
environment department to the local government unit to relocate the residents due to the high risk
from flooding. Despite the high death toll in the aftermath of Tropical Storm Washi, the mayor did
not implement the order of the president of the country to prevent people from returning to areas
that have been proclaimed as No Build Zones (Ginnetti et al., 2013). The report aptly puts it that the
presence of pertinent laws, including the DRRM Act of 2010, is largely not enough if it is not
accompanied by strong political will by authorities to have these laws implemented and complied
with (Ginnetti et al., 2013).
An earlier in-depth study to investigate the causes of death in a series of tropical cyclones in the
eastern part of Luzon in 2004 provide a compelling evidence of the political construct that led to
the disaster that took 1,400 lives in 2004 (Gaillard et al., 2007). The strength and the impact of the
tropical cyclone were magnified by deforestation in the affected areas. Illegal logging has
dramatically reduced the forest cover in the area, yet the cutting persisted because of “widespread
corruption, shortcomings and failures within the government” (Gaillard et al., 2007).
12
Meanwhile, in 2013, Typhoon Haiyan left an even greater destruction for most of the exposed
provinces. In a small island with 1,000 population, which lie between the main island of Cebu
province and the hardest hit province of Leyte, the catastrophic strength of the typhoon flattened all
houses (ABS-CBN, 2013). Yet all 1,000 population survived the onslaught of the typhoon; all it took
was good decision for the local government officials to evacuate all residents to a safe evacuation
site (ABS-CBN, 2013).
With climate change, more tropical cyclones of greater strength and frequency are likely to hit the
country (PAGASA, 2011). The huge historical annual losses of lives and properties depict the glaring
reality that the Philippines have yet to match the increasing strength of natural hazards with
heightened effectiveness in preparedness, and prevention and mitigation measures.
III. Disaster Risk and Vulnerability Frameworks
We briefly review a number of frameworks that have stood out in the literature across disciplines on
the nature of vulnerability in the context of natural hazards. Among others, this enables us to
objectively identify the suitable indicators for inclusion in our model. In addition, we examine the
empirical literature to determine the areas that have been studied, methodologies employed and,
more importantly, to identify the gap in research that we aim to fill.
A. Frameworks on Vulnerability and Disaster Risk Assessment
We briefly present among the more important disaster risk frameworks that appear to have largely
influenced the subsequent empirical works of the disaster risk practitioners, on one hand, and the
various disciplines in the academic community, on the other. The Pressure and Release (PAR)
framework, which was introduced in the book entitled, At Risk (Blaikie et al., 1994; Wisner, Blaikie,
Cannon, & Davis, 2004), provides an interesting depiction of how disasters come about when natural
hazard affects the vulnerable individual or group of people in a particular place at a given time. This
13
model adopts the straightforward equation that disaster risk is a product of hazard and vulnerability,
as follows:
𝑅𝑖𝑠𝑘 = 𝐻𝑎𝑧𝑎𝑟𝑑 𝑥 𝑉𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦 (1)
The proponents of this model put their definition of vulnerability quite clearly, and from a relatively
positive point of view. That is, vulnerability is defined as “the characteristics of a person or group in
terms of their capacity to anticipate, cope with, resist, and recover from the impact of a natural
hazard” (Blaikie et al., 1994). Vulnerable individuals are faced with two pressures; on one hand, is
pressure arising from their vulnerability, and on the other hand, is the pressure arising from the
occurrence of a hazard9.
The Progression of Vulnerability Framework further elaborates the PAR Framework (Wisner,
Gaillard, & Kelman, 2012). As in the PAR, it distinguishes among the three levels of progression of
vulnerability (Figure 1). The first level of the progression is “root causes,” which includes social and
economic structures that determine the distribution of resources, wealth, and power; ideologies in
governance; and, history and culture. An emphasis is made on the need to determine the historical
origin of these structures and to explain the underlying ideologies that give ground for the legitimacy
of these structures (Wisner et al., 2012). This implies that root causes may be distant in space and
time relative to the location of present vulnerability (Wisner et al., 2012).
9 As Birkmann (2006) assesses, the PAR model is one of best frameworks, the great value of which is coming from its usefulness in addressing the causative factors that contribute towards hazards becoming disasters. The approach emphasizes that the imperatives towards successfully reducing vulnerability and disaster risks are the adjustments to the existing economic and political systems, given that these are the underlying causes of rapid urbanization and population growth (Birkmann, 2006).
14
Figure 1. The Progression of Vulnerability Framework
Source: Wisner et al. (2012)
The second level of the progression comprises of “dynamic pressures” (Blaikie et al., 1994; Wisner et
al., 2012). These are grouped into the deficiencies of society’s social, economic and political
processes, and the macro-forces, such as rapid population growth and rapid urbanization,
deforestation, decline in soil productivity, among others. Accordingly, the dynamic pressures serve
as channels through which the root causes result to fragile livelihoods and unsafe locations, which
are the specific forms and manifestations of peoples’ vulnerability in a given time and space, vis-à-vis
a particular hazard (Blaikie et al., 1994; Wisner et al., 2012). We note the unlike in the PAR, the
Progression of Vulnerability Framework further elaborates on the multidimensional nature of
vulnerability.
The UNDP-UNDRO (UNDP-UNDRO, 1992) adds “elements at risk” to the earlier risk equation
(Equation 1), which effectively identifies who or what are at risk. Hence, disaster risk now comprises
three important components, namely: hazard, elements at risk, and vulnerability that need to be
15
quantified separately (UNDP-UNDRO, 1992). This risk framework has been adopted in prospective
or probabilistic disaster risk assessment methodologies in the following general form:
𝑅𝑖𝑠𝑘 = 𝐻𝑎𝑧𝑎𝑟𝑑 𝑥 𝐸𝑥𝑝𝑜𝑠𝑢𝑟𝑒 × 𝑉𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦 (2)
where, Risk is the expected number of fatalities and expected cost of damage per year; Hazard is
the probability of occurrence (expressed as reciprocal of the return period) of a hazard of a given
severity; Exposure, alternatively called as Elements at Risk, is the estimated number of people and
value of properties exposed to such hazard; and Vulnerability is the degree of loss, expressed from 0
to 100 percent, of the elements at risk to a hazard of given severity (NEDA, 2008; Peduzzi et al.,
2009; UNDP-DHA, 1994; UNISDR, 2013)10. We note that this is the framework adopted in the
The exposure sub-index refers to physical susceptibility. The indicators include population growth
rate, urban growth, population density, poor population, value of capital stock, share of net exports
to GDP, share of gross domestic investment to GDP, and share of arable land and permanent crops
11 Apart from the PVI, the system of indicators developed includes Disaster Deficit Index (DDI), the Local Disaster Index (LDI) and Risk Management Index (RMI)(Cardona, 2006).
17
to total land area. It is noted that the PVI explicitly takes exposure and susceptibility conditions as
necessary for the presence of disaster risk (Cardona, 2006). The socioeconomic fragility composes
dependency of GDP growth on the agriculture sector, debt-GDP ratio, and human-induced soil
degradation (Cardona, 2006).
The index for lack of resilience (or the adequacy of it) is represented by measures of human capital,
human development, community and environmental protection, governance, economic
redistribution and financial protection. The specific indicators include the Human Development
Index, Gender-related Development Index (GDI), social expenditure to GDP ratio (health, education,
pension), index of governance, percentage of value of insured structures to GDP.
As Pelling (2004, 2006) points out, PVI has a number of limitations. First, it measures solely intrinsic
vulnerability; there is no consideration made to hazard type, scale of hazard impact and capacity for
disaster response. Second, the PVI is inductive in nature, and as such, the results cannot be verified
against actual disaster outcomes. Hence, it pales in comparison to deductive measures, which are
based on actual historical disaster damage data. Third, there is some degree of subjectivity in the
selection of the indicators that will finally be included in the indices, as well as in the assignment of
weights. On a positive note, there are opportunities for complementarity between inductive and
deductive methods, including for verification purposes (Pelling, 2004).
Meanwhile, in the SoVI, social vulnerability is considered as a result of social and place inequalities
that affect the individual’s susceptibility to harm and capacity to respond (Cutter, Boruff, & Shirley,
2003). Unlike the PVI that measures vulnerability at the national level, the Social Vulnerability Index
(SoVI) is a measure of social vulnerability at the sub-national level (Cutter et al., 2003). The SoVI also
adopts an inductive approach and share the first two limitations of the PVI pointed out earlier.
However, unlike the PVI, the assignment of weights in the SoVI is more objective and systematic; the
18
weights of each indicator within a composite indicator are based on the factor loadings of the
principal component analysis.
