For Eng.Kelvin only

natural_hazards_unnatural_disasters_pt_1.pdf

Home >Reading homework help >For Eng.Kelvin only

MEMORANDUM TO A CONDERNED FINANCE MINISTER

Subject: Natural Hazards, UnNatural Disasters: The Economics of Effective Prevention

his memo introduces a report that you may find useful and interesting. Focusing on preventing death and destruction from “natural” disasters, it concludes that

governments can appreciably increase prevention. The good news is that prevention is often cost-effective. It requires many actions,

and some important ones are under government control. But they are not always obvious. Improving the public delivery of some services, like reliable public transport, allows people to move from unsafe areas close to work to safer locations. Reducing deforestation prevents heavy rains from washing mud, rock, and debris into populated areas. This report suggests how such measures and related spending could be identified and made effective. Effective spending is complex, and cost-benefit analysis (underused) helps, but

institutions that increase the public’s involvement and oversight are vital. Large benefits result from greater transparency in all aspects of government decision making. How the public responds to such prevention measures depends on its trust in the government. Such trust flows from credible institutions, which the report persistently underscores. Prevention pays, but you do not always have to pay more for prevention. A

relatively easy and effective measure is for governments to make information about hazards and risks easily accessible (such as maps of flood plains and seismic fault lines). Allowing markets to work better also helps because much information is embedded in prices. Controls on prices, trade, and the like and excessive tax rates have harmful effects, and correcting them goes a long way in increasing prevention. Effective prevention cannot rest on laissez-faire alone, for markets need to be

complemented with appropriate government actions. Greater spending on some items is warranted: many countries are not taking advantage of the technological improvements in weather and related forecasting. Even modest increases in spending —and greater sharing of data internationally—can have enormous benefits, especially to warn people of impending hazards. Several countries, some very poor, have found large and quick gains from such spending. The gains can also spill beyond borders, enhancing regional cooperation.

The, World Bank, Bank World, and Nations (UN) United. Natural Hazards, UnNatural Disasters, edited by World Bank The, et al., World Bank Publications, 2010. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/apus/detail.action?docID=606027. Created from apus on 2017-04-04 13:20:21.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Effective prevention cannot rest on a single measure or simple slogan either. Ensuring government’s adequate funding of infrastructure, basic services, early warning systems, and the like will have high payoffs. But the financing of infrastructure has to be matched by adequate maintenance. Funding early warning systems is only as useful as the “last mile” of successful evacuation and response. Bangladesh shows that such a response can be effective even in poor countries, while some rich countries (such as the United States in its response to Hurricane Katrina) can stumble over this last step. Despite adequate prevention measures, hazards will strike and funds will be

required for recovery and reconstruction. So knowing disaster’s effects on fiscal sustainability is important for making informed decisions. While the government can borrow, it must ultimately pay it all back from taxes or spending cuts elsewhere. And although donors provide external aid after disasters, studies show they often do so by re-labeling funds without increasing aggregate amounts. You will have to rely on the ability to tax—and spend accordingly. Finally, one message about the future: cities will grow, especially in developing

countries, increasing exposure of lives and property to disasters, but not uniformly or monotonically. Though exposure will rise, better managed cities can reduce vulnerability and risk. Although you do not run the cities, you control many aspects of their financing and can do much to reduce new risks. Damage from hazards— particularly tropical cyclones—are also likely to increase because of climate change. Your successors will have to deal with these more difficult issues, but they will benefit from the steps you take now. If you help correct the problems of the present, generations to come will welcome the future.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Overview

he adjective “UnNatural” in the title of this report conveys its key message: earthquakes, droughts, floods, and storms are natural hazards, but the

unnatural disasters are deaths and damages that result from human acts of omission and commission. Every disaster is unique, but each exposes actions—by individuals and governments at different levels—that, had they been different, would have resulted in fewer deaths and less damage. Prevention is possible, and this report examines what it takes to do this cost-effectively. The report looks at disasters primarily through an economic lens. Economists

emphasize self-interest to explain how people choose the amount of prevention, insurance, and coping. But lenses can distort as well as sharpen images, so the report also draws from other disciplines: psychology to examine how people may misperceive risks, political science to understand voting patterns, and nutrition science to see how stunting in children after a disaster impairs cognitive abilities and productivity as adults much later. Peering into the future, the report shows that growing cities will increase exposure to hazards, but that vulnerability will not rise if cities are better managed. The intensities and frequencies of hazards in the coming decades will change with the climate, and the report examines this complicated and contentious subject, acknowledging all the limitations of data and science.

Four main findings First, a disaster exposes the cumulative implications of many earlier decisions, some taken individually, others collectively, and a few by default. A deeper questioning of what happened, and why, could prevent a repetition of disasters. Several factors usually contribute to any disaster, some less obvious than others. The immediate cause of a bridge or building collapse may be a mudslide, though poor design or construction may have also contributed. But the underlying cause may be denuded hillsides that increased sediment flows (as in Haiti), or poor urban planning that put the bridge or building in harm’s way. Symptoms are easily mistaken for cause: denuded hillsides may result from desperately poor people depleting the vegetation to survive or from logging concessions that encourage tree cutting but not planting. Effective prevention measures are therefore not always “obvious.” Second, prevention is often possible and cost-effective. Studies for the report

examined the costs and benefits of specific prevention measures that homeowners The, World Bank, Bank World, and Nations (UN) United. Natural Hazards, UnNatural Disasters, edited by World Bank The, et al., World Bank Publications, 2010. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/apus/detail.action?docID=606027. Created from apus on 2017-04-04 13:20:21.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

could take in hazard-prone areas of four low- and middle-income countries. Prevention pays for assumed (but reasonable) costs and discount rates. Other prevention measures are embedded in infrastructure (such as adequate drainage ditches). The report examines government expenditures on prevention and finds that it is generally lower than relief spending, which rises after a disaster and remains high for several subsequent years. But effective prevention depends not just on the amount but on what funds are spent on. For example, Bangladesh reduced deaths from cyclones by spending modest sums on shelters, developing accurate weather forecasts, issuing warnings that people heeded, and arranging for their evacuation. All this cost less than building large-scale embankments that would have been less effective. Third, many measures—private and public—must work well together for effective

prevention. Low-lying areas around Jakarta illustrate the complexity of ensuring this: residents raise the plinth of their houses to protect against floods, but they also draw water through borewells causing the ground to subside. Even knowing this, a person has no choice if the government does not provide piped water. So, the prevention measures an individual undertakes also depend on what the government does—or fails to do—and vice versa. That many measures do not work well together in poor countries explains why they

have more disasters. The poor may know the hazard risks they face but depend more on public services that are often inadequate. They live near work on cheaper land exposed to hazards if buses are unreliable, while the rich with cars have better alternatives. The poor would willingly move to safer locations if their incomes rose or if public transport became more reliable. Many governments in poor countries struggle to provide such services, and until they do, the poor will remain vulnerable. Fourth, the exposure to hazards will rise in cities, but greater exposure need not

increase vulnerability. Large cities exposed to cyclones and earthquakes will more than double their population by 2050 (from 680 million in 2000 to 1.5 billion in 2050). The increase will differ by country and region. Vulnerability need not increase with exposure if cities are well managed, but the projected increase in exposure underscores the enormous task ahead. Urban growth is not the only concern. Climate change has received much attention,

and there are urgent calls for immediate action because the effects of climate change are cumulative and felt much later. The 2010 World Development Report discusses the implications of climate change in detail; this report is limited to its direct effects on hazards. One estimate of the increase in damage associated with changed tropical cyclone activity as a result of climate change is between $28 billion and $68 billion annually by 2100. This represents an increase of between 50 and 125 percent over no climate change. There is considerable uncertainty around these long-term projections, reflecting the limits of the data and the climate models that generate

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

them. The damage is in “expected value” terms, but averages hide extremes: a very rare and powerful cyclone could strike a highly vulnerable location causing extremely high damages. And the effects are likely to be concentrated: several small island countries in the Caribbean are particularly vulnerable. These four findings are not actionable prescriptions. Many people must do

numerous things better, but getting them to do so is the challenge. A successful policy response for effective prevention includes information, interventions, and infrastructure. Underpinning this policy response is the role of “institutions,” without which any policy response would be ineffectual. Governments can do much to promote prevention—in line with the policy implications outlined next.

Four policy implications (plus one for donors) First, governments can and should make information more easily accessible. People are often guided in their prevention decisions by information on hazards, yet the seemingly simple act of collecting and providing information is sometimes a struggle. While some countries attempt to collect and archive their hazard data, efforts are generally inconsistent or insufficient. Specifically, there are no universal standards for archiving environmental parameters for defining hazards and related data. Data exchange, hazard analysis, and hazard mapping thus become difficult. Figure 1 shows how few countries collect and archive data on hazards—even though technological advances such as the abundance of free, simple, and open source software (for example, PostGIS, Geoserver, Mapserver, the GeoNode.org project) should make collecting and sharing information easy. And where information is collected, it is not always shared, even though sharing

information on hazards involves relatively little expense because some government agencies already collect and analyze data on hazard risks. Those preparing background papers for this report had difficulties obtaining disaster and related data from various public agencies and universities, even though donors often funded the collection and automation of disaster data. Sometimes “security, commercial, and defense” reasons are invoked, but only a few are legitimate. Sometimes commercial interests take precedence over public good aspects. So, the importance of making information about hazard risks available cannot be

overemphasized. Perhaps because of this significance, the political will to not have information on rising levels of risk publicized is often strong. For example, even though the Federal Emergency Management Agency (FEMA) in the United States has updated coastal flood maps for the U.S. Gulf, it cannot get coastal communities to accept them because the information would reduce property prices. Systematic mechanisms for tracking information related to the changing nature of risk, and translating it into risk-related property valuations, would go a long way to increase the incentives for prevention. Making maps of flood plains and seismic fault lines easily

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

http://GeoNode.org

accessible would make developers and property owners more aware of the risks— and more motivated to build appropriately. Collecting data on weather and climate is also integral to producing accurate forecasts.

Figure 1 Number of countries that archive data for specific hazards

Source: World Meteorological Organization 2006.

Second, governments should permit land and housing markets to work, supplementing them with targeted interventions when necessary. When land and housing markets work, property values reflect hazard risks, guiding people’s decisions on where to live and what prevention measures to take. Detailed empirical work for this report matched some 800,000 buildings in Bogota that differed in their exposure to seismic risk to a range of characteristics (such as size, construction quality, distance from the city center, and whether residential, commercial, or industrial). Because the only difference among comparable properties is their level of hazard risk, this allowed assessing whether property values are lower in riskier areas. figure 2 shows that they are, suggesting capitalization of disamenities from hazard risk. But markets, when smothered, dampen the incentives for prevention. In Mumbai,

where rent controls have been pervasive, property owners have neglected maintenance for decades, so buildings crumble in heavy rains. Rent controls are not unique to Mumbai or developing countries. Rent control laws have remained in place in some form in New York City since 1943, where there are currently about a million rent-regulated and 50,000 rent-controlled apartments. As recently as 2009, legislation was passed in New York that limits the ability of landlords statewide to increase rents. Such laws are expected to return to regulation many household units previously attracting market rents. They exist in about 40 countries, including many developed countries. And rent controls are not the only market distortion. Real estate transactions in many countries incur a tax on sales, not on owning property. But taxing

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

transactions reduces property sales and encourages undervaluation. And restrictions on cement prices and imports can create black markets and exorbitant prices, so that adulterated cement ends up weakening structures.

Figure 2 Property prices for comparable properties are higher in locations farther from earthquake risk in Bogota

Source: Lall and Deichmann 2009.

Getting land and rental markets to work can go a long way to inducing people to locate in appropriate areas and take preventive measures. But this will not be a straightforward task. Nor will it be easy to remove the panoply of market distortions because many benefit vested interests. And knowing what to change first is not obvious. Past policies weigh heavily on the present: many structures now standing were built earlier, and defects are difficult to detect and harder to remedy. A corollary is that correcting policies now will not result in immediate improvements, though correcting them sooner would be better than delaying. Where new construction dominates, as in developing countries’ urban areas, this legacy is less of an issue, but wealthier countries also bear this burden: mispriced insurance (premia too low because of populist pressures on a regulated industry) has led to overbuilding along the hurricane-prone U.S. coastline. The poor bear the brunt of the cumulative effects of such policies (tax structure,

city financing arrangements, and so on) which produce only a limited and unresponsive supply of affordable, legal land sites for safer housing. Governments could greatly expand the choices of the poor—who often locate in dangerous areas and slums exposed to hazards—but this is more subtle than dictating what they should choose. Poor households prefer to have easier access to jobs, even though this may imply living in slums on riverbanks prone to flooding or on hilltops subject to mudslides. In some cases, security of property (clear titles often help) allows people to invest in prevention measures. When the social consequences of settling in hazardous zones are so adverse, the correct response is for governments to make targeted interventions. This could include making land available in safer locations—

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

along with adequate and reliable public transport and other services so that people remain connected to their jobs. Third, governments must provide adequate infrastructure and other public

services, and multipurpose infrastructure holds promise. Much prevention is embedded in infrastructure, but effectiveness depends on quality. Infrastructure needs maintenance: fixing potholes in the road before the winter or the rains; painting steel bridges before they weaken through corrosion; inspecting and fixing cracks in concrete bridges. All engineers know this, but they do not always obtain budget appropriations—even in the United States, where the 2007 bridge collapse in Minneapolis drew attention to such neglect. Spending should go down a list arranged in descending order of (economic) rates

of return. But when subject to arbitrary budget spending limits and lumpiness, low- return spending often gets put ahead of postponable high-return spending. Since maintenance can be postponed, it gets deferred—repeatedly—until the asset crumbles. Drainage ditches, once built, are not adequately maintained and become clogged; so rains result in floods that drown the poor. Other less obvious public services include reliable city transport, and these require better—not always more— public spending. For example, about 30 percent of infrastructure assets of a typical African country need rehabilitation, and just $0.6 billion on road maintenance would yield $2.6 billion in annual benefits (figure 3). Governments must ensure that new infrastructure does not introduce new risk. This

is particularly important since, in many developing countries, infrastructure investment —long-lived capital stock—is likely to peak in the coming few decades. Locating infrastructure out of harm’s way is one way of doing so. Where that may not be possible, another way is to execute multipurpose infrastructure projects, such as Kuala Lumpur’s Stormwater Management and Road Tunnel (SMART). Floods from heavy rains are a hazard, and the 9.7 kilometers long $514 million tunnel has three levels (figure 4), the lowest for drainage and the upper two for road traffic. The drain allows large volumes of flood water to be diverted from the city’s financial district to a storage reservoir, holding pond, and bypass tunnel. Combining the drain with the road has two advantages: it ensures maintenance of a drain that otherwise would be used only sporadically, and it costs less than building each separately.

Figure 3 Underspending on maintenance implies an enormous infrastructure rehabilitation backlog in Sub- Saharan Africa

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Note: The rehabilitation index shows the average percentage across countries of each type of infrastructure in poor condition and thus in need of rehabilitation. Source: Briceño-Garmendia, Smits, and Foster 2008.

Figure 4 Three modes of operation of the SMART Tunnel

Source: Mott MacDonald Group 2009.

Infrastucture, even when well designed, constructed, and maintained, cannot always prevent disasters. Governments must, therefore, pay heed to a subset of “critical infrastructure” that once selected, is subject to higher than usual “margins of safety” (the extra strength that engineers build into designs). Such critical infrastructure must be identified before a disaster to ensure its adequacy. But what is critical is situation specific—safe schools serve as cyclone shelters in Bangladesh, but hospitals (not schools) may be more critical in Turkey to treat crushed limbs when buildings collapse in earthquakes. And governments must be careful about keeping the list short: when it includes too many assets, the costs rise without commensurate benefits. Even the United States encounters difficulties in keeping critical infrastructure manageably small, and other governments will undoubtedly discover this as well. Fourth, good institutions must develop to permit public oversight. Good institutions

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

both reflect and create prosperity, and one robust finding of this report is that countries with well-performing institutions are better able to prevent disasters, including reducing the likelihood of disaster-related conflict. But institutions transcend specific entities. Parliaments, media, business associations, and the like function differently across countries—even if they have similar legal authority and responsibilities. Fostering good institutions means letting evolve a messy array of overlapping

entities (the media, neighborhood associations, engineering groups) that may not all have lofty motives but nevertheless allow divergent views to percolate into the public consciousness. Permitting dissent allows the public to be informed and involved when alternative proposals and opposing views compete for their support. Public involvement and oversight ensure that good ideas are considered even if they are unusual (Kuala Lumpur’s dual-use drain and car tunnel). Such oversight also encourages communities to experiment with, and to devise, their own sustainable arrangements that promote prevention. Where institutions have been suppressed, results are discouraging. Storm damage

is more severe in Haiti than in the adjoining Dominican Republic. Deforestation is the visible difference (figure 5) but the quality of institutions is the less visible one. Haiti’s institutions and communities have withered from decades of misrule. Vibrant communities help ensure that trees are not thoughtlessly felled and that saplings planted will grow. Even if the interest of uplanders who cut the trees may diverge from lowlanders who get the mud flows, communities bridge these differences and manage the fair use of the commons. Prosperity ultimately depends on rebuilding the trust and social capital that was lost even before the earthquakes and hurricanes struck. Often, institutions are linked to democracy, but this report finds that it is not the

label of democracy or dictatorship that matters. Good institutions are associated with political competition more than voting alone (the conventional understanding of democracy). Across both nondemocracies and democracies, the existence of “institutionalized” political parties—parties that allow members to discipline leaders who pursue policies at odds with member interests—is significantly associated with reductions in disaster mortality. The mortality from earthquakes falls by 6 percent for an additional year of competitive elections, and by 2 percent when the average party age rises by a year. Such systems are therefore more likely to respond to citizens’ needs.

Figure 5 The visible border between Haiti and the Dominican Republic

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Source: National Geographic.

Preventing disasters requires many public and private agencies to work well together, and governments could play an institutional role in this. But there is no single recipe for strengthening institutions; a wide variety of political systems can serve the purpose. But encouraging a diverse set of organizations that facilitate collective action by large groups of citizens will allow them to press more effectively for the spread of information, the availability of prevention measures and alternatives, and their cost- effectiveness. And fifth, donors have a role in prevention as well. The report’s overarching theme

is that not enough is being done on prevention. Donors usually respond to disasters after they strike: about a fifth of total humanitarian aid between 2000 and 2008 was devoted to spending on disaster relief and response (figure 6). The share of humanitarian funding going to prevention is small but increasing—from

about 0.1 percent in 2001 to 0.7 percent in 2008. However, prevention activities often imply long-term development expenditures whereas the focus of humanitarian aid— already a tiny part of official development aid—is immediate relief and response. Donors concerned with prevention could earmark official development aid (rather than humanitarian aid) for prevention-related activities. And such aid, if used effectively, could reduce issues arising from the Samaritan’s dilemma: the inability to deny help following a disaster to those who have not taken sufficient prevention measures.

Figure 6 Disasters receive about a fifth of total humanitarian assistance

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Note: Humanitarian aid is “an intervention to help people who are victims of a natural disaster or conflict meet their basic needs and rights,” while official development assistance (ODA) is “money spent on development (education, health, water supply and sanitation, agriculture, and so on) and humanitarian assistance by members of the OECD Development Assistance Committee.” Source: World Bank staff based on data from the Financial Tracking System (FTS) of the UN Office for the Coordination of Humanitarian Affairs.

