economic consequences of catastrophes triggered by natural hazardsrs471cr7450/tr143... ·...

Department of Civil and Environmental Engineering

Stanford University

ECONOMIC CONSEQUENCES OF CATASTROPHES TRIGGERED BY NATURAL HAZARDS

by

T.L. Murlidharan

and

Haresh Shah

Report No. 143

March 2003

The John A. Blume Earthquake Engineering Center was established to promote research and education in earthquake engineering. Through its activities our understanding of earthquakes and their effects on mankind’s facilities and structures is improving. The Center conducts research, provides instruction, publishes reports and articles, conducts seminar and conferences, and provides financial support for students. The Center is named for Dr. John A. Blume, a well-known consulting engineer and Stanford alumnus. Address: The John A. Blume Earthquake Engineering Center Department of Civil and Environmental Engineering Stanford University Stanford CA 94305-4020 (650) 723-4150 (650) 725-9755 (fax) earthquake @ce. Stanford.edu http://blume.stanford.edu

©2003 The John A. Blume Earthquake Engineering Center

ECONOMIC CONSEQUENCES OF CATASTROPHES

TRIGGERED BY NATURAL HAZARDS

A DISSERTATION SUBMITTED TO THE

DEPARTMENT OF CIVIL AND ENVIRONMENT ENGINEERING

AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

T. L. Murlidharan

March 2003

i Abstract

Abstract ________________________________________________________________________

Several important questions related to catastrophes are addressed in this

dissertation. How closely are catastrophes and developmental process related? How is the

post event economic growth related to the losses from catastrophic events? How

important and how long lasting are the various effects likely to be? What is the record of

past catastrophes and what regularities can be inferred from them? Can theoretical

models explain some of these regularities? How will a regional economy behave after a

catastrophic event? How is the effect of a catastrophe propagated to an interacting

region? What pre-event conditions are crucial in explaining the fact that some economies

do better after an event? Can the models explain other post-event behaviors too?

The purpose of this dissertation is twofold. On the one hand, it seeks to detect empirical

regularities in the behavior of economies affected by catastrophes. On the other hand, it

develops various models to study the effect of catastrophes on an economy, which

explain some of the empirical regularities.

Appropriate economic models are developed to explain the observed phenomena.

Theoretical simulations start by perturbing the Ramsey’s model to study the effect of

sudden changes in capital and the post event changes in the productivity. Two extensions

of this model are examined. The first of these studies the effect of efficiency of post-

event reconstruction on subsequent behavior. The second extension studies the effect of a

catastrophe on interacting economies. The behavior of the models from numerical

simulations is corroborated with empirical regression results.

A cross-country study with data from countries from various income groups affected by

different types of natural hazards (earthquakes, floods, hurricanes, and droughts) is

presented. Results based on an econometric model imply that direct losses as a result of

catastrophes are negatively correlated with the post event growth. It is seen that only by

modeling the fact that after a catastrophe reconstructed capital takes time to become

productive can one explain the negative correlation of the direct loss with post-event

ii Abstract

growth. Evidence point to the fact that catastrophes increase the external debt, budget

deficit and inflation. However, these effects are only temporary. Two years after the

event the effect of the catastrophe on the economic growth is statistically insignificant.

A standard regional economic model was used to simulate post event economic behavior

for three historical events - the 1989 Loma Prieta earthquake, 1992 Hurricane Andrew,

and 1994 Northridge earthquake. The model was then used to study the effects of

scenario earthquakes in the Bay Area and the Silicon Valley. The gross regional product,

consumption, investment, and local government spending show declines during the first

two years and then recover depending on the external aid and the efficiency of the

reconstruction process.

One of the main messages of this dissertation is that catastrophes cause myriad problems

in the short term after an event. But efficient reconstruction policies should help the

affected communities to emerge as less vulnerable, more productive and hence

economically stronger regions in the long run. To achieve this efficiency catastrophe

management has to be intimately linked with development policies.

Acknowledgments

iii

Acknowledgments ________________________________________________________________________ The research reported in this dissertation was partially supported by the Shah Family

Fellowship and by the Stanford University Department of Civil Engineering. The author

would like to place on record his gratitude to the many individuals who helped bring the

project to fruition. Professor James Sweeney and Professor Edison Tse offered advice

and support consistently throughout the development of this work. Discussions with

Professor Charles Jones and Mr. Rishi Goyal were extremely useful.

Table of Contents iv

Table of Contents ________________________________________________________________________

ABSTRACT iv

ACKNOWLEDGMENTS vi

CONTENTS vii

LIST OF TABLES xii

LIST OF FIGURES xiv

CHAPTER 1. INTRODUCTION

1.1 CATASTROPHES AND DEVELOPMENT PROCESSES 2

1.2 THEORETICAL MODELS OF ECONOMIES AFFECTED BY CATASTROPHES 3

1.3 EMPIRICAL EVIDENCE 4

1.3.2 Data on Catastrophes 5 1.3.3 Theoretical Models and Evidence on Post-Event Economic Behavior 11

1.4 CATASTROPHES AND REGIONAL ECONOMIES 12

1.5 OUTLINE OF THE DISSERTATION 13

CHAPTER 2. CATASTROPHES AND DEVELOPMENT

2. INTRODUCTION 19

2.1 WHAT IS A CATASTROPHE? 21

2.2 DEVELOPMENT PROCESSES, VULNERABILITY AND CATASTROPHE 24

2.2.1 Macro-Level Determinants Of Vulnerability 29 2.2.1.1 Openness To World Economy 29 2.2.1.2 Development Induced Investment In Large-Scale Projects 30 2.2.1.3 Development And Population Growth 33 2.2.1.4 Development And Urbanization 34 2.2.1.4 Development And Urbanization 35 2.2.1.5 Development And Poverty 36 2.2.1.6 Vulnerability As A “Phase Of Development” 38 2.2.1.7 Development And Government 39 2.2.2 Micro-Level Determinants Of Vulnerability 42

Table of Contents v

2.2.2.1 Presence Of Uncertainties And Development Processes 43 2.2.3 Development, Households And Vulnerability 45 2.2.3.1 Health, Nutrition, And Education 47

2.3 HOW DO CATASTROPHES AFFECT DEVELOPMENT? 52

2.3.1 Macro – Level Effects 53 2.3.1.1 Effects On Development 53 2.3.1.2 Effect On Trade And Investment 54 2.3.2 Effects At A Household Level 57 2.3.2.1 Savings And Investment 57 2.3.2.2 Identifying Transitory Income 59 2.3.2.3 Risk Pooling And Consumption Smoothing 60 2.3.2.4 Effect On Human Capital Investments 61

2.4 METHODS OF COPING- RISK, INSURANCE, CREDIT AND SAVING 63

2.4.1 Households, Groups, Community, Villages 65 2.4.2 Insurance, Savings And Credit 66 2.4.3 Credit, Insurance And Long-Run Development And Growth 68

2.5 AID AND RECOVERY 70

2.5.1 Disaster Aid At The Macro Level 71 2.5.2 Disaster Aid At The Micro Level 73

2.6 POLICY ISSUES 74

2.7 SUMMARY 76

CHAPTER 3: SHORT TERM ANALYSIS USING THEORETICAL MODELS

3.1 INTRODUCTION 78

3.2 MODELING A CATASTROPHE 80 3.2.1 Impact On Consumption And Investment 83 3.2.2 Impact On Welfare 86 3.2.4 Numerical Experiments 87

3.3 MODEL INCLUDING THE EFFECTS OF EFFICIENCY OF POST-EVENT RECONSTRUCTION 98

3.3.1 Model 98 3.3.2 Numerical Experiments 101 3.4 MODEL FOR REGIONAL EFFECTS

3.4.1 Numerical Experiments 114

3.5 CONCLUSIONS 128

Table of Contents vi

CHAPTER 4 EMPIRICAL ANALYSIS 4. INTRODUCTION 131 4.1 PREVIOUS STUDIES 133 4.2.1 Change In Indicators Due To Catastrophes 138 4.3 GENERAL FRAMEWORK AND ECONOMETRIC MODEL 141 4.3.1 Approximation 142 4.3.2 Summary Statistics And Discussion Of The Sample 143 4.3.2.1 Economic Growth 143 4.4.2.2 Effect On Consumption, Investment, Government Expenditure, Net Exports And

Income 145 4.4 EFFECT ON THE ECONOMIC GROWTH 150 4.4.1 Primary Variables 150 4.4.1.1 Direct Physical Loss 150 4.4.1.2 Percentage Affected 151 4.4.1.3 Type Of Hazard 152 4.4.2 Control Variables 153 4.4.2.1 Pre-Existing Economic Conditions 153 4.4.2.2 Health 154 4.4.2.3 Poverty And Inequality 154 4.4.2.4 Government, Bureaucracy, And Institutions 156 4.4.2.5 Infrastructure 158 4.4.2.6 Education 159 4.4.2.7 Trade 160 4.5 INTRODUCTION TO ECONOMETRIC ISSUES 161 4.6 PROBLEMS WITH THE DATA 163 4.7 LIMITATIONS OF CROSS-COUNTRY REGRESSION STUDIES 164 4.8 RESULTS FROM REGRESSION ANALYSIS 165 4.8.1 Growth Rates – Short Term 165 4.8.2 Growth Rates – Average 168 4.9 EFFECT ON MAJOR ECONOMIC INDICATORS 171 4.9.1 Consumption 171 4.9.2 Investment 171 4.9.3 Government Expenditure 171

Table of Contents vii

4.9.4 Inflation, And Interest Rates 173 4.11 CONSUMPTION SMOOTHING AND SAVINGS BEHAVIOR 179 4.12 CONCLUSIONS, EXTENSIONS, AND LIMITATIONS 181

CHAPTER 5 REGIONAL IMPACT OF CATASTROPHES

5. INTRODUCTION

5.1 METHODOLOGIES USED TO STUDY REGIONAL IMPACTS 188

5.2 MODELING PROBLEMS 190

5.4 DESCRIPTION OF EVENTS 191

5.4.1 Loma Prieta Earthquake 191 5.4.2 Hurricane Andrew 192 5.4.3 Northridge Earthquake 193

5.5 A COMPARISON OF THE IMPACTS OF THE EVENTS 193

5.5.1 Effects On The Components Of Personal Income 198 5.5.2 Effects On The Components Of Net Earnings By Place Of Work 198 5.5.3 Dampening Out Effect 201

5.6 SIMULATION OF THE EFFECTS WITH THE REGIONAL MODEL 203

5.6.1 LOMA PRIETA Earthquake 204 5.6.2 Hurricane ANDREW 206 5.6.3 NORTHRIDGE EARTHQUAKE 214 5.7 SIMULATION OF IMPACT OF PROBABLE EARTHQUAKE

SCENARIOS IN THE BAY AREA 218

5.8 MODEL BEHAVIOR WHEN CRUCIAL PARAMETERS ARE VARIED 222

5.8.1 Transfer Payments Effects (Fig. 5.12) 222 5.8.2 Consumer Spending Effects (Fig. 5.13) 222 5.8.3 Government Spending Effects (Fig. 5.14) 224 5.8.4 Labor Supply Effects (Fig. 5.15) 224 5.8.5 Migration Effects (Fig. 5.16) 226 5.8.6 Production Or Fuel Costs (Fig. 5.17) 227 5.8.7 Business Taxes And Credits (Fig. 5.18) 227 5.8.8 Consumer Prices (Fig. 5.19) 228

Table of Contents viii

5.9 SUMMARY AND CONCLUSIONS 230 CHAPTER 6 CONCLUSIONS AND FUTURE WORK

6. INTRODUCTION 231 6.1 CONCLUSIONS 231 6.2 FUTURE WORK 234 REFERENCES 236 APPENDICES CDROM Appendix A - Loss and economic data

Appendix B - Determinants of vulnerability Appendix C - Mathematica© Code for Simulation of Perturbed Ramsey’s Model

Appendix D - Mathematica© Code for Simulation of Model Including Effects of Reconstruction

Appendix E - Mathematica© Code for Simulation of Interacting Regions Model Appendix F - Details of regression for discerning the effect on economic growth Appendix G - Details of regression for discerning the effect on consumption,

investment, government expenditure, and net exports using Penn World Tables

Appendix H - Details of regression for discerning the effect on real interest rate Appendix I - Details of regression for discerning the effect on other indicators

including inflation

List of Figures

viii

List of Tables ________________________________________________________________________

Table

4.1a Disasters in the Caribbean can have significant impact on GDP and growth

(World Disasters Report, 1997)

4.1b Description of variables and their data sources

4.2 Summary statistics for short-term growth

4.3 Summary statistics for average growth

4.4 Summary statistics for external debt

4.5 Summary statistics for budget deficit

4.6 Summary statistics for resource balance

4.7 Specifications and regression analysis describing the effect of catastrophes on

short-term economic growth


average economic growth


external debt


resource balance


budget deficit

4.12 Summary statistics for income, consumption, and savings in the five years

enveloping the disaster year

4.13 Summary statistics for percentage growth rates for income, consumption, and

savings in the five years enveloping the disaster year

4.14 Estimates of consumption changes using lagged changes in income

4.15 Estimates for income changes using lagged savings

4.16 Estimates for consumption changes using lagged savings

4.17 Catastrophic events, associated direct losses, and percent population affected

List of Figures

ix

5.1 Observations on the effects of the Loma Prieta earthquake, Northridge

earthquake, and Hurricane Andrew

5.2 Effect on county’s components of personal income

5.3 Growth rates of the components of net earnings

5.4 Effect on county’s components of net earnings by place of work

5.5 Main economic indicators before the events

5.6 Comparison of model predictions with observed values – Loma Prieta

Earthquake

5.7 Comparison of model predictions with observed values – Hurricane Andrew

5.8 Comparison of model predictions with observed values – Northridge

Earthquake

5.9 Earthquake scenarios and assumptions about regional capacity

List of Figures

x

List of Figures ________________________________________________________________________