With the US counties as case studies, 42 socioeconomic and demographic indicators were selected
from more than 250 variables. Using the principal component analysis, the 42 indicators were
further reduced to 11 composite factors, as follows: personal wealth, age, density of built
environment, single-sector economic dependence, housing stock and tenancy, race, ethnicity,
occupation and infrastructure dependence.
The proponents of the index made an initial test of its reliability and usefulness. The test was done
through a simple correlation between the number of presidential declaration of disaster by county
and the individual county SoVI. The test results showed no strong relationship.
This test results suggest that vulnerability in the context of the natural hazards, regardless of the
vulnerability dimension (social, economic, environmental, political, etc) of interest, cannot be
measured totally independent of the type and magnitude of the hazard, as well as extent of
exposure to the said hazard. There is no social vulnerability, say, to a storm surge if in the first place
the county is landlocked and has high elevation which makes the occurrence of, and being exposed
to a storm surge an impossibility. Moreover, as Schneiderbauer (2006) notes, the degree of
vulnerability changes with type and severity of hazards.
b. Econometric Models and Estimation Methods
Unlike the indices presented earlier that are largely inductive in nature, the econometric models of
vulnerability assessment are mainly deductive using actual historical data. As Pelling (2006) points
out, a deductive approach provides more realism of the results than an inductive approach.
Moreover, the use of historical data captures the dynamic nature of vulnerability (ISDR, 2004).
In these models, the underlying causes of vulnerability are indirectly determined using different
variants of the risk equation and frameworks presented earlier. For instance, the Disaster Risk Index
19
or the DRI, which is designed to assess exposure and vulnerability to disasters (Peduzzi, 2006),
adopts a definition of risk that is influenced by hazard, exposure and vulnerability. Specifically, the
DRI equation is expressed, as follows:
𝑅 = 𝐻 × 𝑃𝑜𝑝 × 𝑉𝑢𝑙 (4)
where R is disaster risk, measured in terms of number of deaths, H is the hazard, measured in terms
of its frequency and strength, Pop is the number of people living in the exposed area, and Vul is
vulnerability, which is the variable of interest, is influenced by socioeconomic and environmental
context of the exposed population12.
While the DRI adopts a cross section approach, the related statistical empirical works that followed
adopted panel data analysis. These works likewise explore a general hypothesis that development
plays a significant role in determining vulnerability and the disaster results (Refer to Appendix 2 for
the summary of empirical econometric models, estimation methods and variables used). Despite
the varied models and estimation methods in these researches, the results give light on the more
important factors or determinants of vulnerability. In a broad sense, they provide evidence that
indeed the level of socioeconomic development, and certain key aspects of the development
processes, and institutions significantly contribute in the resulting number of deaths, affected
persons, and costs of damage. We briefly present the empirical researches with similar focus of
inquiry to this paper. We organize the discussion according to our focus areas, namely:
socioeconomic development, urbanization and development governance.
The cross-country empirical studies are unanimous in the findings that indeed a country’s level of
economic development affects its vulnerability to disasters (Anbarci et al., 2005; Kahn, 2005;
Raschky, 2008; Toya & Skidmore, 2007). However, there is difference in the findings as to the
12 Equation 4 is further simplified and translated into an econometric model. The details are found in Appendix 3, which presents the structure of the econometric model and estimation methods used in the related works whose results are discussed in the succeeding section of the text.
20
direction of relationship between the level of economic development and disaster, as well as the
extent at which the level of development influences vulnerability between developed and
developing countries and/or regions. Overall, developing countries are more vulnerable and face
graver disaster impacts than developed countries.
Using GDP per capita as proxy for economic development, Peduzzi et al. (2009) find that it is
negatively correlated with the deaths across all types of hazards considered, namely: tropical
cyclone, drought, and floods. This finding is supported by that of Kahn’s (2005). In all three
regression models ran, he finds that developed countries have fewer deaths from earthquakes than
those of developing countries. As such, he concludes that economic development serves as an
“implicit insurance”, which cushions the negative impact of disasters. Cavallo and Noy (2011)
attribute this to the investments made by developed countries on prevention and mitigation
measures. These measures are lacking in developing countries given the limits of available resources
(Anbarci et al., 2005; Cavallo & Noy, 2011).
In a similar light, Toya and Skidmore (2007) find that as economies develop, they experience fewer
deaths. This is further confirmed by the lower damage cost-to-GDP ratios among developed
countries than those in developing countries (Toya & Skidmore, 2007). It is interesting to note that
while they find that income is also an important factor in determining the number of fatalities
among developing countries, the magnitude of effect is lower than those in developed countries.
While not completely refuting these findings of a linear disaster-economic development
relationship, Kellenberg and Mobarak (2008) argued the presence of a distinct correlation pattern in
the case of developing countries. They find that economic development may actually increase the
risk people face by “changing micro behaviour in such a way so as to increase aggregate exposure to
disasters” (Kellenberg & Mobarak, 2008).
21
Meanwhile, it is interesting to note that Toya and Skidmore (2007) find that among developing
countries, social conditions matter more than the level of income in reducing the number of deaths.
Specifically, they find that the magnitude of effect of education in reducing fatalities is greater than
that of income. They assert that a more educated citizenry are better able to make informed
decisions along ensuring their safety.
Kellenberg and Mobarak (2008) suggest that risk to disasters is also determined by the
vulnerabilities that come about as consequences of development processes, such as urbanization.
Pelling (2003, in Wamsler, 2006) considers urbanization as powerful contributing factor of social
change that can eventually lead up to disaster risk. Wamsler (2006) substantiates this argument by
asserting that this is largely because urban growth, whether planned or otherwise, happens without
due consideration to reducing disaster risk (Wamsler, 2006).
However, Kellenberg and Mobarak (2008) argue that urbanization can, in different contexts, have
varied effects on risk to disasters. That is, urbanization may reduce or increase vulnerability
depending on the context within which it occurs. Specifically, they found that countries with
comparable levels of income but with different degrees of urbanization have different risk levels. On
one hand, in contexts with competent urban planning, where structures are appropriately designed
and where there is adequate capacity to provide economic and social services, urbanization may not
necessarily increase vulnerability to disasters. On the other hand, where the capacity of urban areas
to deliver key services cannot cope with the rapid influx of population (as is the usual case in
developing countries), urbanization may lead to increased exposure and vulnerability to disasters.
They argue that better employment opportunities in dense urban areas attract low income families,
even if such transfer means increased exposure to disasters. Hence, urbanization in this case
increasingly entices people with inherent vulnerability (because of relatively fewer resources and
weaker capacities to adapt and cope in times of disaster) into harm’s way.
22
The effects of several aspects of governance on disaster deaths and damages have likewise been
explored. Kahn (2005) examined the influence of the form and quality of institutions using several
proxy measures. These include the country’s level of democracy, ethnic fragmentation, income
inequality, and good governance indicators.
Kahn (2005) finds that democratic countries experience relatively fewer deaths from disasters than
those with other forms of governance. Under a democracy, governments adopt intervening
measures to mitigate the adverse consequences of hazards (Kahn, 2005). Likewise, he finds a
negative correlation between good institutional quality and fatalities from disasters. This is
consistent with Raschky’s (2008) view and findings that a country’s institutional framework is a key
determinant of vulnerability to disasters. There are fewer fatalities among countries with better
institutions because resource allocation is better, and laws and legislations are in place, and
effectively enforced (Raschky, 2008).
Anbarci et al. (2005) use inequality, measured in terms of Gini coefficient, as a proxy for quality of
governance and institutions. They argue that a political economy that has low income and high
inequality experiences difficulty in generating collective action to undertake preventive measures.
Against this backdrop, these economies suffer more deaths from disasters. In like manner, Kahn
(2005) finds that, all else equal, countries with higher inequality suffer more fatalities from
earthquakes than countries with lower inequality.
An earlier work by Adger (1999) shows similar results. With Vietnam as a case study, which is in
transition from a centrally planned economy, he finds that the increasing inequality and the
breakdown of collective community action that results from the transition have partly contributed to
greater vulnerability. However, he asserts that the resulting institutional change and economic
restructuring towards a market system augers well in terms of reducing vulnerability because the
informal coping mechanism has re-emerged .
23
Toya and Skidmore (2007) also examined other aspects of development that affect disaster deaths
and damages. They find that countries with stronger financial systems (using the percentage of
money supply to GDP as proxy) posted lower fatalities from disasters. They note that the magnitude
of effect of financial systems is higher compared to that of income level. Meanwhile, in terms of the
effect of the public sector size, they find different results between OECD and non-OECD countries.
They note that among OECD, a large public sector is correlated with more deaths; while among
developing countries, the size of the government is not a significant determinant of vulnerability and
disaster risk.