In addition to these policy implications, readers may find much of interest and use in the report, summarized in the rest of this overview.

Disaster data patterns revisited There have been 3.3 million deaths from natural hazards since 1970, or about 82,500 a year, with large year-to-year fluctuations and no pronounced time trends. Droughts are the deadliest of the four hazard categories (earthquakes, floods, and storms are the others) and poor countries suffer disproportionately—almost 1 million people died in Africa’s droughts alone. Poor countries withstand the worst of disaster deaths (map 1). Despite the avoidable deaths, the absence of a pronounced upward trend suggests

the picture is less bleak than it first appears: exposure is rising rapidly (such as poor countries’ population, both total and urban) yet deaths would trend down if scaled by the relevant population. So, there has been some effective prevention. Data on property damage are less comprehensive than those on deaths, but

damage from all hazards between 1970 and 2008 totaled $2,300 billion (in 2008 dollars), or 0.23 percent of cumulative world output. Damages fluctuate with a modest but discernable upward trend even when adjusted for inflation. They vary by hazards, with earthquakes and storms causing the most damage. And they are disproportionately high in middle-income countries. Again, the data suggest some effective prevention: if damages are scaled by GDP (globally or by country), they generally trend down.

Map 1 Deaths shrink Asia and the Americas—but expand Africa

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Note: Areas reflect cumulative deaths from disasters for 1970 to 2010 (February). Source: World Bank staff based on EM-DAT/CRED.

Even when scaled by output, poor countries with few assets incur little damage, and rich countries (with more capital) effectively prevent damage. Middle-income countries incur the greatest proportional damage (map 2), suggesting why absolute damage has been rising. Institutions that prevent damage develop more slowly than assets as countries

urbanize and prosper. But this is not immutable: even poor countries can undertake effective prevention, and more can rise to the challenge of doing so.

Disasters’ many effects A disaster obviously hurts those affected. It also spares many in the affected area, yet those spared may be indirectly affected. The village tinker’s and tailor’s businesses suffer when a cyclone spares their premises but destroys their customers’ crops. And such indirect effects extend beyond the affected area, which is linked to undamaged areas through commerce. These indirect effects are often—but not always—adverse. Disentangling the effects is difficult, but clarity of concepts can help, starting with measurement. How much output falls in the affected area, and for how long, has been

controversial. Many factors (simultaneous changes in commodity prices, terms of trade, exchange rates) affect output, and studies differ on whether and how these are taken into account when measuring the effect of a disaster. A disaster may affect only a small part of a country, so it may not reduce national output to the same extent as in the affected area. Studies for this report find that national output always falls after a severe disaster, but (depending on the hazard) sometimes rises after a mild one. An earthquake reduces output, but subsequent reconstruction increases economic activity—though people are obviously worse off. Economic growth is output’s rate of change, so even if output recovers only to its former level after falling, growth (for a brief period) would be higher than pre-disaster rates.

Map 2 Damages shrink Africa but expand middle-income countries

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Note: Areas reflect cumulative damage from disasters scaled by GDP for 1970 to 2008. Source: World Bank staff based on EM-DAT/CRED.

Output does not measure peoples’ well-being, especially following a disaster. And not everyone is affected equally—even in the affected area. Farmers who have not lost their crops get higher prices if overall harvest were lower. So the indirect effects —especially in the area outside the disaster zone—are not all adverse. Governments often assess the damage after a disaster, and such assessments

differ in scope, purpose, and technique. The report discusses the conceptual and practical issues in measuring damage and the direct and indirect effects from a disaster. Measuring damage is tricky, prone to both overestimation (for example, double counting) and underestimation (it is difficult to value loss of life, or damage to the environment). Biases also affect the accuracy of estimates, especially when the prospect of aid affects incentives. Accurate measurement is more likely when its purpose is clear, though some items

of interest cannot be measured. Damage assessments have multiple and often overlapping purposes. They could guide government relief (such as how much to spend on alleviating the victims’ suffering, knowing that other spending must be cut or taxes raised). They could show how to hasten economic recovery or identify specific measures to improve prevention. This report examines the conceptual and practical feasibility of meeting each purpose. People do not wait for help to begin repairing their homes and rebuilding their lives,

but the poor, with nothing to fall back on, may require help. The government often provides transfers in cash and kind, but “compensation” is a misnomer because the amounts (typically less than twice per capita GDP) are usually less than what people have lost. The government’s fiscal situation limits these transfers because even if it could borrow, the debts must be later serviced. So, knowing the disaster’s medium- term fiscal implications would be more useful than measuring the damage to private property. If relief and recovery spending displace maintenance of infrastructure, as they often do, the deaths and damage from future disasters would rise. Recovery requires that commerce resume, and this involves restoring the affected

area’s links with the rest of the economy. It is in the self-interest of people and private The, World Bank, Bank World, and Nations (UN) United. Natural Hazards, UnNatural Disasters, edited by World Bank The, et al., World Bank Publications, 2010. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/apus/detail.action?docID=606027. Created from apus on 2017-04-04 13:20:21.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

firms up and down the supply chain to repair these links (banking, trucking)—but physical infrastructure (roads, bridges, railways) is often the government’s responsibility. Assessing damage to public infrastructure is urgent, and governments must quickly decide what, where, and whether to rebuild. This decision will in turn affect individual decisions to rebuild. Who in government decides depends on the country’s administrative structure, and the people affected are best placed to guide the choice of which road or bridge to repair first. A disaster’s effect on an economy’s output, or on the government’s budget, is not

the same as its effect on people’s health and well-being. A disaster undoubtedly reduces the well-being of those affected—and even if survivors recover and consume at their earlier levels, they will have suffered in the immediate aftermath. Many studies have examined how disasters affect people in the short run, and this

report complements those studies with others that find longer lasting adverse effects on schooling, cognitive abilities, and mental health. Some survivors are pushed over the edge and never completely recover: widespread droughts in Africa result in stunted and malnourished children, with permanent adverse effects. An effective safety net can reduce these consequences, but not every safety net is effective. The literature has long noted that disasters and conflicts are connected. Hazards,

particularly earthquakes and droughts, tend to prolong conflicts, but good institutions reduce the likelihood of their erupting. Such institutions are typically associated with democracy and good governance—factors also associated with prosperity. This report finds that the link is through political competition rather than voting alone. Do disasters increase scarcity and thus conflict? Or do they create an opportunity for peace, as in Aceh? Either is possible, and good institutions make the better outcome more likely.

Prevention by individuals The analytical framework of prevention, insurance, and coping has proved useful in many settings, and the report is structured around these concepts, distinguishing individual choices and collective decisions (at different levels of government). People choose how much prevention to undertake (consciously or by default), how much insurance to purchase, and how much residual risk to bear through coping. Is individual prevention adequate and effective?

Figure 7 Private preventive measures pay

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Note: Key prevention measures for which benefit-cost ratios were calculated are: elevating a house by 1 meter to reduce damage from floods (Jakarta); protecting windows and doors and upgrading roofs to prevent hurricane damage (Canaries and Patience, St. Lucia); retrofitting buildings to increase quake resiliency (Istanbul); and flood-proofing a brick house (Rohini River Basin, Uttar Pradesh, India). Source: IIASA/RMS/Wharton 2009.

People undertake prevention to the point where expected benefits (avoiding losses) exceed the costs—subject to their budget constraint (figure 7). But people differ and everyone chooses differently. Such differences do not necessarily imply that some choose badly, but it is reassuring when large numbers take prevention measures that seem well justified. Rates of returns for several commonly-used prevention measures —such as raising the plinth in flood-prone Jakarta, or protecting windows and doors against wind and rain damage in the Canary Islands—show that some measures are warranted but not all. One person’s choice may puzzle another: many live in exposed areas known to be

hazardous—whether in poverty in Bangladesh or in affluence along the Florida coast. Recent theories and experimental findings show that people sometimes misperceive risks and may not always act in their own best interests. But there also are more prosaic explanations involving tradeoffs such as proximity to work and access to such conveniences as public transport, given limited budgets. Living in riskier locations is cheaper for the individual and allows spending on other

necessities (food, children’s schooling), so the poor face difficult choices. Safer

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

structures could be built in risky areas (on hill slopes, in seismic areas) with sufficient knowledge, care and expense. But when a person’s ownership of property is not secure, the possibility of eviction or demolition erodes the incentive to invest in safe structures. A study of 1.2 million land titles distributed in 1996 in Peru finds that land titling is associated with a 68 percent increase in housing renovation within four years. Insecurity of land holdings is not the only disincentive to build well: rent controls or

other similar regulations erode a landlord’s incentive to maintain buildings. The situation in Mumbai, India, where neglected buildings collapse in severe storms, killing occupants, is described in some detail. Mumbai has had rent controls and distorting taxes whose adverse effects have accumulated over decades. Buildings were restricted to being only a few stories tall, hindering agglomeration, and decaying industries occupy land that could be put to better use. Such policies also contribute to the dearth of good housing and to the poor living in unsafe shanty towns that mushroom in and around prospering cities. They have also starved cities of tax revenues, so the needed infrastructure is not built, or is built badly. Structures are also shoddy because people do not always know the hazards they

face or what it takes to build well. Detailed accounts from Italy, Pakistan, and Sri Lanka illustrate the challenge of improving building practices, the importance of information (about hazards and how to build better), and the limited role of building codes. Calls for stronger building codes reverberate after a disaster, and stricter

enforcement becomes the siren call. But there are few improvements if private owners and builders view these codes as yet another hurdle to overcome, or if officials are corrupt or complacent. Like any regulation, codes are also susceptible to capture by vested interests (California’s first building code of 1933 sought to prevent the use of steel to protect the jobs of bricklayers, even though unreinforced brick structures are highly vulnerable in seismic areas). Codes work through “institutions,” and are one cog in a complex system of inculcating better building practices. They are most effective when they contain accurate and accessible information about hazard risks and the properties of newer building materials, and when there are incentives to build sound structures (for example, private owners having clear title). Good building practices can be fostered even without a code, as the rebuilding after the 2005 earthquake in the remote and mountainous region of Pakistan shows.

Prevention through governments Governments can help in effective prevention, but struggle to do so. It is difficult to measure how much governments spend on prevention because this is not a specific budget item. Detailed analysis in Colombia, Indonesia, Mexico, and Nepal found that prevention spending was less than post-disaster spending except in Colombia (figure 8). But this does not imply that it was “too little,” for it is hard to isolate what

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

constitutes prevention and even harder to determine adequate spending. Effective prevention measures are often embedded in other spending (in such

infrastructure as an embankment), and there are indications that reversing the past neglect of maintenance (painting bridges to reduce corrosion and subsequent failure) and investing in intangibles (tallying decrepit structures) has large benefits. So, why does it not happen, and who determines government spending? Some assert that politicians are short-sighted, but competition in the market for votes, like other competition, would generally provide what the public wants. In the United States, voters favor relief spending over prevention, leading some to conclude that voters (not politicians) either are myopic or misperceive hazard risks. The findings are equally consistent with far-sighted voters being skeptical (perhaps justifiably) of politicians’ ability to organize prevention effectively. The challenge for governments is to translate spending into effective prevention,

and cost-benefit analysis is a useful tool—but one that must be used with care. The benefits of prevention are understated if human lives are not valued, but attaching a value to life has enormous moral and ethical implications. Most government spending, especially on prevention, has distributional implications: a dam protects one group but may increase the flood risk of another. Cost-benefit analysis attaches implicit weights —and while these could be explicitly changed, officials lack the moral authority to decide unilaterally. Such decisions require a political consensus that countries with good institutions possess. Cost-benefit analysis is a filter that can rank alternatives, not a scoop that can

generate options. Prosperous countries have better prevention because they also have good institutions that oversee government decisions. Such oversight cannot be only through legislative bodies. And broader involvement requires the government to fully disclose what it knows and does—transparency not just about a decision but the entire process—and to encourage (not just grudgingly tolerate) dissenting views.

Figure 8 Post-disaster spending fluctuates more than pre-disaster spending

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Source: de la Fuente 2009.

The report identifies three specific spending items desirable for prevention. An early warning system can save lives and property. There have been many advances in weather prediction technologies, but few countries have taken full advantage of that. The report outlines these technological developments and how a modest but well allocated increase in spending—and sharing real-time data internationally—would benefit countries. Critical infrastructure that functions during and after a disaster reduces the loss of

life and property. While all infrastructure must be well designed, constructed, and maintained, designating a subset as “critical” allows the government to pay special attention to it. What is critical depends on the situation and the hazard. Critical is not synonymous with the importance of some infrastructure in normal times: the choice requires informed judgment. Environmental buffers offer protection from hazards within physical limitations.

Forests and wetlands offer little protection from extreme floods when soils are already saturated. Similarly, mangrove belts a few hundred meters wide can significantly reduce the destruction from a small tsunami but not a big one. Protecting the environment is cheaper than restoring it, but knowing what to protect is hard because development involves change, and many changes are unforeseeable. But

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

some who seek to protect the environment may have also exaggerated the benefits in cost-benefit analysis: careful analysis is important but difficult. Again, good institutions help: when more people observe—and question—what is happening, better things get done. When governments make what they know freely available and what they decide transparent, the benefits invariably follow.

Insurance and coping People do not take prevention measures to eliminate all risks—nor can they. Insurance and other measures (borrowing, setting funds aside, remittances) “soften the blow” when disasters unfold. But these measures, even though designed for and executed in ex-post situations, also affect prevention and are examined from that perspective. Insurance transfers risk to those willing to bear it. It clearly increases a person’s

choices and thus well-being, but softening the blow dilutes the incentive to prevent, unless the insurance premium reflects the risk and the prevention measures that a person undertakes. The premium must also cover the considerable costs of administration, marketing, and monitoring. Many people forgo insurance if the premia are too high, so commercial insurance develops only for some risks—and in countries where enough people want it. Parametric insurance (where the payout is specified, so that incurred damages do not have to be ascertained) reduces some of the monitoring costs. But such schemes have low penetration rates in developing countries, partly because of a lack of detailed data on the frequencies and intensities of hazards and exposed assets. Insurance invariably draws in the government—as regulator, as provider (in many

countries), or as reinsurer—inevitably adding a political dimension. This often results in attempts to lower the premia through subsidies (as with flood insurance in the United States), or, conversely, to favor insurers by keeping premia high or keeping out competition. An inappropriate premium has adverse effects that are difficult to rectify later: too low a premium encourages construction in hazard-prone areas (vacation homes in Florida). Whether governments should buy insurance against disasters is not as clear-cut as

it may seem: the alternatives are to set funds aside in reserve or to borrow so that they have funds to spend after a disaster. Many governments are already indebted, and even those with low debts may find it difficult to borrow when they most need to. While individuals are risk-averse, there are good reasons for some governments acting on their behalf to be risk-neutral. A risk-neutral entity would buy insurance only if the premium were lower than the probability times the expected loss (which leaves nothing to cover the insurer’s costs). This argues against governments buying commercial insurance. But a disaster that is large relative to their economy’s size (as in the Caribbean, where the main unknown is the island that gets hit) may make some

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

governments risk-averse, and insurance could be beneficial. The Caribbean Catastrophe Risk Insurance Facility, which pools disaster risks

regionally, helps countries in such circumstances purchase insurance less expensively than otherwise. Prices that insurance firms offer may differ from prices in capital markets, and such price comparisons can produce large savings, as Mexico found when issuing catastrophe bonds. The World Bank’s Catastrophe Risk Deferred Drawdown Option is a loan that disburses quickly, to provide immediate liquidity if and when the borrowing government declares an emergency. What cannot be prevented or insured against must be borne, and a variety of

coping mechanisms (“informal insurance,” as distinct from market insurance) have developed over the centuries, many embedded in tradition and custom. Private individuals and groups abroad send remittances directly to those they know, and such remittances surge after a disaster, even when there is no media coverage. The funds arrive quickly to help people cope. While remittances are routinely spent on consumer durables, some improve the

quality of housing. Houses made sturdier could be considered a prevention measure, though the situation varies. In Turkey, 13 years after the 1970 Gediz earthquake, the reconstructed area was peppered with improperly reinforced concrete houses— mostly paid for by the earnings of family members in Germany. Better building practices are needed to ensure safe buildings. But not all who need help receive remittances, and there are sometimes impediments to such flows that the government could remove (controls on capital flows, dual exchange rates). Private remittances also help develop banking and money transfer facilities that strengthen an area’s commercial ties with other parts of the country and the world. Aid also has a role in prevention, but it can be double-edged: while some aid is

warranted, it can also give rise to the Samaritan’s dilemma. Some observers have noted the disincentives donor programs can create—they can, for example, erase a country’s incentive to provide its own safety nets. Nicaragua declined to pursue a weather indexing program after it had been priced in the global reinsurance market: it cited international assistance following Hurricane Mitch in 1998 as an indication of dependable alternatives. Some new but not very strong evidence suggests that post- disaster aid reduces prevention. It may be unfair, though, to blame only countries for neglecting prevention: Mozambique, anticipating major floods in 2002, asked donors for $2.7 million to prepare and got only half the amount, but $100 million were received in emergency assistance following the floods, with another $450 million pledged for rehabilitation and reconstruction. Vibrant communities, however, use aid well. The main lesson for donors is to be

aware of the potentially adverse effects of their actions. Governments in recipient countries can do much to prevent waste that may result from a sudden flow of uncoordinated aid or from inappropriate aid in kind.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Game-changers? Burgeoning cities, climate change, and climate-related catastrophes That urban areas and population will grow is certain; but which cities will grow, and how fast, is less predictable. Most growing cities are in developing countries, and growth increases exposure to hazards (map 3). The growing density of people and economic activity will change the economics of

effective prevention. But greater exposure need not increase vulnerability if cities are well managed. Climate change complicates this further. The scientific models to forecast weather

do not allow confident projections at the local level, but the intensity, frequency, and distribution of hazards will change with the climate. The expected annual damage from climate-change induced tropical cyclones alone could be in the $28 billion to $68 billion range. These estimates, sensitive to various parameters and assumptions about the future, are in “expected value” terms per year. But the damages are not expected to come in a steady stream. Climate change is expected to skew the damage distribution of tropical cyclones and is likely to cause rare—but very powerful —tropical cyclones to become more common. This report finds that for the United States, destructive storms that would come every 38 to 480 years given the current climate, would come every 18 to 89 years with future climate change. Climate change “fattens the tail” of the tropical cyclone damage distribution (figure 9). Even though very rare and damaging storms are part of today’s climate, they will become more frequent in a warmer climate.

Map 3 Exposure to cyclones and earthquakes in large cities may rise from 680 million people in 2000 to 1.5 billion people by 2050

Source: Brecht and others 2010.

Figure 9 Climate change shortens the return period of large storms

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Note: The figure shows the return period for tropical cyclones of different intensity in the United States for one specific climate model (MIROC). A $100 billion storm is estimated to happen once in a 100 years in the United States given the current climate. With a future warmed climate, it is expected to happen once in about 56 years. Source: Mendelsohn, Emanuel, and Chonabayashi 2010a.