Figure

1.1 Overall layout of the thesis

2.1 Determinants of Macro- and Micro- Vulnerability

3.1a Assumed changes in the capital share in the production function

3.1b Changes in initial consumption for various productivity levels

3.1c Evolution of consumption with various levels of productivity using Ramsey’s

model

3.1d Evolution of capital with various levels of productivity using Ramsey’s model

3.1e Evolution of output with various levels of productivity using Ramsey’s model

3.1f Growth of the economy with various levels of productivity using Ramsey’s

model

3.1g Phase space plot of consumption and capital with various levels of

productivity evolution using Ramsey’s model

3.2a Assumed changes in the capital share in the production function

3.2b Changes in initial consumption for various levels of productivity

3.2c Evolution of consumption with various levels of capital loss using Ramsey’s

model

3.2d Evolution of capital with various levels of capital loss using Ramsey’s model

3.2e Evolution of output with various levels of capital loss using Ramsey’s model

3.2f Growth of the economy with various levels of capital loss using Ramsey’s

model

3.2g Phase space plot of consumption and capital with various levels of capital loss

evolution using Ramsey’s model

3.2h Changes in the overall welfare for various levels of capital loss

3.3a Assumed changes in the conversion of maturing capital to productive capital

List of Figures

xi

3.3b Assumed changes in the capital share in the production function for extended

model and the evolution of external aid

3.3c Change in maturing capital with various rates of conversion from maturing

capital to productive capital using extended model

3.3d Change in productive capital with various rates of conversion from maturing


3.3e Change in consumption with various rates of conversion from maturing


3.3f 3D Phase space plot of consumption, maturing capital and productive capital

evolution using extended model

3.3g Change in output with various rates of conversion from maturing capital to

productive capital using extended model

3.3h Growth of the economy with various rates of conversion from maturing


3.4a Assumed changes in the conversion of maturing capital to productive capital

3.4b Assumed changes in the capital share in the production function for extended

model and the evolution of external aid

3.4c Change in maturing capital with various levels of loss using extended model

3.4d Change in productive capital with various levels of loss using extended model

3.4e Change in consumption various levels of loss using extended model

3.4f 3D Phase space plot of consumption, maturing capital and productive capital

evolution using extended model

3.4g Change in output with various levels of loss using extended model

3.4h Growth of the economy with various levels of loss using extended model

3.4i Initial changes in consumption with various levels of loss using extended

model

3.4j Overall welfare changes due to various levels of loss of loss using extended

models

3.5a Change in consumption in affected region with various levels of aid using

regional model

List of Figures

xii

3.5b Change in consumption in unaffected region with various levels of aid using

regional model

3.5c Change in capital in affected region with various levels of aid using regional

model

3.5d Change in capital in unaffected region with various levels of using aid using

regional model

3.5e Phase space plot of consumption and capital for the affected region with

various levels of aid using regional model

3.5f Phase space plot of consumption and capital for the unaffected region with

various levels of aid using regional model

3.5g Change in output in affected region with various levels of aid using regional

model

3.5h Change in growth of output in affected region with various levels of aid using

regional model

3.5i Change in output in unaffected region with various levels of using aid using

regional model

3.5j Change in growth of output in unaffected region with various levels of aid

using regional model

3.6a Change in consumption in affected region with various levels of loss using

regional model

3.6b Change in consumption in unaffected region with various levels of loss using

regional model

3.6c Change in capital in affected region with various levels of loss using regional

model

3.6d Change in capital in unaffected region with various levels of using loss using

regional model

3.6e Phase space plot of consumption and capital for the affected region with

various levels of loss using regional model

3.6f Phase space plot of consumption and capital for the unaffected region with

various levels of loss using regional model

List of Figures

xiii

3.6g Change in output in affected region with various levels of loss using regional

model

3.6h Change in growth of output in affected region with various levels of loss using

regional model

3.6i Change in output in unaffected region with various levels of using loss using

regional model

3.6j Change in growth of output in unaffected region with various levels of loss


3.6k Change in overall welfare in affected region with various levels of loss using

regional model

3.6l Change in overall welfare in unaffected region with various levels of loss


4.1 Comparison of pre- and post-event growth rates (short term)

4.2 Comparison of average pre- and post-event growth rates

4.3a Effect on short term external debt

4.3b Growth of external debt

4.3c Effect on average external debt

4.4 Effect on budget deficit

4.5 Effect on resource balance

5.1 Effect of Loma Prieta Earthquake on personal income of San Francisco-San

Jose CMSA

5.2a Effect on gross regional product (Loma Prieta)

5.2b Effect on consumption (Loma Prieta)

5.2c Effect on capital stock (Loma Prieta)

5.2d Effect on employment (Loma Prieta)

5.2e Effect on consumption deflator (Loma Prieta)

5.2f Effect on government spending (Loma Prieta)

5.3 Effect of Hurricane Andrew on personal income of Dade county

5.4a Effect on gross regional product (Andrew)

5.4b Effect on consumption (Andrew)

List of Figures

xiv

5.4c Effect on capital stock (Andrew)

5.4d Effect on employment (Andrew)

5.4e Effect on price index (Andrew)

5.4f Effect on government spending (Andrew)

5.5 Effect of Northridge Earthquake on personal income of Los Angeles County

5.6a Effect on gross regional product (Northridge)

5.6b Effect on consumption (Northridge)

5.6c Effect on capital stock (Northridge)

5.6d Effect on employment (Northridge)

5.6e Effect on government spending (Northridge)

5.6f Effect on consumption deflator (Northridge)

5.7 Effect of future scenario earthquakes on gross regional product of San

Francisco – San Jose CMSA

5.8 Effect of future scenario earthquakes on personal income of San Francisco –

San Jose CMSA without aid

5.9 Effect of future scenario earthquakes on personal income of San Francisco –

San Jose CMSA with reconstruction aid

5.10 Effect on gross regional product – probable scenarios with no external aid

5.11 Effect on gross regional product – probable scenarios with ant without aid

assuming 10% capital loss







5.15 Effect on gross regional product – transfer payments 10% of loss

5.16 Effect on gross regional product – 1% increase in consumption spending

5.17 Effect on gross regional product – 10% increase in government spending

List of Figures

xv

5.18 Effect on gross regional product – 10% decrease in occupational employment

5.19 Effect on gross regional product – 1% increase in migration of population

5.20 Effect on gross regional product – Increase in relative production costs by

10%

5.21 Effect on gross regional product – 10% decrease in business taxes or 10%

increase in tax credits

5.22 Effect on gross regional product – 10% increase in consumer prices

5.23 Effect on gross regional product – 1% decrease in wage rate

5.24 Effect on gross regional product – 1% decrease in housing prices

5.25 Simulation of personal income of Santa Clara County 1989-1997

5.26 Effect of scenario earthquake on gross regional product of Silicon Valley

5.27 Effect of scenario earthquake on personal income of Silicon Valley

Chapter One: Introduction

1

Chapter One

Introduction ________________________________________________________________________ 1. Introduction What are the effects of a catastrophe on the macro-economic processes? How closely are

catastrophes and developmental process related? What are the socioeconomic

determinants of vulnerability of countries to catastrophes? Do catastrophes actually retard

economic growth? How important and how long lasting are the various effects likely to

be? What trends do past data on catastrophes suggest and can theoretical models replicate

these trends? How will a regional economy behave after a catastrophic event? What

measures will best help the affected community to recover? This dissertation is an

attempt to answer these questions. The purpose of the dissertation is to detect empirical

regularities in the behavior of economies affected by catastrophes and to develop models

to study the effect of catastrophes on a typical economy, which explain some of the

empirical regularities.

A direct tangible consequence of a catastrophe is the huge economic loss, often in the

order of billions of dollars, suffered by the affected region. The vulnerability of a

community to natural hazards depends on various socioeconomic conditions. In addition,

a catastrophe disrupts and changes the complex web of interactions between ongoing

economic, social, and political processes. An intriguing question is whether the

development of the economy of the region is significantly altered by the occurrence of a

catastrophe.

1.1 Catastrophes and Development Processes Catastrophes are not caused by the extremes of nature alone. A catastrophe is

fundamentally a social phenomenon; it involves the intersection of the physical processes

of a hazard agent with the various on-going economic, social, and political processes. For

large segments of the world's underdeveloped population, occurrence of a natural hazard

may worsen an already deteriorating or fragile situation. In order to study the effect of a


2

catastrophe on an economy the factors that describe socioeconomic conditions prior to

occurrence of the hazard event have to be identified. Socioeconomic conditions in a

region are mainly as a result of the developmental processes. The effect of a catastrophe

on the developmental process is complex, especially for developing regions.

Socioeconomic processes including development affect the vulnerability of a community

to natural hazards in subtle ways. The determinants of vulnerability are qualitatively

derived based on empirical observations connecting socio-economic indicators to the

observed losses both at macro - (community) and micro - (household) levels. Socio-

economic indicators that are used include per-capita income, population growth,

education, infrastructure, and quality of governance. This study attempts to bring together

the data and results in disparate fields of study such as growth and development

economics, and sociology of disasters to provide a firm foundation for the connections

between catastrophes and socio-economic processes. The complex ways in which the

occurrence of catastrophe affects the development process is then explained. Methods of

coping with the economic effects of catastrophes are examined. Some implications for

policy design are mentioned.

1.2 Theoretical Models of Economies Affected by Catastrophes

The evolution of an economy after a catastrophic event is investigated in order to analyze

the dynamic effects of a catastrophic event that destroys substantial capital stock.

Ramsey’s model and its extensions are used to address various aspects of the problem.

For these three models, a catastrophe, due to the occurrence of an earthquake or a

hurricane, is modeled by a discontinuous change in the capital stock. The models

simulate the behavior of a typical economy when perturbed by an unanticipated and large

change in the capital stock followed by an arbitrarily complex change in the affected

region’s productivity. The results indicate the initial impact on investment, consumption,

and production.


3

The simulation results point to the importance of modeling the efficiency of the

reconstruction processes after an event. Unless the process whereby the maturing capital

is converted to productive capital is modeled, the fact that post-event growth rate is

negatively correlated with the magnitude of loss cannot be explained. Empirical evidence

presented in Chapter 4, based on data from 43 countries in which catastrophes have

occurred, strongly suggest that greater loss is associated with smaller post-event growth

rates. In addition, Chapter 4 presents evidence regarding an extensive set of pre-event

conditions that are important in post-event recovery. These factors are indicators for the

changes in productivity and the conversion factor that are assumed in simulation models.

Models also indicate the fall in consumption levels after the event. Empirical evidence

indicates losses are negatively correlated with post-event income and consumption

changes are positively related to changes in income. Greater changes in productivity are

reflected in the post-event changes in output.

1.3 Empirical Evidence

Catastrophes triggered by natural hazards such as earthquakes, floods, storms, volcanoes

and droughts are a major global problem. Between 1968 and 1992 major disasters have

affected an average of 113 million people and killed 140 thousand annually (IFRCRCS

1994). More major natural disasters occurred in 1998 than in any other year on record

(MunichRe, 2000). The World Meteorological Organization (WMO) confirmed 1998 as

by far the warmest year since records began. Most catastrophes occur in poorer countries

of the Third World: some 97 percent of deaths and 99 percent of people affected between

1971 and 1995 were in least developed countries (LDCs) (Twigg 1997). Physical

destruction, in absolute terms and the economic consequences of disaster can be very

great. Environmental refugees account for some 58 per cent of all refugees worldwide

(IFRCRCS 1999: 20). Elo (1994) estimated that in 1992 alone the world economy lost

more money from catastrophes triggered by natural events in the LDCs (US $62 billion)

than it spent on development aid (US $60 billion). Three million people per year are

made homeless by flooding (IFRCRCS 1999: 11). During the 1980s the total economic


4

losses from natural disasters exceeded US $10 billion (at 1990 prices) and the average

cost from a single major disaster is now probably about $500 million at mid-1990s prices

(Smith 1996). Catastrophes (post 1970) that have resulted in causing more than the

average cost of $500 million from a single disaster are chosen for the present study.

1.3.2 Data on Catastrophes

Data regarding major catastrophes that occurred around the world between 1971-

1998 is obtained from Center for Research on the Epidemiology of Disasters (Sapir and

Misson, 1992). This includes data on the type of the event, the time when an event

occurred, the place of occurrence, the approximate estimated direct losses, and the

number of people affected. Data regarding economic indicators such as per capita

income, gross domestic capital formation and its growth, gross domestic savings, the

resource balance, and government consumption and its growth are obtained from the

World Development Indicators (World Bank, 2001). Data on institutions, bureaucracy,

education, life expectancy, health, infrastructure are obtained from the web download-

able databases maintained by Easterly and Levine (1997) and Barro and Lee (1995).

Table 1.1 lists the sources of data used for the present study. The complete data set is

provided in electronic form in Appendix A.

It should be noted here that the quality of data associated with catastrophes is not as good

as data on macro-economic indicators. Only in the recent past have efforts been made to

document data from disasters like the EM-DAT database from CRED (Sapir and Misson,

1992). Recognizing that most reporting sources have vested interests and figures may be

affected by sociopolitical considerations, CRED manages conflicts in information by

giving priority to data from governments of affected countries, followed by UNDHA, and

then the US Office of Foreign Disaster Assistance. Agreement between any two of these

sources takes precedence over the third.