We note that our review of literature revealed an apparent lack of research work at the subnational
level that employed econometric methods to deduce the underlying causes of vulnerability. A
subnational study using a modified approach of these models may prove to be more appropriate in
explaining in a more in-depth manner the underlying issues of vulnerability (Birkmann, 2007).
Moreover, as noted earlier in Section I, as subnational study is of practical usefulness planning and
policy-decisions pertaining to DRRM.
IV. Model, Estimation Methodology and Dataset
A. Risk Framework and Model
We reiterate that the critical task of this empirical work is to determine the aspects of development
that influence people’s vulnerability to tropical cyclones at the subnational level. We adopt a
retrospective and deductive approach; we translate the disaster risk framework expressed in
Equation 2 into a disaster impact framework, expressed in Equation 5 below:
Disaster impact = Hazard x Exposure x Vulnerability (5)
24
Adopting a similar generalization by Peduzzi et al. (2009)13, we take the logarithmic transformation
of this multiplicative model. Hence, our econometric disaster impact model is as follows:
where Impactijt is the measure of actual direct impacts on the population in province i of a past
tropical cyclone j, in year t; Hazardijt is a vector of physical characteristics that measure the strength
of a particular past tropical cyclone; Exposedijt, is a measure of the extent of population exposure in i
to j; Geographyi, which is a vector of time-invariant geographic attributes of i; and, Vulnerabilityit is
the vector of control variables we hypothesize as either positively or negatively affecting people’s
vulnerability to tropical cyclones. These are the level of socioeconomic development, urbanization,
and quality of local development governance.14 By controlling for tropical cyclone strength and
exposure to it, we deduce the factors affecting people’s vulnerability.
Since both our dependent and independent variables are log-transformed, each coefficient is
therefore interpreted as elasticity of the dependent variable with respect to the particular regressor.
We note that the logarithmic transformation of the dependent variable addresses the heavy skew
and makes its distribution approximately normal. This likely results to a more normally distributed
error term, thereby, allowing for valid inferences.
We built a new provincial level panel dataset from datasets collected from different sources, and
estimate Equation 6 using pooled OLS and random effects methods. We justify our use of random
effects method both on technical grounds and practical considerations. We make use of a good set
of explanatory variables, including measures of hazard strength15 and geographic variables, to
13 Details presented in Appendix 3. 14 The next sub-section provides details on the individual variables comprising each component. 15 The inclusion of proxy measures for hazard strength in our model also directly addresses the issue on exogeneity assumption pointed out by Noy (2009).
25
represent each component in the disaster risk framework. This allows us to plausibly make the
assumption of exogeneity, as follows:
Cov (Xijt, αi) = 0 (7)
That is, the unobserved heterogeneity or the unobserved variation across provinces, i is
uncorrelated with all of the explanatory variables, the vector Xijt, in all time periods. Hence, ijt is a
composite error term comprising of the unobserved heterogeneity, i, and the idiosyncratic error,
ijt. That is,
𝜀𝑖𝑗𝑡 = 𝛼𝑖 + 𝜂𝑖𝑗𝑡 (8)
The use of random effects estimation method also allows us to control for time-invariant geographic
variables16. Given the intent of this study of informing physical and land use planning, geographic
factors are key variables of interest, hence, the need for these to be purposely included in our
model. Earlier studies also find the necessity to control for these variables in the context of the
Philippines17.
B. Variables and Sources of Data
To our knowledge, this study is the first subnational work using panel data and econometric method
to answer the question we posed. Our choice of proxy indicators for each component of the risk
framework are based on the existing related work, along with the consideration of the specific
circumstances of the Philippines. We build a new provincial-level panel dataset. We cover the
period 2005-2010, which is dictated by data available for our use at the time of study. As will be
16 We likewise use fixed effects for comparison purposes. Fixed effects cannot be used to estimate the full specification of Equation 6 since our geographic variables are time-invariant. 17 In the context of the Philippines, geography is deemed as critical physical factor in explaining exposure to hazards, and indirectly, people’s vulnerability, given the influence of geography on human and socioeconomic development (PHDN, 2013). The 2012/2013 Philippine Human Development Report finds that “geography explains a significant portion of the variation in life expectancy, education, per capita income, and poverty incidence across areas in the country” (PHDN, 2013). Overall, about 34% of the variation in the Human Development Index of the provinces is due to varying geographic features, namely, climate type, slope, elevation, and whether sea/landlocked (PHDN, 2013).
26
presented in the succeeding sections, data on the number of fatalities with provincial breakdown is
from 2005-2013. Meanwhile, our data for rainfall volume per province is only until 2010. Hence,
our panel dataset covers a period of six-years, from 2005 to 201018.
a. Impacts
We consider two direct disaster impacts to the population. Our first measure of disaster impact is
the number of fatalities per thousand population19 in province i, that is affected by tropical cyclone j
in year t (Fatalityp1kijt), while the second is the number of affected persons per 1,000 population
(Affectedp1kijt)20. We note that none of the existing econometric inter-country study along our line
of inquiry presented in Section III attempted to identify what determines the number of affected
persons. Our attempt is the first one we know, both at the intra and inter-country level.
By scaling the number of fatalities and affected persons using total provincial population, we
account for the varying size of the provinces. Doing so also has the added advantage of
comparability of these measures across areas. We run two sets of regressions, one each for our two
measures of direct impact of disaster on the population.
We make use of the tropical cyclone disasters impact dataset from the National Disaster Risk
Reduction and Management Council (NDRRMC) of the Philippines. Data with provincial breakdown
is available only for tropical cyclones and only from 2005 to 201321. Unlike the more commonly used
18 We intend to extend the analysis later to cover the years 2011 to 2014; we undertake this once we complete dataset. 19 We note that this manner of presenting our dependent variable on fatalities may also be used as measure of vulnerability for the
Philippines’ probabilistic disaster risk assessment methodology . The measure for fatality in the said methodology is called the factor of fatality, expressed as a percentage of deaths to the number of affected persons (NEDA, 2008). We note that in the Philippine guidelines, exposure among the population and the number of affected persons are used interchangeably, apparently using the latter as proxy indicator for the former. 20 Following Kahn (2005), Kellenberg and Mobarak (2008), and Raschky (2008), ln (Impactijt) in Equation 6 is ln(1+fatalityp1kijt) in the first set of regression, and ln(1+affectedp1kijt) in the second set of regression. By doing this, the observations with zero values for fatalityp1k and affectedp1k are not dropped from sample when the logarithmic transformation is done, but are instead given a value of zero. 21 The NDRRMC has no available dataset with provincial disaggregation of disasters impacts brought about by other hazards that occurred in the country.
27
EM-DAT database, the dataset we use includes all tropical cyclone disasters regardless of scale22 and
provides provincial level breakdown of disaster impacts on people and properties.
Majority of the related cross-country empirical researches used the EM-DAT database. While the
said database lists the affected provinces per tropical cyclone, it does not provide provincial
disaggregation of disaster impacts. In addition, we observed a few inconsistencies in the dataset for
the Philippines. For instance, the record of a tropical cyclone in 1995 that passed the northernmost
part of the Philippines reflected as among the affected provinces those which are located in the
southern part of the country where no unusual 24-hour rainfall volumes were recorded.
b. Hazard
We use two measures of hazard strength considering that tropical cyclones can trigger other
hazards, namely: floods, landslide, coastal flooding, and storm surge. While the first three are
induced more by heavy downpour of rainwater than strong winds, the opposite is generally true for
storm surges where high wind speeds are a major contributing factor.
We use the amount of maximum 24-hour rainfall volume as our first measure of hazard strength.
The data on rainfall volume that correspond to the destructive tropical cyclones recorded by the
NDRRMC was generated from the PAGASA. For a given tropical cyclone, the exposed provinces
experience different magnitude of the hazard, depending on whether they are directly under the
tropical cyclone path or along the periphery. To account for this, the rainfall volume assigned to
each province per tropical cyclone (Rainijt) is based on the maximum 24-hour volume recorded in the
nearest rain gauge station during the entire duration that the tropical cyclone is in the Philippine
Area of Responsibility. We use the daily rainfall recorded of 30 stations across the country.
22 The EM-DAT database includes only disasters that meet any of the following criteria: 10 or more fatalities, 100 affected persons, declared a state of emergency, or issued a call for external assistance (Guha-Saphir, Below, & Hoyois, 2014).
28
A number of earlier related cross-country empirical work on tropical cyclones have often used
number of occurrences within the country in a given year as the proxy for the hazard magnitude.
We deem that rainfall volume is a far better measure of tropical cyclone magnitude, and of the
tropical cyclone’s capacity to create destruction.