Scientists have identified several catastrophes that a changing climate might trigger: drastic sea level rise, disruption of ocean currents, large-scale disruptions to the global ecosystem, and accelerated climate change, for example, from large releases of methane now trapped by permafrost. Catastrophic risks and costs need to be weighed differently than less severe events. Prudence in responding to catastrophic threats calls for a portfolio of measures that emphasizes learning and mid-course corrections. A broad portfolio is desirable because the potential effectiveness of individual measures is uncertain. Cities, climate, and pending catastrophes are altering the disaster prevention

landscape. While hazards will always be with us, disasters show that something has failed. But determining what has failed and deciding on the corrective measures are not always obvious. And arguing whether Hurricane Katrina or Cyclone Nargis occurred as a result of climate change detracts attention from policies that continue to misprice risk, subsidize exposure, reduce individuals’ incentives to reduce risk, and promote risky behavior in the long run. People rise out of poverty through better technology, greater market access, and

more investment in activities that spill benefits from one set of economic actors to others through greater interdependence, higher productivity, and stronger institutions. Living in cities facing serious risks of inundation is undesirable, but a failure to significantly reduce poverty would be even more undesirable. Fortunately, neither is inherently necessary. People acting individually and through responsive governments can prosper and survive. Progress requires and results in better institutions: those, after all, are the basis of sustainable development.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

CHAPTER 1

Fluctuating Deaths, Rising Damages—the Numbers

arthquakes, storms, and other hazards killed about 3.3 million people between 1970 and 2010, an annual average of 82,500 deaths worldwide in a typical

year, a small fraction of the roughly 60 million who die every year and of the 1.27 million killed in traffic accidents alone (WHO 2009). Disasters kill many simultaneously and affect many more but evoke more attention than the numbers warrant. For example, for every person who dies in an earthquake, more than 19,000 people must die of food shortage to receive the same expected media coverage, all else equal (Eisensee and Strömberg 2007). That the attention comes from sensational media coverage (“sell newspapers”) is a circular explanation. Psychologists, sociologists, anthropologists, and others offer different explanations for our emotions: how one dies matters, and our reactions differ whether a person drowns while fishing for a living, surfing for fun, or in a flood that washes a home away. Our emotional reaction may be accentuated by a perceived lack of control over the

event (Acts of God). But natural disasters, despite the adjective, are not “natural.” Although no single person or action may be to blame, death and destruction result from human acts of omission—not tying down the rafters allows a hurricane to blow away the roof—and commission—building in flood-prone areas. Those acts could be prevented, often at little additional expense. This report is about prevention—measures that reduce the risk of death, injury, and

damage from disasters—and how to ensure it cost-effectively. Post-mortems of disasters often find that much of the death and destruction could have been averted fairly inexpensively, but this is misleading. Consider $2 billion in damage from a disaster that could have been prevented by measures costing “only” $20 million annually. If the hazard occurs only once in 200 years, the expected annual loss is $10 million (= $2,000,000,000 × 1/200), and the $20 million on prevention could be better spent elsewhere. Prevention is economical in this numerical example only if the event were more frequent, the damage greater, or the prevention cheaper.

Box 1.1 The framework for the report

Disasters occur when households and assets are both exposed and vulnerable to natural hazards. Preventing disasters thus means undertaking measures that reduce exposure and vulnerability to contain deaths and damages. Not all disasters can be prevented however, and impacts depend on how individuals and governments react and cope.

Box figure 1.1. The framework for the report

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Chapter 1 describes the distribution of damages and deaths from natural disasters by type of hazard and countries, and what this may imply. Chapter 2 looks at the effects of disasters on welfare, and on economic aggregate output and government finances, and how these are measured. Chapter 3 presents a simple framework to understand how individuals manage risk. Chapter 4 starts with a discussion of disaster spending priorities and who determines them. Various collective preventive measures are then discussed (early warning systems, protecting critical infrastructure, and environmental buffers). But people cope, and how they do and the role of insurance and aid flows are covered in chapter 5. Exposure, vulnerability, and hazard patterns change over time, and chapter 6 provides some perspective on the future in the context of urbanizing cities and a changing climate.

Source: World Bank staff.

Prevention measures differ in cost and effectiveness. A person can undertake some unilaterally, such as building a house on a higher plinth—and others collectively, such as building an embankment. Some individual measures substitute for the collective (a house on stilts instead of flood embankments) and others are complements (cholera may increase during floods, but installing a septic tank is pointless if others do not). What people do affects others: those behind an embankment, for example, are protected from floods, but the redirected waters could increase damage elsewhere. And even those behind the embankment would incur greater damage if there were a breach; so embankments lower the risk of modest damage and increase the (low) risk of severe damage. These complexities are examined in later chapters. This chapter simply presents the related data and patterns (box 1.1).

Box 1.2 Understanding the terms in the report The terms in this report are used differently across disciplines.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Hazard is a natural process or phenomenon (floods, storms, droughts, earthquakes) with adverse effects on life, limb, or property. Hazards differ in severity, scale, and frequency and are often classified by cause (such as hydro-meteorological or geological).

Exposure is the people and property subject to the hazard.

Vulnerability is a characteristic that influences damage: some communities absorb and recover more readily than others because of physical assets (building design and strength), social capital (community structure, trust, and family networks), and political access (ability to get government help and affect policies and decisions). Measures to reduce vulnerability include mitigation (which reduces the hazard’s likelihood, as in reforesting the slopes to prevent rapid runoff and floods or reducing greenhouse gas emissions to reduce the frequency and intensity of extreme weather events), prevention (measures to reduce damage, as with higher plinths for floods), preparedness (evacuation plans), and relief (help after a disaster).

Disaster is the hazard’s effect on society as a result of the combination of exposure and vulnerability. So strictly, disasters, not hazards, cause deaths and damage.

Disaster risk is often calculated as a multiplicative function of hazard, exposure, and vulnerability. It is multiplicative because for disaster risk to exist, all three—hazard, exposure, and vulnerability—have to be present.

Deaths are readily counted, but injuries require some judgment about their seriousness. Those with broken limbs are included, but what about those with mere scratches—or major mental depression—that go untreated? Differences in criteria, and how data are gathered in practice, make comparisons across countries (and time) difficult. The numbers of affected persons (injured, homeless, and in need of immediate assistance) often measure the scale of the disaster; but adding the homeless to those whose farmland was temporarily flooded implicitly accords each equal importance.

Note: For formal definitions, please see http://www.unisdr.org/eng/terminology/terminology-2009-eng.html. Source: World Bank staff.

Some reports on disasters have noted a rising toll that has set off alarms with calls for action. While some actions may be appropriate, it is important to know how the numbers are collected and analyzed and what they may imply. (Box 1.2 explains the terminology and box 1.3 discusses the various data this report uses.)

Box 1.3 Global natural hazard databases: Varied purposes, varying details The three main global sources for data are EM-DAT, NatCat, and Sigma. EM-DAT is the acronym for data that the Center for Research on the Epidemiology of Disasters (CRED) has been collecting by country since 1988 (going back to 1900). CRED collects data on deaths, injuries, and damage from news accounts and other (unspecified) sources for earthquakes, hurricanes, floods, and other disasters that killed 10 or more people, affected at least 100, or resulted in a “state of emergency” or a call for international assistance. Other databases are by event, not country, with different inclusion criteria (and so are not strictly comparable). Munich Reinsurance Company maintains NatCat, and Swiss Reinsurance Company (Zürich Re) maintains Sigma (fewer events but includes both insured and uninsured damages). For 1988–2002, EM-DAT reports 756 million people affected, NatCat reports 277 million, and Sigma 19

million (Guha-Sapir 2002). All these databases have differing levels of detail and have their own strengths and weaknesses.1

Larger numbers do not necessarily attest to more comprehensive data because there are, sometimes, exaggerated reports of mortality; and insurers ignore countries with few commercial prospects. An example cited in box 2.3 of the UN 2009 Global Assessment Report on Disaster Risk Reduction is the 1999 landslides in

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

http://www.unisdr.org/eng/terminology/terminology-2009-eng.html

Venezuela. World Bank reports put fatalities at 50,000; EM-DAT at 30,000; but Universidad Central de Venezuela anthropologist Rogelio Altez, who carefully examined detailed death records in each state, concluded that fewer than 700 died (Altez 2007). CRED’s data (EM-DAT) is the only publicly available global disaster database. It is used to portray trends in

this chapter, and for some cross-country empirical analyses in other parts of the report. EM-DAT records disasters (a disaster triggers the inclusion into the hazard category; for example,

earthquakes in uninhabited areas are not recorded), and trends related to these records are analyzed and shown further in this chapter for the 40 years between 1970 and 2010. For presentational purposes, disasters are divided into five hazard categories: droughts, earthquakes (which also include tsunamis, volcanoes and dry mass landslides), extreme temperatures (heat and cold waves), floods (which include wet mass landslides caused by rains), and storms (including cyclones and typhoons).

Source: World Bank staff.

3.3 million deaths in the 40 years to 2010 Some 3.3 million people died between 1970 and 2010, and the deaths fluctuate considerably: some years with many deaths punctuate several years with few deaths (figure 1.1). Some short spans suggest a trend (1973 to 1975, or 1993 to 2005), but statistical tests attach a low confidence level for an overall upward trend.2 The absence of a pronounced upward trend in mortality when population and exposure (those living in the hazard prone areas) have risen dramatically suggests that some prevention measures have likely been effective. More people were affected in the two recent decades than earlier. This increase

may reflect greater exposure to hazards, or better reporting in recent years, or both. Half the world’s people now live in cities up from 30 percent in 1950, and most large and rapidly growing cities in poor countries struggle to provide public services, including disaster prevention. Individuals build shacks in the flood plain or on steep hillsides vulnerable to hazards, an issue examined in greater depth in chapter 3.

Figure 1.1 Deaths fluctuate—the number of people affected is on the rise

Source: World Bank staff based on EM-DAT/CRED.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Disasters can strike anywhere Disasters affect all regions (figure 1.2). Floods and storms are the most common, while droughts much less so (except in Africa) (figure 1.3).3 Deaths are more concentrated: droughts in Africa are the deadliest; storms in East and South Asia also take many lives (figure 1.4). Differences across countries suggest that some countries prevent disasters better

than others. The contrast in the death toll in Haiti and the Dominican Republic, sharing the same island and storms, underscores the point that disasters are manmade, not natural. We are capable of reducing the death toll even in poor countries: moving food averts a famine despite droughts; early warning systems reduce deaths caused by storms and floods. Clearly, more can be done to reduce deaths, but property cannot flee from an imminent hazard, so we turn to damage.

Damages are rising The annual global damage from disasters between 1970 and 2010—adjusted for inflation—fluctuates like deaths but is also rising in spurts.4 Damage in the recent two decades is significantly greater than in the earlier decades (figure 1.5). This could reflect greater exposure, or better reporting, or both. Most of the damage is from storms, earthquakes, and floods—in that order.

More so in rich countries, less in poor Rich countries (North America, Europe, and increasingly Asia) incur greater absolute damage (though not relative to GDP). The damage is least in Africa, where the poor possess little (figure 1.6). Earthquakes and storms are the most destructive, again not surprising because they affect valuable structures, often in richer countries.5

Figure 1.2 Disasters affect all regions

Note: In this report, Africa corresponds to the African continent and islands; East Asia to Asia except for south and central Asia; South Asia to countries south of the Himalayas plus Afghanistan; Europe and Near East to Europe and former Soviet republics, and to Turkey, Iran, and countries east of Suez, including the Arabian peninsula; Northern America to Canada and the United States; Central and Southern

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

America to Mexico and countries south, including the Caribbean; and Oceania to Australia, New Zealand, and the Pacific islands. Countries are grouped by continent and shared seismic characteristics and other common hazard characteristics. Number of events by region, 1970–2010 (February).

Source: World Bank staff based on EM-DAT/CRED.

Figure 1.3 Disasters almost everywhere (1970–2010)

Source: World Bank staff based on EM-DAT/CRED.

Figure 1.4 Droughts deadliest in Africa, earthquakes elsewhere

Note: Number of events by type of event and by region 1970–2010 (February).

Source: World Bank staff based on EM-DAT/CRED.

Figure 1.5 Damage on the rise in the last two decades (global damage from hazards, 1970–2010)

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Source: World Bank staff based on EM-DAT/CRED.

Small island economies are hit hard Because absolute damage is larger in rich countries (with more assets), the numbers are often scaled by GDP (a flow, though damage is of the asset stock) to allow comparisons among countries. That disasters have a higher cumulative impact on small economies is already known, but this exercise quantifies that impact more accurately.6 Many of the 25 countries with damages more than 1 percent of GDP (in a sample of 175 countries) are small island economies (figure 1.7). Even a single event can adversely affect the economy of a small, vulnerable country. And even though damage is less than 1 percent of GDP for 86 percent of countries, a country’s GDP is irrelevant to the victims who may lose all they possess. A high ratio of cumulative damage to GDP suggests when help from international donors could be useful, but most are “repeaters” that need prevention more than relief.

Figure 1.6 More damage in rich countries, mostly from earthquakes and storms

Note: Damages by type of event and by region, 1970–2010 (February).

Source: World Bank staff based on EM-DAT/CRED.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Figure 1.7 Many small island developing states are among the 25 countries with damages above 1 percent of GDP

Source: World Bank staff based on EM-DAT/CRED.

Deaths expand Africa—damages shrink it Some countries are barely recognizable when a map’s areas reflect deaths (map 1.1). Africa looms as large as Asia, and the Americas shrink (the north to almost nothing). And when those areas reflect damages, Africa shrinks and middle-income countries expand (map 1.2).

Map 1.1 Deaths shrink Asia and the Americas—but expand Africa

Note: Areas reflect cumulative deaths from disasters for 1970 to 2010 (February).

Source: World Bank staff based on EM-DAT/CRED.

Map 1.2 Damages shrink Africa but expand middle-income countries

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Note: Areas reflect cumulative damage from disasters scaled by GDP for 1970 to 2008.

Source: World Bank staff based on EM-DAT/CRED.

Multiple hazards, clustering in different ways Each hazard afflicts countries differently, but many countries are subject to multiple hazards, though the importance of each differs. Earthquakes and volcanoes (geophysical hazards) tend to cluster along fault boundaries characterized by mountainous terrain. Floods, cyclones, and landslides (hydro-meteorological hazards) affect the eastern coastal regions of the major continents as well as some interior regions of North and South America, Europe, and Asia. Drought is more widely dispersed across the semiarid tropics. Areas subject to hazards fall primarily in East and South Asia and in Central

America and western South America (map 1.3). Many of them are also more densely populated and developed than average, leading to high potential for casualties and damage. But geography is not destiny. Many countries in harm’s way have managed to protect their population over time, and this report examines how this has happened. And for countries that have not dealt with disasters effectively, this report asks why and explores ways for doing so. Rich and poor countries are subject to hazards, but most of the 3.3 million deaths

over the last 40 years were in poor countries. Damage, however, may be rising in absolute terms, with earthquakes and storms causing the most damage. And middle- income countries are particularly vulnerable. Rising absolute damage is plausible, considering the increased exposure from urbanization (examined in chapter 6).

Map 1.3 Where hazards have struck

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Source: World Bank staff. Earthquakes above magnitude 6 on the Richter scale for 1950 to February 2010 (from Northern California

Earthquake Data Center, www.ncedc.org); tropical storm tracks for 1975–2007; droughts based on standardized precipitation index (SPI,

larger values indicate a higher probability of precipitation deficits) compiled for the Global Assessment Report 2009 (from

www.preventionweb.net/english/hyogo/gar).

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

http://www.ncedc.org

http://www.preventionweb.net/english/hyogo/gar

Spotlight 1 on Bangladesh The Antecedents of Lives Saved

Cyclone Sidr was first observed southwest of the Andaman Islands in the Bay of Bengal six days before it made landfall on November 15, 2007. Tracking its path and growing strength, the Bangladesh authorities had time to prepare a well-rehearsed response: they issued warnings and activated 44,000 volunteers who helped evacuate roughly 3 million people from their homes and accommodate 1.5 million in shelters. Few were surprised and unprotected when Sidr hit, but its immense force was

devastating. The category 4 cyclone (5 is the most severe) with a 1,000 kilometer diameter and winds up to 240 kilometers an hour whipped up 5.5 to 6.0 meter waves that surged over embankments designed to withstand 2.5 meters. Sidr’s forces were moderated when passing over the Sundarbans, a large wetland of mangrove trees, but such wetlands have diminished over the years, and vast unprotected areas were severely damaged. Rescue and relief efforts began immediately after the cyclone abated. The 12

worst affected districts, though less densely populated and poorer than the national average, had 18.7 million people: 55,000 injured, and 4,400 dead or missing. The government estimated that assets worth $1.16 billion were damaged, almost all in housing and other infrastructure. Losses of $517 million were expected. But it could have been far worse if the country had not learned from earlier tragedies.

Endemic hazards

Bangladesh is prone to many hazards (spotlight map 1). Cyclones are frequent and occur before and after the monsoons (April–May and October–November are when most cyclones occur): 508 formed in the Bay of Bengal over the past century, 17 percent making landfall in Bangladesh, others in adjoining India and Myanmar, and several dissipated over the ocean. In November 1970, a cyclone killed over 300,000 people and fed the discontent that led to Bangladesh’s separating from Pakistan in 1971. The parliamentary elections of 1970 gave East Pakistan’s Awami League an absolute majority, but the outcome was not respected. The political turmoil and street protests complicated the government’s handling of the cyclone, and the disaster added to the growing discontent that culminated in Bangladesh’s independence. Cyclones are not the only hazard: there are also frequent floods, infrequent

earthquakes, occasional droughts (19 between 1960 and 1991 and a severe one in July 1983 that affected 20 million), and tornados (in April, the hottest month, and kal- baishakhi pre-monsoon storms, with winds up to 100 kilometers per hour). The Himalayas are rising and seismically active as the Indian subcontinental plate is thrust

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

under that of Tibet. A major earthquake (over 7 on the Richter scale) is a 1 in 50 year event in Bangladesh; but there is little awareness and few precautions. The 1947 partition of India and the further subdivision of Pakistan in 1971 left Bangladesh with a single seismic monitoring station that detected the 4.2 Richter scale tremors in February 2001 but could not determine the epicenter without data from neighboring India to help triangulate it. Spotlight map 1 shows where the three hazards are most prevalent. Most of

Bangladesh is a flood plain, and high ground is scarce in the flat delta formed by three heavily silt-laden rivers (Ganges, Meghna, and Brahmaputra) that split into more than 700 waterways emptying into the Bay of Bengal. Eighty percent of the waters arrive in a few months: the rivers’ combined catchment area of 1.76 million square kilometers is 12 times the size of Bangladesh and includes much of the subcontinent (northern India, Bhutan, Nepal, and parts of China). Of the inflow, 95 percent (844 billion cubic meters) is between May and October, and more than 80 percent of the rain falls between June and September. Unlike violent cyclones, flood waters rise slowly but inexorably and turn deadly only when everything is submerged. Even if people survive by clinging to trees and rooftops, they may later starve if their livestock drown; so men often stay back with their cattle and evacuate reluctantly. This habit serves people poorly in coastal districts where sudden storm surges wash away those who do not heed the warnings to evacuate.

