Distribution of emergency relief in post hurricane mitch nicaragua

Introduction 1.1 Hurricane Mitch and Nicaragua 1.2 Natural Disasters in Developing Countries 2.. Combining household characteristics from EMNV98 and responses to questions about the hurr

The Distribution of Emergency Relief in Post Hurricane Mitch Nicaragua

by Catherine Ambler

A thesis submitted in partial fulfillment

of the requirements for the Degree of Bachelor of Arts with Honors

in Economics

Williams College Williamstown, Massachusetts

May 9,2005

Abstract

This thesis assesses the distribution of emergency relief in Nicaragua after

Hurricane Mitch It describes the household and community characteristics that

increased a household's probability of receiving aid in the first six months after the hurricane In all cases analyses are broken down by three categories of aid-giver: the government, NGOs and international donors, and religious organizations The thesis finds little evidence of systematic targeting towards the most vulnerable households There were well-defined differences among the practices of the different aid-givers and a strong overall emphasis on region, particularly for rural areas

Aclcnowledgements

In researching and writing this thesis I have accumulated debts to many people

On a personal note I would like to thank my friends for providing encouragement and distraction and my parents for their unconditional support and enthusiasm I would like also to thank the many members of the Economics Department at Williams College from whom I solicited help and feedback Honors coordinator Professor Ralph Bradburd and the Class of 2005 honors students deserve a special mention I owe special gratitude to two faculty members who guided me in my work: Professor Alan de Brauw led me to this topic, got me started, and provided technical assistance and substantive feedback throughout the process; and, finally, my advisor, Professor Anand Swamy, has tirelessly offered ideas, advice, reassurance, and prompt responses to more emails than I (and I'm sure he) would care to count

Table of Contents

1 Introduction

1.1 Hurricane Mitch and Nicaragua

1.2 Natural Disasters in Developing Countries

2 Describing the Dataset

CHAPTER 1: INTRODUCTION

The promises of billions of dollars in international money for emergency relief and reconstruction in the wake of the staggering human and material costs of the tsunami that devastated coastal areas of Indonesia, India, Sri Lanka, and Thailand in late

December 2004 provides a stark example of the importance of effective responses to natural disasters Notwithstanding the continued vulnerability of developing countries to natural disasters, the literature on the economic impacts of such disasters is remarkably sparse and focuses mainly on institutional mechanisms for the management and reduction

of risk Very little scholarly work has been done on the allocation and effectiveness of the emergency relief money that flows into affected countries in the aftermath of a

disaster Yet no level of risk mitigation in the affected region could have effectively prepared people to cope with the effects of such a powerful disaster as the East Asian Tsunami Emergency relief is an essential component of the recovery process, and the question of whether or not aid is reaching those who are most vulnerable is crucial to the development of affected countries Thus, economic studies of the distribution and impact

of disaster relief are potentially extremely important for the development of future

policies and strategies

This thesis examines the targeting of emergency relief in Nicaragua in the six months following the impact of Hurricane Mitch in October and November of 1999 The poor countries of Central America have long been some of the most vulnerable countries

in the world; their small populations have repeatedly been subjected to hurricanes,

earthquakes, and volcanic eruptions Before the tsunami Hurricane Mitch had been the worst natural disaster in recent memory Almost 3,000 people were killed in Nicaragua

alone (PAHO 1998) and damage estimates for the country topped $1.5 billion (Lane 2000) Emergency relief was substantial OECD countries sent $105.7 million in just bilateral emergency assistance in 1999, only one of many sources of money to Central America in that year (OECD 2004) However, evidence of government corruption during the relief process and a lack of preparedness and coordination among other aid

organizations indicate that a systematic analysis of where that aid went is important

This thesis is based on data from the Encuesta de Hogares sobre Medicidn de

Nivel de Vida (EMNV), a livings standard survey supported by the World Bank LSMS

program A full survey was taken in 1998 just before the hurricane, and a more limited survey of only affected areas was taken six months after Mitch in 1999 Combining household characteristics from EMNV98 and responses to questions about the hurricane and receipt of emergency relief from EMNV99, I investigate the extent to which poverty, the level of damage suffered from the hurricane, and a variety of other household

characteristics affected the probability that a household received emergency aid

This study is the first empirical investigation of the Hurricane Mitch relief effort

in Nicaragua It contributes to the general literature by expanding on previous work in two ways In my analysis of which households were more likely to receive aid I control for community, recognizing that some households would be more or less liltely to receive aid simply because of where they were located I take this geographic analysis further by also examining which regions of the country were most targeted and what characteristics

of communities affected a household's probability of receiving aid Throughout the thesis I also break aid receipts down by aid-giver to analyze the distinct targeting patterns

of the three main categories ofaid-giver: the government, NGOs and international donors, and religious organizations

The rest of this chapter describes the hurricane and relevant political and

economic conditions in Nicaragua, as well as providing a general background to natural disasters in developing countries Chapter 2 describes the data set and reports summary statistics Chapter 3 describes my regression models and discusses their results Chapter

4 concludes

1.1 : Hurricane Mitch and Nicaragua

From October 22 to November 2, 1998 Central America was besieged by the deadliest hunicane in living memory and one of the three deadliest in modem history.' The storm began as an ordinary tropical depression, one of many that are identified and named by National Hurricane Center in Miami every year Most eventually lose strength and fade away long before they ever hit land, but Mitch persisted On October 24 it was classified as a full blown hurricane with winds of up to 90 miles per hour Despite still being many miles out at sea Mitch's fierce winds claimed their first seven casualties in Costa Rica The hurricane changed course several times; first Cuba and Jamaica and then Belize and Mexico braced themselves for the impact But on October 26, with its winds now reaching a monstrous 180 miles per hour, an unrelated high pressure system in the atmosphere interacted with Mitch and left it idling off the Honduran coast for three days, devastating the island of Guanaja and causing intense rainfall across the rest of the

region

1

Unless otherwise noted the information in these paragraphs is taken from the New York Times article