Catastrophes are relatively rare events in any given country by definition. In order to

obtain a broad understanding of the effects of catastrophe in different countries, data has


5

been compiled for catastrophes that have occurred in 43 countries. The World Bank

classifies countries according to the income per capita. The four main categories in which

the nations are classified are the high income (>$9266 GNP per capita), upper-middle

income ($2996 - $9265), lower-middle income ($755-$2995 GNP per capita), and low

income (<$755 GNP per capita). In the sample, there are thirteen countries belonging to

the high income group (Australia, Canada, Denmark, France, Greece, Italy, Japan, Korea,

Rep., Netherlands, Spain, Switzerland, United Kingdom, United States), five to the upper

middle income group (Argentina, Brazil, Chile, Mexico, South Africa), fifteen to lower

middle income group (Algeria, Colombia, Dominican Republic, Ecuador, El Salvador,

Guatemala, Indonesia, Iran, Islamic Rep., Jamaica, Mongolia, Peru, Philippines, Russian

Federation, Thailand, Turkey), and ten to lower income group (Bangladesh, Burkina

Faso, China, Honduras, India, Nepal, Nicaragua, Pakistan, Vietnam, Zimbabwe). A

distribution of per capita income of the countries in the sample at the time of occurrence

of a particular event is shown in Fig.1.1. The figure illustrates the fact that the sample

gives a sufficiently general representation of all income groups. Fig.1.2 gives the region-

wise classification of the loss ratios (defined as the total economic loss from all events in

a particular year for country as a proportion of its GDP). It is clear from the figure that

high loss ratios (greater than 1% of the GDP) are concentrated in developing regions of

the world whereas in the developed world majority of the loss ratios are below 1%.

Fig. 1.1 Distribution of per capita GDP in the event set

0%10%20%30%40%50%60%70%80%90%

100%

$100 $1,000 $10,000 $100,000

GDP per capita (USD current)


6

Generality of the inferences depend on including different hazard types in the sample.

The present sample includes various types of natural hazards. It includes 24 earthquake

events, 62 floods, 57 hurricanes or cyclones or typhoons or storms, 20 droughts, and 6

other events such as bush fires, volcanoes, and landslides. It is important to find out

whether the type of hazard has a significant effect on the nature of macro-economic

changes.

The sample has concentrated on post-1970 catastrophic events. This is because no cross-

country study of the macro-economic effects for those events has been attempted. The

study by Albala-Bertrand (1993a, b) concentrates on a few events that occurred in the

1970s. Fig 1.3 is a plot showing the number of events in a year in the sample that caused

more than US $500 million (current) in losses. From the figure it is clear that there is an

increasing trend of such events worldwide. However, if the loss-GDP ratios are graphed

as shown in Fig. 1.4, there is no trend. The loss ratio is therefore an appropriate

normalized indicator of the catastrophe magnitude that can be used to study the

phenomenon over time. By normalizing the loss with its current year GDP, the issues

related to the present values of the past losses has been addressed. Comparison of a loss

of $500 million in 1970 with the same loss in 1990 would have otherwise been

problematic. A similar reasoning is applied to the proportion of a population of a country

affected by a catastrophe since Fig. 1.5 does not show a trend in time. These two

measures of catastrophe, namely economic loss ratio and the percentage of people

affected, are inter-related as Fig. 1.6 illustrates. Higher loss ratios are strongly associated

with more number of people being affected (as a proportion of the total population).

The data on socio-economic indicators are compiled based on the events. For example,

for the 1987 earthquake in Ecuador, all the relevant indicators such as the GDP per capita

are collected for ten years before the event date. The average of this data is used in the

study on the determinants of vulnerability in Chapter 2. Indicators are also compiled three

years around the event year i.e. three years before and after the event. This is used in the

study on the economic consequences of a catastrophe in Chapter 4. In the following two


7

sections, a summary of the results, which will be discussed in detail in the subsequent

chapters of this dissertation, will be presented.

Fig. 1.3 Number of events causing more than US$ 0.5 billion economic loss are increasing

R2 = 0.53

0

2

4

6

8

10

12

14

16

1970 1975 1980 1985 1990 1995 2000

Year

Num

ber o

f eve

nts

Fig. 1.2 Regionwise distribution of the loss to GDP ratios

0.0%

0.1%

1.0%

10.0%

100.0%

Category of events that occurred in a year (1970-'98) in a nation

Loss

as

a pr

opor

tion

of G

DP


8

Fig 1.4 Loss ratios do not exhibit any trend with time

-4.0-3.5

-3.0-2.5-2.0

-1.5-1.0-0.5

0.00.5

1970 1975 1980 1985 1990 1995 2000

Year

Loss

rat

io (l

og s

cale

)

Fig. 1.5 Population affected do not exhibit any trend with time

0.00

0.05

0.10

0.15

0.20

0.25

0.30

1970 1975 1980 1985 1990 1995 2000

Year

Popu

latio

n af

fect

ed (a

s a

ratio

of

tota

l pop

ulat

ion

log

scal

e)


9

Fig. 1.6 Smaller economic loss is associated with smaller percentage of population affected

R2 = 0.30, N= 118

0.000%

0.000%

0.000%

0.001%

0.010%

0.100%

1.000%

10.000%

100.000%

0.0% 0.1% 1.0% 10.0% 100.0% 1000.0%

Economic loss as a percentage of GDP

Perc

enta

ge o

f pop

ulat

ion

affe

cted

(log

log

scal

e)


10

1.3.3 Theoretical models and evidence on post-event economic behavior

Chapter 3 develops dynamic models that simulate the effect of catastrophes. The results

of model simulation suggest various hypothesis that are tested using cross-country study

data from 43 countries from all income groups affected by different types of natural

hazards (earthquakes, floods, hurricanes, and droughts). These included more than 155

events (in some countries, more than one event may have occurred in a year).

Based on an econometric model, detailed in Chapter 4, statistical regularities are inferred

that corroborated the theory-generated hypothesis. Important inferences from the

empirical and theoretical studies are summarized below.

The magnitude of economic loss (as a proportion of GDP) is:

• negatively correlated with the post-event annual percentage economic growth,

• negatively associated with the post-event income.

• associated with increase in inflation and the real interest rates.

• negatively associated with the post-event income and this results in changes in

consumption.

• associated with changes in ex-ante saving behavior at least temporarily after the

event.

A simple cause-effect relation cannot explain the interaction between the occurrence of a

catastrophic event and its impact. The empirical results enumerated above generally

imply that catastrophes retard economic growth and savings, and increase the real interest

rates, inflation, and government spending. However, these effects are only temporary,

since two to three years after the event the effect of the catastrophe on the economic

indicators is statistically insignificant. Nations and regions affected by catastrophe start

rebuilding immediately after the event. However, the recovery process may be complex.

The pre-event socioeconomic conditions to a large extent determine the magnitude of


11

impact and the ‘coping’ strategy of the affected community. The strategies adopted for

coping, in-turn, determine the post-event socioeconomic conditions.

1.4 Catastrophes and Regional Economies Having discerned patterns of economic slow down immediately after the event followed

by growth, a regional economic model is used to explain the post event behavior of an

affected region, in Chapter 5. A standard regional economic model was used to simulate

three historical events. The three events were the 1989 Loma Prieta earthquake, 1992

Hurricane Andrew, and 1994 Northridge Earthquake. Actual observed personal incomes

of the affected counties, as reported by Bureau of Economic Analysis, were compared

with the model generated personal incomes for validation. The model performed well

with a mean absolute percentage error not exceeding 3%.

The model was then used to simulate the effects of hypothetical scenario earthquakes in

the Bay Area (comprising of eleven counties of the San Francisco – San Jose – Oakland

Combined Metropolitan Statistical Area) that might have occurred in the year 2000.

Various direct loss and job loss levels were studied. Simulation results indicate, that for a

$30 billion capital loss and 25,000 - job loss scenario, the Bay Area’s gross regional

product would be down by 14% without any reconstruction and aid (worst case scenario)

during the year of the event. With minimal aid and reconstruction assumptions, the gross

regional product will be lower by 7% and would have totally recovered by the year 2002.

Consumption, investment, and local government spending would show declines during

the first two years and then rapidly grow as the economy recovers. These simulation

results concurred with the simulation of the theoretical model of interacting regions. Two

policy alternatives were simulated – business credits for new investment after the event

and increased local governmental spending. It was concluded that business credits

resulted in assisting rapid recovery after a catastrophe.

One of the main messages of this dissertation is that catastrophes cause myriad problems

in the short run, including slower growth, lower levels of savings and increases in

consumption. But efficient reconstruction policies result in better production techniques


12

for the affected communities. This results in the affected communities to emerge as less

vulnerable and economically stronger regions in the long run. Reconstruction policies

have an important role to play but these will again depend upon extant socioeconomic

factors such as preparedness. To achieve this efficiency catastrophe management has to

be intimately linked with development policies.

1.5 Outline of the Dissertation Fig. 1.7 provides an overview of the layout of the dissertation (chapter and section

number appear on top right corner of the boxes). The diagram brings out the

interdependencies of the chapters. The chapter following this introduction provides a

comprehensive review of literature on economic effects of catastrophes. There are few

studies that address the issues related to catastrophes triggered by natural hazards.

Chapter 2 maps statistically significant associations between socio-economic indicators

and the loss-GDP ratios in the chosen disaster data set based on empirical observations.

These empirical regularities are used to relate literature from development economics and

disasters and to determine the indicators of vulnerability of a nation to natural hazards.

Chapter 3 presents theoretical dynamic economic models that simulate the effect of

catastrophes. Chapter 4 investigates evidence from the consequences of major disasters

that have occurred around the globe and relates it to the theoretical models presented in

Chapter 3. Chapter 5 focuses on the effects of catastrophes at a regional county level. The

simulation results from a standard regional economic model are compared with the

theoretical model presented in Chapter 3. Chapter 6 concludes with pointers towards need

for future research.


13

Table 1.1 Data Description and Sources

Variables Description Source

Primary Loss Current US dollars CRED Number of people affected

Includes persons dead, injured, homeless or otherwise affected

CRED

Type of disaster and its year of occurrence

Earthquake, floods, storms, hurricanes, cyclones, drought, forest fires, and avalanches

CRED

Control Growth of GDP Average annual growth of real GDP per

capita 2001 WDI, World Bank

Institutions Bureaucratic quality 0-6 index Dates : 1982, 1990

PRS; ICRG

Government Rule of law 0-6 index Dates : 1982, 1990

Easterly Levine (1997)

Institutions Freedom from Corruption 1-7 index Dates : 1982, 1990

Easterly Levine (1997)

Government Repression of Civil Liberties 1-7 index Dates : 1980, 1990

Gastil (1990), Gastil (1987).

Literacy Percent of population literate Dates : 1980

Banks (1984)

Illitercy Percentage of "no schooling" in population

Barro and Lee (1993)

Enrollment Gross enrollment ratio for higher education


Enrollment Gross enrollment ratio for secondary education


Enrollment Gross enrollment ratio for primary education


Life expectancy at age zero

Life expectancy at age zero Barro and Lee (1993)

Health Daily calorie intake. Could be used as a measure of poverty.

World Bank's BESD database

Health Daily protein intake (grams). World Bank's BESD database

Health services availability

Number of hospital beds per thousand inhabitants Dates : 1980, 1990

World Bank's BESD database

Roadways Paved Roads/Highways Dates : 1980, 1990

2001 WDI, World Bank

Railroad Mileage per square mile 2001 WDI,


14

Variables Description Source

Dates : 1980, 1990 Average

World Bank

Income Inequality

Gini coefficient for income that can range from a low of 0 to a high of 100

Deininger and Squire (1996)

Poorest Bottom quintile in income distribution 1980, 1990; Average


Richest top quintile in income distribution 1980, 1990; Average


Gender Inequality

Female to male average schooling years, age 26+ 1980, 1990;


Balance of Payments

Balance of Payments as a percentage of GDP


Debt Debt as a percentage of GDP 2001 WDI World Bank

Trade Trade as a percentage of GDP 2001 WDI World Bank

Money Average annual growth rate of the money supply during the last five years minus the potential growth rate of real GDP

Derived from 2001 WDI, World Bank

Inflation Standard deviation of the annual inflation rate during the last five years

Derived from 2001 WDI, World Bank

Consumption Household, government, as a percentage of GDP and their growth


Genuine savings Savings as a percentage of GDP 2001 WDI, World Bank

Interest Rates Real, nominal, interest rate spreads 2001 WDI, World Bank

Government Size

(Government consumption/GDP) 2001 WDI, World Bank

Takings Transfers and subsidies as a percent of GDP

Gwartney and Lawson, 1997

International exchange

Difference between the official exchange rate and the black market rate


International exchange

Actual size of the trade sector compared to the expected size



15

Fig. 1.7 Layout of the thesis and interdependencies among the chapters

2.2How do developmentprocesses determine

vulnerability?

3.2Model 1

Ramsey's growthmodel

3.3Model 2

Model includingefficiency of reconstruction

3Theoretical models

Economic growth, outputand consumption

4.2-6Effect of direct loss on

economic indicatorsincluding growth

6Summary

ConclusionsFuture work

4Empirical Data

Macro-economicIndicators

2.3How do catastrophesaffect development

processes?

3.3Model 3

Interacting Regions

5.3-7Case Studies

Loma Prieta, Hurricane AndrewNorthridge Earthquake

5-5.2Simulation

results froma regional model

3 and 5Catastrophes and

interaction betweenregional economies

2.4Methods of

CopingAid and Recovery

2.1Catastrophe

DefinitionPerspectives

1Introduction


16

Chapter Two: Catastrophes and Development

17

Chapter Two

Catastrophes and Development

________________________________________________________________________

2. Introduction

One of the objectives of this chapter is to show the subtle ways in which various

socioeconomic processes including development affect the vulnerability of a community

to natural hazards. Reversing the causal arrow, the complex ways in which the

occurrence of catastrophe affects the development process is also explained.