Spotlight map 1 Bangladesh is prone to disasters

Source: Bangladesh Space Research and Remote Sensing Organization, Bangladesh Water Development Board, Geological Survey of

Bangladesh

Traditional adaptation

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Why has this hazard-prone delta been populated for centuries? Because the settling silt makes the land very fertile. People grew rice and jute, accommodating the river’s seasonal rhythm. Long-stalked rice varieties that survive the post-monsoon floods beginning in June constitute the aman crop. The aus crop is planted in the premonsoon months of March and April and harvested during July and August. And the boro crop is planted in the dry season and harvested in March-April (later for high-yielding varieties). Farmers choose rice varieties best suited to their local area’s rainfall and flood

patterns, and build their houses (plinth heights) as safely as their budgets and technology permit. The 1947 partition put some jute mills in India, and the movement of goods and people across the border largely stopped: jute’s importance continued to decline as synthetic fibers replaced it, and rice remains the main crop.

Prevention: Expensive embankments …

The 1970s and 1980s were deadly (spotlight figure 1). Various flood abatement schemes were proposed: the Master Plan of 1964 called for massive embankments preferred by the engineers in the Water Development Board. The proposals languished because donors (who became important after Bangladesh’s independence) were split over their choice. The World Bank financed the construction of some embankments, but its 1971 Land and Water Study urged small scale developments, especially low lift pumps to tap ground water for irrigation in the dry season that allowed more of the high-yielding short-stem rice varieties to grow. The government restricted the use of tube wells when the water table was found to be falling.

Spotlight figure 1 Mortality from floods and storms in Bangladesh

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Source: World Bank staff.

Over 5,700 kilometers of embankments (3,400 in coastal areas), 1,700 flood control/regulating structures, and 4,300 kilometers of drainage canals were built over 30 years. The experience was sobering. Embankments merely redirect the water flow and are effective only when they are well located, designed, constructed, and maintained—but many were not. The resulting breaches rendered the entire embankment ineffective, and some farmers, seeking to protect their crops and fields, also created some breaches intentionally. The farmers often were neither consulted when the embankments were built nor compensated when their more vulnerable fields flooded. Embankments act as dams impeding the flood waters from draining rapidly, and protracted immersion increases damage to standing crops. Local authorities had also built some embankments, ignoring the larger delta’s

hydrology. But rivers change course, often with little warning, as silt scours their banks. During the 1966 flood season, the river moved 1,500 meters (almost 1 mile)

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

laterally downstream from Faridpur, digging a new 30-meter-deep channel. This shifting river course confounds land ownership and increases fatalities from floods when farmers stay put to preserve their land claims. The 1988 floods were not particularly deadly—although they claimed 2,440 lives

that year—but they affected Dhaka, the capital, galvanizing the government (and donors) into action. The 1989 Flood Action Plan dusted off the 1964 proposals for embankments along the entire length of the river, but donors balked at the staggering cost, prompting additional studies. Millions living between the river and the planned embankments would remain exposed. Resettlement was impossible: many were fishermen needing ready access to the river, and these unprotected farmers and fishermen found advocates at home and abroad to voice their concerns.

… to cost-effective measures

As doubts over the embankments’ merits grew, there was a thoughtful search for better alternatives that took account of the delta’s complex hydrology and agronomy. The World Bank 1971 study’s benefits of underground aquifers for drinking (reduced water-borne diseases) and irrigation began to be appreciated. The 1987 National Water Plan had estimated underground aquifers’ capacity at 69

billion cubic meters, but a more careful estimate in 1991 raised it to 78 billion. The declining water table was found to be localized around Dhaka, which drew water from wells for its growing city population; so restrictions on drilling for irrigation were lifted elsewhere. Tube wells proliferated, especially after private agricultural investment was deregulated and import tariffs (on pumps and the like) were lowered. Agriculture was transformed: low-yielding varieties in the aus and aman cultivation

gave way to high-yielding (irrigated) varieties that rose from 14 percent higher yield in 1973 to 54 percent higher yield by 1993. But there were also unexpected setbacks. In some areas, tubewells led to arsenic poisoning when the substrate’s naturally occurring minerals leached into the water. A remedial program to test and treat potable water was begun. But the merits of groundwater use and agriculture’s reduced vulnerability were apparent after the severe 1998 floods: rice harvests that were expected to fall by 11 percent actually rose by 5.6 percent. After the 1970 cyclone and independence, and building on the early cyclone shelter

construction that started in the late 1960s, the government, in partnership with the Bangladesh Red Crescent Society, established the Cyclone Preparedness Program in 1972. Working with local communities, a system appropriate to the area was developed to transmit hazard warnings—radio broadcasts complemented by flags of various colors, hoisted for all to see. People were taught what they signified, and what to do. Cyclone shelters began to go up in the late 1960s, and the livestock refuges in the early 1970s. But after 138,000 people died in the April 1991 eastern

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

coastal zone cyclone, the Multipurpose Cyclone Shelter Project began to increase the number of shelters. Each district’s deputy commissioner chaired a disaster management committee that included local representatives, both elected and from nongovernmental organizations (NGOs). The May 1997 cyclone, of similar magnitude, claimed 111 lives—far fewer than the

cyclone in 1970. But a cyclone’s severity is not the only determinant of fatalities, just as lives saved are not automatically the result of shelters built. Other factors matter. In 1970, large numbers of migrant workers were in the area for the harvest, and the 1997 cyclone struck the less densely populated hilly districts of Chittagong. How many people are exposed depends on the place, the season, and even the time of day. Better preparedness has helped, and cyclone shelters have reduced cyclone risks

for millions. More remains to be done: shelters have space for about 2.8 million people, or 7 percent of the coastal area’s population, but many shelters are not functional. The government has built 2,133 shelters and 200 livestock refuges in 15 of the 19 coastal districts, but the estimates of those functioning vary between 1,639 (Centre for Environmental and Geographic Information Services 2004) and 1,868 (Local Government Engineering Department). Almost a thousand schools were constructed to double as shelters, but many are not suitable; because their location and livestock facilities are inadequate.

Continuing complexities

While more shelters would help, they will not be enough. Rivers continue to bring down silt and the effects of upstream neglect: increased flows from glacier melt (reflecting deforestation and climate change), and poor effluent treatment (sewage and toxic waste). Development is also changing Bangladesh’s vulnerability. The growing share of manufacturing in output increases exposure in rapidly growing cities: Dhaka with 8 percent of the country’s population accounts for 15 percent of the GDP, and the port of Chittagong is a world “hotspot.” Well-engineered embankments around densely populated towns may be cost-effective, but the settling silt raises the ground level, putting the town at a lower level and increasing the likelihood of—and damage from—an embankment breach. So while the country may be less susceptible to minor floods, it is more exposed to major storms, floods, and earthquakes. Tackling these new challenges requires greater cooperation with neighbors. The

delta’s complex hydrology requires having and sharing data on river flows and hydro- met conditions—in real time, if people are to be warned of imminent danger. Without upstream water level data, Bangladesh could not forecast floods with sufficient accuracy and lead times until recently. Now satellite data based on global weather models allow 10-day forecasts. A proposal to link the Brahmaputra with its huge

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

water flows with the Ganges languishes because each country is suspicious of the other’s data and motives and because the engineering, ecological implications, and economics remain unexamined. Such differences go back to when Bangladesh was the eastern province of a

hostile Pakistan. India signed a treaty in 1960 to share the Indus waters with Pakistan, dividing the eastern and western waters, which could then be harnessed by each country separately. But Pakistan’s role in and after India’s 1962 conflict with China prevented a similar agreement over the Ganges. In the late 1960s, India began building a barrage at Farakka (completed in 1974) to keep the port of Calcutta (India) open and the Hooghly River navigable by diverting water during the dry season. Following a short-lived agreement after Bangladesh’s independence, disputes over the barrage’s effects on (Bangladesh) Khulna’s agriculture and on other northwestern districts continue, underscoring the complexities of the legal and hydrological issues. The water dispute extends to other issues and complicates disaster prevention.

Talks have begun in early 2010 between Bangladesh and India to attempt to resolve outstanding issues relating to water sharing and protecting banks of common rivers. Bangladesh’s population continues to grow (though its rate of increase has slowed), and some 35 million people, a quarter of the country, now inhabit coastal areas exposed to cyclones. The mangroves of the Sundarbans (which reduced Sidr’s destructive force) have shrunk in half over the last 50 years. Cities and manufacturing normally attract the growing numbers, but Bangladesh’s

cities are not in safe locations, and an international border cordons off the low-lying delta. Migration is a thorny issue, especially when India’s central government struggles with the grievances of border hill tribes and Myanmar remains closed. How many more can the crowded Gangetic delta safely accommodate before prevention becomes prohibitively expensive? These are not questions for Bangladesh alone. Governments that created the

borders could make them more permeable. Should donors nudge them to do so, much as they did with cyclone-proof shelters? Donors with funds and good intentions also make faulty suggestions, and decision-making was unlocked only after a disaster outraged people. Better institutions that enable sound and timely decisions come with development, so disasters are a barometer of development. While this message echoes through this report, Bangladesh shows how even poor countries can prevent disasters, thereby nourishing such institutions.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

CHAPTER 2

Measuring Disasters’ Many Effects

ohn Stuart Mill, the English philosopher and economist, wrote “what has so often excited wonder, the great rapidity with which countries recover from a state of

devastation; the disappearance, in a short time, of all traces of the mischiefs done by earthquakes, floods, hurricanes, and the ravages of war” (Mill 1872). Is what Mill wrote in 1872 still applicable in today’s context? And even if he was correct in asserting the “great rapidity with which countries recover”—subsequently consuming at their earlier levels—what about the welfare of those affected? Economists typically use individuals’ incomes or a nation’s output to measure

prosperity. Income—or output—is surely an important but imperfect determinant of welfare. Indeed, if output were a perfect measure of welfare, one would rejoice the birth of a farm animal and bemoan that of a child (Bauer 1990).1

In the context of disasters, measuring changes in output is an imperfect measure of changes in consumption,2 and it cannot fully capture the pain and suffering inflicted by personal injury, the injury or death of loved ones, or the anxiety engendered by dislocation and uncertainty about the future. Even so, given the frequency of calculating and using output measures of disasters’ effects, it is important to understand the approaches and pitfalls. This chapter first attempts to assess disasters’ effects on aspects of well-being

measured by health, nutrition, education, and mental state of mind. It then examines assessments of disaster’s local and economy-wide effects on output (gross domestic product, or GDP). Certain findings confirm and validate what we know and might expect, but others could be surprising. Most studies of disasters’ effects focus on the immediate aftermath. The chapter

begins by complementing such studies with others that find longer lasting effects on various aspects of well-being such as schooling, cognitive abilities, and mental health. Disasters, even if short-lived, can have long-term consequences: some survivors are pushed over the edge and never completely recover. Droughts, particularly widespread in Africa, result in stunted and malnourished children with permanent adverse effects. The chapter discusses the association between disasters and conflicts. Do disasters increase scarcity and thus conflict, or do they create an opportunity for peace, as in Aceh? The chapter then turns to disasters’ effects on economic output, growth, and a

government’s budget. If and how much output falls, and how long it takes to recover The, World Bank, Bank World, and Nations (UN) United. Natural Hazards, UnNatural Disasters, edited by World Bank The, et al., World Bank Publications, 2010. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/apus/detail.action?docID=606027. Created from apus on 2017-04-04 13:20:21.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

from a disaster, are controversial issues because some distinctions (such as affected versus unaffected areas; those directly affected versus others) are not always clear. Physical damage and disruption reduce output in the affected area, and because the area usually is linked to undamaged areas through commerce, people elsewhere would also be affected. These indirect effects are often, though not always, adverse: those relying on the affected area for supplies or markets may be hurt, but others who offer alternative supplies may be able to increase output. So, national output may not fall as much as that in the affected area. Studies also differ over whether and how they correct for the effects on commodity

prices, terms of trade, and exchange rates, which also affect output. New studies that correct for such factors find that national output always falls after a severe disaster but sometimes rises after a mild one. This may surprise those who think of output as the sole measure of people’s well-being, because people are obviously worse off. The seeming disparity arises because output and welfare are not the same. Repairs and rebuilding add to economic activity. And economic growth is output’s rate of change, so growth could exceed pre-disaster rates as output catches up to its pre-disaster levels. These outcomes in measures of material economic activity are thus consistent with potentially severe losses in overall well-being. Governments often assess damage after disasters, and such assessments

typically have different and multiple purposes. Keep in mind the intended purpose(s) of assessments, particularly since accurately measuring a disaster’s effects is tricky: making informed decisions requires reliable estimates of the relevant measurement concepts of damage and loss. Moreover, biases such as double counting could sneak in unless one is careful. The accuracy of damage estimates is also affected by biases in measurement, especially with the prospect of aid. But what is valuable is not always valued, such as the effect of disasters on intangibles. What could make damage assessments more accurate and helpful? While people

do not wait for help to repair their homes and rebuild their lives, the recovery is faster with appropriate and timely help from others (family, suppliers, customers, nongovernmental organizations [NGOs]) and from the government. Commercial ties with other individuals and firms help the recovery. But businesses and individuals also rely on public infrastructure (roads, bridges, railways). The government must therefore quickly decide on the sequence of repairs and on whether to change the location and resilience of structures. These decisions will affect what firms and people do. So, assessing the damage to public infrastructure, and the costs of repairs and rebuilding is urgent, to efficiently implement public measures for recovery. And this requires quickly assessing the impact of a disaster on a government’s fiscal position. But repairs to public infrastructure cannot be instantaneous—the costs are spread

over time. Governments in developing countries struggle to raise taxes for the 10 to

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

20 percent of GDP they typically spend, so even if costs of repairing public infrastructure are spread over time (or financed through borrowing), damage assessments should examine the disaster’s fiscal implications for the public sector and the ability to finance recovery—keeping in mind that fiscal revenues depend on national output, which will not fall as much as that in the affected area. A frequent purpose of damage assessment is compensation: with much of the

damage to private property, governments may wish to compensate at least the poorest for the damage they incurred. Whether it is useful to try to comprehensively value damage to private property is questionable. Setting aside the complexities of measurement and biases, compensation is seldom linked to damage. While it may be desirable to limit such transfers to those who are both poor and have incurred damage (a subset of those in the affected area), distinguishing between the chronically and the temporarily poor is difficult. And it would be unfortunate if such spending displaced that on adequate infrastructure and its maintenance—especially since their neglect increases vulnerability to future disasters. Not everything needs to be measured or valued in a desire to be comprehensive for governments to help people directly. Indeed, damage assessments could be more useful if they were simpler. Finally, damage assessments are often conducted as a prelude to foreign aid.

However, if donors seek instead to help a country achieve more than a recovery to the status quo ante, then damage estimates, especially if based on pre-disaster measures of output and asset values, may not be that informative. Recognizing the limits of damage assessments would also enhance their value.

Individuals over the edge Studies on the short- to medium- term effects of a disaster on poverty abound.3 Many survivors of disasters, rich and poor, recover fully, but a few do not. Healthy people survive temporary deprivation, but elder people and women are particularly vulnerable. Even temporary malnourishment could permanently stunt growth and lower cognitive abilities among children younger than three. While much has been written on short-run effects, panel data to examine the longer term effects on human welfare, some more subtle than others, are scant; but the absence of data does not mean the absence of a problem. Some new studies explore the sufferings of survivors, particularly children.

Declines in schooling and health

Enrollment rates for children aged 7–15 fell by about 20 percent in parts of Côte d’Ivoire with extreme rainfall changes between 1986 and 1987 relative to unaffected regions (Jensen 2000). Earthquakes had similar effects: school attendance fell by almost 7 percent among households heaviest hit by the two strong earthquakes that

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

affected El Salvador in 2001 (Santos 2007). Children in households most affected were about three times more likely to work than attend school. Temporary withdrawal from school sometimes becomes permanent: children

withdrawn from schools during droughts in Central Mexico between 1998 and 2000 were about 30 percent less likely to resume their studies (de Janvry and others 2006). Boys in Tanzania worked longer hours after a drought: a 5.7 hour increase in work reduces their schooling by a year, observed 10 years later (Beegle, Dehejia, and Gatti 2006). Complementing these country studies in a background paper for the report,

Cuaresma (2009) conducts a cross-country analysis of the link between disasters and human capital accumulation (measured by secondary school enrollment). The findings are that those more exposed to earthquakes between 1980 and 2000 have lower secondary school enrollment rates: 1.65 percentage points lower for a country with mean occurrence of quakes compared with a country with no quakes.4 Another study finds that households with a higher probability of experiencing floods in Bangladesh are more likely to “hold” extra years of schooling relative to land (Yamauchi, Yohannes, and Quisumbing 2009a, 2009b). In Ethiopia and Malawi, exposure to highly frequent droughts in some cases reduced schooling investment. And asset holdings prior to disasters, especially household human capital stock, help maintain schooling investments. Disasters reduce school enrollment: parents want education for their children but

may pull them out temporarily after a disaster to help with more pressing tasks, or because schools have been disrupted. Resuming education requires effort, and a permanent loss or decline may be because many children (or parents) give up or because teaching remains disrupted. In either case, something must be done; what, depends on the details. Moreover, cognitive and analytical abilities—only imperfectly related to schooling—could be affected even without reductions in school enrollment. Visits to the doctor decline after a disaster, but with little effect on health. After

Hurricane Mitch in 1998, sick children in affected areas were 30 percent less likely to be taken to clinics but with no significant difference in the prevalence of illness. The larger point: outputs, like cognition or health, are harder to measure than declines in school enrollment or doctor visits.5

Increased stunting …

Malnourishment has adverse effects, especially on young children, and this occurs during extended droughts, especially in Africa. Children who lose weight may catch up later (Foster 1995), but while “wasting” (low weight-to-height ratio) is reversible, “stunting” (low height-to-age ratio) is almost always permanent (figure 2.1).

Figure 2.1 Undernourished children become shorter adults

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Note: Z-scores (standard deviation scores) are a system by which a child or a group of children can be compared to the reference population. For population-based assessment—including surveys and nutritional surveillance—the z-score is widely recognized as the best system for analysis and presentation of anthropometric data. (WHO, http://www.who.int/nutgrowthdb/about/introduction/en/index4.html)

Source: Victora and others 2008.

In a group of 400 rural households, children aged 12 to 24 months at the time of the 1982–84 droughts in Zimbabwe were 2.3 centimeters shorter in late adolescence (Alderman and others 2006). In the Kagera region of Tanzania, children younger than 5 years exposed to a drought in 1991–94 were roughly 1 percent shorter than the population’s median height 10 years after (Alderman and others 2009). In Ethiopia, children either in uterus or younger than 36 months affected by the 1984 famine were 3 centimeters shorter than other comparable children 10 years later (Porter 2008). In China, rural adults who were children in the 1959 and 1962 famines were 3.03 centimeters shorter (Chen and Zhou 2007). And in Indonesia, females born in a year with 20 percent higher rainfall are 0.14 centimeters taller (Maccini and Yang 2008).

… and diminished cognitive abilities

Malnutrition that causes stunting also diminishes cognitive skills by inhibiting learning (reducing schooling) and productivity. In rural Zimbabwe and rural Tanzania, malnutrition reduces the years of schooling completed. In both cases, after finding that droughts reduce children’s height, their educational achievements as adolescents were regressed on their height when they were younger. In Zimbabwe, the 12- and 24-month stunted children during the 1982–84 droughts had delayed school enrollment (3.7 months) and lowered grade completion (0.4 grades) 13 to 16 years later. And in Tanzania, schooling at adolescence in 2004 was nearly a year more for a boy in the 95th percentile of height distribution than for another in the 80th percentile, when they were under 5 and exposed to the 1991–94 drought.