When Nature Rages by James C McKinley Jr., November 9, 1998

Beginning on October 29 Mitch moved inland and, despite being downgraded to a tropical storm, it roamed across Honduras, dunlping huge amounts of water wherever it went Floods and mudslides wiped out entire villages In five days as much rain fell over Honduras as usually falls over the course of an entire year The heaviest rains moved into northern Nicaragua, and on October 30 the department of Chinandega in the Pacific region of the country experienced the wrath of Mitch at its greatest intensity The

remarltable rainfall of that day, combined with five days of already heavy rain, proved to

be too much The crater lalte of the Casitas volcano collapsed, causing a mudslide of mammoth proportions An estimated 14 villages were covered in mud As Mitch lost steam it moved up through Guatemala and El Salvador before fizzling out as it entered Mexico

The aftermath of the hurricane was gloomy Official death figures for the region range from 9,000 to 11,000 (Charvkriat 2000), and almost 3,000 people died in Nicaragua alone; almost a thousand more were unaccounted for (Pan American Health Organization [PAHO] 1998) 567,752 were listed as injured and 36,368 more as affected (Sistema de

la Integration Centroamericana 1999 from Lane 2000) And human losses were only one part of the story Nicaragua's already poor infrastructure did not stand up well against the power of the hurricane More than 250,000 homes were affected and 870,000 people were displaced (approximately one fifth of the population) 500 schools and 300 health centers were damaged or disrupted and 25 percent of the transportation system was

severely affected; almost 60 bridges were "partially or totally damaged" (International Monetary Fund 1999) Agricultural effects include the loss of one third of the 1998-99 banana crop and over half of the expected tobacco yields for that period (Foreign

Agriculture Service 1999).~ Productive agricultural assets such as livestock were also destroyed in many cases Estimates from the government of Nicaragua put total

hurricane damages at $1.5 billion-6 1 percent of the country's gross domestic product (Lane 2000)

In the November 8, 1998 Boston Globe article "Storm sweeps away the gains of a decade in Central America" David Marcus and Richard Chacon highlight the problems of development left by the Mitch catastrophe Though agriculture had been growing at an average annual rate of 3.2 percent since 1987, for farmers like the featured Herman Pasillos, who lost half of his 17 acre field, those gains meant nothing in the days after Mitch Pasillos still had to feed the eleven members of his family The article comments,

"Pasillos' problem will be the country's dilemma: How do you start over when you have almost nothing?"

The political situation of Nicaragua

Civil war, dictatorship, and U.S intervention have plagued the recent history of Nicaragua The U.S backed Somoza family held dictatorial power for much of the mid twentieth century, but their dynasty collapsed in 1979 when it was deposed by the

Sandinistas, a socialist party that solidified its power with elections in 1984 The

Sandinistas attempted to implement socialist reforms but rising civil war intensified by the illegal American support of the Contra rebel group thwarted many of their political plans An agreement was finally reached to hold democratic elections in 1990 Violeta Chamorro, the candidate of a coalition of 14 opposition parties, was elected president and served for seven years, ending the civil war and stabilizing the economy with market reforms She was succeeded by Arnoldo Alemiin, another member of the anti-Sandinista

More information available at h~://www.fas.usda.govipecad2/articles/99-03/nicaragua.htm,

9

political coalition, who was president in 1998 when Hurricane Mitch hit Shortly after he left office in 2002 Aleman was arrested on coi-ruption charges including money

laundering, embezzlement, and electoral crime In 2003 he was sentenced to a 20 year prison sentence

Though the new Nicaraguan government had been relatively stable, it had also been permeated by corruption, raising questions as to how much a president like Aleman could be trusted to carry out a relief effort that was transparent and honest Reports in La Prensa, an independent daily newspaper in Managua, highlight the widespread fears of corruption during that time Large international NGOs such as Oxfam International attributed delays in the reconstruction process specifically to corruption in the Alernan administration Similar concerns were voiced at the May 1999 co~lsultative group

meeting in Stockholm held to review the reconstruction process Although Aleman returned from the meeting with the promise of $1.3 billion in aid from the international community, that aid was conditional on total transparency, to be achieved through in- country supervision of the relief process by the donating countries The Boston Globe reported that many people were comparing the Aleman government to that of the

Somozas, which stole millions in aid money after a 1972 earthquake These considerable criticisms of the Aleman administration, both internal and external, are good reason to suspect that the distribution of aid was far from politically blind

The economic situation in Nicaragua

Decades of civil war, dictatorship, and foreign dependence have left Nicaragua a very poor country Consistently rated with Haiti as one of the two poorest countries in the Western Hemisphere, purchasing power parity GDP per capita was only $2,300 in

2004 Nicaragua also has one of the highest debt to GDP ratios in the world,

accumulated during decades of reliance on foreign aid In 2003 external debt was 17.6 percent of GDP and in 1997, the year before Mitch, the government spent over half its total revenue servicing the debt (Oxfam 1998) Yet, despite its debt burden, the

Nicaraguan macroeconomy had made some progress during the 1990s because of

diversification and a red~lced dependence on foreign aid GDP growth had been positive

through the mid 1990s; it was 5 percent in 1997 and the upward trend was expected to

continue The vicious impact of Mitch caused growth to drop to 4 percent in 1998, but international aid flows allowed growth to rebound to 6 percent in 1999, a pattern typical

of developing countries hit by a large natural disaster (USAID 2002)

The 1998 EMNV survey, completed just before Mitch struck, provides a picture

of the poverty situation in Nicaragua at the time of the disaster The first column of Table 1 shows the 1998 poverty rates by region 47.9 percent of total households were living in poverty and 17.3 percent were living in extreme poverty,3 about a 2 percent improvement in both categories from the 1993 survey (World Bank 200 1 c) Poor

households were (and still are) characterized by low levels of education, large families, and reliance on agriculture Poverty in Nicaragua also displayed regional and rural patterns Figure 1 shows the broad regions of Nicaragua: Pacific, Central, and Atlantic