We also make use of the data on maximum wind speed per tropical cyclone (Windjt) as an added
measure of hazard magnitude. However, we are unable to assign province-specific maximum wind
speed per tropical cyclone.
We note that due to data limitations, including the absence of maps on areas prone to other hazards
induced by tropical cyclones, we are unable to further detail our vulnerability assessment according
to each of these other hazards associated with tropical cyclones. In the Philippines, a 90 mm 24-
hour rainfall volume can trigger a flood, while a 360 mm 24-hour rainfall volume can induce a
landslide23 (NEDA, 2008, 2012a).
c. Exposed
At the time that we conducted this study, the closest and readily available proxy indicator for the
exposed population for our first set of estimation (with Fatalityp1kijt as dependent variable) is the
number of affected people, which we are aware is another measure of risk to people24. We note
that in a similar study by Raschky (2008), he likewise use the number of affected persons as one of
the explanatory variables “to control for the social magnitude of the disaster”. For the second set of
23 On a national scale, floods induced by a 90 mm 24-hour rainfall volume are classified as a “Frequent” flood events, while those triggered by 360mm and 480 mm of 24-hour rainfall volume are classified as “Likely” and “Rare” flood events, respectively. However, due to the varied climatology and geographic features (among others) across provinces, the rainfall threshold per event varies from province to province (NEDA, 2008, 2012a). 24 While “exposed population” and “affected population” are used interchangeably in some related work (such as in NEDA (2008)), we make the distinction between the two. Exposed population refers to those persons exposed to the hazard but who were not adversely affected. Affected population refers to those persons exposed to the hazard and who were adversely affected; that is, affected population are among the exposed population who have vulnerabilities.
29
estimation (with Affectedp1kijt as dependent variable), we use the provincial population as our proxy
measure for exposed25.
d. Geography
The geographic control variables commonly found in related empirical work are location and land
area. Given the distinct and complex geographic features of the Philippine archipelago, we use
several geographic control variables. We use spatial datasets from various sources. We use
Geographic Information System (GIS) overlay analysis tools to generate the values of several
geographic controls. These variables are province-specific and do not change over time.
Instead of using total land area as is usually done in related empirical work, we disaggregate the
provincial land area by slope category. We use two broad slope categories26. Slopeflat is the area of
land within a given province with a slope range of 0 to 18%. Slopesteep is the area of land with a
slope above 18%. To determine the overall direction of influence of the slope variable, we also run a
separate preliminary regression using the average slope of each province (slopemean). From a land-
use planning perspective and based on the Revised Forestry Code of the Philippines, areas with slope
of above 18% is not suitable for settlements use, and must not be used for such purpose (GOP,
1975; NEDA, 2007). These land use policies are supposed to be embodied in the land use plans of
the local government units, and their corresponding zoning ordinances.
Similarly, we also use two elevation variables. The variable elev0300 is the area within the province
that are 0 to 300 meters above sea level. Meanwhile, the variable elev300a is the area within the
province with elevation of more than 300 meters above sea level. As we did for slope, we also run a
25 The estimation of exposed population may be done by overlaying the path of each tropical cyclone with the population map of the provinces with the use of GIS tools. 26 There are six slope categories in the Philippines, as follows: (1) 0 to 3 percent – level to nearly level; (2) 3 to 8 percent – gently sloping; (3) 8 to 18 percent – undulating to rolling; (4) 18 to 30 percent – rolling to moderately steep; (5) 30 to 50 percent – steep; and, (6) above 50 percent – very steep. We only make two broad categories here to distinguish between the areas that are suitable for settlements use and those otherwise.
30
separate preliminary regression using the average elevation of each province (elevmean) to
determine the overall direction of influence of the elevation variable.
For location, we use dummy variables indicating the country’s major island groups, and for provinces
located in the east-most coastline. The dummy variable dli has value of 1 if a given province is part
of Luzon island group, value of zero (0) otherwise, and dvi has value of 1 if a given province is part of
Visayas island group, value of zero (0) otherwise27. The dummy variable deasti has a value of 1 if a
given province is located in the east-most part of the country (along the eastern shoreline), value of
zero (0) otherwise as tropical cyclones always arrive on the east. We identified 19 provinces along
the eastmost, shown in Map 8 in Appendix 1.
We also use geographic control variables to determine whether or not the presence bodies of water
in a given province explains disaster impacts. Riveri is the area of a river that traverses the
province28, while landlockedi is a dummy variable that has a value of 1 if a given province is
landlocked, value of zero (0) if the province has coastal areas. There are 15 landlocked provinces
(Map 8, Appendix 1).
e. Vulnerability
This comprises our main variables of interest. To test the hypothesis that the level of socioeconomic
development reduces people’s vulnerability, we disaggregate the components of the Human
Development Index (HDI) to examine separately the influence of economic development and social
development. We use real per capita income (percapit), which is a common measure of economic
development, and average educational attainment (in years) of the population (schoolmit) and
average life expectancy (lifeit) together as measures for health and education, or overall social
27 As presented in Section II, the Philippines is composed of three major island groups, we use only two dummy variables to avoid the dummy-variable trap. The omitted island group is Mindanao, and, hence, is the comparison group. 28 We note that our data on the area for river is incomplete. All provinces in the Philippines area traversed by rivers but our existing dataset do not have values for 13 provinces.
31
development29. Furthermore, to provide a contrast to the adequacy of income (as proxied for by
percapit), we also separately model the influence of lack of resources using poverty incidence
(povincoit) as proxy. Due to the high correlation between per capita income and poverty incidence (-
0.87), we enter them into the model one at a time.30
For our inquiry on the nature of the influence of urbanization on population vulnerability, we use
both the overall population density in the province (popdenit), and population density in built-up
areas (builtdenit). The former is computed as provincial population divided by total provincial land
area, and the latter, provincial population divided by the total built-up areas only in the province31.
We deem that the latter is a superior indicator for urbanization than the overall population density
of the province as it considers only the areas of the province that actually have human settlements.
It indicates adequacy of space or lack of it in the built environment.
We also derived an indicator for quality of local development governance. Given the provincial
resolution of this study, we use public finance data of the local government units to construct a
governance variable. We use the percentage of locally-generated tax revenues to the total income
of local government units (LGU) (taxinctit) within the provincial geographic boundary. These include
the provincial, city and municipal local government units32.
Our choice for this indicator warrants a brief background. In the Philippines, the total annual income
of a local government unit (LGU) comes from two major sources, namely: 1) incomes from local
sources earned through the efforts of the local government unit; and 2) incomes provided by the
central government, mainly in the form of Internal Revenue Allotment or the IRA. The annual
29 These three variables are used in the computation of the Philippine HDI as measures for standard of living, knowledge, and longevity, respectively (PHDN, 2013). 30 Sources of data are the 2012/2013 Philippine Human Development Report (PHDN, 2013) and the Philippine Statistical Agency. 31 Sources of data and maps include the National Statistics Office-Philippine Statistical Agency, National Economic and Development Authority, Department of Agriculture, and a number of provincial local government units. 32 The sources of basic data are the annual Statements of Income and Expenditures of LGUs prepared by the Philippine Bureau of Local Government Finance (BLGF, 2014).
32
provision of IRA seemingly provides disincentive for the LGUs to undertake local revenue generation.
“LGUs have generally been unwilling to raise their own revenues, particularly through potentially
rich sources such as property tax. The IRA has effectively substituted for own-source revenue
generation” (Balisacan & Hall, 2006). Meanwhile, the collection of taxes, which is the main source
of local income of LGUs, is one of the country’s more problematic areas of local governance. This is
due to the problems of tax evasion and avoidance by the taxpayers, coupled with corruption among
the tax collection bodies.
Given this specific circumstance of the Philippines, taxinct serves as a good indicator of institutional
quality, particularly along local development governance. 33 An LGU that raises its own revenues has
increased ability to provide public goods, including human capital and services and for public safety.
Hence, a high value of taxinct reflects high level of integrity, commitment, accountability and
effectiveness of the local government units in performing their mandated roles. A low value
indicates otherwise.
The results of the Breusch-Pagan Lagrange Multiplier (LM) test suggest the use of random effects
over OLS to estimate the model for our first set of regressions, except that for our inquiry on
urbanization. For the second set of regressions, the test indicate the use of pooled OLS. We conduct
sensitivity tests to check the robustness of results. The first two robustness checks involve varying
the set of control variables, particularly by dropping variables other than the vulnerability variables.
The objective here is to examine the consistency of the sign and/or significance of the coefficients of
the vulnerability variables given the change in the set of control variables (Leamer, 1983). If the sign
and significance of the coefficient of the vulnerability indicator of interest do not change, then the
33 We also mention that the Philippines has an indicator of the quality of governance called the Good Governance Index or the GGI (PSA-NSCB). We do not use the GGI as it is basically an average value of socioeconomic indicators, including those that we individually use as proxy for the different aspects of development that we examine in this study. We note, however, that the GGI includes local government finance indicator, which strengthens the validity of using taxinct as our proxy.