About 25 percent of world’s population lives in areas at risk from natural hazard. But the

most vulnerable people are the poorest. 40 of the 50 fastest-growing cities are in

earthquake zones (IFRCRCS 1999: 18). It has been estimated that the richest billion

people on the planet have an average income about 150 times that of the poorest billion

people, who have little choice but to locate in unsafe settings, whether these be urban

shanties or fragile rural environments. The Intergovernmental Panel on Climate Change

(IPCC) says that 60 per cent of the world’s population will be living in potential malarial

zones by 2100. There could be an extra 50 to 80 million cases of malaria and 3.5 million

cases of river blindness (IFRCRCS 1999: 14).

What determines vulnerability of communities to natural hazards? In LDCs broad and

complex socioeconomic problems combine with insecure physical environments to create

a high degree of vulnerability. For vulnerable people in the LDCs, access to resources at

either a household or individual level is most critical factor in achieving a secure

livelihood or recovering effectively from disaster (Blaikie et al. 1994). In addition, risk

varies according to occupation, social class, ethnicity, caste, age, and gender making

vulnerability determination a complex question.


18

In developed countries even major events rarely cost more than 0.1 percent of GNP but

according to Zupka (1988), the negative impact on poor countries can be 20-30 times

greater. In the Commonwealth of Independent States natural hazards have regularly been

responsible for taxing the economy 3-4 times more than in the USA (Porfiriev, 1992).

Some countries have been highly vulnerable. For example, the GNP of the five countries

of the Central American Common Market was reduced by 2.3 percent between 1960 and

1974 as a result of disasters triggered by natural events (Smith 1996). Similarly, small

island countries in the Caribbean and the Pacific Oceans that depend on a narrow range

of primary products (the Dominican Republic in 1979, Haiti, Saint Lucia and Saint

Vincent in 1980, Fiji in 1993) have suffered damage from Hurricanes equivalent to 15

percent of their GNP. Smith (1996) reports that whilst the number of disasters claiming

at least 100 deaths has more than doubled in 30-year period from 1963-1992, disasters

creating economic damage equivalent to 1 percent or more of GNP have risen well over

four-fold.

There has been marked fall in the fatalities for some hazards in many of the wealthier

countries. But the world trend is towards more disaster-related deaths and damages

driven mainly by increased vulnerability in the LDCs. As of 1999, half the world’s

population lives in coastal zones. Ten million are at constant risk of coastal flooding

(IFRCRCS 1999: 11). One of the chief reasons for disproportionately large numbers of

deaths in case of sudden onset hazards like earthquakes or flash floods is that the poor

live in most vulnerable environments – in structures that are either non-engineered or

semi-engineered which might be located in low lying and vulnerable areas. The United

Nations estimates that 80 per cent of the world will live in developing countries by 2025,

more than half of which will be "highly vulnerable" to floods and storms (IFRCRCS

1999: Chapter 2). Though technology exists for constructing even non-engineered

structures to withstand moderate levels of hazards, this technology is not adopted by the

poorest. Uncertainty of threshold whereby a non-engineered structure becomes unsafe

and almost negligible probability of occurrence of a severe hazard is partly responsible

for this behavior.


19

Many other equally important reasons can be elicited by investigating the structures of

vulnerability generated by ongoing socioeconomic processes. Smith (1996) cites several

reasons why disaster impact is growing, even if frequency of geophysical events is

unchanged and despite the many positive steps being taken to reduce disasters. The

reasons Smith cites are population growth, land pressure, urbanization, inequality,

climate change, political change, economic growth, technological innovation, social

expectations, and global interdependence. Using data from historical catastrophes, these

factors are shown to be determinants of vulnerability in this chapter.

The organization of this chapter is as follows. The next section describes various

perspectives about catastrophe. The macro- and micro-level determinants of vulnerability

are described next. The connections between development processes and vulnerability to

catastrophes are presented using arguments based on empirical regularities observed from

data on disasters and indicators of socio-economic processes. How do catastrophes affect

development? This question is examined in the Section 2.3. Results from literature are

reviewed. Theoretical and empirical results, from the research presented in later chapters,

are discussed. The mechanisms used for coping with risk, such as insurance, credit, and

saving, are then discussed. Section 2.5 examines how external aid helps in recovery. This

review chapter concludes with a discussion of policy issues.

2.1 What is a Catastrophe?

In this section various perspectives of catastrophes are briefly reviewed. This

becomes imperative for determining the parameters and issues that would be of relevance

to the discussion, to understand the reasons of emphasis placed by disaster researchers on

seemingly different issues, to design well-balanced policies, and if possible to gain a

holistic picture of disaster research.

Gilbert (1998) lists three main paradigms for studying disasters. In the first paradigm

catastrophes is imputed to an external agent. The affected human population is a passive

“victim of the environment”. An extreme version of this paradigm sees catastrophes as


20

“acts of God”. The second paradigm views disaster as the result of underlying community

logic, of an inward social process. Catastrophes result from the interaction of physical

hazards with ongoing vulnerable socioeconomic processes. The third paradigm views

disaster as an entrance into state of uncertainty. In this paradigm catastrophe is tightly

tied into the impossibility of defining real or supposed dangers, especially after the

upsetting of the mental frameworks we use to know and understand reality.

Russell Dynes (1998) indicates - “a disaster is a normatively defined occasion in a

community in which extraordinary efforts are taken to protect and benefit some social

resource whose existence is perceived as threatened.” Robert Stallings (1998) points out -

“disasters are fundamentally disruptions of routines.” For Anthony Oliver-Smith (1998)

disaster is “a process/event involving the combination of a potentially destructive agent

from the natural, modified and/or constructed environment and a population in a social

and economically produced condition of vulnerability, resulting in a perceived disruption

of the customary relative satisfactions of individual and social needs for physical

survival, social order and meaning.”

Uriel Rosenthal’s (1998) re-conceptualization of the notion of sudden onset disasters is

interesting. A dam collapse is usually thought of as sudden onset catastrophe. But is it

really so? Its vulnerability is determined by the quality of construction, the politics, path,

and the kinds of channelization for drainage, among many other factors. There are both

structural and non-structural aspects mentioned here, each with different places in social

time, and virtually all taking place long before the dam failed. Catastrophes therefore

have complex and interrelated origins as well as consequences. What determines the

structure of vulnerability is as important as what determines the vulnerability of the

structure. We need to look no further than ongoing socioeconomic process to understand

the structures of vulnerability.

For the purposes of this study, a catastrophe is a low probability high consequence (in

terms of either lost lives or direct physical damage) economy wide event that acts as a

strain on the affected region’s resources and socioeconomic processes. As a result, low-


21

income countries may be forced to either borrow or dis-save huge amounts to recover.

Considerable time may be required to bring the community to pre-disaster conditions.

Potential losses from disasters are usually classified as: (i) direct or capital (ii) indirect or

income, and (iii) negative secondary or output effects. The financial value of damage to

and loss of capital assets – constructed facilities including buildings, infrastructure,

industrial plants, and inventories of goods including crops, account for direct losses.

Direct losses also include loss of lives and include measures of the total number of

affected people who are rendered homeless.

Direct losses are usually the most readily assessed after a catastrophe has struck. In

economic terms direct losses can be equated with stock losses. It is important to

distinguish between financial estimates of loss from economic loss in case of constructed

facilities. The financial loss would involve the replacement value of the lost asset,

independent of its condition or age. Thus the replacement of a collapsed bridge as a result

of an earthquake would involve the replacement cost of a new bridge, independently of

the age or condition of the collapsed bridge. On the other hand, if the destroyed bridge

were near the end of its useful economic life, the economic cost of its destruction might

be very small if replacement would soon have been necessary. Of course, the loss of a

new bridge would impose far greater economic losses, which would approximate the

financial losses incurred.

Indirect losses arise from interrupted production and services, measured by loss of output

and earnings. For example, damage to roads and ports can hold up exports, imports, and

distribution of basic necessities affecting health and education, as well as other

productive sectors. Depending on the magnitude of the direct loss, the impacts of

disasters may or may not affect the country’s GDP. In some cases effects can spread

beyond national borders. For example the 1985 Mexico earthquake destroyed the central

telephone exchange. Many Central American countries were affected as their

transmission lines ran through Mexico City. Such indirect losses can be equated with

flow losses.


22

Secondary effects of disasters are felt through longer-term impacts upon economic

performance including the development processes. Secondary effects are not easy to

estimate, which is reflected by the fact that not many research studies have concentrated

on this question. In the following sections we examine the complex ways in which

development processes and catastrophes are interrelated. In Chapter 3, various models are

presented that simulate the changes in productivity that arises after a catastrophe has

struck a region. The capital loss and the consequent changes in productivity due to

reconstruction result in permanent changes in overall welfare of the affected region.

Chapter 3 explains how this welfare loss can be used to quantify the secondary losses.

Development processes change the structures of vulnerability of a population.

Development decisions without adequately addressing the question of sustainability lead

to creation of inefficient facilities and services that contribute increasingly to disaster

impact (Kreimer and Munasinghe, 1991). One of the unintended consequences of the

development-induced change is that when a natural hazard strikes a highly vulnerable

region it may result in a catastrophe. The following questions are now examined:

• What are the socioeconomic determinants of vulnerability? How do development

processes increase vulnerabilities of some people to natural hazards? (Section 2.2)

• How does the occurrence of a catastrophe act adversely for the development process in

the affected region? (Section 2.3)

2.2 Development Processes, Vulnerability and Catastrophes

Vulnerability of a population to natural hazards can be summarized at both the

economy-wide (macro-) and household (micro-) levels. The main purpose of this section

is to examine the various factors that determine macro- and micro-level vulnerability and

their interdependencies. Fig. 2.1 gives an overall picture of these interactions. The

interactions are explained in the following.


23


24

Fig 2.1 Determinants of Macro- and Micro- Vulnerability

OverallVulnerability

Vagaries ofinternational

markets

Openess toworld economy

Low levelsof disasterawareness

and preparedness

Inadequatelabor supplyafter event

Low levels ofheath

Susceptible todisease after event

Health,Nutrition, and

Education

Poverty andpopulation

growth

Sloppyresidential

construction

HouseholdVulnerability

Poor construction quality

UrbanizationForces poor

to occupy vulnerableregions

Lack ofdisaster

recovery facilities

Poorinfrastructure

facilities

Large scaledevelopment

projects

Lack ofinsurance, credit

Inadequate savings

Fragilesocial

networks

Phase ofdevelopment


25

Change, in the widest sense of the term includes development processes and it is hard to

separate them. Development, viewed positively, brings with it: (i) increases in per capita

output; (ii) a shift of labor out of agricultural sector and into relative security of

manufacturing and services; (iii) the integration of regional markets, assisted by

improved transportation and communication networks; (iv) increased trade with the

outside world; and (v) improvements in governments services aimed at alleviating or

mitigating poverty.

One of the main indicators of economic development is per capita income and an

important observation from the compiled loss data is illustrated in Fig 2.2. It is clear from

Fig.2.2 that higher loss ratios (annual economic loss/GDP) are associated with countries

with low per capita income. In this figure and for all the subsequent figures in this

chapter, the economic indicators are an average of ten years before the occurrence of an

event. The power-law relationship between the loss ratio and the per capita income

illustrated in Fig. 2.2 has clearly many applications, which will not be addressed in this

dissertation. What is relevant for the present purposes is that increases in per capita

income are associated with lower loss-GDP ratios. Per capita income thus constitutes a

significant indicator for vulnerability of a nation to natural hazards. Fig 2.3 brings out a

similar the relationship between the percentage of people affected and the per capita

income. As a consequence of low per capita income many people in most third world

countries that are vulnerable either lack preparedness measures, or the level of protection

is inadequate, or their livelihood level lacks resilience to economy wide catastrophes. It is

often the case that they are unable to provide themselves with self-protection, and the

state is unable or unwilling to offer much relevant social protection against economy

wide catastrophes. In developed industrialized countries, preparedness levels may be high

and in general livelihoods are more secure and insurance makes them more resilient.

Per capita income is certainly one of the indicators that determine vulnerability, but it

should be noted that development, in reality, is a very uneven affair, with some people

benefiting, often at the expense of others. Distributional issues and equity are major


26

problems that many regions have yet to deal with satisfactorily. It is therefore not

surprising to note that development may increase vulnerability of some to natural

hazards. A highway construction project certainly brings about changes to the community

it serves. The highway presumably develops trade between the various regions it passes

through. But it also increases the probability that skilled labor from less developed

regions migrates to more developed regions. If the regions are rich in natural resources,

then the exposure of these natural assets to exploitation is increased. Exploitation is one

of the more common unintended consequences of the highway project. Another example

is the construction of high-rise buildings in earthquake zones without using earthquake-

resistant techniques or building on flood plains. More generally, development projects

result in changing the vulnerabilities of population to natural hazards. Besides per capita

income, other indicators of the development processes that are determinants of

vulnerability as discussed in the following sections.

Analysis of various political economies and the way they structure societies such that

similar hazards lead to very different impacts on one society compared to another is

required to unravel the exact nature of the phenomena. In United States, the vulnerability

of people to hurricanes is much less than in Bangladesh (or the countries of the

Caribbean) because of generally higher levels of income (which enable recovery more

easily), and the high degree of preparedness. The socioeconomic framework of self-

protection and social protection has reduced vulnerability to natural hazards of many. But

there exist sizable groups even in the wealthiest parts of the world that are still

vulnerable. Class, gender, race and ethnicity are likely to be very significant indicators of

the variable impact of hazards. For instance in the US not everybody enjoys social

protection (preparedness and mitigation measures) against hurricanes or earthquakes.