As one moves across the country away from the capital near the Pacific coast, poverty increased The two semi-autonomous, largely rural regions that make up the Atlantic region had the highest levels of poverty while Managua (the capital city) and the rest of the Pacific region had the least (World Bank 2001~)

3

The general poverty line is C$4,259 annual (US$l 10 per day) and the extreme poverty line is C$2,246 annual (US$0.58 per day) A person below the extreme poverty line is unable to meet his caloric

requirements

Source: web.idrc.ca/en/ ev-27985-201-1-DO-TOPIC.hN

The Atlantic region is where most of Nicaragua's indigenous population lives, groups that make up the bottom of the country's ethnic hierarchy The region's

development has also been different from that of the rest of the country; incorporated into Nicaragua in 1894, its residents now enjoy a degree of autonomous rule because of their indigenous heritage (Gonzalez 2001) Though it covers about half the country its sparse population makes up only 10 percent of the total (World Bank 2 0 0 1 ~ ) Its geography is mostly humid tropical forests and swamps, and its economy is plagued by poor

infrastructure and underdeveloped markets The second poorest area, the Central region, has been fully integrated into the development of the country, but it is more mountainous than the Pacific coast, more rural, and more dependent on agriculture Its infrastructure, though better than that of the Atlantic region, ranks after that found in Managua and the rest of the Pacific region These poorer areas are also the most rural; Nicaragua's urban

centers (including Managua) are concentrated in the Pacific region (Corral and Reardon 2001) Rural poverty was higher than urban poverty in every region; overall, more than two thirds of the rural population was poor in 1998, while less than one third of urban inhabitants were living below the poverty line (World Bank 2 0 0 1 ~ )

Table 1: Poverty rates by region

Central Urban 39.4 39.4 37.6 Central Rural 74.0 77.6 75.1

Atlantic Urban 44.4 44.4 43.0 Atlantic Rural 79.3 80.6 76.7 Source: World Bank Poverty Assessment, EMNV

1998, 1999,2001

To understand changes in poverty rates in the 1999 and 2001 surveys it is

instructive to appreciate where the impacts of the hurricane were the most intense The areas most visibly affected were in northwest Nicaragua, in the Pacific Rural region near the mudslide at the Casitas volcano The Central Rural region was also acutely affected, and the Atlantic region, the poorest part of the country, suffered the least amount of damage

Columns 2 and 3 of Table 1 show the 1999 and 2001 poverty rates The 1999 post Mitch EMNV suggests that Nicaragua's overall poverty situation was not greatly affected by the disaster Between 1998 and 1999 the national poverty rate remained unchanged; the only significant changes occurred in the Pacific Rural region where extreme and general poverty both fell by about four percentage points, and the Central Rural region where poverty rates in both categories increased by three to four points

(World Bank 2001a) These patterns continued through the 2001 EMNV; poverty fell slightly across all regions and by more than 10 percent in the Pacific Rural area The Central Rural region is the only Mitch affected area that saw poverty rates rise between

1998 and 2001

The World Bank Poverty Assessment attributes these changes in poverty rates to the greater vulnerability of the Central Rural region because of its high dependence on agriculture and the influx of investment into the Pacific Rural region as a part of the post Mitch reconstruction process (World Bank 2001c, 2003) The Pacific Rural region was, for example, the only rural region that improved the percentage of its main access roads that were paved between 1998 and 2001 The Poverty Assessment further argues that national poverty reduction in 2001 was driven by broad economic growth supported by increased investment related to the post Mitch reconstruction efforts Additionally, crops such as rice and beans, staples that make up large percentages of the diets of the poor, have been supported by post Mitch programs Consequently, their prices have risen relatively less than other foodstuffs, driving the decline of extreme poverty These effects can all be attributed to the flow of money into affected areas aimed at repairing the devastation left by the h ~ r r i c a n e ~

However, this analysis does not identify affected and unaffected households, making it unclear whether the investment that was driving the reduction in poverty was actually accruing to those poor Nicaraguans who had been most affected by the storm or whether the gains were coming from relatively undamaged households benefiting from community based aid and reconstruction projects Evidence from Benjamin Davis and

It should be noted that these poverty numbers are based on consumption expenditure Though aid may have been able to reduce poverty measured in this way, it is possible that asset levels were unable to recover A poverty line based on assets could tell a different story

Marco Stampini's (2002 and 2003) panel study of movements in and out of poverty using the 1998 and 2001 EMNV finds that the small reduction in poverty over this time masks significant movements in and out of poverty; some households were able to break out of poverty while others, not previously poor, fell into it Overall rural poverty fell by 6 percent, from 60 percent to 54 percent, but this is just the net of 16 percent of households moving up out of poverty and another 10 percent falling into poverty Additionally, 9 percent jumped from extreme rural to moderate rural poverty, while 7 percent moved in the reverse direction The trend is more rural than urban; only 20 percent of urban

households shifted categories (Davis and Stampini 2002) This evidence indicates a great deal of variability in the economic performance of households during this time period Identifj4ng which households were direct beneficiaries of disaster aid is an important component of understanding why households experienced such different economic

results in the wake of the hurricane

Aid in Central America after Mitch

Anecdotal evidence about the quality of the disaster relief effort in Central

America after the hurricane is helpful in constructing a context for my empirical analysis

of aid distribution in Nicaragua In the days and months immediately following the hurricane governments across the world pledged approximately $9 billion to reconstruct Central America Most of the money was destined for Nicaragua and Honduras, the two hardest hit countries (Thompson and Fathi 2005) The United States was one of the largest donors; by February 1999 the government had already supplied $300 million mostly in direct in-kind emergency relief (White House 1999) Bilateral emergency assistance for Nicaragua from OECD countries was $34.8 million 1998, rising to $105.7

million in 1999 and falling to $17.1 million in 2000 (OECD 2004) The World Food Program distributed 2,329 metric tons of food in November 1998 in Nicaragua and continued to meet the needs of 400,000 people during December They pledged to continue these obligations for six months and countless other organizations made similar promises The unprecedented magnitude of the relief effort makes an analysis of its effectiveness important