Table 2.1 Stunted children have lower cognitive scores

Philippines South Africa Indonesia Brazil1 Peru Jamaica2

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

http://www.who.int/nutgrowthdb/about/introduction/en/index4.html

Cognitive score (8 years, n =

2489)

Ravens Matrices (7

years, n = 603)3

Reasoning and arithmetic (9 years,

n = 368)

Attained grades (18 years, n =

2041)

WISC IQ (9 years, n =

72)

WAIS IQ (17– 18 years, n =

165)3

Reading and arithmetic (17–18

years)3

Not stunted 56.4 0.17 11.2 8.1 92.3 0.38 0.4

Mildly stunted 53.8 (–0.21) 0.05 (–0.12) 10.3 (–0.26) 7.2 (–0.4) 89.8 (– 0.20)

Moderately or severely stunted

49.6 (–0.54) –0.23 (–.040) 9.7 (–0.43) 6.5 (–0.7) 79.2 (– 1.05)

–0.55 (–0.93) –0.60 (–1.00)

Note: Data are mean (effect size as unadjusted difference from non-stunted children in z scores). 1. Males only. 2. The sample comprised stunted (< –2SD) children participating in an intervention trial and a non-stunted (> –1SD) comparison group. 3. SD scores. WISC = Wechsler Intelligence Scale for Children. WAIS = Wechsler Adult Intelligence Scale. Source: Grantham-McGregor and others 2007.

Children between 12 and 36 months who are moderately or severely stunted compared with not stunted (height-for-age greater than one negative standard deviation) have reduced cognitive skills (measured through IQ tests) in later childhood (Grantham-McGregor and others 2007).6 For example, in the Philippines, reading and math test scores for children at age 8 who were stunted during childhood were 0.75 standard deviations below that of children not stunted (table 2.1). Malnourished children become less productive adults: their lower body mass makes

manual labor less productive, and their lower cognitive skills make skilled work more difficult.7

… reduce subsequent earnings

Children malnourished during the 1982–84 drought in Zimbabwe had a 7 percent loss in (extrapolated) lifetime earnings (Alderman and others 2006). The 1991–94 drought in the Kagera region of Tanzania also reduced lifetime earnings by about 1 percent, a smaller but still significant effect because the sample included older children who were less vulnerable. Similarly, the 1959–61 birth cohort (malnourished in famine) earned less as adults: the 1959 cohort in areas where the death rate is higher by 1 in 1,000 earns 2 percent less per capita (Chen and Zhou 2007).8

This effect of nutrition on earnings may act through cognitive skills. Poor nutrition in Guatemala lowered cognitive skills and reduced earnings (Hoddinott and others 2008). For two groups with 25–42 year olds, those who got nutritional supplements as 0 to 3 year old children had higher wages.

Mental health falls—but can recover

Income, consumption, and health are poor proxies of welfare, and Amartya Sen suggests measuring welfare by functionings and capabilities—what people accomplish with income, health, and education (Sen 1987). Physical or psychological trauma diminishes welfare even if earnings do not fall. There is justified concern with the psychological effects of disasters. Norris (2005)

reviews 225 studies in developing and developed countries and finds that many suffer The, World Bank, Bank World, and Nations (UN) United. Natural Hazards, UnNatural Disasters, edited by World Bank The, et al., World Bank Publications, 2010. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/apus/detail.action?docID=606027. Created from apus on 2017-04-04 13:20:21.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

from post-disaster psychological disorders. But most of these studies address small samples (150 people on average) and very few studies have systematically followed larger samples of affected people over several years. A background paper for this report, using household data employing a baseline collected ten months prior to the 2004 Aceh tsunami, examines the mental health of its adult survivors (Frankenberg and others 2009). Annual follow up surveys in the subsequent four years provided a “before and after” indication of mental wellbeing. For 9,000 adult survivors of the 2004 tsunami in Aceh, post-traumatic stress reaction (PTSR) scores for each respondent in affected areas were high in heavily affected areas as much as 6 to 14 months after the tsunami. But even without treatment, these scores declined with time (figure 2.2). This analysis also addressed the association between disaster-induced PTSR and

key socioeconomic outcomes such as physical health, demographics (widowhood), work, income, and household wealth. It controlled for community-fixed effects and damage area as well as the age, gender, and education of each respondent, finding little influence of initial PTSR on most outcomes.9

Figure 2.2 Post-traumatic stress reaction (PTSR) scores fall over time in all tsunami-affected areas

Source: Frankenberg and others 2009.

It may not be possible to generalize the Aceh findings, especially because mental health issues are culture and condition sensitive. But if these findings hold in other contexts, they have important policy implications. And though mental health counseling was relatively unavailable in the aftermath of Aceh, much PTSR faded over time and no long-term socioeconomic effects were reported. This suggests that it may be better to channel scarce resources in a disaster’s immediate aftermath toward traditional relief activities as opposed to early-stage mental health interventions. The enduring response, however, may say otherwise. As reconstruction begins, PTSR declines, but new cases of adverse behavior may emerge among some people. And these people may benefit from targeted mental health interventions. Clearly though, more work is needed to explore the links between mental health and disasters. Future

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

research should consider the composition of this more vulnerable group and how to help them lead better lives after a disaster.

Conflicts: Cause or consequence? Some claim that disasters lead to conflict,10 particularly in Africa with its droughts and earthquakes (Wisner and others 2004). Earthquakes occur much more in countries where there is civil war (Brancati 2007). This association prompted plausible theories that invoke greater scarcity of resources: Homer-Dixon (1999) argued that environmental scarcity drives conflict, and many empirical studies examine how droughts are related to conflicts. In Africa, a 1-percent annual increase in rainfall reduces the probability of serious

conflict by about 6 percent (Miguel and others 2004). Exceptionally low rainfall makes conflict more likely. Similar results come from different rainfall data: a 20-percent decline in rainfall raises the probability of civil war by 3.6 percent after controlling for climatic conditions and land degradation (Hendrix and Glaser 2007). These patterns could capture the timing of the conflict more than the cause: the effect is more significant when the drop is from an exceptionally high rainfall year, consistent with the difficulties of fighting in such weather (Ciccone 2008). So weather matters for conflict, even if there is little evidence that it causes it. Regardless of whether their fields are irrigated, farmers and pastoralists often

dispute claims over land and water, sometimes leading to conflict.11 After the Sahelian drought of the early 1970s—when Côte d’Ivoire’s government supported Fulani pastoralists moving to areas where Senofo peasants lived—the Senofo households lost roughly 20 percent of income from crop damage by Fulani cattle (Bassett 1988). But other Senofo cropping patterns also changed, and encroachments on the manure-rich Fulani lands led to conflict. Similarly, of 800 households interviewed in the Sahel region of northern Nigeria, 200

experienced conflicts, more than half related to resource access; 60 percent occurred in the dry season, and the most violent ones were in the fertile flood plains (Nyong and Fiki 2005). Pastoralists claim that farmers cultivate along cattle paths during droughts, while farmers said pastoralists watered their cattle at their wells and allowed them to graze on their crops. Qualitative studies show that drought in Afghanistan and volcanic eruptions in eastern Congo exacerbated conflict in 2002 (Wisner and others 2004). Conflicts are not only strongly correlated with earthquakes—but also last longer

when earthquakes occur (Brancati 2007). Countries in conflict experience earthquakes roughly six times as often (every four years while those without civil war had one every 25 years).12 And the duration of the 44 conflicts where there were no earthquakes was 8.8 years, a little more than half the 15.4 year duration of the 19 conflicts where there was at least one earthquake.13 The analysis ensured that this

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

relationship is not simply the spurious result of longer wars increasing the temporal window for earthquakes to occur. The probability of an earthquake in a conflict year (0.25) is greater than when

countries are at peace or where there is no conflict. Civil wars obviously do not cause earthquakes, nor earthquakes civil wars. Instead, earthquakes prolong conflicts, perhaps reducing the advantage of the stronger power, the government. Take the 1999 earthquake in Colombia: 1,000 died, thousands were injured, and 35,000 lost their homes. Coffee production suffered, and survivors, frustrated at the government’s slow disaster response, clashed with the police and looted establishments. This diverted government security forces, and the rebels took advantage of the situation to renege on an agreement to withdraw from the demilitarized zone, increasing attacks and prolonging fighting.

Relief

Relief (which foreign donors also provide) is often another weapon in the conflict, and those who control its distribution provide it to victims who support them, victims who could be won over to their side, or those expected to remain neutral (victims or non- victims). It all depends on how the war is fought. In Sri Lanka, the 2004 tsunami pummeled the contested areas of Ampara and

Batticaloa. How was relief for housing reconstruction allocated across districts?14 Of the 5,300 Muslim and 5,260 Tamil homes destroyed in Ampara, 2,080 homes were rebuilt, and about 2,560 of the 8,600 Tamil homes destroyed had been rebuilt in Batticaloa (Kuhn forthcoming). In contrast, in the largely Sinhalese districts of Galle, Matara, and Hambantota, about 9,120 of the 9,350 homes destroyed were rebuilt. These findings suggest that the government assists only its committed supporters when assisting potential supporters in contested areas is difficult. Political considerations are important even in areas under government control, which is not surprising. There is evidence of political bias in the distribution of disaster aid in the United States and elsewhere (Keefer and others 2009). The same 2004 tsunami devastated Aceh, where the conciliation and peace that

followed is a refreshing contrast. Aceh, the stronghold of the GAM rebels, was largely outside the control of the Indonesian government that administered the assistance that followed in the tsunami’s wake. But some of the assistance was used to re-integrate the GAM insurgents into peaceful civilian life. Elsewhere, combatants use disaster relief to gain a military advantage. A 1976

earthquake in Guatemala killed at least 20,000 people outright and many more from illness and injury. The government allowed international assistance, including religious groups and other NGOs and bilaterals, unfettered access to the damaged area. But the western highlands, where the quake struck, were not yet embroiled in the fighting. The government used quake relief to gather intelligence and squelch any incipient

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

rebellion (Hinshaw 2006). Aid was part of the effort to prevent the earthquake from becoming a recruiting tool for insurgents. Spotlight 4 describes Ethiopia’s use of food aid as another weapon in its long

drawn out civil war and nearby Sudan’s similar response to the same drought. The fighting disrupted aid delivery, the government made little effort to assist the three southern provinces in 1984, and theft (including that by government forces) hobbled logistics at Port Sudan (Burr and Collins 1995). As the drought persisted and aid finally trickled in, the insurgent Sudan People’s Liberation Army (SPLA) blocked aid until the villages they controlled (not necessarily the most drought ravaged) also received aid. Only after disease (visceral leishmaniasis and meningitis) broke out in the south and spread to Khartoum in 1987, did food aid flow. But only for a short time. Insurgent success facilitated relief flows to areas that supported them: by April 17,

1989, the SPLA had taken 11 government garrisons and three district capitals, and donors delivered more aid between January and February 1989 than they had in the five years between 1983 and 1988. Both economic and military assistance to the Sudan government dropped, the defense and finance ministers resigned, inflation in Sudan approached 80 percent, and bread shortages emerged in Khartoum. The SPLA wouldn’t cooperate on land transportation, and foreign donors moved 40 percent of relief by air at a huge cost of $700 a ton. Even these restricted flows ended, and the war between the government and SPLA continued.

Breaking the cycle of conflict

Could Aceh be the example, not the exception? Could disasters break the conflict cycle? Pakistan and India have fought long and hard over Kashmir. But they cooperated to provide assistance even in disputed areas after the 2005 earthquake, though each feared that aid could provide an advantage to the other and that it restricted the use and staffing of relief aircraft (Renner and Chafe 2007).

Table 2.2 Civil war, rainfall, and the rule of law

Dependent variable: Probability of civil war Not controlling for rule of law Controlling for rule of law

Rainfall growth from last period (t–1) to current period (t) –0.11 –0.05 (0.04) (0.34) Rainfall growth, (t–2) to (t–1) –0.08 –0.03 (0.07) (0.5) Rule of law (t–1) –0.17 (0.001) Rule of law (t–2) 0.1 (0.03) Number of observations, countries 451,32 451,32 R2 0.08 0.14

Note: Ordinary least squares with clustered standard errors are used because rainfall is not significant controlling for country-specific fixed and year effects, with or without controls for the rule of law. Nearly all variation in rule of law is cross-country. Other control variables are initial income per capita, ethnic fractionalization, religious fractionalization, whether a country is an oil exporter, how mountainous the

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

country is, and the log of population. P-values are in parentheses. Source: Keefer and others 2009.

Such cooperation may be of interest to both countries: while the Kashmir conflict is militarized, both Pakistan and India want to win the Kashmiris’ hearts and minds. The governments competed while cooperating with aid logistics, but they did not address territorial claims. Goodwill is short-lived, so disasters spur incipient dispute resolution efforts but rarely stop conflicts. Kelman notes how Cuba and the United States lost four opportunities to thaw frosty relations through post-hurricane aid: Hurricane Michelle in 2001 and Hurricanes Dennis, Katrina, and Wilma in 2005 prompted one country to offer aid that the other rejected (Kelman 2007). Weak though such glimmers of hope may be, they should not be missed. The empirical association between disasters and conflicts, and the episodes just

outlined, suggest that both may result from something else—the missing variable of “institutions of good governance.”15 When proxies for such institutions are included in regressions, they are statistically significant. The effects of rainfall shocks on income are strongest in Africa (Fiala 2009). And in Sub-Saharan Africa, rainfall declines can trigger conflict (Miguel and others 2004). This result is strong, and the effects of rainfall on growth appear to be entirely conditional on the rule of law. Rainfall does not significantly affect the chance of war when a proxy is included

(table 2.2). The two law proxies have opposite signs because they summarize two effects: the improvements in the rule of law, and the high rule of law. If the rule of law were unchanged, the law proxy reduces the probability of civil war; if the rule of law improves, the probability of civil war falls even more.16

The likelihood that disputes turn into conflicts after a disaster depends on what the government does. Governments that do not take measures to prevent disasters do not protect their peoples’ rights to property or protect their people against insurgency. So a disaster could ignite conflict. The quality of governance and institutions matter in two ways: private investment in recovery does not take place, and people scramble to seize what is not theirs. Collier and Goderis (2007) note that this also happens in what they call the “natural resource curse.” The analysis of conflicts produces three points. First, disasters have an adverse

effect on conflict only in places where the rule of law is already weak, so a disaster could ignite conflict. Second, there is a strong incentive to divert disaster relief during conflicts. And third, disasters can occasionally break the conflict cycle, as shown in Aceh after the tsunami, but such goodwill is short-lived. Disasters undoubtedly reduce the well-being of victims and their surviving families.

But their effect on victims is not synonymous with their effect on an economy’s output or output growth.

Welfare falls, but what are the effects on output? And for how long? The, World Bank, Bank World, and Nations (UN) United. Natural Hazards, UnNatural Disasters, edited by World Bank The, et al., World Bank Publications, 2010. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/apus/detail.action?docID=606027. Created from apus on 2017-04-04 13:20:21.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

A disaster could reduce output (certainly in the affected area and possibly nationally) both because of physical damage and because of a disruption in normal economic activities (figure 2.3). There are two related questions. How long will it take before output recovers, if it

does? And what, if anything, can the government do to hasten the recovery? The answers have generated much discussion, in large part because studies come to different conclusions. Many find that disasters have adverse longer term effects, but some find little or no effect, and a few even find that national output increases—a result that does not contradict a drop in welfare (box 2.1).

Figure 2.3 A possible post-disaster GDP path

Note: Other paths are also possible. For example, output could also rise above pre-disaster levels, but this can be an artifact of a reconstruction boom, as discussed later in the chapter. Source: World Bank staff based on Hochrainer 2006.

Box 2.1 Previous studies find a medley of effects of output and growth Previous studies find ambiguous effects of disasters on national output. There are many reasons for this, but the underlying one is that it is particularly difficult to identify causal effects of disasters on output. Differing findings suggest that models may be misspecified, often because relevant variables are omitted. It is also difficult to include all relevant factors, especially when some (like the network of links between affected and unaffected areas) cannot be measured. Results also vary because studies differ in the period they examine and the techniques they use. Otero and Marti (1995) found that disasters decrease national income and tax revenues, raising the fiscal

and the trade deficits (exports fall and imports rise) in the short run. In the longer run, post-disaster spending increased debt service payments, reducing development and creating persistent external and fiscal imbalances. Benson (1997a, b, c) and Benson and Clay (1998, 2000, 2001) examined the short-term effects of several

disasters in Dominica, Fiji, Vietnam, and the Philippines. Agriculture was most strongly and adversely affected, and poverty and inequality rose, but the effects of the disaster could not be isolated from other adverse developments (such as terms of trade). Murlidharan and Shah (2001), in examining the effect of disasters on medium-term economic growth, found

that growth was reduced. Disasters were also associated with growing external debts, budget deficits, and inflation. Hochrainer (2006) examined 85 disasters in 45 countries and found that GDP fell in the disaster year and

that growth subsequently did not rise to more than compensate (a fall in cumulative GDP).

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Noy (2009) found that the country’s ability to mobilize resources for reconstruction and global financial conditions helps explain disasters’ effects on GDP growth. Cuaresma and others (2008), in one of the few longer term studies, found that disaster risks reduce knowledge spillovers from developed to developing countries. One plausible reason behind this finding is the importance of institutions in a nation’s absorptive capacity for foreign technologies: disasters tend to affect technology absorption positively only in countries with relatively high levels of GDP per capita. Not all studies found adverse effects. Albala-Bertrand (1993) looked for, but did not find, significant longer

term effects of disasters in developed countries. And the effects in developing countries faded away after two years, though some adverse effects on income distribution persisted. He concluded that disasters are “a problem of development, but essentially not a problem for development” (p. 202). Albala-Bertrand questioned many of the assumptions and estimating techniques used in the literature.

Attributing all the change in output and economic growth to the disaster would be misleading because other factors also influence growth: studies that do, find effects to be small and to differ in sign. This suggests that economies and disasters differ so much that any effect on growth and output depends on the details: some are adverse, others not; some are ephemeral, others long lasting. Caselli and Malhotra (2004) found that disasters did not reduce GDP, fiscal deficits, or inflation in either the

short or medium term. Fatalities and damage appear to depend on the stage of a country’s development, not the disasters per se. Caselli and Malhotra also concluded that the loss of capital and labor did not affect short- term economic growth. Skidmore and Toya (2002) go even further: they found the frequency of disasters to be positively correlated with long-run economic growth after conditioning for other determinants, which they interpret as Schumpeterian creative destruction (which is unconvincing because disasters are not selective in what they destroy). More recent studies (Hallegatte and Dumas 2009; Hallegatte and Ghil 2008) find the results are sensitive to elasticities of substitution in the production function and whether the disaster occurred during an upturn or downturn of the business cycle.

Source: World Bank staff.