Hurricane Andrew affected the Black community to the significantly greater level than

others (Morrow 1997). Black and non-Cuban Hispanic households, across income levels,

were much more likely than Anglo and Cuban households to report insufficient

settlements and this was in part due to differential access to policies with larger

corporations.


27

Fig. 2.2 Greater per capita income is associated with smaller annual economic loss as a proportion of GDP

Loss/GDP = 2.3614*(GDP/capita)-0.7111

R2 = 0.4143

0.01%

0.10%

1.00%

10.00%

100.00%

1000.00%

100 1,000 10,000 100,000

Per capita income (current US$)

Ann

ual e

cono

mic

loss

as

a %

of

GD

P

Fig. 2.3 Greater per capita income is associated with smaller population affected

R2 = 0.20, N= 117

0.0000%

0.0001%

0.0010%

0.0100%

0.1000%

1.0000%

10.0000%

100.0000%

100 1,000 10,000 100,000

Per capita income (current US$)

Perc

enta

ge o

f pop

ulat

ion

affe

cted

(log

log

scal

e)


28

Generally speaking, vulnerability to a catastrophe is the result of people’s positions

within various political, social and economic fields, and the manner in which various

institutions in these fields respond to hazard in terms of awareness, emergency and crisis

management, and reconstruction. The reasons why catastrophes happen can be inferred

only by taking an objective and holistic view of the determinants of vulnerabilities and

damage levels in economies under stress. In the following sub-sections the determinants

of vulnerability will be enumerated first at a macro-level and then at a micro- or

household level. Links will be established between these macro- and micro parameters.

Data used in the subsequent regressions are presented in electronic form in Appendix B.

2.2.1 Macro-level determinants of Vulnerability

The determinants of vulnerability to natural hazards at a macro-level are: (i)

openness to the world economy, (ii) large-scale investment in development projects, (iii)

stability of the monetary system, (iv) urbanization, (v) population growth, (vi) poverty,

and the phase of development.

2.2.1.1 Openness to world economy

Increasing globalisation of the economy as a consequence of development implies

that regions, nations, and sub-national regions are directly connected to the rest of the

world. Whilst this globalisation brings with it unprecedented access to global trade and

resources, the risks from a natural hazard have also gained new transmission channels.

The 1995 Hanshin-Awaji earthquake forced the closure of the major Kobe port. As a

result many firms in US that rely on imports for manufacturing suffered due to delay in

shipping. For the affected region, though, globalisation provides a means to recover from

the disaster.


29

The pattern of financial relationship between the industrialized North and the Third

World has altered with decolonization. The Third World has traditionally depended on

agricultural and mineral exports the prices of which are falling. Simultaneously, prices of

imported energy and technology have increased. Most of the Third World countries have

little opportunity to process and market what they produce and are dependent on imports

from industrialized nations which are often highly priced or tied to aid packages. This has

created circumstances in which many Third World nations are faced with great difficulty

in maintaining their balance of payments. Therefore, a viewpoint often expressed is that

the functioning of world economy is against the LDCs thereby reinforcing hazard

vulnerability. Current account balance is an indicator of the level to which a country is

dependent on imports. Fig 2.4 illustrates a clear negative relationship between

dependence on imports and the level of economic losses from catastrophes.

Countries faced with severe debts usually resort to national policies favoring export

production. As a result land degradation may result from destruction of forest, soil,

wetlands, and water sources. In order to service debt, new lands are cleared for ranching

or commercial cropping. Coastal areas are drained, mangrove forests cut, in order to

accommodate the expansion of tourist hotels and other foreign installations that hold out

the hope of hard currency earnings. Population growth and urbanization increase demand

for energy and in many countries dams (often large-scale) are built to produce electricity.

These dams flood vast areas of forest and other lands, forcibly displacing the inhabitants

to more vulnerable areas. The result of this debt-induced activity is an increase in the

vulnerability of the exposure. A severe hurricane on a coastal tourist resort in the

Caribbean results in huge property losses.

2.2.1.2 Development induced investment in large-scale projects

Economic growth in the developed countries has increased the exposure to

catastrophic property damage. Due to shortage of prime land in urban areas, extremely

vulnerable sites are chosen for the development of real estate such as coastal Florida and

resorts in Hawaii. Intensive capital development has increased the probability that a


30

hazard like a hurricane will encounter an increasing amount of constructed facilities

unless steps are taken to reduce risks within cities and on industrial sites.


31

Fig. 2.4 Larger negative external balance on goods and services is associated with larger losses

R2 = 0.2201

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

(35.0) (30.0) (25.0) (20.0) (15.0) (10.0) (5.0) - 5.0 10.0

Econ

omic

loss

as

a %

of G

DP

Pre-event external balance on goods and services % of GDP

Fig. 2.5 More government repudiation of contracts are associated with larger losses R2 = 0.3848

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 2.0 4.0 6.0 8.0 10.0 12.0

Econ

omic

loss

as

a %

of G

DP

Government repudiation of contracts


32

For the LDCs, development planners often introduce technology at the so-called “leading

edge” of whatever version of rapid, systemic change they define as “development”. This

may be irrigation technology in the form a large dam that displaces thousands of families

in what economists call “the short run”. It might take the form of low-income housing or

the development of an industrial complex. Such development initiatives, though well

intentioned and useful, can have a series of unintended, unforeseen consequences, the

most detrimental of which is an increase in vulnerability of the poorest.

In many developing countries, as the economy grows, production and exchange become

increasingly complex and the institutional structure of the economy changes accordingly.

Traditional institutions, like extended households that are important mechanisms for

surviving shocks, may no longer be optimal. Since the risk-sharing functions may be

performed by other institutional arrangements that vary in their abilities both to fit local

circumstances and to perform their respective tasks, transition to a developed society

leaves many groups marginalized thus increasing their vulnerability. In developing

countries for instance, insurance is not a well-developed risk-sharing institution.

Financial inter-mediation is usually poorly developed in the LDCs. Financial inter-

mediation is very important for economic development because individuals live in risky

environments, which makes savings, insurance and consumption credit yield direct

benefits in coping with risk. Another reason is that the development of credit and

insurance should enhance an economy’s investment efficiency and, possibly growth.

Failures of inter-mediation are intimately linked with misallocation of capital and

inefficiencies. Individuals who have the most productive investment opportunities may

be denied access to funds.

The interaction of production risks with information asymmetries may increase

opportunistic behavior and hence reduce production. In developing countries, institutions

such as courts, various kinds of bonding mechanisms, social norms, and the structure of

incentives in future contracts are not strictly implemented and thus they are not able to


33

limit the practice of opportunistic behavior by different agents. This leads to further

inefficiencies in the economy. In fact, data from past record of catastrophe suggest that

higher rate of government repudiation of contracts introduces more risks in the economic

environment and hence is associated with higher losses (Fig 2.5). Constant threat of

occurrence of a natural hazard may increase the uncertainties and limit investment

decisions, thus dampening economic growth.

2.2.1.3 Development and Population growth

Economic development and population growth affect the relationship of people

with their environment in several complex ways, most of them negative. In the opinion of

Dando (1980, pp. 105-9), the pressure to produce ever more food is creating “an agro-

environment conducive for eco-catastrophes.” The debate about environmental

degradation is frequently linked to the Malthusian notion of a regional “carrying

capacity”, this being defined as the number of people and animals a specified area can

maintain over a period of time. When populations exceed this limit, a cycle of over-

exploitation of the land is set in motion, which ultimately degrades the natural resource

base to such an extent that human and animal survival is unsustainable.

Continued population growth outstrips the ability of governments to invest in education

and other aspects of social development including disaster preparedness measures.

Higher population growth has detrimental consequences on disaster susceptibility. This is

clearly indicated in Fig 2.6. Higher population growth is positively associated with higher

loss-GDP ratios. Population growth also creates further competition for land resources

and in urban areas this increases the vulnerability to natural hazards. In the very poorest

countries, the human use of natural resources has created a problem of food security and

fragile livelihoods. Only quarter of the people in Africa have access to safe drinking

water. As a result even a hazard of mild magnitude can have catastrophic consequences.


34

Fig. 2.6 Higher population growth is associated with larger losses

R2 = 0.3359

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Econ

omic

loss

as

a %

of G

DP

Average population grow th

Fig. 2.7 Larger urban population growth is associated with larger losses

R2 = 0.3471

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

0 1 2 3 4 5 6 7 8

Econ

omic

loss

as

a %

of G

DP

Urban population grow th (annual %)


35

2.2.1.4 Development and Urbanization

As a result of various processes associated with development, people from the

countryside move into cities seeking better job opportunities. This urbanization process

results in land pressure as migrants from outside move into already overcrowded cities.

The new arrivals are forced to occupy disaster susceptible regions. Catastrophes being

rare events by definition, the migrants rarely make their decision to migrate to vulnerable

regions based on natural hazard probabilities. Slum residents often incur greater risks

from natural hazards (especially landslide and fires) as a result of having to live in very

closely spaced unsafe shanties that are located in low-lying areas. Current projections

indicate that within the coming decade there will be twenty cities with populations

between 10 to 25 million. Of these, fourteen are in the Third World, eleven in hazardous

zones (Blaikie et al. 1994).

Moreover, these cities are expanding rapidly, with the obvious risk of sloppy construction

standards (Tyler 1990). Maskrey (1994) cites the example of Peru, which has become

more hazard-prone with the post-colonial shift of population from mountain communities

to high-risk urban centers. This is especially the case in the capital, Lima, which is

located in a seismic zone where houses of Spanish design with heavy roofs are crowded

with low-income families. Record from past catastrophic events indicates that higher

urban population growth is associated with larger losses as a proportion of GDP (Fig.

2.7).

Lipton (1977:18-19) argues that rural poverty and vulnerability to famine are often a

function of government policies which are biased in favor of the interests of urban elite,

and which therefore discriminate against the interests of the agricultural sector in general,

and of the rural poor in particular.


36

2.2.1.5 Development and poverty

People become disaster victims because they are vulnerable. Because people have

different degrees of vulnerabilities, they suffer differently. Inefficient development

policies are associated with increasing inequalities between communities and households,

resulting in rising vulnerabilities for some groups of people. The more the inequalities

that exist in a community more pronounced are the effects of a hazard. Growing poverty

creates greater vulnerability to natural hazards for the population for several reasons.

Farmers may be dispossessed of land and compelled to grow cash crops rather than

subsistence food. Urbanites may be forced to live in most dangerous built up areas. A

catastrophe simply reinforces the growing gap between rich and poor. Even in “normal”

times the poorest sections of society are pressured to over-use the land, and when disaster

strikes, the conventional responses may merely accelerate the continued

underdevelopment and marginalisation. Vulnerability is also the result of third world

impoverishment perpetuated by technological dependency and unequal trading

arrangements between rich and poor nations (Susman et al., 1983).

It is important to note that vulnerability and poverty are not synonymous, although they

are often closely related. Vulnerability is a combination of characteristics of a person or

group, expressed in relation to hazard exposure, which derives from the social and

economic condition of the individual, family, or community concerned. High levels of

vulnerability imply a catastrophic outcome in hazard events. Vulnerability is a complex

combination of both the qualities of the hazards involved and the characteristics of the

people. Poverty on the other hand describes people’s lack or need. Vulnerability is a

relative and specific term, always implying a vulnerability to a particular hazard. A

person may be vulnerable to loss of property or life from floods but not to drought.

Poverty may or may not be a relative term, but there are no different types of poverty for

any one individual or family depending on the causes.


37

As many as 850 million people live in areas suffering severe environment degradation

(Smith, 1996). In many LDCs more than 80 per cent of the population is dependent on

agriculture but many are denied an equal access to land resources. Poverty forces many

people to adopt unsustainable land use practices. Countries with a legacy of deforestation,

soil erosion and over-cultivation find their environment more vulnerable to natural

hazard, especially floods and droughts. In fact the record of historical catastrophes

suggests a statistically significant relation between deforestation and the loss levels as

shown in Fig. 2.8

Fig. 2.8 Greater net forest depletion is associated with larger lossesR2 = 0.1595

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

Econ

omic

loss

as

a %

of G

DP

Pre-event Genuine savings: net forest depletion (% of GDP)


38

2.2.1.6 Vulnerability as a “phase of development”

Poor rural households may experience increased vulnerability during the

transition from a peasant or semi-subsistence society to a market economy. Traditional

mechanisms of coping against natural hazards are disrupted due to the capitalist

penetration of subsistence economies. Most development schemes do not substitute

traditional coping mechanisms since the incentives for preparing against natural hazards

may be missing. Unless the traditional coping mechanisms are substituted with alternate

mechanisms, large segments of the population will be made more vulnerable to natural

hazards. Sen (1981, p.173) has written that: “The phase of economic development after

the emergence of a large class of wage laborers but before the development of social

security arrangements is potentially a deeply vulnerable one.” As Clay (1986, p.180) puts

it: “freeing the hidden hand where the basic needs of the majority of people is not assured

is potentially a recipe for disaster.” When entire communities are made vulnerable during

“the phase of the pure exchange system transition” and are further destabilized by

occurrence of a natural hazard coupled with adverse processes of development, such as

changes in modes of production, the result can be catastrophic.

In many rural areas in developing countries transition from a village economy to a market

economy brings with it hastily adopted building techniques. People invest in residential

buildings that are unsafe because they are not built according to standard safety

guidelines. For example, as far as seismic safety is concerned, a poor household living in

light roof building may be much safer than a middle-class household who live in a semi-

engineered building with heavy roofs. Most of the deaths in the 1993 Latur (Maharashtra,

India) and the 2001 Bhuj (Gujarat, India) earthquake resulted from building collapse and

damage.