As the emergency phase ended, and the cameras began to leave, international assistance began to run out Much of the promised $9 billion was meant for the lengthy reconstruction process-the rebuilding of roads, homes, and schools But most of the money has yet to materialize, even six years later Less than a third of the promised aid was ever raised, and much of what did come was offered to the already debt strapped countries of Central America in the form of cheap loans Though there are some signs of development, many roads and bridges in inland Nicaragua and Honduras are still in ruins, agriculture remains devastated, and planned housing projects have not been completed (Hiscock 2005) Those houses that have been constructed are often unrealistically far from jobs, and, in many cases, the government has yet to provide basic services

(Thompson and Fathi 2005) Though this study will focus on the dynamics of emergency relief in the six months following the hurricane it is important to understand the even greater challenges that the victims of the hurricane faced The best way to set the stage for a healthy reconstruction process after a catastrophe like Mitch is to implement a well organized, fair system for the distribution of emergency relief And with the reality that promised reconstruction aid may never come, emergency aid is even more important

In the days immediately following the hurricane more than 320 NGOs, various church groups, and the Nicaraguan government were attempting to work together to provide relief (The Non-Governmental Emergency Relief and Reconstruction Coalition 1998) Both the country and these organizations were unprepared; the major problems that were encountered were the result of a continual lack of coordination and timing among the various actors (Red Cross 2000) The International Red Cross, one of the largest distributors of aid, admitted that its efforts were hampered by its failure to

coordinate relief contributions and the late arrival of much of its staff and basic supplies (Hiscoclc 2005) The aid was mostly supply driven, meaning that instead of indicating to the international community the real needs of their populations, Honduras and Nicaragua adopted a policy of accepting all forms of aid, regardless of its cultural appropriateness or necessity.% demand driven aid process in which the countries made specific requests based upon the realities of their situations could have been more efficient As is often the case following a major disaster, data was severely lacking and no criteria existed for determining who was affected, how badly they were affected, and who was eligible for relief The public did not receive good information on who qualified for aid and how to obtain it (World Bank 2004) In such a confusing situation it is likely that some of the neediest people were also the most unable to access the emergency resources that they needed

The successes that did come from the relief process mainly came from work done

at the community level.6 The Red Cross's model of "community centered" rehabilitation and relief is similar to that employed by many of the different aid organizations operating

Examples of culturally inappropriate aid include canned food without can openers and American baby formula indigestible by unaccustomed infants

6

Most of these examples come from Honduras

in Central America (Red Cross 1999) The municipal level was described as the "key" to the recovery process Local assistance was much more effective when municipal leaders (mayors) were committed and active or when there were established NGOs in the

community prior to Mitch (World Bank 2004) This suggests that the most helpful relief work was done by those that understood the needs of the different communities that they were working in It also indicates that aid may be geographically variable, even within regions, as those communities that are more effectively organized are better able to distribute aid to those in need Even the U.S military adopted a community based

approach in Honduras Supplies were sent to "hub" towns and then distributed to

"spoke" villages on the basis of assessed need (US Department of Defense 1999) This particular method of aid distribution could lead to certain gravely affected households not benefiting from aid simply because they were located in a less affected village Remote areas are also subject to discrimination Because of the importance of this community based approach, geographic considerations are an important dimension of my analysis of the allocation of disaster relief

1.2: Natural Disasters in Deve1opin.g Countries

The impact of a natural disaster on a household can represent a major economic shock through the reduction of income and destruction of assets Coping mechanisms include cutting expenditures, increasing the time devoted to work, borrowing against future earnings, drawing on remaining assets or insurance, help from friends and family, and disaster aid (Carter and Castillo 2004) However, all of these strategies are

contingent on their availability and in a post disaster emergency situation food may be

scarce, prices high and work unavailable The cost of institutional safeguards like insurance coverage and functioning credit markets is often prohibitive for affected households in the developing world Drawing on an asset base only serves to reduce the wealth of a household and make it more susceptible to the next environmental shock In light of these problematic coping mechanisms the role of disaster aid becomes important;

it is the main financial instrument utilized by developing countries for rehabilitation and recovery The success or failure of households in recovery dictates whether they can recover to their pre-disaster level or whether they become trapped in their post-disaster condition at a low asset base For many households receipt of disaster assistance is the only mechanism that can realistically return them to their pre-disaster level of well- being.7

Disaster assistance comes in three stages: emergency, rehabilitation, and post rehabilitation In the first phase the focus is on survival-providing temporary shelter, medical treatment, food, clothing, and other necessities of life The rehabilitation phase

is about restoring normality to the country Housing is rebuilt and communities achieve self sufficiency The goal of post rehabilitation is development and the reduction of vulnerability to the next disaster (Kent 1987) My analysis focuses on the emergency stage, vitally important not only because it ensures survival but because it sets a

household up for rehabilitation The emergency phase is often chaotic; with so many people suffering it is hard to analyze needs Consequently, emergency relief efforts generally suffer from an inability to determine effectively who needs aid and what kind

' Carter and Castillo (2004) find that in Honduran communities with high levels of trust and altruism the effects of Mitch were less drastic, while those in "social isolation" were less likely to recover

of aid they need, as indeed seems to have been the case during the post Mitch emergency phase

37 percent of the total housing stock provided adequate protection against potential natural disasters (Charvkriat 2000) Less stringent environmental regulations that are the norm in the developing world often also contribute to the magnitude of the damage The effects of Hurricane Mitch are thought to have been more intense because of widespread deforestation in Central America The destroyed forests could have reduced flooding and absorbed some of the excess rainfall that made Mitch so powerful (Farber 1999)."