33
said indicator is robust; otherwise it is fragile (Leamer, 1983). The other two tests involve using
reduced datasets; the first dataset excludes the four observations with very high values of fatalities
(more than 100 fatalities), while the second dataset excludes both the observations with very high
values of fatalities and zero fatalities.
f. The Dataset
As noted earlier, the data available allows us to construct a panel dataset, covering six years, from
2005 to 2010. As of 2005, the Philippines has seventy-nine provinces, all of which experienced
tropical cyclone disasters 34. Within this period, a total of 104 tropical cyclones passed the Philippine
Area of Responsibility (PAGASA, 2014). Of which, 57 were reported by the NDRRMC as destructive
tropical cyclones. Together, these destructive tropical cyclones claimed a total of 2,625 lives and
affected 35,885,883 persons.
These 57 destructive tropical cyclones make a total of 722 provincial observations in the dataset,
indicating that, on average, 13 provinces were affected per tropical cyclone. During the six-year
period, each province, on average, was affected by nine tropical cyclones. Maps 3 to 5 in Appendix 1
depict the distributions of the total number of events, number of fatalities and number of affected
by province during the period covered. Visual inspection reveals that the number of events and
impacts on population vary across provinces, regions and major island groups. This is because
climatology vary across the country. As shown in Map 2, tropical cyclones typically pass the
northern part of country, which comprise the northern part of the Luzon major island group (PIDS,
2005). Of these total observations, 550 are for the provinces in the Luzon island group, 118 in the
Visayas island group, and the remaining 54, in the Mindanao island group. The provinces of La
34 To date there are 81 provinces in the country. The 81st province, Davao Occidental, was created only in 2013, while the 80th, the province of Dinagat Islands, was created in the last quarter of 2006. During the period 2005-2010, there are no separate records of disaster impacts as well as socioeconomic data for Dinagat Islands.
34
Union, Pampanga and Zambales each have 21 observations; all three are located in the Luzon island
group.
Table 2 below shows the descriptive statistics of the indicators used in the model, covering the
period 2005-2010. The indicators are presented in their original form in the table but are entered
into the model in their respective logarithmic transformation. Relative to the affected province’s
population, the highest fatalities recorded is 0.508 per 1,000 population, or around 1 fatality per
2,000 people. This is brought by Typhoon Reming in 2006 that caused widespread flooding and
mudslides to Bicol Province (in Luzon). The typhoon also recorded the highest wind speed of 320
kph, more than double the average of 120 kph for all the tropical cyclones covered in the study.
Meanwhile, in terms of maximum 24-hour rainfall volume, the average was 102 mm. The highest
recorded of 685 mm was brought by Typhoon Pepeng in 2009.
Average real income per capita range from a minimum of USD578 (Tawi-Tawi) to a maximum of
USD2,710 (Benguet Province), and an average of USD1,430 across provinces. Interestingly, poverty
incidence range from a nil rate (Batanes) to a high of 59.5 percent (Zamboanga del Norte). In terms
of the average educational attainment (in years) of the population, provincial values range from 7.1
years (Sulu) to 11.99 years (Batanes). The lowest average life expectancy is 52.8 years (Tawi-Tawi),
while the highest is 76.4 years (La Union).
35
Table 2. Descriptive Statistics Variable Description Obs Mean Std. Dev. Min Max
fatalityp1k Number of fatalities per 1,000 population per tropical cyclone 722 0.0065647 0.0277813 0.000242 0.5075603
affectedp1k Number of affected persons per 1,000 population per tropical cyclone 722 50.74525 121.4238 0.0002474 976.9595
rainfall Maximum 24-hour rainfall volume per province per tropical cyclone 722 101.201 97.05351 0 685
wind Maximum wind speed per tropical cyclone 722 128.5568 47.52271 0 320
pop Population 722 1148584 901142.2 18800 4132500 slopeflat Area in the province with slope 0-18% 722 1178.411 949.6761 11.74 3638.24
slopesteep Area in the province with slope above 18% 722 1898.186 1230.904 112.25 6389.82
slope_mean Average slope of the province 722 11.21645 4.10167 4.390004 22.53294
elev300 Area in the province with an elevation of at least 300 meters above sea level 722 910.618 682.834 24.48784 3588.287
elev300a Area in the province with an elevation of above 300 meters above sea level 722 1257.017 1017.029 0.021052 8108.857
elevmean Mean elevation of the province 722 327.2028 256.6644 32.96931 1227.091
river Area of river within the province 722 32.47636 59.59135 0 1004.78
landlocked Dummy variable with a value of 1 if a given province is landlocked, value of zero (0) if province is coastal 722 0.2493075 0.432912 0 1
dl Dummy variable with a value of 1 if a given province is part of Luzon island group, value of zero (0) otherwise 722 0.7617729 0.4262941 0 1
dv Dummy variable with a value of 1 if a given province is part of Visayas island group, value of zero (0) otherwise 722 0.1634349 0.3700183 0 1
deast Dummy variable with a value of 1 if a given province is located in the east-most part of the country (along the eastern shoreline), value of zero (0) otherwise 722 0.2493075 0.432912 0 1
percap Real per capita income (in USD) 722 1429.963 464.626 578 2710
schoolm Average years of schooling of the population 722 10.02605 0.7077236 7.1 11.99
life Average life expectancy 722 68.73269 3.901539 52.8 76.4
taxinct Percentage of tax revenue to total LGU income 722 11.42541 9.612789 0.1363333 43.68192
builtden Population density in built-up areas 722 11595.54 11607.09 2468.133 95691.28
popden Population density in the province 722 409.9637 444.4856 28.00835 2336.07 *The omitted island group is Mindanao
Population density in built-up areas range from 2,468 persons per square kilometre (Tarlac) to a high
of 95,691 persons per sqkm (Lanao del Sur), which is over eight times higher than the average of
11,596. Meanwhile, the ratio of provincial tax revenue to total LGU income range from a high of
43.68 percent (Laguna) and a low of less than 1 percent (Sulu), which practically indicates a full
reliance to the internal revenue allotment from the national government.
36
It is worthwhile to note that the worse off provinces measured in terms of the different the
socioeconomic and governance indicators (low per capita income, high poverty incidence, etc) are
those located in Mindanao, while the better off province are those located in Luzon. Conversely, the
provinces in Mindanao, on average, experienced the least number of tropical cyclone disasters. As
earlier noted, this is because tropical cyclones more frequently pass through the northeastern part
of the country.
V. Results and Discussions35
Table 3 shows the estimation result under full model specification for our first set of estimates
(dependent variable is lfatalityp1k), while Table 4 shows the result for the second set (dependent
variable is laffectedp1k)36. Contained in Columns 1 to 5 of these tables are the estimates using
pooled OLS method, while Columns 6 to 10 are those using random effects method. As can be
gleaned, the two method yield very similar results in terms of the significant vulnerability variables.
As earlier noted, the results of the Breusch-Pagan Lagrange Multiplier (LM) test require the use of
random effects over pooled OLS to estimate the model for our first set of regression, except for our
inquiry on urbanization; the test require the use of pooled POLS for second set of regression. Hence,
in discussing the results for the first set regressions, we focus on the random effects estimates; we
refer to the OLS results for the second set.
Column 6 in Table 3 shows that the coefficient of per capita income is negative and highly significant,
indicating that fatality is a decreasing function of income, a result consistent with related empirical
literature that income provides an insulating effect from the adverse impacts of disasters (Khan,
35 We note again that all variables are entered into the model in their respective logarithmic transformation. For brevity in the analysis, we simply refer to the name of the indicators and dispel with repeatedly indicating they are in logarithmic form. 36 We note, however, that for the first set of estimates (with lfatalityp1k as the dependent variable), we do not control for lriver. Based on data gathered, of the 79 provinces included in our sample, only 67 provinces have data on the area of river. We ran regressions for the first set of estimates where we included lriver as control variable. We find that the coefficient of lriver is not statistically significant. Also, by including lriver as a control variable, there are only 636 observations, indicating that 96 observations are dropped. Hence, we dropped lriver in the first set of estimates to maintain the number maximum of observations.
37
2005). Conversely, from the standpoint of inadequacy, the coefficient of poverty incidence
(lpovinco) in Column 7 in Table 3 is positive, and significant. This quantitatively validates the earlier
assessments that in the Philippines, poverty is a critical factor in determining vulnerability to
disasters (Shepherd et al., 2013).