39

2.2.1.7 Development and Government

The basic functions of the state are to provide law and order and to protect the

property rights. Such services are generally exchanged in return for tax payments. As the

single group in society to which all belong and from which exit may be least possible, the

state may solve a variety of coordination problems and overcome externalities plaguing

other institutions. Usually it is the government that has to provide assistance to disaster

areas. In the LDCs, the government may be plagued with inefficiencies and corruption.

Development aid is mostly used for betterment of those in power, making the poor more

vulnerable. Many catastrophes occur because governments fail or have limited capacity

to provide basic transportation, communication, health and other infrastructure needs of

the poorest. Poor infrastructure with almost no maintenance, lack of welfare programs,

which results in inadequate housing and health provision combined with low nutritional

status results in increasing the vulnerability of poor in the community. The evidence from

past record of catastrophes clearly establishes the link between infrastructure and the

levels of losses experienced. Fig. 2.9 illustrates the fact that better infrastructure as

indicated by the availability of electricity is strongly associated with lower loss-GDP

ratios.

Fig. 2.9 Larger electric power consumption per capita associated with

smaller lossesR2 = 0.4922

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Econ

omic

loss

as

a %

of G

DP

Electric pow er consumption (kw h per capita)


40

Market and non-market (including governmental) institutions, which work quite

adequately, though not perfectly, in normal times can easily turn even a moderate

aggregate shock into a catastrophe. According to Sen (1981) markets and institutions

determine the factors, which bear on the likelihood that an exogenous shock will turn into

mass entitlement failure and hence a catastrophe. The factors include quality and

distribution of endowments, the structure of prices, and the pattern of transfers. The

factors that can transform a shock into a catastrophe appear to be intrinsic features of

quite normal economies – rather than peculiar features of highly distorted or badly

damaged economies. They are always present, but normally hidden from view. And they

do surface in any number of ways after a catastrophe has occurred. The devastating

earthquake that struck the most densely populated and industrialized area of Turkey on

August 17, 1999 can be seen as a recent example. Corruption was rampant in the building

industry leading to substandard construction. The earthquake brought forth these inherent

societal deficiencies (corruption) to full public view by causing severe damage to

buildings built using sub-standard techniques. Bureaucracies (Fig 2.10) and increasing

governmental intervention with mostly untrained civil servants result in inefficient

organizational structures, creating vulnerable societies

Fig. 2.10 Inefficient bureaucracies are associated with larger lossesR2 = 0.2931

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 1 2 3 4 5 6 7

Econ

omic

loss

as

a %

of G

DP

Bureaucratic quality (1-poor, 6-best)


41

Weak social infrastructure as indicated by poor law enforcement (Fig. 2.11), weak

unprepared government, corrupt bureaucracies (Fig. 2.12), and a relatively closed

political regime, all enhance vulnerability to hazards.

Fig. 2.11 Better rule of law is associated with smaller lossesR2 = 0.271

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 1 2 3 4 5 6 7

Econ

omic

loss

as

a %

of G

DP

Rule of law (1- Poor, 6-Good)

Fig. 2.12 Higher prevalence of corruption is associated with larger losses R2 = 0.1743

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 1 2 3 4 5 6 7

Econ

omic

loss

as

a %

of G

DP

Corruption (1-High, 6-Low )


42

2.2.2 Micro-level determinants of Vulnerability

But what are the root causes of vulnerability to natural hazards? How is the

severity of the catastrophe determined by the initial values of the micro (household) level

parameters affecting vulnerability? These are some of the questions that we will address

in this section.

According to Cannon (1994), vulnerability may be divided into three aspects: the first is

the degree of resilience of particular livelihood system of an individual or group, and

their capacity for resisting the impact of a hazard. This reflects economic resilience,

including the capacity for recoverability (another measure of economic strength and

responsiveness to hazards). This can be called “livelihood resilience”, and has some

affinity with Sen’s concept of entitlement (Sen, 1981). Sen introduces the concept of

entitlement failure as a primary reason for the occurrence of famines. Markets and

institutions determine the factors, which bear on the entitlement failure and hence famine.

The factors are – distribution of endowments, income opportunities, the structure of

prices, and the pattern of transfers. Even a small entitlement shock to poor people can

induce large changes in their survival prospects.

The second component is the degree of self-protection that includes “health”. The health

of individuals, the operation of various social measures for hazard protection including

preventive medicine and quality of constructed facilities including residential structures.

The third component is the degree of preparedness of an individual or group. This is

determined by the protection available (for a given hazard), something that depends on

people acting on their own behalf and on social factors. The notion of precautionary

savings is of relevance here as will be explained in a later section. Communities

repeatedly affected by hazards use various schemes like precautionary savings, social

networks, loans, and credits to help themselves insulate from these adverse income


43

fluctuations. The presence of such schemes and their efficiency in smoothing

consumption greatly reduces the vulnerability to natural hazards.

An additional component is the type of hazard. An earthquake, for example, may result in

a catastrophe if it strikes a community that is otherwise well prepared for a hurricane.

One of the reasons for this might be that buildings designed solely to withstand

hurricanes might be vulnerable to earthquakes.

2.2.2.1 Presence of uncertainties and development processes

Risk and uncertainty problems are very important in LDCs since the sources of

risk and their magnitudes are sufficiently varied and large and the relevant probability

distributions of alternative outcomes are unknown. Higher uncertainty in the growth of

income (as measured by the standard deviation of economic growth) leads to more

vulnerability of the affected society. Fig 2.13 associates higher uncertainty in growth

rates to the larger loss-GDP ratios. The return period, intensity, and magnitude of natural

hazards can at best be estimated in probabilistic terms. Moreover, given a hazard of

particular intensity, the potential direct loss in the affected region can be estimated only

probabilistically. Development processes continually change the vulnerabilities of the

regions and the exact way in which vulnerability of a region evolves is complex. This in

turn increases the inherent risk.


44

Fig. 2.13 Greater uncertainty in economic growth is associated with larger losses R2 = 0.1324

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

1.0 10.0

Econ

omic

loss

as

a %

of G

DP

Standard deviation of annual economic grow th

Fig. 2.14 Larger volatility in inflation is associated with larger losses

R2 = 0.115

0.01%

0.10%

1.00%

10.00%

100.00%

1000.00%

(0.5) - 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Econ

omic

loss

as

a %

of G

DP

Log of Pre-event standard deviation of inf lation


45

Development processes in the LDCs may be retarded because of increased risk aversion

at the household level. The degree of risk aversion may be larger in LDCs than

elsewhere, in part because incomes are nearer to the minimal subsistence level and in part

because these risks are more related to other problems. For example, an observed

production shortfall can be explained with several reasons. It may be more difficult with

the available information to distinguish among the following alternative explanations: (1)

the direct effects of bad weather on output, (2) the indirect effects of bad weather on

output via the effects on health and the effective labor supply, (3) producer mistakes in

resource allocation, and/or (4) shirking of the workers. Among a cross-section of

countries it is useful to compare some indicator of the overall risk with the loss level. One

such indicator is the volatility in inflation. Higher volatility in inflation implies higher

volatility in future expectations. This in-turn implies more risky environments for

investors as well as households. Fig 2.14 clearly brings out the association between

higher inflation volatility and higher losses. This in-turn implies that inflation volatility is

a key determinant of vulnerability. But it is also sometimes argued that the poor are less

likely to be risk averse, since they have little to lose even if they fail. Development is

retarded because of the limited number of choices the poor are faced with. This denial of

accessibility to the poor is partly responsible for making them vulnerable to hazards.

In the following section we establish broad relationships between development processes

on the household level and its connection to vulnerability. The determinants of

vulnerability at a household level are (i) health, food, and nutrition, (ii) education and

consequent disaster awareness, (iii) endowments, (iv) infrastructure including sanitation

and availability of drinking water, (v) preparedness measures, and (vi) quality of

residential structures.

2.2.3 Development, Households and Vulnerability

It is well recognized that in many developing economies, the household and

family are key economic decision-makers and intermediaries, whereas, as development


46

progresses, the market or the state takes over some of these roles. The capacity to self-

organize in times of crisis is extremely important for absorbing the impact of a

catastrophe. In the very early stages of economic development, where income levels

hover around the subsistence level, the risk-reduction element of the basic economizing

function may be the dominant one. It is this element, therefore, which may provide the

basic rationale for some of the most important institutions, such as the family, the tribe,

or the kin group. In such contexts, institutions, such as the need to offer hospitality to

everyone in the village and to allow every individual to obtain much knowledge about

each other, may be very efficient. This is especially true where production techniques are

simple and most exchanges are personal and repeating. Even poor societies, which are

well organized and cohesive, can withstand, or recover from, a catastrophe better than

those where there is little or no organization and people are divided. Similarly, groups

who share strong ideologies or belief systems, or who have strong experiences of

successfully cooperating to achieve common social goals, even when struck by a

catastrophe, may be better able to help each other and limit some kinds of suffering than

groups without such shared belief. Opportunistic behavior is rare because the

aforementioned institutions make for its detection.

Some of the most important choices households make revolve around the human capital

of children and adults. The fact that human capital investments are associated with higher

standards of living and welfare has been repeatedly demonstrated both in aggregate data

and in studies that have used individual or household level micro data. The vulnerability

of households in LDCs to natural hazards is due to low level of human capital

investments in health, education and disaster awareness programs. Low levels of disaster

awareness may lead to scant attention to building standards resulting in sloppy

construction vulnerable even to moderate intensity hazards. The cyclone that struck an

impoverished eastern Indian state of Orissa brought to full view the abysmal levels of

human capital investments that existed in the region prior to the event.


47

2.2.3.1 Health, nutrition, and education

What are the factors that determine a household’s ‘health’ and hence its

vulnerability to natural hazards? Studies have focused on the effects of wages, food

prices, health programs, and family planning, on child health, schooling and fertility

outcomes. In areas in which the government carried out agricultural intensification

activities, Rosenzweig (1982, 1990) finds that market returns to primary schooling and

school enrollments are found to be higher and fertility lower for farm households in those

areas. Higher enrollments (Fig 2.15) are important for easier communication of disaster

awareness programs and lower fertility reduces pressures on vulnerability created by

unchecked population growth. Human capital investments also depend on infrastructure,

such as water and sanitation quality; measures related to price and quality of health,

education and family planning facilities; and prices of other health or education inputs

such as foods.

Based on a few studies conducted on the impact of health infrastructure on health

outcomes that are relevant to our concerns, we can conclude the positive effects of

development processes that try to increase the access to health facilities of the poorest.

This improved access to health facilities results in contributing towards reducing the

vulnerability to natural hazards, as Fig. 2.16 illustrates. Rosenzweig and Wolpin (1982)

and Hossian (1989) show a negative relationship between clinics per capita and child

mortality in India and Bangladesh respectively, also using density measures. Thomas,

Lavy and Strauss (1992) show a positive relationship between doctors and child height in

Cote d’Ivoire, and Deolalikar (1992) finds a positive relation between health expenditures

per capita and child weight among low-income households in Indonesia. The data on past

catastrophes reveals that higher child mortality is associated higher loss-GDP ratios (Fig.

2.17).

Health of the members of a household determines its vulnerability to epidemics such as

typhoid or malaria after a flood as well its ability to recover after an event. These factors

are also dependent on availability of adequate sanitation infrastructure at the community


48

level. Most LDCs lack the basic facilities of sanitation and health rendering the

households vulnerable to adverse health consequences after a catastrophe. Data on past

catastrophes reveals that there is positive association between access to health

infrastructure and the loss-GDP ratios (Fig. 2.18).

Fig. 2.15 Higher secondary school enrollment is associated with smaller lossesR2 = 0.4001

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 0.5 1.0 1.5 2.0 2.5

Econ

omic

loss

as

a %

of G

DP

Pre-event secondary school enrollment

Fig. 2.16 Availability of physicians is associated with a decrease in the losses R2 = 0.3035

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

(2.5) (2.0) (1.5) (1.0) (0.5) - 0.5 1.0

Econ

omic

loss

as

a %

of G

DP

Pre-event physicians (per 1,000 people, log)


49

Fig. 2.17 Higher infant mortality rate is associated with greater losses

R2 = 0.4048

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 0.5 1.0 1.5 2.0 2.5

Econ

omic

loss

as

a %

of G

DP

Pre-event mortality rate, infant (per 1,000 live births)

Fig. 2.18 More number of hospital beds is associated with smaller losses

R2 = 0.3896

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

1 10 100 1,000 10,000

Econ

omic

loss

as

a %

of G

DP

# of hospital beds per thousand


50

2.2.3.2 Labor and food intakes

The determination of returns of labor plays a central role in models of development since

labor is by far the most abundant resource in low-income countries. Most of the income

for households in LDCs results from labor-intensive employment. Current nutrient

intakes play a crucial role in enhancing productivity for most of these jobs; for instance

calorie intake increases maximum oxygen uptake (which is related to maximum work

capacity; Spurr, 1983). On the other hand many jobs, in the service-oriented industries of

the developed countries do not require maximum physical effort. It seems likely that the

impact of health on income depends on the nature of work. A laborer, for example, may

suffer a larger decline in income because of physical injury than would a more sedentary

worker. Therefore a catastrophe that seriously undermines the physical ability of a labor-

intensive household may have a drastic effect on its earning capacity. In other words the

vulnerability of households is critically dependent on its food intakes. Studies indicate

among the poor households there is a positive correlation between expenditure and

calorie intakes. As income (or expenditure) rises, households switch to higher valued

foods, not necessarily with higher nutrient content (Behrman and Deolalikar (1987),

Strauss and Thomas (1990), Subramanian and Deaton (1992)). From these studies it is

clear that at low levels of expenditure, the calorie intakes and expenditure are positively

correlated but when per capita calories reach about 2000 per day, the curves flatten out.