Reconstruction costs from disasters in the developing world are also

proportionally much higher than those in the developed world, forcing countries to stretch their resources and colnpromise on important aspects of reconstruction The physical damage caused by Mitch was estimated at $1.3 billion, an incredible 61 percent

of Nicaragua's GDP (IMF 1999) In contrast, when Hurricane Andrew, the costliest hurricane in history, hit Florida in 1992 it left behind $26.5 billion in damage, a mere 0.4 percent of American GDP (Charvkriat 2000) The incomprehensible death and

8

The article Central America: a disaster the was waiting to happen can be found at

http://www .zmag.org/zmag/articles/jan99faber.htm

destruction left in the wake of the East Asian tsunami is a sober reminder of how

vulnerable poor countries are to the power of severe weather and geologic events

Between 1988 and 1997 natural disasters across the world claimed 50,000 lives per year and caused yearly damage estimated at $60 billion 94 percent of the world's recorded natural disasters during the 1990s and more than 97 percent of the total deaths caused by these disasters were in developing countries (World Bank 2001d) While average death rates due to natural disasters fell by 75 percent in developed countries through the 1960s, 70s, and 80s, that statistic rose by 400 percent for the developing world Though economic costs from natural disasters were rising across the globe, the growth rates in poor countries were at least twice (and often much more) that of rates for developed nations (Berlte and Beatley 1997) In Latin America and the Caribbean over the past 30 years there has been an average of 32.4 disasters annually resulting in 7500 deaths a year, with average annual costs ranging between $700 million and $3.3 billion (Charvkriat 2000)

Among the Central American countries Nicaragua has been the hardest hit Its rate of cumulative losses to GDP is 338.4 percent, the highest in all of Latin America and the Caribbean except for the tiny island nation of Montserrat (Charvkriat 2000) This illustrates how Nicaragua is vulnerable to natural disasters not only because of its status

as a poor country, but also because of its small size When Nicaragua is hit by a severe weather event the entire country is likely to be affected to some degree; there are no resources from unaffected regions to draw on for recovery And the situation only

promises to deteriorate The incidence of natural disasters across the globe has increased over the past ten years (World Bank 2001d) and some evidence indicates that the number

of severe weather events is likely to rise (Charvkriat 20001, again highlighting the

importance of understanding how effectively the international community is able to respond to these events

Developing countries are vulnerable to disasters at a macro level because of the acute vulnerability of their large poor populations Poor households suffer more because

of insufficient financial resources, poorly built housing in risk prone areas, inadequate transport options for evacuation, insecure employment, and the lack of insurance The inability of poor households to achieve full recovery also increases their vulnerability to future disasters (Morrow 1999) Reducing this vulnerability is essentially about reducing risk But the realities of risk reduction are costly-purchasing insurance or relocating a household, measures that are often too expensive for the poor Additionally, the low frequency nature of the natural disaster risk makes it unlikely that even those households with some resources will choose insurance over a more visible need like investment in agriculture or school fees (Charvkriat 2000)

The inability of the poor to recover has the potential to lead to a situation of increasing poverty in the aftermath of a large natural disaster In Honduras after Mitch poverty rates increased by an average of four percent in affected areas Floods and droughts in Chile have been shown to increase inequality both across communities and across households within communities (Charvkriat 2000) If disaster aid is meant not only to provide for the basic needs of an affected population but also, as most scholars argue, to act as a vital component of the development process, then aid agencies should make a concentrated effort to target poor households in the distribution of benefits Yet some evidence exists that this is not always the case A study of reconstruction after a

1970 earthquake in Peru found that landowners were the first to benefit from new

housing (Charvkriat 2000) Similarly, a study focusing on Hurricane Mitch in Honduras (to be extensively reviewed below) found little evidence of pro poor targeting policies (Morris et a1 2002) Wealthier households are often seen as needier after a natural

disaster because they lose absolutely more, an inevitability of their larger asset stock (Wisner et a1 2004) But the poor lose far more as a percentage of their total property, leaving them with very few options for recovery, and making it essential to design

policies that give poor households the extra help they need As the 200012001 World Development Report concludes:

These responses [of poor households to natural disasters] perpetuate chronic poverty, possibly preventing future economic growth because of the irreversible losses in human and physical capital And that is why, when they occur, among the top priorities should be to protect poor people (World Bank 2001d, 161)

Poverty is not the only household characteristic that increases disaster

vulnerability Age, gender, number of dependents, and access to social and political networks can all affect a household's ability to recover from disasters and merit study in

my analysis of aid distribution (Morrow 1999) Of these gender is particularly important Women, more vulnerable themselves, are also the source of the vulnerability of their children Female headed households are already economically and politically

disadvantaged, and without special attention have limited recovery potential In

Honduras female headed households composed 41 percent of the families in shelters in the capital of Tegucialpa in the first few weelts after Mitch, but as the population in the shelters fell, that proportion rose to almost 60 percent, strongly suggesting that

households headed by women had not been as successful in the recovery process (IADB

1999) Consequently, to achieve the protection of both women and children, relief programs should be designed to favor women

The role of the government is also crucial when considering who may be more likely to get disaster aid Bureaucracy and corruption have the potential to both delay aid and target it away from those who need it most High levels of governmental corruption were present not only in Nicaragua but also in Honduras In both countries NGO and bilateral aid programs were thought to be more transparent (World Bank 2004) Studies

of disaster aid in the United States during the 1990s have found that almost half of the disaster money distributed by the Federal Emergency Management Agency (FEMA) was politically rather than need motivated (Garret and Sobel 2002) Additionally, disaster payments from the Department of Agriculture were higher in states represented by public officials who played a role in the allocation of those payments (Garret, Marsh and