Table 3. Factors Affecting People’s Vulnerability to Tropical Cyclones Full model specification Set 1: Dependent variable is lfatalityp1k
Pooled OLS Random Effects (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
38
Table 4. Factors Affecting People’s Vulnerability to Tropical Cyclones Full model specification Set 2: Dependent variable is laffectedp1k
Pooled OLS Random Effects (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
In accordance with the findings of Toya and Skidmore (2009), Column 8 in Table 3 reveals that, social
development matters in ensuring safety from the adverse impacts of disasters. We find that high
level of education (lschoolm) and good quality of health (llife) are inversely correlated with fatalities
(lfatalitiesp1k). This implies that apart from structural interventions, adequate public investment for
building human capital is likewise warranted to ensure permanent and long-term impact in reducing
people’s vulnerability and increasing their resilience. This is especially important for the Philippines
as there is widespread underinvestment in health and education among the poor. As Toya and
39
Skidmore (2007) find in their comparative assessment between OECD and developing countries,
income and social development measures are both negatively correlated with disaster impacts,
however, social development carries greater importance among developing countries.
We next examine the influence of urbanization, which is closely linked with economic growth. In
general, urban areas in the Philippines exhibit the benefits from the agglomeration of people and
economic activities (Corpuz, 2013). However, our result reveals a positive and significant coefficient
of lbuiltden, as shown in Colum 9 of Table 3. This hints of the diminishing safety of people as the
existing built-up areas become more population dense. This may partly reflect the burgeoning of
settlements in hazard prone areas and the lagging provision of adequate services for the additional
population, particularly in areas exhibiting high population growth rate. It is interesting to note,
however, that lpopden has a negative and significant coefficient; that is, that an increase in overall
population density in a province is negatively correlated with fatalities. This indicates that a more
even distribution of the population across the province is favourable for reducing disaster fatalities.
In terms of local development governance, the coefficient for our proxy indicator (ltaxinct) is
significant and inversely correlated with fatalities (lfatalitiesp1k), as shown in Column 10 in Table 3.
Our result and those of related inter-country assessment (Adger, 1999; Raschky, 2008) earlier
presented together denote that good governance, whether at the national or subnational level, is
important in reducing vulnerability, and consequently, disaster impacts.
Meanwhile, most of the hazard and exposure control variables are significant and their signs are as
expected. Overall, disaster impacts on the population are an increasing function of the tropical
cyclone strength. On one hand, the proportion of fatalities increases with increases in rainfall
volume (Table 3). On the other hand, both rainfall volume and wind speed are important in
explaining the proportion of affected persons (Table 4). It is worthwhile to note that in our
preliminary regression that only controls for hazard, the strength of the tropical cyclone only
explains a small proportion of the variation in the number of fatalities and affected persons (R2 of
40
less than 0.1 in Table 9 in Appendix in 2). This conveys that grave disaster impacts are not mainly
results of the destructive hazard strength; thus, these disaster impacts can largely be prevented
through appropriate DRRM measures to reduce exposure and vulnerability.
In terms of exposure, fatality is an increasing function of exposed population, as proxied for by the
proportion of affected persons (Columns 1 to 5 in Table 3). However, there is an insignificant
coefficient for population (Columns 6 to 10 in Table 3), which is our proxy for exposure for the
second set of estimates (where the proportion of affected person is the independent variable). This
is not surprising given the limitation of using as proxy the total population of the province, as some
provinces may not have been entirely exposed to the hazard.
We discuss with some length the geographic control variables, given the importance of the
implications of results to policies, particularly along land use planning. The results in Columns 6 to
10 in Table 3 generally reveal that both slope categories are important in explaining the fatalities
resulting from tropical cyclones. It is noted that while lslopeflat has a negative and significant
coefficient, lslopesteep has a positive and also significant coefficient37. A plausible explanation for
these is that settlements on flatlands have better mitigation measures in place than those on steep
slopes. It has been noted however that in the Philippines, communities in steeps slopes are also
increasing. Gaillard et al. (2007) find that when the traditional areas for settlement in the lowland
are already reaching there carrying capacity, many Filipinos resort to taking residence in marginal
areas, such as those with steep slopes that are prone to rain-induced landslides, particularly those
already denuded of trees (Gaillard et al., 2007).38
The results shown in Table 4 reveal that lriver has a positive and statistically significant coefficient,
for each of the regression presented. This indicates that the larger the area of the river traversing
37 The preliminary regressions that only control of geographic variables (Table 11 in Appendix 4) shows a positive and significant coefficient for the ln mean slope (slopemean), and a negative and significant coefficient for ln mean elevation (lelevmean). 38 From a land-use planning perspective and based on the Revised Forestry Code of the Philippines, areas with slope of above 18% are not suitable for settlements use, and must not be used for such purpose (GOP, 1975; NEDA, 2007).
41
the province, the greater is the number of affected persons. One plausible explanation for this is
that riverbanks in the Philippines are often densely populated, including the river buffer zones,
particularly by informal settlers. With heavy downpour, the occurrences of riverine flooding is
common, particularly in urban areas where river drainages are blocked with debris and garbage, and
even human settlements. The improper anthropogenic practices within the watershed, such as
illegal logging and mining, add on to the destructive potential of the hazards faced by communities
along the riverbanks, and subsequently exacerbate the disaster impacts (Gaillard et al., 2007;
Ginnetti et al., 2013; NEDA, 2012b). Likewise, the coefficients of landlocked in Table 4 are negative
and significant. The result implies that there are relatively fewer affected persons in landlocked
provinces than in coastal provinces. The above significant results indicate the presence of
settlements in close proximity to yet another hazard prone areas, namely along riverbanks and
coastlines, likely a manifestation of the low enforcement or compliance to the Philippine Water
Code39.
We subject our vulnerability variables for robustness checks, the first of which involves dropping
selected geographic control variables. Table 5 and 6 show the estimation results where the variables
for slope category are dropped, while Tables 7 and 8, the results where variables for slope and
elevation categories are dropped. The results in these tables are compared with those in the
corresponding columns in Table 3 and 4, respectively. As noted earlier, we want to determine
whether the sign and/or significance of the coefficients of the vulnerability indicator of interest
changes.
39 The Philippine Water Code states that banks of streams and rivers, and shores of lakes and seas along urban areas are subject to a three-meter easement of public use (GOP, 1976). Similarly, the pertinent provisions of the code are supposed to be embodied in land use plans and zoning ordinances, to prohibit human settlements on these easement areas. Subsequently, the plans and ordinances have to be implemented and compliance to these has to be regularly monitored. However, this is not often the case. In addition to poor enforcement and monitoring of plans and policies, the continued increasing density in these areas is also a result of a complex set of socioeconomic factors and processes, as presented in Section II.
42
For the first set of regressions, we find that the level of education (proxied for by lschoolm) loses
significance when subjected to the second robustness check (Column 8 in Tables 7). For the second
set of regressions (with laffectedp1k as dependent variable) we find no changes in the sign and
significance of the vulnerability variables.
Table 5. Robustness Check 1: Dropped the variables on slope categories Set 1: Dependent variable is lfatalityp1k
Pooled OLS Random Effects (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
43
Table 6. Robustness Check 1: Dropped the variables on slope categories Set 2: Dependent variable is laffectedp1k
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
44
Table 7. Robustness Check 2: Dropped the variables on slope and elevation categories Set 1: Dependent variable is lfatalityp1k
Pooled OLS Random Effects (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation
45
Table 8. Robustness Check 2: Dropped the variables on slope and elevation categories Set 2: Dependent variable is laffectedp1k
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
46
Tables 15 to 18 in Appendix 4 show the results of our further sensitivity tests that involve using
reduced dataset. We briefly present herein those for the regressions with lfatalityp1k as the
dependent variable. In the first reduced dataset, we removed four provincial observations with
more than 100 fatalities. The results shown in Table 15 indicate that all the vulnerability variables
earlier found to be important determinants of fatalities remained significant, and retained their
signs. In the second reduced dataset, we only include provincial observations that 1 to 100 fatalities.
Likewise, coefficients of the various vulnerability variables remain significant and retain their signs.
VI. General Conclusions and Next Steps
We find significant quantitative evidence of the strong linkage between several important aspects of
development and disaster impacts on the population. Consistent with existing cross-country
evidence, we find that the level of socioeconomic development at the provincial level is critical in
reducing human vulnerability, and subsequently, in reducing disaster impacts to the population. The
positive coefficient for built-up density supports the earlier findings that the consequences of
unplanned and rapid urbanization are the generation of vulnerabilities and growth of settlements in
hazard prone areas. Our results confirm that the quality of governance in the Philippines can alter
disaster impacts. In addition, geography is an important determinant of disaster impacts.