When a catastrophe strikes a poor household living on subsistence diet, loss of access to

food results in effects much more severe than for relatively richer households who

consume more than 2000 calories per day. In fact the observations connecting losses to

calorie or protein intakes seems to support this hypothesis. Fig 2.19 and Fig 2.20 relate

the calorie or protein intake respectively to the loss-GDP ratios. More food intakes are

positively associated with smaller loss-GDP ratios.

An earthquake that damages key infrastructure facilities of an urban conglomerate

may leave many sedentary workers unemployed. It is a common observation that there is

a spurt in construction activity after an earthquake. Though it is not obvious that energy

or other nutrient intakes should be correlated with either productivity or labor supply,

there is some evidence that the body adapts to changes over some range in energy intakes


51

in such a way as to keep functioning intact. Therefore only at extremely low levels of

calorie intakes, a common feature in chronically poor nations productivity or labor supply

suffers. In times of a catastrophe, it is the poor households that cannot supply the required

labor because of their low calorie intakes. This may result in a slow recovery process in

LDCs.

Fig. 2.19 Larger daily calorie intake is associated with smaller losses

R2 = 0.296

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

1,000 10,000

Econ

omic

loss

as

a %

of G

DP

Daily calorie intake

Fig. 2.20 Greater daily protein intake is associated with smaller lossesR2 = 0.304

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

- 20 40 60 80 100 120

Econ

omic

loss

as

a %

of G

DP

daily protein intake (grams)


52

The connections between the macro- and micro-level determinants of vulnerability of a

group to natural hazards are shown in Fig 2.1. To explain one of these connections,

consider how the fact that an economically backward community will have majority of

households that have low levels of investments in health, nutrition, and education. At

household level this in turn implies low levels of disaster awareness and preparedness,

vulnerable health, and inadequate labor supply when it is most needed, i.e. immediately

after the event. All these factors contribute towards increasing the vulnerability of the

household. Other factors contributing towards macro- and micro- vulnerability are shown

in Fig. 2.1, which is a summary of the discussion above. These complex relationships that

determine vulnerability explain the fact that a hazard of similar intensity can cause

different levels of damage to two different communities. Determinants of vulnerability

are important in choosing the environmental and control variables used for examining

empirical evidence regarding post-event economic behavior, as will be discussed in

Chapter 4.

2.3 How do catastrophes affect development?

Having examined the complex ways through which development changes the

vulnerability of a community to natural hazards, this section focuses on the effect of the

occurrence of a catastrophe on socioeconomic processes. Catastrophes disrupt

socioeconomic processes of the affected communities and consequently it behooves us to

relate the adverse effects to development process. As has been previously mentioned, the

main purpose of this dissertation is to investigate the economic consequences of

catastrophes. By reviewing the literature that examines the effects of economy-wide

shocks on various socioeconomic processes, the work reported in the following chapters

can be placed in the right context.

In a historical context, Jones (1987) conjectures that the contrasting paths of development

between East and West were caused by a different incidence of disasters. According to

Jones (1987), with respect to changes over time the aspects of natural environment that

seem most to influence economic history are the very sharpest category of changes


53

including catastrophes, immediate adjustments to which were hard to make. On the other

hand, incremental changes of the kind documented in climatic history seem to possess

little independent explanatory power. Economies adjusted to them. Climatic and other

incentives were greater in the Orient than in the Occident. This, in turn, gave more

incentive for the peasants in the East to invest in larger families and less incentive in

physical capital. Consequently, development in the Occident led to rapid industrialization

while the Orient relied on agriculture for growth and industrialization process lagged

behind. Though such an extreme view of the effects of different incidence of disasters is

debatable, it nevertheless shows us the importance of natural hazards in explaining at

least some aspects of the variations in growth rates observed globally.

2.3.1 Macro – Level Effects

Occurrence of a catastrophe triggered by a natural event has potential

consequences for ongoing socioeconomic processes of the affected society at a macro-

level including: (i) development processes including growth, (ii) balance of payments

deficits, (iii) budget deficits, (iv) poverty and inequality, (v) trade and investment and (vi)

sudden movement of population.

2.3.1.1 Effects on Development

One important consequence of a catastrophe especially for developing countries is

the disruption to well-laid development plans when investment resources committed to

long-term programs are reassigned to emergency disaster operations. Long-term

development goals might be undershot or foregone altogether as the original development

programs lose their resources. To quote a recent example, Hurricane Mitch was the most

devastating hurricane of the 20th century. The statistics are staggering: 9,346 known

dead, 9694 missing, 2 million affected, and over $4 billion dollars (US) in damage.

Hurricane Mitch, it is claimed, has decimated the economies of Honduras, Nicaragua,


54

Costa Rica, El Salvador, Guatemala, and other Central American countries, setting them

back several decades.

Mary Anderson and Peter Woodrow (1989: 107) cite four examples of development

projects that were interrupted by a catastrophe. In two of these examples (Burkina Faso

and Kordofan, Sudan), basic developmental and environmental work was interrupted by

drought, diverting the NGOs effort into emergency feeding programs. In Joyabaj,

Guatemala, several volunteer agencies involved in developmental activities had to join

together to respond to the earthquake of 1976. In Santo Domingo in the Bicol Region of

Philippines, a village development project of the International Institute of Rural

Reconstruction was pre-empted by a sudden volcano eruption in the area of the intended

work.

Results based on simulation of theoretical models in Chapter 3 indicate that there are

overall welfare losses after a catastrophic event. The post-event consumption is lower

than the pre-event levels as the affected region invests in rebuilding. Chapter 4 presents

empirical evidence showing that post–event growth rates are negatively correlated with a

measure of direct losses. Growth being an indicator of development, the studies in

Chapters 3 and 4 corroborate with the anecdotal evidence presented in the above

paragraph.

2.3.1.2 Effect on Trade and Investment

Increasing globalization in the world economy has resulted in developing various

channels through which a shock is transmitted worldwide. Peek and Rosengren (1997)

investigate the extent to which the sharp decline in Japanese stock prices was transmitted

to the United States via U.S. branches of Japanese parent banks and identify a supply

shock to US bank lending that is independent of U.S. loan demand. They conclude that

binding risk-based capital requirements associated with the Japanese stock market decline

resulted in a decrease in lending by Japanese banks in the US that was both economically

and statistically significant.


55

Japan’s capital Tokyo is a central player in worldwide economic activity. The Tokyo

region accounts for roughly 30% of the nominal GNP of Japan and the stock exchange

ranks third in the value of the world trading volume handled each day. Most large

Japanese companies base their headquarters in greater Tokyo and several key industries

are heavily concentrated in the area, including banking, insurance, transportation, oil

refining, printing and publishing, and telecommunications

Eight of the world’s 10 largest banks have their headquarters in Tokyo. The ramification

of an earthquake striking at the political and economic center of Japan would be

tremendous. An earthquake in the Sagami Trough of similar magnitude that occurred in

the Tokyo metropolitan area in 1923 would be truly catastrophic. For the Tokyo

metropolitan area, including the Tokyo, Chiba, Kanagawa, Saitama, and Shinuoka

prefectures, Risk Management Solutions, Inc. (1995) estimated a total economic losses

ranging from $2.0 to $2.7 trillion with property loss due to shaking and fire alone ranging

from $1.0 to 1.2 trillion. The earthquake would cause 40,000 to 60,000 deaths. Hadfield

(1992) reports a study made by Tokai bank, which projected that such an earthquake in

the Tokyo region could cause a crash in the US stock and bond markets, a decrease in the

flow of Japanese funds to foreign countries, and a resulting international financial crisis.

In a region known to have a high strike rate for natural hazards there is an undermining of

business confidence and discouragement of investment. Investors require unattainable

high rates of return from their projects in order to compensate the risks of operating in

disaster-prone areas. Also, regression analysis (Chapter 4) point to the fact that

catastrophes cause increases in inflation and the real interest rates. This further

discourages investment. As a result the region’s economic development may be

dampened.

If as a consequence of a catastrophe, an export crop is destroyed by storms and flooding,

the balance of payments deficits may become unmanageable. The effect can be very

important in small economies where cash crops are the main source of foreign exchange.


56

It is a factor taken into account by IMF when awarding emergency funding to countries

suffering from the effects of disasters.

If governments are forced to overspend on disaster recovery and reconstruction there is a

growing public sector deficit and debt. Empirical data presented in Chapter 4 indicate an

increase in budget deficit after an event. Development projects are assigned lower

priorities after a catastrophe. Also, since the poor are unable to bid themselves out of

disaster-prone situations as higher income households do, it results in increase of level of

poverty. Cross-country regression analysis presented in Chapter 4 indicates that there is

an increase in the rate of debt growth after an event.

Catastrophes may result in unexpected movements of population especially from

devastated rural areas to unaffected towns in search for a means of livelihood or

employment. This makes the development of already crowded towns more difficult. For

example, after the recent volcanic eruption in Boma, about 400,000 people migrated to

Rwanda and safer regions in Congo. Regional model simulation of the effects of

migration indicates that recovery process slows down (Fig. 5.33).

In economic terms, secondary losses such as these can be counted among the negative

externalities of disasters. There can also be positive effects of disasters, which provide

unexpected opportunities to upgrade plant and machinery or renew aging infrastructure.

In most cases, however, benefits are unlikely to outweigh the costs of the losses. In

Chapter 3, welfare changes after a catastrophe is quantitatively examined. Results

indicate that a greater loss in capital results in greater overall welfare losses.

Analyses of the effects of economy-wide shocks on human capital outcomes suggest a

mixed picture. It is to be expected that the effects of aggregate shocks will vary by

country. This may result from differences in levels in socioeconomic development,

market structure and sectoral mixes and also levels of publicly and privately provided

safety nets. The impacts are also likely to vary depending on the particular outcomes.

Palloni and Hill (1992) find in Latin America that macro shocks do affect infant and child


57

mortality from respiratory tuberculosis and diarrhea. Hill and Palloni (1992), Palloni, Hill

and Aguirre (1993) suggest a Malthusian response of age at marriage or marital fertility

to changes in aggregate income.

These studies of economy-wide shocks have not examined specific differences in country

characteristics that may explain differential responses. In Chapter 4 we present evidence

and study the consequences of catastrophes for macro economic factors. Literature has

also not differentiated between aggregate-level shocks and more local shocks, let alone

the possibility that household’s ability to adjust to shocks may be associated with their

characteristics. For example, are the poor more vulnerable to the impact of adverse

economic shocks? It is necessary to turn to micro-level evidence to answer these

questions.

2.3.2 Effects at a Household Level

In this section we briefly review the research that focuses on both household

responses ex-post to adverse shocks and ex-ante to perceived-risks. Much of this recent

literature has addressed the question of whether households are able (both by themselves

and using community-level mechanisms) to smooth their consumption perfectly against

all risks. Other studies include household saving behavior in presence of shocks and

modeling effects on human capital related outcomes.

2.3.2.1 Savings and Investment

Studies such as Deaton (1992a,b) and Paxon (1992), attempt to test the permanent

income hypothesis. According to the permanent income hypothesis, people base

consumption on what they consider their "normal" income. In doing this, they attempt to

maintain a fairly constant standard of living even though their incomes may vary

considerably from month to month or from year to year. As a result, increases and

decreases in income which people see as temporary have little effect on their


58

consumption spending. The idea behind the permanent-income hypothesis is that

consumption depends on what people expect to earn over a considerable period of time.

People smooth out fluctuations in income so that they save during periods of unusually

high income and dis-save during periods of unusually low income. Deaton (1992a,b) and

Paxon (1992) find little support for the strong form although Paxon finds a weaker

version does rather well for Thai farm households. Using estimated impacts of regional

time-series shocks in rainfall on current household income to identify transitory income,

she finds that large fractions of transitory income are saved and so household

expenditures are little affected. Nevertheless, a strict form of the hypothesis that all

transitory income is saved is rejected. Evidence presented in Chapter 4 indicates that

unanticipated change in income due to occurrence of a catastrophe results in changes in

consumption. Evidence is presented to show that catastrophes change ex-ante saving

behavior at least for two years after the event.

In addition to strong assumptions regarding credit markets, permanent income models

treat both permanent and transitory income as exogenous. As pointed out by Besley

(1995), this is very restrictive when there exist many potential income sources and more

so in dynamic models. As an alternative, dynamic models have been used that allow for

endogenous income or credit market imperfections. Rosenzweig and Wolpin(1993) and

Fafchamps (1993) specify dynamic programming models of farmer behavior response to

shocks. Rosenzweig and Wolpin model bullock investment and dis-investment decisions,

incorporating the tradeoff in livestock sales between the potential need for current cash

and foregoing future output due to the loss of livestock for traction.

For those studies that model the effects of shocks through income, a key issue is how

income is measured. The concept of permanent income assumes that all income is

exogenous. Including endogenous components in the income measure will result in the

volatility of transitory income being systematically understated. Rejecting risk pooling in

tests such as the Townsend-type test (Townsend 1994) may then be more difficult.


59

2.3.2.2 Identifying Transitory Income

Measuring exogenous swings in income is extremely difficult to achieve in

practice. For studies of risk pooling labor income along with asset sales, transfers and

remittances from temporary migration and farm profits net of the value of family labor

are likely to be endogenous. For example, Morduch (1994) and Rosenzweig and

Binswanger (1993) find that poor Indian farmers are likely to adjust their farm

investments in ways that lower expected profits, but decrease profit variation. Fafchamps

finds that weeding labor is adjusted to rainfall received earlier, at planting time.

In view of the difficulty associated with measuring purely exogenous swings in income,

instrumental variables and household fixed effects techniques have been used to free the

model of unobserved error. Wolpin (1982) uses regional, time-series data on rainfall to

construct long-run moments to instrument current income (which measures permanent

income with error) in a savings equation. Paxson (1992) extends Wolpin’s (1982) study

by using deviations of rainfall from its long run mean (and functions thereof) to construct

a measure of transitory component of income for use in her savings equation. Measure of

permanent income and the (expected) variance of income are likewise constructed.