Aggregate and sectoral effects on economic output and growth in the long term Debates over disasters’ effects arise because, as box 2.1 shows, the findings vary: studies use different data and estimation techniques, and include different disasters. Several background papers were commissioned for this report to build on these studies to look past the immediate effects of a disaster (always adverse). These new studies correct for other factors to isolate the effect of disasters; each technique has its advantages and limitations that are briefly discussed along with its findings. Despite statistical care, the limitations of cross-country regressions in some of the studies reported here must be emphasized, and the conclusions here reflect those caveats. Hochrainer (2009) considers 225 large disasters between 1960 and 2005, and

compares the country’s post-disaster GDP with what he projects (had the event not occurred).17 GDP is on average 2 percent lower even five years later (however, with large deviations around the mean), and a nonparametric test including detailed uncertainty analyses finds this difference to be statistically significant. But the GDP is measured against projections based on recent growth without correcting for the many other factors that influence the economy (the difference between observed and projected output is explained using two techniques).18

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Two background papers examine the issue from another perspective by adjusting for the effect of the many factors that also influence output in the medium term (5 years) and the short (1 to 3 years).19 Loayza and others (2009) estimate the medium-term effects of different hazards simultaneously on economic growth using a model with three main sectors (agriculture, industry, and services) and with the whole economy, correcting for two sets of variables that also affect growth.20 The first set comprises structural and institutional variables such as education, financial development, monetary and fiscal policy, and trade openness. The second, external conditions such as terms of trade and period-specific dummy variables. They calculate rates of growth (not levels of output to make the series stationary that econometric techniques require) in discrete five-year periods using data for 94 countries (68 developing) over 45 years (1961–2005); so each country has at most nine observations (table 2.3).21

Table 2.3 Growth effect of a “typical” (median) disaster

Effect on:

GDP growth Agricultural growth Industrial growth Service growth

Median intensity of:

Droughts –0.6%*** –1.1%*** –1.0%** –0.1% Floods 1.0%*** 0.8%*** 0.9%*** 0.9%***

Earthquakes –0.1% 0.1% 0.9%* –0.1% Storms –0.1% –0.6%*** 0.8%* –0.2%

Note: The effects on GDP growth rates—the rate of change of output—and not on output levels. So, a typical drought could reduce overall GDP growth by 0.6 percent; agriculture growth by 1.1 percent, and so on. *significant at 10%; **significant at 5%; ***significant at 1%. Source: Loayza and others 2009.

Five-year non-overlapping rates of growth do not capture short-run effects (hence the parallel study summarized in sequence). The main findings are that medium-run economic growth is generally lower after a disaster. But the effect depends on the type of hazards and is not always statistically significant or uniform.

Overall growth falls by 0.6 percent after a drought of typical (or median) intensity, with the most adverse effect on agricultural and industrial growth. Overall growth barely falls after a typical earthquake, but industrial growth rises, perhaps because of reconstruction. Agricultural growth falls by 0.6 percent after a typical storm, but industrial growth rises, again perhaps because of reconstruction. Interestingly, overall growth rises by a statistically significant 1 percent after a flood of typical intensity. This is plausible because although floods disrupt farming and other activities, they may also deposit nutrient-rich silt and may increase hydroelectric power, which boosts industrial growth. For example, in Norway, an unexpected glacial lake outburst flow in 2001 allowed the Norwegian utility Sisovatnet to produce an additional year of hydropower.22

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Capturing such gains depends partly on having the right infrastructure in the first place (here, a reservoir capable of holding excessive water).

But severe disasters (limited to only 10 percent of all disasters) have adverse effects regardless of type. The adverse effect on agricultural growth doubles for severe droughts; the rise in growth after severe floods becomes statistically insignificant; and severe storms are more damaging, particularly for industrial growth. Table 2.4 shows the results. In the second parallel background paper, Fomby and others (2009) trace the

annual growth in the year of and the year following the events to examine the adjustment path in the shorter run (1 to 3 years). The model pools the experiences of various countries over time to arrive at mean responses of growth to disasters of different intensities. While losing country specificity, the model detects the time pattern of the recovery reasonably and robustly. The full sample has 87 countries, with some from every region, and covers 48 years from 1960 to 2007. The full sample and a developing countries subset (70 percent of the full sample) are analyzed, adjusting for the severity of disasters.

Table 2.4 Growth effect of a “typical” (median) severe disaster

Effect on:

GDP growth Agricultural growth Industrial growth Service growth

From median intensity of severe:

Droughts –1.0%*** –2.2%*** –1.0%* 0.3% Floods 0.3% 0.6% 0.1% 0.4% Earthquakes –0.0% –0.1% 0.3% 0.0% Storms –0.9%** –0.8%** –0.9% –0.9%

*significant at 10%; **significant at 5%; ***significant at 1%.

Source: Loayza and others 2009.

They find that moderate and severe disasters affect growth more in developing countries than in rich countries; but this may reflect their size and diversity rather than their income. Growth typically does not rise after severe disasters, especially in developing countries. But the effects on output still depend on the hazard and the structure of the economy (similar to Dumas and Hallegatte model, which emphasizes the elasticity of substitution in production). This may be why earlier studies that did not adjust for the effects of other (nondisaster) factors found growth effects that differ depending on the period. Particular cases would of course differ from the “average” findings: not every flood

raises agricultural growth (flash floods wash away sediment, but annual floods in Bangladesh deposit rich silt). And the effect is on national output: the affected area may differ, and as the previous section showed, some survivors suffer even long after the national economy recovers. These studies, unlike many earlier ones, have accounted for the many nonhazard factors that influence output (structural and

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

institutional variables, terms of trade). But our understanding of economic growth is incomplete, so not every relevant factor may have been included. Even so, two conclusions are warranted. First, a disaster has a smaller effect on the national economy, especially if the affected area is small in relation to the rest of the country and there are substitute producers and markets in the affected area. Second, the area’s commercial links with the rest of the country (and world) would moderate the effect. Two additional background papers look at disasters’ effects on output from a

different perspective. In a paper commissioned for this report, Lopez (2009) develops a general equilibrium model and shows that while disasters can have dramatic negative effects on the level of per capita income, they may propel a formerly stagnating economy into a virtuous path of continuing growth. Under certain conditions (if disasters reduce the tangible to intangible asset ratio in an economy, and if governments do not repeat past policy biases against intangibles), the rate of per capita income growth could increase over the long run. In another background paper prepared for this report, Dercon and Outes (2009)

examine 240 households in six villages in the Indian states of Andhra Pradesh and Maharashtra over 30 years (1975–2005, with gaps in 1983 and 2001) to empirically test the impact of disasters on income levels in these villages. They predict income over time and find it to be lower than what they project by regressing current income (available for nine waves) on the earlier year’s income using as instruments annual village rainfall (alone and interacted with land area and the number of children per household). Much depends on the accuracy of their projected incomes and the importance of omitted factors like the prices of the commodities the farmers grow. They find that droughts cause some households’ incomes to plummet and not recover, especially in households with lower education and landholdings in the 1970s. They interpret this as a permanent loss. To summarize, even short-lived impacts of disasters on health and education can

have long-term effects on income and well-being. Disasters always reduce well-being of those affected, but may or may not have a negative impact on output growth in the medium term (5 years), which depends, in part, on the severity and type of hazard and level of economic development. Storms and droughts seem to have systematic negative impacts on medium-term growth; not so for floods and earthquakes. But severe disasters (10 percent of all disasters) have adverse effects regardless of their type.

Measuring the damage: Twice over and half under?

Measuring damage twice over

Many estimators, reporters, and aid agencies add damages (to stocks) and losses

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

(of flows) together—which may result in double counting, as noted above.23 Consider the collapse of a building with rented apartments: when rents and building values are observed, one finds that the collapsed building’s value (“damage”) is the present value of lost rental stream (future flow “losses,” adjusting for maintenance and other costs).24 Buildings may not change hands frequently, and space may not always be rented out. But even if prices and rents are not readily observed, conceptually, the lost asset value from physical damage equals the present value of the lost income flow from the affected assets. While this relationship is clear with privately owned assets, valuing damage to

public infrastructure is more complicated. Why? Because these assets do not have market-based valuations. Valuing the flow of lost economic benefits is harder; and the economic rate of return on the public asset may be far greater than the rate of return on private capital (especially in countries where infrastructure is insufficient to begin with).25 Even so, a damaged asset generates a smaller income flow, and the economic value of the physical damage is the present value of this reduced flow, which may not equate to the capital loss or to the cost of repair and reconstruction. This implies that adding measures of the lost social benefits from damage to a public hospital (due to reduced access to care), and the cost of reconstruction (as a crude proxy for the lost value of the asset), would double count the output losses. The discussion applies to lost output from affected physical capital. But output

could also fall without damage to physical assets—for two very different reasons.26 Take droughts: without water (an input), harvests decline, reducing aggregate output in agricultural economies, although the long-term value of the land may not be affected. The effect is not limited solely to agriculture. And it is not just agricultural growth that is affected: droughts, again through a direct effect on input (water), could reduce industrial output—as in Kenya—through reduced hydroelectric power generation.27

Disruption is the second reason for a decline in output without damage to physical capital. SARS in East Asia disrupted travel and the supply chain that spanned countries reducing output—though there was no physical damage to assets and very few died from the disease (Brahmbhatt and Dutta 2008). So output could fall without physical damage. But a disaster often results in both physical damage and disruption, and keeping the two conceptually separate avoids error in measurement such as double counting. Lost rentals of a destroyed building (either explicitly observed or implicit) are the

direct effects, but may be indirect effects as well. Displaced people may travel longer distances to work and food grains for consumption and cement for repairs may be costlier because roads have been washed away. To measure all the indirect effects, however, indirect benefits should also be estimated. Tourism to the affected area may decline, but output elsewhere would rise when the tourists travel to other

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

destinations.28 These effects, perhaps significant, are more difficult to measure—and this is not done systematically and consistently, perhaps because they accrue to non- victims and are diffused over a wider area. The loss estimates therefore rarely measure reduced national output. Even within the affected area, overall loss measures mask the fact that not everyone is adversely affected (those with undamaged fields and silos benefit from higher prices of grain). Damage measurement is highly sensitive to the measurement concept. Consider

estimating the value of physical damage when Cyclone Sidr knocks down a thatched hut in Bangladesh (for which there is neither a rental nor a property market). Is the damage what the farmer had spent in materials with or without the (forgone) value of his time in building it? This “acquisition cost” (what it cost the farmer) could differ substantially from “replacement cost” (what it would now cost to rebuild the hut) or from the conceptual asset value of the structure (what the lost structure could have fetched in exchange).29

These are different concepts, but there is no record of many of these measures, so the estimator makes an educated guess that depends on the purpose at hand. Donors (domestic and foreign) may want to know, “What do I have to spend to replace the farmer’s hut?” Foreigners would consider the landed price of imported materials (such as sheet iron and steel), adding the local labor costs (at “fair” wage rates) if they intend to build the structure before giving it to the farmer. The local NGOs may consider the prices of the locally available bamboo, and consider the prevailing wage rates to arrive at a lower number—and the two may differ by much more than the transport cost—because the “law of one price” does not hold internationally (Isard 1997). For the victims, few of whom wait for governments or donors to rebuild their house or provide the materials, the relevant measure of damage is, “what is the cheapest way for me to get it working/habitable again?” If the Bangladeshi farmer could recover some of the scattered material and rebuild

while waiting for his flooded fields to dry (when his time may be worth little because there are no competing farming demands), the expense incurred (the replacement value of the “damage”) would be far less than what enumerators estimate.30 And this amount cannot be ascertained by asking victims (usually through translators with local officials also present) because the prospect of aid may influence their answer. Questioning these claims would add insult to victims’ injuries; and estimators, being human, are moved by the very visible deprivation of the victims.31

Reported estimates mix many concepts. Moreover, such estimates from earlier assessments are not compared with subsequent output declines. To make this comparison correctly, one must also take account of other factors that affect output (as summarized earlier). Nor do the loss estimates measure the decline in victims’ well-being. Sometimes what the government provides is added (in cash or in kind, as with food or tents). But this fiscal cost is only a transfer (from taxpayers to the

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

beneficiaries) and not an output loss. The fiscal cost may be relevant when requesting aid; but the effect on output should not be confused with its effect on the budget. The point is that accurate estimates are more likely when the purpose of measurement is clear.

Measuring damage half under

Biases in measurement can also go the other way, leading to underestimates of damages. Although the dead are counted, damage estimates ignore the value of lives lost (the difficult conceptual and ethical issues of valuing consequences of risks to life are discussed in chapter 4). The destruction of “the commons”—environmental buffers, forests—is rarely included because they are difficult to value and have no well-defined claimants. Such effects could be substantial: Markandya and Pedroso- Galinato (2009) find that disasters (earthquakes, storms, and floods) destroy natural capital (cropland, pastureland, and protected areas) and that the destruction is greater when disasters last longer.32 The effect on natural capital is further complicated because it is not possible to distinguish disasters that have positive side- effects (floods that increase fertility or forest fires that sustain forests) from those that do not. Clearly, what is valuable is not always valued. Recognizing that GDP is not a perfect metric of welfare, another background paper

goes beyond the effect of disasters on output to estimate the effect on “genuine savings,” (Mechler 2009).33 This is an alternative welfare indicator based on concepts developed for green national income and wealth accounting (see Hamilton and Atkinson 2006). Genuine savings aims at better measuring the “true” national savings by adding investments in human capital and subtracting the consumption of capital stock, the depletion of natural resources, and the adverse effects of air pollution. Disasters, by reducing genuine savings, could affect medium to longer term welfare (as measured by changes in consumption expenditure over 5 to 33 years). The findings, though tentative, suggest that including all disaster asset losses may better explain variations in post-disaster welfare, and these findings are most pronounced for low-income countries, perhaps because of their greater dependence on natural capital. This is likely to be an underestimate because of the limited number of observations, mainly because of a lack of genuine savings data for some highly vulnerable countries, such as for many disaster-prone Caribbean countries.

Improving measurement, clarifying purposes

An assessment can have several purposes. But clarity is needed about who makes what decision and which estimate is most relevant for accurate measurement. This section makes three points. First, comprehensive damage assessment of public infrastructure is useful, especially if decisions on repairs and priorities are made centrally. Second, decisions for the sequence of repairs and the funds require estimating a disaster’s fiscal effects, which is different from estimating property

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

damage. Third, whether it is useful to try and value damage to private property in an attempt to be comprehensive is questionable, especially if compensation is not linked to damage: it is unclear what decision requires it, and if possible biases could be avoided. If the reason is to determine where and whether the government should rebuild, gauging the extent of damage (as opposed to valuing it) may be a better option. Likewise, the merit of estimating output declines by sector is unclear because of high sectoral interdependence. Projecting sectoral output correctly is highly complex and useful only in a few situations, and market prices may be sufficient signals of shortages.34

One purpose of post-disaster efforts is to promote a quick recovery. Supply chains and services (such as banking) are often disrupted, and it is in peoples’ self-interests to restore these services by drawing on established family and commercial ties. In a background paper for this report, de Mel and others (2008) examine the post-tsunami recovery in Sri Lanka and find that despite the lack of insurance and low aid flows, affected households drew on their own savings and that of relatives and friends to replace 60 percent of lost assets (microenterprise owners two-thirds) by the summer of 2007; three quarters of microenterprise owners had replaced all their damaged housing by April 2008.35 Not everyone recovered as rapidly, but many did recover. People rebuild their lives and livelihoods more quickly and easily when their commercial links with the rest of the country (food, building supplies, telephone, and banking services) and within the region are restored, including public infrastructure (roads, bridges).36

Box 2.2 Revenues and expenditures: Disasters’ fiscal consequences Lis and Nickel (2009) examine the budgetary effect of large weather disasters (droughts, heat and cold waves, floods, storms, and wildfires) in 138 countries between 1985 and 2007. They adjust for the effects of other variables on the deficit such as business and political cycles (parties in power). Large disasters are defined as those affecting 100,000 people or more. Such disasters raise the budget deficit in developing countries by between 0.23 and 1.1 percent of the GDP but they rarely do so in rich (OECD and EU) countries. In a three-year study, Benson and Clay (2004) examine country studies of disasters to assess their

economic effects including those on government finance. They find that accounting systems do not track spending in ways that allow a thorough analysis (as chapter 3 also observes), but they have some interesting insights. The three elements of government finance examined are revenue, expenditure, and external assistance

(typically from international donors). The effect on revenues was most difficult to estimate: the structure of taxes changed in Bangladesh after trade tariffs fell starting in the 1980s, so econometric estimates from past data would be a poor guide. Countries differ substantially in their sources of tax revenues: Montserrat relies more on consumption taxes, and personal incomes and consumption fell after the volcanic eruption in the 1990s. Disasters increase government expenditures almost immediately. Budgets are reallocated and relief

spending rises after a disaster. This happens almost every year in some countries like Bangladesh. Benson and Clay find evidence that such reallocations come at the expense of maintenance in Dominica. Capital spending, largely discretionary, falls. But in some countries (the Philippines), budget headings are too broad to tell what is happening. Disasters’ long-term effects on government spending are also difficult to estimate: budget categories

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

change, and disaster management spending is not a separate category. Moreover, multiple agencies and public enterprises keep different accounts and do so differently (firms are on accrual, budget is on cash basis), and several state-owned enterprises (including banks as in Bangladesh) also provide relief. Benson and Clay find that although donors frequently provide aid after disasters, they do so by re-labeling the

funds without increasing the aggregate amounts. Disasters have had little effect on aid trends in Bangladesh, Dominica, and Malawi, which all received substantial aid even before disasters. This suggests that post- disaster aid may not augment what governments have to spend, so they may ultimately have to rely on their ability to tax their people and spend accordingly.

Source: World Bank staff.

The government must repair damage to public infrastructure to restore severed links. This requires decisions on the sequence of repairs and on the government’s wealth. Estimating a disaster’s fiscal effects therefore takes on some urgency. Forecasting tax revenues (to pay for rebuilding) may be the harder task, and even when revenues fall by a small fraction of national output, the fiscal sustainability of many developing countries may be jeopardized. A wider budget deficit in poorer countries following a disaster underscores the importance of careful spending (box 2.2). While the better off find the resources to rebuild, many are left destitute.