Marshall 2004) The importance of the national government in a relief effort cannot be overstated, and it is therefore critical to understand to what extent these governments tend

to serve the best interests of their citizens

Other empirical work on the targeting of disaster relief

Despite the importance of disaster relief to the natural disaster recovery process most of the literature on natural disasters in the developing world focuses on various aspects of risk and recovery The lack of good data has limited the investigation of the dynamics of disaster aid Saul Morris et a1 (2002) provide the most comprehensive empirical study of aid allocation in their analysis of the effects of Mitch on the rural poor

in Honduras Their survey of 2,398 rural households was designed to capture the effects

of the hurricane on the poorest parts of the population Because of the geographical

distribution of poverty in Honduras their selected villages and towns largely miss the areas of the country where the most visible hurricane damage was present The authors found that these poor households experienced direct loss of income through both

agricultural destruction and the disruption of business activities Though wage labor composes 36 percent of labor income in rural Honduras, just 7.7 percent of households suffered losses in wage income due to the hurricane, indicating that those more reliant on agriculture were more vulnerable to the effects of Mitch Despite the fact that the

absolute amount of wealth lost by the non-poor was much higher than that lost by the poor, the poor actually suffered more because they had so few assets to begin with The authors estimate that households in the lowest wealth quintile lost 18 percent of their pre Mitch assets, compared to only 3 percent for households in the highest quintile They also find that short term (emergency) relief was inadequate for the amount of losses suffered

The survey also collected information on the amount of disaster aid received, and

in a second paper Morris and Quentin Wodon (2003) test both the targeting and

allocation of disaster relief using the probability of receiving aid and the amount of aid received as their dependent variables They employ assets as their indicator of well- being, using both pre Mitch asset levels and asset losses due to the hurricane as

explanatory variables in their regressions Their analysis of aid allocation suggests that the amount of emergency relief did not fluctuate according to need Instead, aid was generally allocated as a lump sum that did not vary according to how much a household lost This is unsurprising given that emergency relief most often comes in the form of food, clothes, medicine, and other necessities that are hard to vary across households

In the next step of their analysis, employing a model that hypothesizes that the receipt of aid depends both on pre Mitch well-being and loss of well-being due to Mitch, Moris and Wodon test whether one was more important than the other in determining who received aid They find that the probability of receiving relief was negatively correlated with pre Mitch assets and positively correlated with asset losses, verifying the conventional logic that the poorer one is and the more one loses the more likely one should be to receive aid However, they also find that the weight placed on asset losses

in targeting decisions was ten times larger than that placed on pre Mitch asset levels, suggesting the very poorest people were not as likely to receive aid largely because they had absolutely less to lose

CHAPTER 2: DESCRIBING THE DATASET

In this chapter I will build the framework upon which my more detailed

regression analysis will rest I first describe the dataset in detail and then explain the variables I will be using in my analysis Finally, I use summary statistics to provide a sketch of some initial patterns in the distribution of emergency relief

2.1 : The Dataset

In 1993, 1998, and 2001 Nicaragua carried out three rounds of a Living Standards Measurement Survey (LSMS) A more limited "post Mitch" survey was conducted in

1999 following the impact of Hurricane Mitch The Encuesta de Hogares sobre

Medicidn de Nivel de Vida ( E M N V ) was designed by the Institute Nacional de

Estadisticas y Censos (the National Institute for Statistics and Census) and supported by

the World Bank LSMS program The purpose of the surveys was to collect data on living standards that could inform policy and shape development strategy The World Bank LSMS program has been operational since 1985 By providing technical assistance as surveys are written and carried out the World Bank has been able to ensure that LSMS data, including that collected in Nicaragua, is of high quality (Kinnon et al 2003) The analysis in this thesis uses the responses to questions about disaster aid from EMNV99 and household characteristics identified in E m 9 8 to analyze the targeting of

emergency relief The 200 1 survey provides an opportunity for some limited analysis of longer term recovery I will not be using the 1993 data because the sample is different

from the one surveyed in 1998 Changes in the administrative units of the country and new census data made a completely new sample more practical (World Bank 2002b)

The sample for EMNV98 aimed to include data from 4,300 households Because

of a higher than expected refusal rate, the final number of interviewed households was 4,209 encompassing a total of 23,208 household members The final number of

households with completed consumption aggregates is 4,040 EMNV98 allows

representative analysis at the national, rural/urban, and regional levels Therefore, the data has been classified into seven distinct geographic groups: Managua, consisting of the capital city of Managua; Pacific Urban and Pacific Rural, consisting of the

departments of Chinandega, Leon, Masaya, Carazo, Granada, and Rivas; Central Urban and Central Rural, consisting of the departments of Nueva Segovia, Madriz, Esteli, Jinotega, Matagalpa, Boaco, and Chontales; and Atlantic Urban and Atlantic Rural, consisting of the North Atlantic Autonomous Region (RAAN), the South Atlantic

Autonomous Region ( U S ) , and the department of Rio San Juan The data can be further broken down by the 147 municipios (counties), and census segments, the last consisting of a maximum of 50-60 households

EMNVOl was designed to be a panel with EMNV98 The surveyors attempted to revisit all the households interviewed in 1998 as well as some new households to account for population growth 4,191 households were visited in the 2001 survey; 301 8 of those were panel from 1998 This represents an attrition rate of approximately 25%, not

uncommon for a panel data set in a developing country (Davis and Stampini 2002)

The EMNV99 post Mitch survey is more limited in scope Because Hurricane Mitch hit very shortly after the completion of field work for EMNV98 it was decided to

return to the field to resurvey households in affected areas in order to update the data The main purpose of EMNV99 was to replace expenditure numbers from households in the 1998 dataset that were located in affected areas, thereby portraying a more accurate picture of the poverty situation and providing for better informed policy choices in the aftermath of the hurricane The survey also included questions specific to the effects of Hurricane Mitch to allow for a more complete understanding of the impact of the

hurricane The fieldwork for EMNV99 was carried out in May and June of 1999

Households were included in the 1999 survey exclusively by virtue of having been included in the original 1998 survey and their presence in a hurricane affected area Efforts were made to find missing households within their original municipio Though EMNV99 cannot be said to be representative of hurricane affected areas in Nicaragua the presence of this data with its link to the pre disaster EMNV98 data provides a unique opportunity to study the targeting of disaster aid My working sample consists of 556 households in EMNV98 and 99 and 439 in EMNV98,99, and 01 .9 The data is fiom five regions, twelve departments, 40 municipios, and 42 census segments