Overall, our results provide support for a development path that is carefully planned and which
deliberately integrates into the process the various components of disaster risk, particularly
exposure and vulnerability because these components can be controlled. This can be facilitated with
an accountable and effective governance, which recognizes the merits of structural measures, as
well as non-structural measures. These include proper land use planning and increasing public
investment in human capital to achieve long-term results in terms of reducing vulnerability, as well
as building resilience to various shocks.
47
We continue the endeavour to improve this study as additional relevant datasets become available
for our use. A more extensive and detailed dataset on historical disaster impacts will give us more
observations needed to undertake not only area specific, but also element (population, property,
economy) and event-specific vulnerability assessment.
In addition, maps in vector format of areas prone to floods, rain-induced landslides and storm surge
will allow us to identify the distinct determining factors of people’s vulnerabilities specific to each of
these hazards induced by tropical cyclones. Due mainly to data limitations, we are not able to
quantitatively explore the impact of environmental degradation on human vulnerability. Among
others, vector maps on the state of environmental quality or degradation (i.e. forest cover, etc.) will
allow as to undertake such an assessment, using both spatial and statistical analysis tools.
Map 3. Number of Destructive Tropical Cyclones, by Province, 2005-2010
Map 4. Total Number of Fatalities per Destructive Tropical Cyclone, by Province, 2005-2010
Map 5. Total Number of Affected Persons per Destructive Tropical Cyclone, by Province, 2005-2010
49
Map 6. Average Number of Fatalities per Destructive Tropical Cyclone, by Province, 2005-2010
Map 7. Average Number of Affected Persons per Destructive Tropical Cyclone, by Province, 2005-2010
Map 8. Landlocked Provinces and Provinces along the Eastmost Coastline
50
Appendix 2. Preliminary Regressions
Table 9. Preliminary Regression: Controls are only Hazard Variables Set 1: Dependent Variable is lfatalityp1k Set 2: Dependent Variable is laffectedp1k OLS RE FE OLS RE FE (1) (2) (3) (4) (5) (6)
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
51
Table 10. Preliminary Regression: Controls are Combinations of Geography Variables Set 1: Dependent Variable is lfatalityp1k Set 2: Dependent Variable is laffectedp1k OLS RE FE OLS RE FE OLS RE FE OLS RE FE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001
52
Table 11. Preliminary Regression: Controls are Combinations of Hazard and Geography Variables Set 1: Dependent Variable is lfatalityp1k Set 2: Dependent Variable is laffectedp1k OLS RE FE OLS RE FE OLS RE FE OLS RE FE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Table 12. Preliminary Regression: Controls are Combinations of Exposed Population and Geography Variables Set 1: Dependent Variable is lfatalityp1k Set 2: Dependent Variable is laffectedp1k OLS RE FE OLS RE FE OLS RE FE OLS RE FE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Exposed Population laffectedp1k 0.0936*** 0.0912*** 0.0915*** 0.0891*** 0.0896*** 0.0915***
Table 13. Preliminary Regression: Controls are Combinations of Hazard, Exposed Population and Geography Variables Set 1: Dependent Variable is lfatalityp1k Set 2: Dependent Variable is laffectedp1k OLS RE FE OLS RE FE OLS RE FE OLS RE FE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001
55
Appendix 3. Estimation Result Using Fixed Effects under Full Model Specification
Table 14. Affectors Afftecting People’s Vulnerability to Tropical Cyclones Full Model Specification Using Fixed Effects (Dependent variables are indicated above the column numbers)
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation
56
Appendix 4. Additional Robustness Checks
Table 15. Robustness Check 3: Dataset excludes observations with more than 100 fatalities
Set 1: Dependent variable is lfatalityp1k Pooled OLS Random Effects
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
57
Table 16. Robustness Check 3: Dataset excludes observations with more than 100 fatalities
Set 2: Dependent variable is laffectedp1k Pooled OLS Random Effects
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
58
Table 17. Robustness Check 4 Reduced Dataset: Only includes observations with 1 to 100 fatalities Set 1: Dependent variable is lfatalityp1k
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
59
Table 18. Robustness Check 4 Reduced Dataset: Only includes observations with 1 to 100 fatalities Set 1: Dependent variable is laffectedp1k
t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001 Note: The “l” attached to each variable name, except for the dummy variables, indicate that the variable is in logarithmic transformation OLS reflects adjusted R-sq
60
References
ABS-CBN. (2013). Why No One Died on Island in Cebu. Manila, Philippines: ABS-CBN Philippines. Adger, W. N. (1999). Social Vulnerability to Climate Change and Extremes in Coastal Vietnam. World
Development, 27(2), 249-269. doi: http://dx.doi.org/10.1016/S0305-750X(98)00136-3 Anbarci, N., Escaleras, M., & Register, C. A. (2005). Earthquake fatalities: the interaction of nature and
political economy. Journal of Public Economics, 89(9–10), 1907-1933. doi: http://dx.doi.org/10.1016/j.jpubeco.2004.08.002
Balisacan, A. M., & Hall, H. (2006). The Dynamics of Regional Development: The Philippines in East Asia.
Bayudan-Dacuycuy, C., & Lim, J. A. (2013). Family size, household shocks and chronic and transient poverty in the Philippines. Journal of Asian Economics, 29(0), 101-112. doi: http://dx.doi.org/10.1016/j.asieco.2013.10.001
Birkmann, J. (2006). Measuring vulnerability to promote disaster-resilient societies: Conceptual frameworks and definitions. In J. Birkmann (Ed.), Measuring vulnerability to natural hazards :towards disaster resilient societies. New York: United Nations University.
Birkmann, J. (2007). Risk and vulnerability indicators at different scales: Applicability, usefulness and policy implications. Environmental Hazards, 7(1), 20-31. doi: http://dx.doi.org/10.1016/j.envhaz.2007.04.002
Blaikie, P., Cannon, T., Davis, I., & Wisner, B. (1994). At risk : natural hazards, people's vulnerability, and disasters London; New York: Routledge.
BLGF. (2014). Local Government Unit Financial Data: Statement of Income and Expenditure of Provinces, Cities and Municipalities. from http://www.blgf.gov.ph/#
Bohle, H. G. (2001, 2/2001). Vulnerability and Criticality: Perspectives from Social Geography. Newsletter of the International Human Dimensions Programme on Global Environmental Change, pp. 1-7.
BSP. (2014). Economic and Financial Statistics Retrieved from: http://www.bsp.gov.ph/statistics/efs_ext3.asp
Cardona, O. (2006). A System of Indicators for Disaster Risk Management in the Americas. In J. r. Birkmann (Ed.), Measuring vulnerability to natural hazards: towards disaster resilient societies. New York: United Nations University.
Cardona, O., Aalst, M. K. v., Birkmann, J., Fordham, M., McGregor, G., Perez, R., . . . Sinh, B. T. (2012). Determinants of risk: exposure and vulnerability Managing the Risk of Extreme Events and Disasters to Advance Climate Change Adaptation. Cambridge University Press, Cambridge UK , and New York, NY, USA: Intergovernmental Panel on Climate Change (IPCC).
Cavallo, E., & Noy, I. (2011). Natural Disasters and the Economy--A Survey. International Review of Environmental and Resource Economics, 5(1), 63-102. doi: http://www.irere.net
Cutter, S. L., Boruff, B. J., & Shirley, W. (2003). Social Vulnerability to Environmental Hazards. Social Science 4, 84(2), 242-261. doi: http://dx.doi.org/10.1111/1540-6237.8402002
Davidson, R., & Shah, H. (1997). An Urban Earthquake Disaster Risk Index. Stanford: The John A. Blume Earthquake Engineering Center, Department of Civil Engineering.
Florano, E. R. (2014). Community Governance for Disaster Recovery and Resilience: Four Case Studies in the Philippines. Philippine Institute for Development Studies Retrieved from http://dirp3.pids.gov.ph/webportal/CDN/PUBLICATIONS/pidsdps1438.pdf.
Gaillard, J.-C. (2008). Alternative paradigms of volcanic risk perception: The case of Mt. Pinatubo in the Philippines. Journal of Volcanology and Geothermal Research, 172(3–4), 315-328. doi: http://dx.doi.org/10.1016/j.jvolgeores.2007.12.036
Gaillard, J.-C., Liamzon, C. C., & Villanueva, J. D. (2007). ‘Natural’ disaster? A retrospect into the causes of the late-2004 typhoon disaster in Eastern Luzon, Philippines. Environmental Hazards, 7(4), 257-270. doi: http://dx.doi.org/10.1016/j.envhaz.2006.11.002
Ginnetti, J., Dagondon, B., Villanueva, C., Enriquez, J., F.T., T., Bacal, C., & Carcellar, N. (2013). Disaster-induced Internal Displacement in the Philippines: The Case of Tropical Storm Washi/Sendong. Norway.