Governments sometimes build temporary shelters and provide relief, but those who have lost everything (say, if what land they had is deemed unsafe) have nowhere to go and may need direct government assistance. It would be a misnomer to call such transfers from the government (land to resettle, or payments in cash and kind) “compensation” because the amounts are typically small (generally less than two times per capita GDP, and the relatively better off get little even if they lost more assets). It may be useful to limit such transfers to those who are both poor and who have

incurred damage (a subset of those in the affected area), though distinguishing between the chronically and temporarily poor is difficult. The difficulties are compounded especially when the help is needed quickly. Morris and Wodon (2003) examine the allocation of relief after the 1998 Hurricane

Mitch in Honduras and argue that “the nature of emergency aid often makes it difficult to allocate aid in a differentiated pattern among beneficiaries.” Examining data from a household survey conducted six to nine months after the hurricane, they find that the chance of receiving relief was related to assets lost but inversely related to wealth (i.e., the rich are less likely to get help); but if one controlled for whether their dwelling was damaged, the amount of relief was neither related to pre-Mitch wealth nor assets lost. Put simply, what a person got in kind (food aid) after a house was damaged was not related to the value of what was lost or owned before. Mauritius, a small country with few people, distributed transfers in a public meeting based on simple observable criteria of house damage (so any deception is deterred by public disclosure). Pakistan, recognizing the challenges after the earthquake (described in

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

chapter 3), gave each person or family fixed amounts for relief and to help rebuild their destroyed houses. If the transfer amounts, whether for relief or rebuilding homes, are far lower than

the damage incurred, why measure damage to private property? Aid could be given to all in heavily damaged districts or counties (as in Pakistan or Mauritius), and these districts could be identified through aerial photographs or satellite imagery. All Africa Global Media (December 3, 2009) reports that the Kenya-based International Livestock Research Institute will analyze freely available satellite data from the National Oceanic Atmospheric Administration that could distinguish live from dead vegetation in northern Kenya. This data could tell insurers whether to pay claims, thereby reducing the costs of verifying the distinction by a visit. Haiti provides a more recent example, where Operation GEO-CAN—Global Earth Observation Catastrophe Assessment Network—estimated and classified building damage based on high- resolution aerial imagery in areas severely affected by the earthquake. The first set of damage maps for the city of Port-au-Prince was produced within 48 hours of the project’s commencing.37 Such assessments would be easier than valuing the damaged assets, which is highly susceptible to incentive and other measurement issues. Providing aid based on assets damaged in such cases would obviate the time and effort to measure and value everything. It is important to understand the limits of damage assessments when using them.

Much of the discussion here applies to disasters that cause destruction on a scale relatively small compared to the rest of the economy, with the economy expected to recover (ultimately) to its former state. But attempts to measure and value damage for such tragedies as the January 2010 Haiti earthquake—where the scale of destruction is such that it rewrites the future landscape—may be misplaced. If a disaster fundamentally altered a whole economy, neither flow nor stock estimates before the crisis would reflect the new long-term equilibrium after it. In such cases, estimating the value of damage matters less than identifying the prevention measures. Measures to move from the depths of the disaster to a new and different post-

disaster resilient state will depend on what that state is envisaged to be. On preventing future disasters, later chapters explain why no single or simple measure exists: effective prevention requires cooperative measures. And the underlying cause of a disaster (and thus the effective prevention measure) is less obvious than its proximate cause. The assessment following the 2009 cyclone in the Lao Democratic People’s Republic found that people were not adequately warned of the impending flood, although such predictions were possible from upstream flows and rainfall measured over the previous several days. Better weather and hazard forecasting (chapter 4) would clearly have helped, but would dams upstream have been more cost-effective? Such searching questions are difficult for a damage assessment to answer.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Spotlight 2 on Turkey Where civilizations and tectonic plates meet

The Marmara earthquakes struck I.zmit, an industrial city some 90 kilometers east of Istanbul, with a 7.4 magnitude early on August 17, 1999, before most people awoke, and struck Düzce, with an epicenter about 100 kilometers east of the I.zmit earthquake, with a 7.2 magnitude on November 12 of the same year. In I.zmit alone, building collapses killed 17,000, injured another 40,000, and made

about 200,000 homeless. Total damage, estimated at $5 billion, could have been worse. The fire that raged for days when a 90-meter-high reinforced concrete heater stack collapsed at the I.zmit refinery did not spread to the adjacent oil storage tanks. In Düzce, close to 700 lives were lost. After caring for the dead and injured, the government considered how to prevent

similar disasters. Seismic fault lines crisscross the country, and many cities are on fault lines, with new faults discovered as detection technology improves. The 1999 Marmara quakes followed a well recorded westward movement of earthquakes along the 1,500-kilometer North Anatolian fault line (spotlight map 1). Scientists think it very likely that between 73,000 and 120,000 people will be injured if a major quake shakes Istanbul, home to 12 million.1 The Turkish government tried three things to preclude further disasters: increase insurance coverage; improve buildings’ quality; and better prepare itself. Following is a discussion of each.

Spotlight map 1 Turkey is at the meeting of three tectonic plates

Source: World Bank staff.

Reducing government liability—promoting insurance

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

The government was financially responsible for rebuilding even privately owned structures that collapse in an earthquake. The 1959 law that stipulated this (no.7269) eroded public finances (it is impossible to pay claims with insufficient tax revenues, especially taking into account that Turkey’s macroeconomic stability is recent). It also eroded the owners’ incentive to construct sound structures. Almost immediately after the 1999 earthquake, the government sought to amend

the law and established a quasi-governmental Turkish Catastrophe Insurance Pool (TCIP under Decree 587) to cover earthquake damage both direct and indirect (e.g. fires, explosions, and landslides that follow). Turkey had many private insurance firms, but the small insurers were distrusted and few homes were insured. The insurance was compulsory for all residential buildings within a municipality (and to qualify for cheap loans under existing schemes) and voluntary for industrial and commercial buildings and for private homes outside the municipality (such as unincorporated villages). Only time will tell if the government’s explicit refusal to cover uninsured losses

would hold in the political aftermath of a disaster, but the attempt is laudable. Chapter 5 discusses the role of insurance in preventing disasters, and commendable efforts were made in Turkey to encourage insurance on commercial terms (premia reflect some risks and prevention measures).2 But the government’s role remains large, and it is TCIP’s insurer of last resort: it explicitly undertook to pay claims that exceeded TCIP’s funds in an exceptionally large catastrophe. Despite insurance being compulsory and TCIP setting affordable insurance premia,

only 22.3 percent of registered urban dwellings countrywide (slightly over 3 million dwellings) were insured as of June 2009. That was similar to the nonmandatory coverage in California, but far lower than the 30 percent market penetration predicted for 2001 and 60 percent for 2006. Poor enforcement is often blamed, but low penetration reflects deeper difficulties. Only legally built buildings with proper permits may be insured when, as in many developing countries, construction without permits is widespread. Squatter communities—gecekondu, literally “overnight settlers”— spring up in areas unsafe for construction. But the difficulties do not stem from absence of clear title alone: many with clear

title to land do not always secure a permit before starting or extending their buildings. Such structures are liable to be demolished; but the threat may also exacerbate the construction of unsafe structures. Those with insecure rights lack the incentive to build safe structures. There is no evidence that those with permits build better structures than those built “illegally,” though the construction inspection process in place for the “legal” buildings should ensure the compliance with the current technical standards. But lack of information on the hazards, such as precise location of fault lines and how to build safer structures, also contributes to poor building practices. Overall, the improvement of building practices—discussed in chapter 3 and briefly

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

mentioned here—is paramount for a disaster-prone country like Turkey.

Improving building quality

The collapse of housing, typically four to eight stories high with many tenants, accounted for most of the deaths and injuries in the Marmara quakes. Many structures that collapsed were on or too near the fault lines. But clearly some structures are better designed and built than others often of the same vintage. The photograph in spotlight figure 1 shows a collapsed building while the adjacent one does not. spotlight figure 2 shows a collapsed bridge with undamaged but displaced spans; so the fault may not lie only with private owners who flouted a building code.

Spotlight figure 1 Damages to dwellings

Source: Archives of the Turkish Photo Reporters’ Association.

With much of Turkey earthquake-prone, less damage and fewer lives lost require better structures. About 30 percent of publicly owned buildings (3,600 of 12,000) in Istanbul are vulnerable to earthquakes. But a major effort to retrofit and reconstruct important public structures is now under way. A World Bank project (The Istanbul Risk Mitigation and Emergency Preparedness Project, ISMEP approved in 2006) has sought to make Istanbul a more seismic-resilient city by helping its provincial administration develop thumb rules to help choose between retrofitting and rebuilding as well as to fix about 840 public buildings initially. Even though this is less than a quarter of all public buildings and a third of the 2,400 buildings deemed critical, the number of buildings retrofitted is likely to be higher given the additional resources attracted. The unprecedented scale of this retrofitting is improving engineering practices, but only careful evaluation after the project’s completion can tell how successful it has been.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

In addition to publicly owned buildings, some 35 to 38 percent of private buildings are thought to be unsafe, and 70 percent to be below the higher current seismic standards (JICA 2002). Retrofitting is expensive, but more worrying is that with poor records and unlicensed construction, it is unclear how much new construction is safer. And there is every reason to be concerned. Because retrofitting and urban transformation go hand in hand, much attention has been given to improving Turkey’s building code and its enforcement.

The role of a code, more than its contents

Turkey has tried to learn from the experience of others, particularly from California’s and the EU’s experiences. Turkey’s seismic code, originally drawn in 1975, was updated in 1998 and 2007. Together with a 1985 development law that defined urbanization principles and other relevant aspects related to structures, the code provided the basis for safer structures and better urban planning. But compliance is poor, despite a new edict on building inspection that parliament enacted in 2000 to improve it. Chapter 3 explains why flouting the code is also not always an enforcement issue, but a symptom of a different sickness: the unfortunate combination of lack of information and incentives.

Spotlight figure 2 Damage to infrastructure

Source: Arifiye Bridge (by Suleyman Arat from Hurriyet 2009).

Much effort went into adapting the content of California’s quake-resistant norms into Turkey’s seismic code, but greater understanding of the role of a code in a

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

country’s institutional setting would have also helped. Municipalities, including Istanbul, have underfunded municipal engineering and staffed planning departments with unaccredited engineers. In such situations, building codes become mere hurdles to overcome, opening the door to corruption: in 2006, 40 municipal officials in three towns in Turkey were arrested for taking bribes in return for allowing unlicensed construction (Escaleras, Anbarci, and Register 2007). Clearly, the role of a code depends on the situation and that differs across

countries and changes over time. It is unfortunate when attention to building codes (regardless of how appropriate or necessary) distracts from what can be done to improve building practices. Owners need both the incentive and information to build well, and chapter 3 shows that the government can do much to correct the former and supply the latter. It is difficult to say how many fewer lives would have been lost in 1999 had all Turkish buildings complied with the code. But even if all new buildings are well constructed, many people will remain vulnerable because of the large existing stock of buildings of dubious quality. Retrofitting will take time even if it is worth doing. So improving preparedness is urgent.

Improving preparedness

Greater Istanbul and the Istanbul Governorate have sought to reduce the city’s vulnerability by increasing emergency preparedness (skills and technical capacities of response units, but also public awareness and training). A World Bank project is helping build and equip new Disaster Management Centers with modern emergency management information and communication equipment. Containers with first aid equipment and supplies are in several localities, and civic organizations know what to do (spotlight figure 3). Forty-six neighborhoods in Istanbul and 73 in the Marmara region have been equipped with materials for first responders, and the District Disaster Support Project (Mahalle Afet Destek Projesi) has trained 3,136 volunteers. Each of the three pieces—advancing insurance, inculcating safer building practices,

and improving preparedness—is a necessary ingredient for a safer Turkey. The government can complement these initiatives by making access to better information easier and restoring owners’ incentives sooner. No two disasters unfold the same way, however, and only when the next one strikes will the adequacy of these arrangements be known.

Spotlight figure 3 Container for emergency medical relief and the cover of the Handbook for Local Disaster Preparedness

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Source: Istanbul Provincial Disaster and Emergency Directorate.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

CHAPTER 3

Prevention by Individuals

his chapter examines how people choose prevention measures individually, and the next chapter, collectively. It begins with a simple analytical framework to

understand how much prevention individuals choose to undertake, how much insurance to purchase, and how much residual risk to bear. It then concentrates on whether individuals undertake enough prevention. People are guided by information—much of it embedded in prices—and limited by

their budgets: they undertake prevention up to the point when the expected benefits (avoiding losses) exceed the measures’ costs. Yet people differ, and their choices are not identical even when confronted with similar budget constraints. Some choices reflect distorted prices and others inadequate knowledge of the hazards or newer technologies of prevention. Individuals also differ in their risk aversion. Many live in exposed areas known to be hazardous—whether in poverty in Bangladesh or in affluence along the Florida coast. Observing this, some conclude that people are fatalistic or myopic. Recent findings that people misperceive risks lend credence to the view that people do not always act in their own interests, but there are also more prosaic explanations. A detailed empirical study finds that property values in Bogota, Colombia, reflect

hazard risks after correcting for proximity to work and access to such conveniences as public transport. This is consistent with risk being perceived correctly and suggests people make informed choices—even if some seem harsh when people live in riskier locations. But structures that are safe could, with sufficient care and expense, be built in risky areas (on hillslopes, in seismic areas). But when a person’s ownership of property is not secure, the possibility of eviction or demolition erodes the incentive to invest in safe structures. Of 1.2 million land titles distributed in 1996 Peru, land titling is associated with a 68 percent increase in housing renovation within four years (Field 2005). Insecurity of land holdings is not the only thing that erodes incentives to build well:

rent controls or other similar regulations diminish a landlord’s incentive to maintain buildings. Neglected buildings collapse in earthquakes and severe storms kill occupants. The harmful effects of such controls and distorting taxes (such as stamp duties on transactions) accumulate over decades. They have led to poor land use and building size and location (decaying industries on land that could be put to better use). They have also contributed to a housing shortage, leaving the poor to live in unsafe shanty towns that mushroom in and around prospering cities. And they have starved

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

cities of tax revenues, so the needed infrastructure is not built, or is built on the cheap.

Prevention, insurance, and coping: A simple framework Ehrlich and Becker (1972) explain how a person chooses how much risk to bear and how to reduce it given the choices they have (De Ferranti and others 2000; Gill, Packard, and Yermo 2005; Baeza and Packard 2006). The person (or family) can take prevention measures (“self-protection” in their paper) that reduce the loss from a hazard (living on an upper floor or building on a higher plinth to avoid losses from a flood), and buy insurance that compensates for losses when they occur. They also distinguish self-insurance, when the person hopes to be able to absorb a loss, from market insurance, which pays a specified sum when the event occurs. Prevention entails measures that have a cost, and insurance entails a premium, and a person chooses the level and combination that best moderates consumption fluctuations. Everyone makes such choices every day in many settings, and each person may

choose differently. Some buy a costlier car built to reduce the risk of a fatal accident, others a cheaper flimsy car—and insurance. Similarly, some farmers self-insure by planting different crops in dispersed plots, sacrificing some yield by doing so. Informal arrangements (reciprocity with neighbors) reduce the losses from a broken leg or the death of an ox, but they cannot fully handle the risk of a disaster that simultaneously affects the entire local community. Market insurance helps in such cases because it extends beyond the local community. When prevention is “excessively” costly, insurance allows people to make transfers in specified “states of the world” (e.g. if an earthquake occurs). Put differently, people generally choose the desired amount of prevention given

their income—but a few may spend excessively to avoid all risks, and others too little. Taking risks implies that they will occasionally have adverse outcomes and must “cope” with them. Table 3.1 summarizes how people prevent, insure, and cope as individuals, communities, and through governments (coping collectively is “relief and recovery”).

Table 3.1 Individuals and governments prevent, insure, and cope with disasters

Measure Individuals/household Community Government and international organizations

Prevention

• Owning multiple assets and with many sources of income;

• Investments to protect and maintain assets (timely repairs);

• Relocating to safer areas as a group;

• Community-training programs;

• Local public goods and services (community-based information systems, small-scale irrigation and infrastructure projects).

• Good analysis and a system to convey information about risk (disaster risk profiles, raising public awareness, early warning systems);

• Public works;

• Well specified and enforced property rights and, by extension, predictable policies and political

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

• Permanent migration.

systems.

Self- insurance

• Owning both financial and nonfinancial assets (livestock, stored grain, durables).

• Local borrowing and savings schemes;

• Rotating access to common property resources.

• Facilitating markets for different assets, including household goods;

• Ready access to prevailing market prices;

• Adequate physical and social infrastructure.

Market insurance

• Property and catastrophe insurance;

• Agricultural insurance.

• Microfinance (semi-formal); • Savings and credit associations;

• Cereal and grain banks.

• Sovereign budget insurance and catastrophe bonds.

Coping (relief and recovery)

• Temporal migration intensification or expansion of household labour;

• Draw on stocks of social capital (credit, food, charity/begging);

• Running down stocks of human and physical capital;

• Reducing or minimizing household expenditures.

• Rotating savings and credit associations (ROSCAs);

• Inter-household transfers and private remittances;

• Public employment guarantee schemes.

• Safety nets (cash transfers and public works);

• Social investment projects (social funds);

• Disaster aid funds or food donor assistance (contingent loans).

Source: World Bank staff, based on Gill and Ilahi 2000.

Prevention: Do individuals do enough? This section takes two approaches, both limited, to attempt to answer the question. The first examines the financial merits of specific prevention measures, and whether they are “widely” undertaken. The second approach examines whether observed market prices reflect known risks: if they do, one is more confident that people act appropriately in their self-interest. A study commissioned for the report examined the costs and benefits of specific

retrofitting measures that homeowners could take against different natural disasters in hazard-prone areas of four low- and middle-income countries (box 3.1) (IIASA/RMS/Wharton 2009). Figure 3.1 shows benefit-cost ratios for the four examples using assumed (but

reasonably typical) costs: elevating a house with mixed wall, concrete floor, and The, World Bank, Bank World, and Nations (UN) United. Natural Hazards, UnNatural Disasters, edited by World Bank The, et al., World Bank Publications, 2010. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/apus/detail.action?docID=606027. Created from apus on 2017-04-04 13:20:21.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

asbestos roof by one meter in Jakarta; protecting windows and doors in a wood frame house in Canaries, St. Lucia; retrofitting a five-story building to increase quake resiliency in Istanbul; and flood-proofing a brick house by building with new brick on a raised plinth in the Rohini Basin, Uttar Pradesh, India. The benefit-cost ratio is shown for a range of assumed discount rates (0–15 percent) and different expected durations of the structure (1, 5, 10, and 25 years). Prevention seems cost-effective for the above measures in all four cases if the structure lasts 10 years or more.1 For shorter time periods, cost effectiveness depends on the discount rate (for high discount rates, the benefit-cost ratio is less than one for some of these measures, implying that prevention is not financially viable). Are people undertaking such prevention? Some do, others do not. A survey of 254

adults from five locales in Istanbul after Turkey’s 1999 earthquake on risk perceptions and attitudes towards prevention found that while people were aware of the risk, only a fifth of respondents said they had taken some preventive action: 13 percent inside the home and 9 percent for the building (Fis¸ek and others 2002). Only about half those who had taken no action invoked high costs (a possible proxy for a tight budget constraint) as a reason for inaction.2 Such seemingly inconsistent behavior warrants an explanation, and many are turning to the recent findings of behavioral economics.

A walk on the behavioral side

Traditional economists explain peoples’ choices invoking prices and incomes, rarely questioning whether people choose wisely. A growing body of work in cognitive psychology lends credence to these doubts. These disciplines have come together as behavioral economics, and its findings have important implications for how we view risk. Kahneman and Tversky (1979) pioneered this field, and the biases that they and

others have since found go by different labels. Rabin (1998, 2002) surveys this vast and growing literature and lists several systematic biases. People have a loss aversion bias: they care more about the costs of undertaking some action (could be retrofitting or buying insurance) than about its gains, even if these are equal-sized. Ricciardi (2007) surveys the behavioral finance literature, which finds the average investor perceives the pain of a loss twice as much as the pleasure from an equivalent gain. This loss aversion bias is related to the status quo bias: people prefer things as they are to changes that involve losses of some goods, even if these losses are more than fully compensated. It is less clear how such biases translate into prevention measures. Is paying for prevention the immediate loss (in which case there may be insufficient prevention), or is the expected damage the loss that weighs on people (in which case prevention measures would be undertaken)?