The 1998 and 2001 EMNV surveys included a household survey administered in two rounds, a price questionnaire administered at the community level, and an

anthropometric survey that measured the heights and weights of household members Because of time constraints the post Mitch survey consisted of only a household

questionnaire conducted in one round The household surveys were similar from year to year including information on education, agriculture, expenditures, household assets, employment, general health, fertility, and time use The EMNV99 survey includes a

These numbers drop to 456 and 390 when accounting for missing values of the household damage variable

number of explicit questions about the effects of Hurricane Mitch on housing, agriculture, and household businesses All three surveys contain questions about social programs, but the EMNV99 in particular includes categories of donations typical of emergency disaster assistance such as food, clothes, shoes, medicine, and water All three datasets contain the geographically weighted consumption aggregate used to determine the poverty levels presented in the World Bank Poverty Assessment as well as dummy variables that

indicate poor and extremely poor household^.'^ The 1999 and 2001 figures have been adjusted for inflation

2.2: Variable Description

Table 2 lists the variables I use in my household and geographic analyses Based upon comparison with control variables included in the Honduras study, the realities of the economic situation in Nicaragua, and characteristics that have been shown to enhance vulnerability to natural disasters I used responses from the EMNV98 (collected before the hurricane) to construct explanatory variables to use in my analysis of aid targeting

The damage variables come from a question in the 1999 survey in which

respondents were asked whether their house had been damaged by the hurricane They were then asked to rate the degree of damage to their walls, floor, roof, plumbing, and electricity I classified any household that indicated that one or more of these categories had been "completely destroyed" as a high damage household A11 other households reporting damage were classified as low damage These damage variables are used as a loss variable, assuming that those households that suffered physical damage were those

10

The poor and extremely poor classifications are derived from the consumption aggregate and the

previously discussed poverty lines set by the World Bank

Table 2: Description of variables

Dependent variables:

aidall* A composite of food, clothing, shoes, medicine, water, and construction

materials donations, and employment programs from all sources aidgov* Dummy for aid composite from the government only

aidNGO * Dummy for aid composite from NGOsIinternational donors only

aidchurch* Dummy for aid composite from religious organizations only

foodall* Food aid from all sources

foodgov* Food aid from the government

foodNGO * Food aid from NGOsIinternational donors

foodchurch* Food aid from religious organizations

Explanato y variables:

Poverty dummy* 1 if the household is poor in 1998

Log per capita expenditure 1998 log per capita expenditure

Damage* 1 if household reports housing damage from Mitch

High damage* 1 if household reports housing heavily damaged from Mitch

Low damage* 1 if household reports housing somewhat damaged from Mitch

Gender* 1 if household head is female

High education* 1 if household head education level is beyond primary school

Low education* 1 if household head education is at the primary school level

Elderly* 1 if there was an elderly person in the household

Number of children Number of children 15 and under in the household

Community organizations Number of community organizations belonged to by members of the

household Religious organization* 1 if someone in household is a member of a religious organization Rurallurban* 1 if household is rural

Explanatory variables for

geographic equations:

Poverty average Proportion of households that were poor within each municipio

Damage average Proportion of households that were damaged within each municipio Community average Average number of community organizations belonged to by a household

within each municipio Religion average Proportion of household within each municipio that belonged to a

religious organization Note: * indicates a dummy variable

that were generally more affected and therefore more in need of aid It is an imperfect measure, and because the houses of the poor tend to be poorly constructed and not well maintained one should especially be worried about a correlation with poverty However, the correlation between housing damage and poverty is less than 0.1 Table 3 shows the percentages of households reporting damage by poverty status These numbers do indicate that a greater proportion of poor than non-poor households experienced damage,

but it is not as overwhelming a difference as one might expect Housing damage was certainly not limited to the poor population

Not damaged 1 5 1.07 59.09

Table 3: Households experiencing hurricane damage

Other possible loss measures in the dataset include loss of animals, crops, and the

Damaged

disruption of household business Because each of these only applies to the fraction of

the households that had animals, crops, or a household business to begin with they are also somewhat unsatisfactory There is an additional benefit to housing damage in that it does not tend to over-represent the effects of the hurricane on those households with more wealth In the Honduras study asset loss was used as the need variable resulting in those with more assets appearing needier since they tended to lose absolutely more assets than the poor By using housing damage I will be able to control for need in a way that is more even across the income distribution Additionally, because this sample, unlike the sample in the Honduras study, is taken exclusively from the areas of the country most affected by the hurricane, it can be assumed that all the households in the sample were affected at some level; the housing damage variable is an indicator of those households that were the most affected The housing damage variable provides the most consistent way to compare need across households, both poor and more affluent, mral and urban

The "community" variables are derived from responses to questions in E W 9 8 that asked whether anyone in the household was a member of one of a variety of

community organizations including local development groups, educational councils, women's groups, savings co-ops, sports clubs, professional organizations, and religious organizations In my analysis membership in these organizations is used as a measure of

social capital People in these organizations are more connected, and these connections may help them to access emergency relief I use these responses to create a dummy variable indicating membership in any of the organizations, a variable that shows the number of community organizations households belonged to, and, because membership

in religious organizations is much higher than membership in the other groups

(accounting for more than half the total), I also sometimes use a dummy variable

representing affiliation with a religious organization The construction of the other variables is self explanatory The gender of the household head, the presence of an elderly household member, the number of children in the household, the age of the

household head, and the household head's education level are included because they are all factors that increase one's vulnerability or could affect ability to access aid A dummy variable for poverty and log per-capita consumption are alternately used as measures of pre Mitch well being Poverty increases vulnerability and should be important in

determining who receives aid

My dependent variables are taken from the EMNV99 aid questions Households were asked whether they had benefited from a wide variety of social services since