GOP. (1975). Presidential Decree No. 705: Revising Presidential Decree No. 389, Otherwise Known As the Forestry Reform Code of the Philippines. Manila, Philippines: Government of the Philippines.
GOP. (1976). Presidential Decree No. 1067: The Water Code of the Philippines, A Decree Instituting A Water Code, Thereby Revising And Consolidating The Laws Governing The Ownership, Appropriation, Utilization, Exploitation, Development, Conservation And Protection Of Water Resources. Manila Philippines: GOP.
Guha-Saphir, D., Below, R., & Hoyois, P. (2014). Emergency Events Database (EM-DAT): International Database. Retrieved from: http://www.emdat.be/database
Guha-Sapir, D., Hoyois, P., & Below, R. (2012). Annual Disaster Statistical Review 2011: The Numbers and Trends. Brussels, Belgium: Centre for Research on the Epidemiology of Disasters (CRED), Institute of Health and Society (IRSS), Universite Catholique de Louvain.
Guha-Sapir, D., Hoyois, P., & Below, R. (2013). Annual Disaster Statistical Review 2012: The Numbers and Trends. Centre for Research on the Epidemiology of Disasters (CRED), Institute of Health and Society (IRSS), Universite Catholique de Louvain. Brussels, Belgium. Retrieved from cred.be/sites/default/files/ADSR_2013.pdf
Guha-Sapir, D., Hoyois, P., & Below, R. (2014). Annual Disaster Statistical Review 2013: The Numbers and Trends. Brussels, Belgium.
ISDR. (2004). Living with Risk: A Global Review of Disaster Reduction Initiatives Vol. 1. Retrieved from http://www.unisdr.org/we/inform/publications/657
Jose, S. R. G. (2012). Preliminary Examination of Existing Methodologies for Allocating and Tracking National Government Budget for Disaster Risk Reduction (DRR) in the Philippines: ADB and UNISDR.
Kahn, M. E. (2005). The Death Toll from Natural Disasters: The Role of Income, Geography, and Institutions. Review of Economics and Statistics, 87(2), 271-284. doi: http://dx.doi.org/10.1162/0034653053970339
Kellenberg, D. K., & Mobarak, A. M. (2008). Does rising income increase or decrease damage risk from natural disasters? Journal of urban economics, 63(3), 788-802. doi: http://dx.doi.org/10.1016/j.jue.2007.05.003
Leamer, E. E. (1983). Let's Take the Con Out of Econometrics. American Economic Review, 73(1), 31-43. doi: http://www.aeaweb.org/aer/
NDRRMC. (2014). Tropical Cyclones and Their Effects, 2005-2013. NEDA. (2005). Regional Development Plan of Northern Mindanao. Cagayan de Oro City, Philippines:
National Economic and Development Authority. NEDA. (2007). Guidelines on Provincial/Local Planning and Expenditure Management. Manila,
Philippines: National Economic and Development Authority and the Asian Development Bank. NEDA. (2008). Mainstreaming Disaster Risk Reduction in Subnational Development and Land
Use/Physical Planning in the Philippines. Manila, Philippines: National Economic and Development Authority, United Nations Development Programme, European Commission Humanitarian Aid Retrieved from www.neda.gov.ph.
NEDA. (2012a). Disaster Risk Assessment Methodologies (Material Presented during the Regional Experts’ Forum on Disaster Risk Reduction and Management and Climate Change Adaptation). Cagayan de Oro City, Philippines: National Economic and Development Authority.
NEDA. (2012b). Strategic Action Plan for the Reconstruction and Rehabilitation of Areas Affected by TS Sendong. Cagayan de Oro, Philippines: NEDA.
Noy, I. (2009). The macroeconomic consequences of disasters. Journal of development economics, 88(2), 221-231. doi: http://dx.doi.org/10.1016/j.jdeveco.2008.02.005
Noy, I., & Vu, T. B. (2010). The economics of natural disasters in a developing country: The case of Vietnam. Journal of Asian Economics, 21(4), 345-354. doi: http://dx.doi.org/10.1016/j.asieco.2010.03.002
PAGASA. (2011). Climate Change in the Philippines. Manila. PAGASA. (2014). List of Tropical Cyclone in the Philippine Area of Responsibility, 1980-2013. PAGASA. (undated). Tropical Cyclone. http://www.pagasa.dost.gov.ph/learning-tools/tropical-
cyclone-learning?showall=1&limitstart= Peduzzi, P. (2006). The Disaster Risk Index: Overview of a quantitative approach. In J. Birkmann (Ed.),
Measuring vulnerability to natural hazards: towards disaster resilient societies (pp. 171-181). New York United Nations University.
Peduzzi, P., Dao, H., Herold, C., & Mouton, F. (2009). Assessing global exposure and vulnerability towards natural hazards: the Disaster Risk Index. Natural Hazards and Earth System Sciences, 9(4), 1149-1159.
Pelling, M. (2004). Visions of Risk: A Review of International Indicators of Disaster Risk and Management: United Nations - International Strategy for Disaster Reducation, United Nations Development Programme.
Pelling, M. (2006). Review of global risk index projects: Conclusions for sub-national and local approaches. In J. Birkmann (Ed.), Measuring vulnerability to natural hazards: towards disaster resilient societies (pp. 151-170). New York United Nations University.
PHDN. (2013). 2012/2013 Philippine Human Development Report. Manila, Philippines: Philippine Human Development Network.
PIDS. (2005). Tropical Cyclone Signals: Bracing for the Wind. Manila, Philippines: Philippine Institute for Development Studies.
PSA. (2012). The 2010 Census of Population and Housing Reveals the Philippine Population at 92.34 Million. http://web0.psa.gov.ph/
Raschky, P. A. (2008). Institutions and the losses from natural hazards. Natural Hazards and Earth System Sciences(8), 627-634.
Schneiderbauer, S. L., Daniele. (2006). Social levels and hazard (in)dependence in determining vulnerability. In J. Birkmann (Ed.), Measuring vulnerability to natural hazards: towards disaster resilient societies. New York United Nations University.
Shepherd, A., Mitchell, T., Lewis, K., Lenhardt, A., Jones, L., Scott, L., & Muir-Wood, R. (2013). The Geography of Poverty, Disasters and Climate Extreme in 2030. London, United Kingdom: Overseas Development Institute.
Toya, H., & Skidmore, M. (2007). Economic development and the impacts of natural disasters. Economics Letters, 94(1), 20-25. doi: http://dx.doi.org/10.1016/j.econlet.2006.06.020
UN. (2014). World Urbanization Prospects: The 2014 Revision. Retrieved from: http://esa.un.org/unpd/wup/CD-ROM/Default.aspx
UNDP-DHA. (1994). Vulnerability and Risk Assessment: Disaster Management Training Programme. UNDP-UNDRO. (1992). An Overview of Disaster Management (2nd ed.): United Nations Development
Programme; United Nations Disaster Relief Office. UNDP. (2004). Human Development Report 2004 Cultural Liberty in Today's Diverse World. 1 UN Plaza,
New York, New York, 10017, USA: United Nations Development Programme. UNISDR. (2005). Hyogo Framework for Action 2005-2015: Building the resilience of nations and
communities to disasters: United Nations Office for Disaster Risk Reduction. UNISDR. (2013). From Shared Risk to Shared Value –The Business Case for Disaster Risk Reduction
Global Assessment Report on Disaster Risk Reduction. Geneva, Switzerland: United Nations International Strategy for Disaster Reduction (UNISDR).
UNU-EHS. (2014). World Risk Report 2014 Focus: The City as Risk Area: United Nations University for Environment and Human Security.
Wamsler, C. (2006). Mainstreaming risk reducting in urban planning and housing: a challenge for international aid organizations. Disasters, 30(2), 151-177.
WB-EASPR. (2003). Natural Disaster Risk Management in the Philippines: Enhancing Poverty Alleviation Through Disaster Reduction: World Bank - East Asia and Pacific Region, Rural Development.
WB. (2014). Philippine Economic Update Pursuing Inclusive Growth through Sustainable Reconstruction and Job Creation. Philippines: World Bank, Philippine Country Office, East Asia and Pacific Region.
Wisner, B., Blaikie, P., Cannon, T., & Davis, I. (2004). At risk : natural hazards, people's vulnerability, and disasters (Second ed.). London; New York: Routledge.
Wisner, B., Gaillard, J., & Kelman, I. (Eds.). (2012). The Routledge Handbook of Hazards and Disaster Risk Reduction. Great Britain: Routledge.