Box 3.1 Evaluating the costs and benefits of structural mitigation measures The commonly used metric for measuring the hazard risk of an asset or portfolio of assets is the exceedance

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

probability (EP) curve. An EP curve indicates the probability that a given loss will occur in a given year. Most risk models involve four main modules:

A hazard module characterizes the hazard in a probabilistic manner. Often the events that can impact the risk are described—estimating location, magnitude, and associated annual probability among other characteristics. An exposure module describes a structure or multiple structures that may be damaged. Key characteristics that describe a structure’s susceptibility to damage are defined. A vulnerability module estimates the damage to the exposure at risk, given the magnitude of the hazard. A financial loss module draws on these first three modules to create loss estimates that have a given probability of exceedance.

Based on these modules, an EP curve can be constructed as depicted in the figure below, where the likelihood that losses will exceed Li is given by Pi. The x-axis shows the magnitude of the loss (for example, in dollars) and the y-axis shows the probability that annual losses will exceed this level (Grossi and Kunreuther 2005, Hochrainer 2006).

Box figure 3.1. Example of an exceedance probability (EP) curve

For each case study, relevant measures were selected for reducing losses from the disaster. EP curves were constructed for a representative house or houses with and without the preventive measure in place. Benefits were quantified through reductions in the gross average annual loss (area under the EP curve) after preventive measures are applied to a structure and discounted over the relevant time horizon. Cost estimates of each preventive measure were derived from various sources. Combining these estimates, benefit-cost ratios were calculated.3 Measures are effective when the benefit-cost ratios exceed one.

Source: IIASA/RMS/Wharton 2009.

Figure 3.1 Private preventive measures pay

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Note: The figure refers to the following examples: elevating a house with mixed wall, concrete floor, and asbestos roof by one meter in Jakarta; protecting windows and doors in a wood frame house in Canaries, St. Lucia; retrofitting a five-story building to increase quake resiliency in Istanbul; and flood-proofing a brick house by building with new brick on a raised plinth in India. Source: IIASA/RMS/Wharton 2009.

Experiments also find that people attach greater value to something they already own (“endowment effect”) than they did before having it—even when selling or buying involves no transaction cost. Kahneman, Knetsch, and Thaler (1990) gave mugs worth $5 each to a group of students, and offered to buy them back. Students exchanged their mugs for almost twice what another (statistically identical) group not given the mugs bid for them ($7 vs. $3.50). People seem to prefer what they already possess, and this endowment effect appears in many settings. It suggests inertia or the psychological cost of change: new efforts at prevention are less likely than protective measures already in place, but it does not say whether existing measures are sufficient. Kahneman and Tversky have also exposed systematic misperceptions of

probabilities and risks: people overestimate low probability events and underestimate large probability events. This would imply that Turks may overestimate earthquake risks and, if these translate into action, would overprotect their properties while Bangladeshis would underestimate the risk of floods and under protect their homes and assets.

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

But the biases are not consistently related to the frequency of events: people underestimate the risks they have not experienced and overestimate those that they have. Those who have driven without incident have a lower perception of the risk of an automobile accident than those who had a recent accident. Similarly, the perceptions of risk rise after an earthquake, an infrequent event, and people take more precautions (Jackson 1981). The perceived risk of an airplane crash or a terrorist attack is especially high after one has occurred, and hearing about an event raises risk perceptions less than experiencing it. Hung, Shaw, and Kobayashi (2007) found that those living outside the river dyke in Hanoi who experienced the catastrophic floods of 1971 expected future floods more than others. People are misled by how questions are phrased in a survey or how information is

presented, so “framing” matters. In the classic “Asian disease” experiment, people were asked to choose between two undesirable options to counteract a disease threatening 600 people. Tversky and Kahneman (1981) showed how people chose different alternatives, even though the choices had the same consequences, depending on how the outcomes were described (saving people or people dying). Yamagishi (1997) found that people generally think a disease to be less dangerous when fatalities are conveyed as percentage probabilities (12.86 percent) than as proportions or fractions (1,286 out of 10,000). Keller, Siegrist, and Gutscher (2006) found that psychology students in the University of Zurich perceived a higher threat of flood when flood was presented as a 40-year risk (with 33 percent probability of flood) as opposed to an annual risk (with 1 percent probability of flood). A survey conducted in the United States in 2006 finds that most respondents

assess their risks as “below average” (Viscusi and Zeckhauser 2006). Those in riskier areas who experienced disasters estimate their risks to be higher, but not as high as they should statistically. Put differently, these people appeared to underestimate their risks even though the survey was conducted when the World Trade Center attacks and Hurricane Katrina were neither fresh nor forgotten.

More prosaic explanations

Behavioral economics is interesting, especially the research under way. But should policies change when we know that answers to a survey depend on how the question is phrased and how behavior in experimental settings is inconsistent? Behavioral economics finds biases in both directions. Did Istanbul’s current residents experience the 1999 Marmara earthquake (in which case they overestimate the likelihood of an earthquake) or hear about it (in which case they underestimate the odds)? And if any bias depends on distance, would perceptions cross international boundaries into neighboring Greece? There are at least three more prosaic explanations for why people may take fewer

prevention measures than others think they should. First, people without security of ownership (this includes renters) will be reluctant to incur the expense of prevention—

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

even if they know the benefits—because they would not benefit if evicted. Insecure ownership is widespread, and the country spotlight on Turkey illustrates the prevalence of buildings without permits, often on land to which they do not have clear title. Similarly, landlords would not incur the expense if rents were controlled or rent increases were restricted (as with laws that limit rent escalation in a lease). Second, if retrofitting capacity were limited, perhaps because only a few have the

resources, skills, or special equipment necessary, it would take several years to retrofit the existing stock of buildings—even if retrofitting were cost-effective. A survey like that of post-quake Istanbul cited earlier would find that only a fraction of the buildings were retrofitted. But it is difficult to infer myopia from a snapshot, and subsequent surveys may find an increase in retrofitted buildings. Third, even if retrofitting were cost-effective now, there is an “option value” to

waiting if retrofitting technology itself changes rapidly and costs are expected to decline. Even if the financial returns to retrofitting were high, the returns from postponing the retrofit may be greater still because lower cost technology may soon be available. Under these circumstances, owners who do not retrofit are being far- sighted, not myopic (though tenants would live with the risk of postponing the retrofit). It would be inappropriate to make “policy recommendations” based on such

ambiguous evidence: more searching studies are needed to know whether people systematically ignore risks and why people appear to neglect prevention.

Prices reflect hazard risks when land and real estate markets work If property values reflect hazard risks correctly, people can make informed choices based on prices that guide their decisions on where to live and what prevention measures to undertake. To examine empirically whether property values indeed reflect such risks, one must correct for other desirable qualities (location, view, and other amenities) that also influence property prices. Moreover, unlike stocks (equities) that trade frequently on a centralized exchange, every house and building is unique and trades infrequently. Even when property changes hands, the recorded price may not be accurate if there are taxes or other adverse consequences. And even if prices were recorded accurately, houses that trade in consecutive periods may differ considerably in size, quality, and location. So it is difficult to construct price indices without making some assumptions. Consequently, the price indices may appear to change sluggishly even if the prices (bid and ask) respond quickly to information and to changing market conditions, and econometric techniques must respect these limits of the data. Even so, many studies find that property prices reflect the risk of hazards. Istanbul property values in 2000 were lower near the fault lines in the Sea of

Marmara than those farther away (Onder, Dokmeci, and Keskin 2004). In contrast, proximity to the fault line did not matter for 1995 property value data. The 1999

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

earthquake may have made people aware of earthquake risks, so more recent property prices reflect this. But as the Turkey Spotlight shows, there have been many quakes through history, and a more likely explanation is that after the 1999 earthquake, many fault lines were newly identified and publicized. Similarly in the United States, flood zone disclosure is mandatory in some areas of

North Carolina, so buyers are aware of flood risk before buying a property. Using a hedonic property price model, Bin, Landry, and Meyer (2009) find that the property market reflects geographic differentials in flood risk, reducing property values on average by 7.3 percent. The market capitalizes risk as flood insurance premia equal the discount in property values. Bin and Polasky (2004) examine the effect of Hurricane Floyd on property values in North Carolina (September 1999, affecting 2 million people and causing $6 billion in property damage). Few properties were insured before the hurricane, and the prices of houses in the floodplain fell between 4 and 12 percent. This decline was more than the capitalized insurance premia, suggesting that home owners bore costs that exceeded the insured value. (The reduction of the property values on average was $7,460 and the increase in premia for flood insurance was $6,880.) A background paper for this report examines whether property prices reflect

seismic risks in Bogota, Colombia (Lall and Deichmann 2009). Hedonic models allow measurement of the extent to which land and house prices capitalize the attractions like size, views, and amenities (Lancaster 1966; Rosen 1974), and this technique could also capture the effect of disaster risks. Some 800,000 buildings in Bogota that differed in their exposure to seismic risk were matched on a range of characteristics (such as size, construction quality, distance from the city center, and whether residential, commercial, or industrial).4 This technique implies that the only difference among comparable properties is their level of hazard risk. This allows us to assess whether property values are lower in riskier areas, and if they are, that suggests capitalizing disamenities from hazard risk.

Figure 3.2 Property prices for comparable properties are higher in locations farther from earthquake risk in Bogota

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Source: Lall and Deichman 2009.

Property values per unit of construction were compared in the 10 most seismically risky neighborhoods, grouped by distance from the riskiest area (Figure 3.2). Properties in areas adjoining the riskiest neighborhoods are valued 13,434 pesos higher than in the riskiest area, with the price difference increasing with distance: 28,265 pesos for the second quintile and 124,533 pesos for the farthest quintile of neighborhoods. So, land and property values reflect seismic risks in a poor country, a remarkable

finding that casts doubt over assertions that people are myopic and ignore hazard risks. Office rents also reflect hazard risks. Gomez-Ibañez and Ruiz Nuñez (2007)

gathered data on office rents in the central business district of 155 cities around the world in 2005, along with information that could affect rentals such as construction wage rates, steel and cement prices, metropolitan populations, and incomes. These data were linked to that on disasters hotspots to see whether office rents are sensitive to disaster risks. Rents are lower (by 30 percent) in earthquake-prone cities, but not in cities prone

to floods and cyclones. The results suggest that, where markets function, prices tend to reflect hazard risk.

But what these studies do not distinguish is whether prices reflect risk stemming from exposure (in a hazardous site) or vulnerability (building characteristics that influence damage). This may well be an artificial distinction since technological advances make it increasingly possible to build safe buildings in hazardous areas. There is suggestive evidence, however, that prices reflect even vulnerability—at least when information about vulnerability (building characteristics) is readily available. Nakagawa and others (2007) use a 1998 hazard map of the Tokyo Metropolitan Area to examine the extent to which rents reflect earthquake risk and seismic-resistant construction. The Building Standard Law amended in 1981 to improve buildings’ seismic resistance applied only to new construction. Rents on older buildings (likely less safe) were lower in the risky areas. In Tehran, Willis and Asgary (1997) found by interviewing real estate agents that earthquake-resistant houses in all city districts are significantly more expensive than others. This evidence suggests that vulnerability-reducing measures also tend to be

capitalized into property values—at least when they are revealed through hazard- location maps or data on building quality: expenditures in such measures are likely to be recovered through increases in property prices. And such investments are likely to increase with economic density because people have more to lose with disruptions from natural hazards. Just as we should be careful about inferring too much about aggregate behavior

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

from individual—and often idiosyncratic—behavior, we ought to be cautious about deducing individual behavior from aggregate analyses. Still, the discussion here underscores the role of markets in capitalizing hazard risk into property prices, and the role of prices and information in helping individuals perceive risks and make informed choices. Tokyo is a city where rental and land markets operate reasonably well. When such markets are stifled—as in many developing countries—that reduces the incentives for individuals to undertake such risk reduction measures.

Smothered markets dampen prevention incentives

Prices incorporate a lot of information—even about hazard risk, as just shown—and people make better decisions when markets are allowed to function. So, the importance of making hazard risk information available cannot be overemphasized. Perhaps because of this significance, the political will to not have information on rising levels of risk publicized is often strong. For example, even though FEMA in the United States has updated coastal flood maps for the U.S. Gulf, it cannot get coastal communities to accept them because the information would reduce property prices. Systematic mechanisms for tracking information related to the changing nature of risk, and translating it into risk-related property valuations, would go a long way to increase the incentives for prevention.5

The markets relevant for safe buildings are those not only for land but also for related goods and services: if cement prices are controlled, a black market emerges where prices exceed what they would otherwise be. And if cement were allocated to selected villages or people (deemed deserving or vulnerable), many would surreptitiously sell and not use it because of the high prevailing prices. Although people know that their mud huts may wash away for want of cement, they make the difficult tradeoff if the proceeds could be better used to feed a starving family or buy medicine for a sick child. Important markets have been smothered in many countries, sometimes

inadvertently. For example, price and rent controls imposed by the British Empire during World War II remain in some cities (such as Cairo and Mumbai).6 Mumbai’s building predicament shows how vested interests became deeply entrenched long after that war ended and countries became independent. Buildings in Mumbai collapse during the heavy monsoon downpours because they have deteriorated for decades and because of feeble attempts to improve the situation. Rent controls in Mumbai may have initially benefited tenants at the expense of

landlords, but over time everyone suffers. Rent controls cause landlords to forgo maintenance and neglect their properties, and tenants not only live in dilapidated buildings but die when they collapse in heavy rains. Even if tenants are willing to either pay higher rents or to maintain the building, each tries to not pay his share of the expense (free riding), especially if appropriate retrofitting involves structural changes to the entire residential structure and not to individual apartments. Tenants also may

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

lack the legal authority to make changes to their building’s structure. And even when tenants overcome free riding (and the tenants of the entire building agree), they may lack the title to obtain a mortgage. Tenants with funds soon move to newer and safer buildings, and those who remain are often poor with few alternatives. Tenants in rent- controlled apartments often sublet without the landlord’s agreement, but they would demand the present value of the lower rent in return (“key money” in New York; “pugree” in Mumbai). Their legal right to do so varies, and in Mumbai the sums are so large and the tax rates are so high that this “black money” is rarely declared. Rent controls are not unique to Mumbai or developing countries (Seligman 1989).

They exist in about 40 countries, including many developed countries (Global Property Guide 2009). Rent control laws have remained in place in one form or the other in New York City since 1943, where there are about a million rent-regulated and 50,000 rent-controlled apartments (Council of the City of New York 2009). As recently as 2009, legislation was passed in New York that limits the ability of landlords statewide to increase rents. Such legislation is expected to return to regulation many household units previously attracting market rent (Peters 2009). Rent controls are not the only market distortion. Real estate transactions in many

countries incur a stamp tax—the same that spurred American colonies to rebel in the 1770s.7 The ad valorem is on sales (at a punitive 20 percent rate until quite recently), not on owning property. But taxing transactions reduces property sales and encourages undervaluation when new owners register their claim in the city office, where registrars often do not dispute it, perhaps in exchange for a bribe. So, true market prices are difficult to discern. The revenues are not large, but they do not accrue to the city that provides the infrastructure and services (water supply, garbage collection). Worse, real estate is often transferred or bequeathed without being recorded, making the land register out of date. So, borrowing against property is difficult. More pernicious than low revenues accruing to a part of government that does not provide city services is the poor land use that results—a particular problem in rapidly transforming cities. Decrepit “sick” industries that barely operate (such as once-profitable textile mills) remain on large land tracts in prime locations with easy access to old roads and railway lines while new industries locate where workers cannot easily commute. A city cannot provide services without revenue, and Indian cities depend on what

the state or central government transfers. And when the city’s residents do not pay taxes directly to the city administration, officials are not always responsive to their needs. To prevent excessive demands on existing and ancient infrastructure, the city of Mumbai restricts a structure’s floor-area ratio or “FAR” (a building’s total floor area divided by the lot size) to 2.0 for a four-story building, preventing the construction of tall buildings. Mumbai planners went against the grain of markets: floor-area ratios were 4.5 when introduced in 1964 and instead of allowing denser development to

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

accommodate urban growth, they were reduced to 1.3 in 1991. Mumbai’s buildings have fewer floors than other major cities: a third that of Shanghai and less than a fifth that of Moscow. The potential gains from denser development are so large that some developers have offered to pay for infrastructure in exchange for being allowed to construct taller buildings. But such deals can easily spawn more corruption. Besides, tackling infrastructure in an ad hoc rather than a well-planned way would result in more difficulties down the road. So land use is dismal: growth is accommodated outward not upward, putting greater demands on transport. These difficulties are not unique to Mumbai. Bertaud and Brueckner examine the

welfare costs in Bangalore, an even faster growing Indian city where traffic congestion threatens continuing prosperity (Bertaud and Brueckner 2004). Other cities have attempted to regulate development densities, reducing housing supply on suitable land. In 1979, the federal government in Brazil established the basic legislation at the national level for developing, approving, and registering urban land subdivisions. Among these parameters: a minimum lot size of 125 square meters, with minimum frontage of 5 meters, and a compulsory donation of 35 percent of development area for public uses and open spaces. This effectively zoned many poor people out of the formal land and housing market (Lall, Wang, and Da Mata 2007).

Disparities, discount rates, and the poor

Poor people face disproportionately high hazard risks: the aggregate statistics in chapter 1 show this, and the developments in Mumbai illustrate why. Evidence from Bogota shows that the poor tend to cluster in more hazardous areas. Map 3.1 shows areas of differing seismic risk. Map 3.2 shows that the poor live in the most earthquake prone-areas: on average, twice as risky. What can one infer from this? Property prices reflect seismic risks, so risky

property is cheaper to rent or buy. Not surprisingly, the poor live in these areas—not just in Bogota but elsewhere. As property prices in the worst affected areas fell after Hurricane Andrew in 1992, more low-income households moved to these locations (Smith and others 2006). This is a pattern repeated around the world: the poor often live in dangerous areas and slums are often at risk for disasters.

Map 3.1 An earthquake risk index for Bogota

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .

Source: Lall and Deichmann 2009.

It is sometimes asserted that the poor are (besides being cash constrained) myopic and have a higher discount rate than the rich. Yet those who have carefully examined how the poor live find they save an impressive proportion of their meager incomes (Collins and others 2009). Using 250 detailed, yearlong “financial diaries,” they show that villagers and slum dwellers in Bangladesh, India, and South Africa maintained that even those living on less than a dollar a day save large proportions of their meager incomes. Such savings are entrusted to friends and relatives and do not directly find their way to banks or other formal financial intermediaries. The poor routinely make huge sacrifices for future gains—moving far from their rural families to squalid urban settings to earn money to send home and provide their children with more food and better education. Scrounging through the rubbish bins as rag-pickers is still working, and living in the drainage ditches may not be an intertemporal choice but a location decision that combines cheap land and housing with proximity to employment centers. The poor’s choices are limited by the absence of adequate public goods, such as public infrastructure: most cities in poor countries lack reliable buses, and many regulations deter collective taxi and mini bus services.

Map 3.2 Poor people live closer to hazard-prone areas in Bogota

C o p yr

ig h t ©

2 0 1 0 . W

o rl d B

a n k

P u b lic

a tio

n s.

A ll

ri g h ts

r e se

rv e d .