Hurricane Mitch Many of the services identified in the survey were unrelated to disaster aid or, like questions dealing with road repair, aimed at communities rather than

individual households I selected the direct donation of food, construction materials, clothes, shoes, medicine, water, and land, as well as participation in an employment program as appropriate for my analysis These categories capture the types of goods and services that are needed during the emergency phase of a natural disaster I constructed two separate measures of aid The first is the dummy variable aidall, which is one if a

household received any of the above types of aid and zero otherwise Because the direct donation of food was the most common type of aid I also conduct my analyses on a dummy variable for receipt of food aid, foodall

The questionnaire also asked households receiving aid to identify the source of that aid, therefore allowing for the analysis of not only what types of households were more likely to receive aid overall, but also an examination of whether those

characteristics varied by category of aid-giver I grouped the aid-givers into three broad categories: the Nicaraguan government, NGOs (non-governmental organizations) and international donors, and church groups and other religious organizations I again have categories for both all aid and food aid: aidgov, foodgov, aidngo, foodngo, aidchurch, andfoodchurch Table 4 gives the percentages of households receiving each kind of aid

Table 4: Aid receipts by households

Number of households that Percentage of households that received aid received aid

Table 5: Descri~tive L Statistics

HHhead high education (%) 11.39 3.02 5.13

Note: EMNV98 contains 4040 households with completed consumption aggregates 1789 of the households are rural and 546 of them are in designated Mitch affected areas (were included

in EMNV99) Per capita expenditures are in U.S dollars and were converted at the 1998

exchange rate of 10.58 Nicaraguan cordobas per U.S dollar

Finally, I illustrate broad regional trends across the whole data set and broken down by categories of aid receipt

Table 5 presents summary statistics for both the entire 1998 sample and my sample of households in areas affected by Mitch There are some consistent differences between the two datasets, the most striking of which is the considerably higher proportion

of poor households in the 1999 sample The post Mitch sample is, by necessity, largely rural because the hurricane had much greater effects in rural areas, largely missing main urban centers like the capital city of Managua As Table 5 shows, rural households have higher poverty rates, fewer female headed households, and lower levels of education across the board The numbers in the post Mitch sample follow these trends

Descriptive statistics for households that received aid are presented in Tables 6, 7 and 8 Table 6 is designed to allow comparisons across categories of households in terms

of aid receipts It reports the percentage of households in various categories that received aid Tables 7 and 8 are designed to allow comparison of the targeting policies of aid- givers Each aid-giver's allocation is broken down by type of recipient These numbers are shown only to illustrate initial broad patterns in aid receipt Regression analysis will tell a more complete story I first briefly discuss overall targeting patterns evident in Table 6 and then draw some limited comparisons between aid-givers using the numbers from Tables 7 and 8 In all cases I focus my discussion on the all aid category because, at this level, patterns in the food aid numbers are similar

The first column of Table 6 presents the percentages of households in various

categories that received aidall The numbers show that aid organizations reached a

greater proportion of poor households (67 percent) than non-poor households (45

percent) providing some initial evidence that poor households were targeted in the

emergency relief process A similar difference is present in the proportions of damaged and non-damaged households that received aid Almost 70 percent of damaged

households received aidall in comparison to just less than half of the non-damaged

households, numbers that indicate that households with housing damage may have been

targeted However, the high percentage of non damaged households that received relief also indicates that need as measured by housing damage was far from a prerequisite for aid relief

Patterns are less clear for the other categories presented in Table 6 Though the education variable appears to be important, it is correlated with poverty and no

conclusions can be drawn about its effect without controlling for other factors Small differences are present in the comparison of households in community or religious

organizations (a measure less likely to be related to poverty), suggesting that social capital may have a role in the distribution of aid The most striking difference comes from the rural urban breakdown 67 percent of rural households received aid compared

to only 36 percent of urban households, a not unexpected contrast given that the impact

of the hurricane was more severe in rural areas

Tables 7 and 8 allow for a better comparison of targeting behavior by aid-givers For all aid-givers the percent of people receiving aid who were poor was higher than the percentage of poor households in the total EMNV99 sample Approximately three quarters of the households receiving aid from both religious groups and NGOs were poor, compared to slightly less than 70 percent of government households The other central variable of interest, the housing damage dummy, indicates that aid organizations were targeting aid towards those households that were materially affected by the disaster The

54 percent of households receiving aid that reported damage is higher than the 46 percent

of those households in the total sample The NGO and international donor category was particularly sensitive to housing damage; a full 60 percent of their beneficiaries

Table 6: Percentages of households that received aid

Aidall aidgov aidngo aidchurch foodall foodgov foodngo foodchurch Poor HHs

Nonpoor HHs

Damaged HHs

Non-damaged HHs

Female headed HH

Male headed HI9

HHs in a community organization 62.12 26.77 28.28 17.68 5 1.52 14.65 17.17 12.12 HHs not in a cornrn organization 56.32 21.55 29.02 10.92 43.68 10.63 18.39 7.76

HHs in a religious organization 62.42 24.2 28.66 19.75 52.87 14.01 17.2 14.65 HHs not in a religious organization 56.8 1 23.14 28.79 10.8 43.96 11.31 18.25 7.2

Note: the numbers are percentages of each demographic category that received aid For example 66.67% of poor households received aidall

Table 7: Descriptive statistics by type of aid (aid composite)

Of those Of those affected receiving aidall receiving aidgov receiving aidngo receiving

HHhead any education (%)

HHhead high education (%)

Table 8: Descriptive statistics by type of aid (food aid)

Of Mitch affected Of those Of those Of those Of those

receiving receiving receiving areas receiving foodall

HHhead any education (%)

HHhead high education (%)