Rosenzweig and Stark also use rainfall, plus interactions with a household’s dry and

irrigated land owning, to instrument for the household-level mean and variance of farm

profits in explaining the variance of food consumption.

Rosenzweig (1988) uses household fixed effects to model the impacts of household full

income surprises on net transfers into the household and on household net indebtedness.

Likewise, many of the full income pooling tests implicitly use household fixed effects by

transforming the estimating equation so that consumption growth is the dependent

variable (Deaton 1992b).


60

2.3.2.3 Risk Pooling and Consumption Smoothing

Given a measure of transitory shock, ex-ante and ex-post strategies used to

smooth consumption have been modeled. To the extent that households are successful in

smoothing consumption, it is less likely that idiosyncratic shocks will affect human

capital investment decisions. A set of studies have tried to estimate the impacts of

unanticipated income changes on various dimensions of ex-post methods of consumption

smoothing, including transfers, credit transactions, asset sales and labor force

participation.

In Chapter 3, effect of sudden decreases in capital due to catastrophes (economy-wide

shock) on consumption is studied using three models. Results, using Ramsey’s model

indicate an instantaneous drop in consumption. An extended model that simulating the

conversion of maturing capital to productive capital (these terms are explained in Chapter

3) indicates that consumption drops but is not instantaneous as predicted by Ramsey’s

model. After the changes in productivity have stabilized, consumption settles to a level

below its pre-event level, if we assume a permanent increase in the capital share in the

production function of the affected region.

A different set of evidence has been provided in the studies prompted by Townsend’s

(1994) work on testing for complete risk pooling against idiosyncratic shocks. The

intuition for Townsend’s test is that if households are able to perfectly smooth their

consumption (or at least smooth against idiosyncratic risk), then conditional on a

household fixed effect and aggregate village consumption (or alternatively a village-

specific time effect), consumption should be unrelated to household income. With panel

data, household fixed effects can be used, so that consumption growth is regressed on a

village-specific time effect and the lagged level or changes in household income.

Many of the empirical tests to date, most using ICRISAT’s India data are consistent with

Paxon’s (1992) tests of permanent income hypothesis. Full pooling is rejected (as is the

strong form of permanent income), as household income is found to affect changes in


61

consumption over and above village-time specific fixed effects. However the estimated

income effects on consumption are small. These studies suggest that using household

expenditure (per capita) as a measure of long-run income human capital studies may be

reasonable.

A central hypothesis in the risk pooling literature is that wealthier households are better

able to pool their risk. This may arise because wealthier households have more access to

credit markets, because they have more assets to sell in case of need, because they may

have more diversified income sources (such as non-farm employment), or because

relatives living apart may be better able to afford help during times of distress. For

instance, Townsend (1994) and Morduch (1993) stratify their analyses on land owned

and find that perfect risk pooling is likely to be rejected among the landless or small

farmers. Rosensweig and Stark (1989) find that inherited assets help mitigate the impact

of farm profit variability on the variability of food consumption. It should be noted that

these tests are for risk pooling against idiosyncratic risks. Catastrophes, on the other

hand, are economy-wide events. The next section presents some empirical results of

studying economy-wide shocks.

2.3.2.4 Effect on human capital investments

Studies of responses of human capital investments to shocks have examined

expenditures on human capital inputs, such as food (Rosenzweig and Stark 1989,

Morduch (1993, 1994), Rosenzweig and Binswanger (1993), individual nutrient intakes

(Behrman and Deolalikar 1990), child growth (Foster, 1995); schooling attendance

(Jacoby and Skoufias, 1992), and infant mortality (Ravallion, 1987, 1990, 1997;

Razzaque, Alam, Wai and Foster 1990). A different set of studies has decomposed effects

of child mortality on fertility into expected (hoarding) and shock (replacement) effects

(Olsen and Wolpin 1983).

Ravallion and Razzaque et al. estimate the impact of the 1974 Bangladesh famine on

subsequent child mortality. Ravallion shows that in the Matlab area, time-series mortality


62

rates closely track rice prices. Rice prices rose by 50 percent over a very short (3-month)

period at a time when Bangladesh did not have any public safety net programs, such as

the food-for-work program adopted in later years. Household, family and village

mechanisms were, in many cases, overwhelmed. Using vital event data linked to census

records from the Matlab area, Razzaque et al. demonstrate that higher mortality was not

uniformly distributed: smaller increases were registered among wealthier households.

Foster (1995) examines the impact of a major flood in rural Bangladesh on growth in

child weight during the subsequent three months. He derives an Euler equation that

represents changes in household utility associated with changes in child’s weight.

Changes in child weight are expressed as a function of changes in rice prices, changes in

the rate of change of rice prices, changes in the incidence of illness (instrumented by

lagged illness incidence), child age and gender and the elapsed time between the initial

weighing and the follow up. Interest rates are captured by the amount of borrowing by the

household and in aggregate by the village. Foster reports that a higher price of rice is

associated with significantly lower growth for children in landless households, but the

effect is not significant for children of better off households; this is consistent with the

hypothesis of differential ability to smooth. However this ability to smooth may be put to

test in the case of economy wide catastrophes. It is in such situations that the health of the

affected community is seriously undermined.

Is education affected by the occurrence of a catastrophe? Esther Dufflo (1993), based on

evidence from Indonesia, concludes that increased investment in education infrastructure

results in increases in percent of primary educated people in the population. Using

Dufflo”s result, albeit negatively, if the catastrophe results in major destruction of

education related infrastructure, or considerable loss of life, then this may result in a

negative impact on the community’s long term literacy rate.

Hannan Jacoby and Emmanuel Skoufias (1992) show that investment in children’s

education in India is responsive to adverse shocks. Jacoby and Skoufias (1992) use four

years of the ICRISAT data, divided into two cropping seasons, in each year, to examine


63

whether changes in the time allocated to school attendance responds to changes in

measured full income, controlling for changes in the local child wage and for village

season effects. Changes in income do have a significant impact on school attendance, net

of the opportunity cost of children’s time, which the authors interpret as indicating a lack

of perfect consumption smoothing. One implication of this study is that if there is a

significant change in income after a catastrophe it may have an adverse affect on school

attendance.

Jacoby and Skoufias then use Paxon’s method to decompose full income into permanent

and transitory components, conditional on village-season-year effects. They find that

transitory income is only significant for landless households, while anticipated effects are

not significant for either type. This evidence is consistent with landless households using

their children as assets to borrow against bad times.

In sum, there are rather few studies that have attempted to measure effects of resource

shocks on human capital outcomes and how existing assets may condition those effects.

This is certainly a question, which is raised directly by the issue of whether economic

adjustment hurt the poor disproportionately as claimed by Cornia, Jolly and Stewart

(1987). The few studies reviewed here suggest that there are some impacts, certainly in

the case of major events such as a flood or famine, and the effects seem to hit the poor

hardest, consistent with intuition. Whether smaller changes have negative impacts,

though, or whether households are able to adjust in ways not detrimental to human capital

investment is still unclear. Furthermore, the aggregate time-series evidence indicates

differences among countries, which may be partly a function of the existence and quality

of social safety nets as well as of the level of market and human capital development and

thus households’ ability to adjust.

2.4 Methods of Coping- Risk, insurance, credit and saving

In LDCs disaster is often accepted as a “normal” part of life. In this situation

group coping strategies in the forms such as extended households are important. Nomadic


64

herdsmen in semi-arid areas have tended to accumulate cattle during years with good

pasture as an insurance against drought (Smith 1996). In developed countries technology

and engineering design have provided a high degree of reliability for most urban services

against natural hazards. But a severe earthquake can easily disrupt road networks, electric

power lines or water systems. This can have damaging consequences because, when such

systems fail, there is frequently no alternative source of supply.

Claude Gilbert (1998) and Hewitt (1998) point out that disasters have often been

considered as the affairs of the public authorities rather than the affairs of citizens. In

doing this, the perception of the population as a whole is merely passive and bound to be

directed and commanded in cases of disasters. But nearly all studies carried out on deep

crisis situations (Gilbert 1998) show that human communities do participate in the

management of disasters. It is well known that in case of earthquakes, such as the one

that happened in Mexico City in 1985, the assistance to the victims comes first of all

from other survivors, with the means used by the authorities contributing to emergency

only to very little extent. In short, what is interesting in the empirical studies of disasters

is the surprising capacity of the reaction and self-organization of people outside any usual

public or institutional structure. Given the capacity of the citizens to react formidably in

disaster situations, there has been little effort to link studies in risk and disaster-coping

behavior in development economics literature with disaster studies. The disaster literature

has concentrated on investigations on functioning, good or bad, of public powers and

official emergency systems.

In the following section we present literature from development economics that presents

evidence regarding economic behavior in adversely affected low-income settings. People

and in particular disaster victims rely on various social coping systems: households,

groups, community, villages, government and non-government agencies, insurance,

credit, and international institutions.


65

2.4.1 Households, groups, community, villages

Risk and consumption smoothing problems condition structure of rural

households. Indeed, the structure of households in the low-income settings appears to

differ distinctly from that of high-income industrialized countries characterized by more

organized markets, governmental social insurance schemes, more predictable income

sources, and technological change, where the dominant household form is the nuclear

family.

In the post-disaster recovery period each family’s social and economic position and

connections within the larger community are critical factors influencing outcomes for

their household (Drabek et al. 1975; Bolin 1982). Comparative disaster studies have

revealed three principal modes of family recovery: first, the autonomous use of personal

resources (such as savings or insurance); second, reliance on informal kinship support

systems; and third, the utilization of institutional resources (such as government

assistance) (Morrow, 1997). While victims often make use of all three, the extent to

which one dominates is primarily determined by the larger political end economic setting

(Bates and Peacock 1989). In modern industrialized settings, kinship assistance is less apt

to be the primary source of help, but still is important (Bolin 1982; Nigg and Perry 1988).

Morrow (1997) reports those families in Hurricane Andrew’s path provided a unique

opportunity to study the role of kin networks in disaster preparation and response. Only

14 per cent of households in the survey reported in Morrow (1997) received assistance

from relatives when preparing their homes and, among those reporting having kin in the

area, the rate was only 16 per cent. Using logistic regression models Morrow shows that

minority (Black and Hispanics) families are more apt to have been helped by relatives for

pre-disaster preparations as compared to Anglos households. Overall, kin networks

appear to be under-utilized during hurricane preparation: While nearly 75 per cent of the

respondents had relatives living nearby, less than 20 percent reported assisting or being

assisted by them.


66

During the storm about 27 per cent of the total sample reported having relatives stay in

their home during the storm and the rate was significantly higher for Hispanics. After the

storm 24 per cent of those with relatives in the area reported receiving major assistance

with such things as supplies, debris removal, and repairs. Similarly, 30 per cent of those

with family in the area reported assisting relatives after the storm. 44 per cent among the

sub-sample from South Dade received help from relatives in the area. These findings

suggest that under severe conditions family networks become an important source of help

in the aftermath of a disaster, even in an industrialized country.

2.4.2 Insurance, Savings and Credit

The functions of savings, credit and insurance are intimately connected with one

another in most developing economies. However, at first sight, there ought to be a

division of roles between transactions that transfer resources across time, as with savings

and credit, and those that transfer resources across states of the world, as with insurance.

Moreover, decisions about how to allocate resources over time and states ought to be

separable in the sense that the consumer’s decision about how much to borrow and save

is independent of uncertainty about future events. This viewpoint is appropriate only

under the most restrictive of circumstances, when we are in a competitive economy with

a complete set of Arrow-Debreu securities and no externalities. An insurance scheme

may be approximated in the actual economy by various risk-sharing opportunities and

markets. Some possible sources of insurance include stocks in securities markets,

borrowing and lending in credit markets, unemployment insurance, contracts between

employer and employee, crop insurance for farmers, and insurance among family

members or close communities. The separation of the functions of insurance and

saving/borrowing no longer holds when markets are incomplete. It is limitations on

insurance possibilities that make it essential to treat savings, credit and insurance in a

unified way.


67

Two features of developing economies are particularly germane to the link between

savings and insurance. First, the absence of markets for trading in risks is particularly

noticeable. Many types of insurance possibilities taken for granted in developed

countries, are simply not traded. This is especially striking given the relative importance

of risk in the lives of many inhabitants of LDCs, such as the risk of suffering certain

infectious diseases. Second, a large fraction of population is typically dependent on

agricultural income for their livelihood. The latter may be subject to drastic weather

shocks and commodity price fluctuations. Rosenzweig and Binswanger (1993) bring out

the relative importance of risk in an agricultural economy by the finding that the

coefficient of variation of income from south India is 137. For white males aged 25-29

surveyed in Longitudinal Survey of Youth in the US in 1971, the number is just 39.

When insurance markets are incomplete, saving and credit transactions assume a special

role by allowing households to smooth their consumption streams in the face of random

fluctuations. In a purely autarkic model, the savings decision is made in isolation. While

in a developed country we might naturally think of saving using demand deposits that

earn interest, this is not necessarily a good model for LDCs (Besley, 1995). There is

evidence that because of high transactions cost, low levels of literacy and numeracy,

mistrust of financial institutions, individuals will often accumulate savings in forms other

than demand deposits such as assets. The national accounts statistics of India, for

example, report that in 1987-88 households accounted for more than 80 percent and less

than 52 percent of household savings was in the form of financial assets, the rest being

direct saving in physical assets. Bevan, Collier, and Gunning (1989), discusses the

importance of accumulation in non-financial assets after the Kenyan coffee boom in the

late 1970s. The role of these non-financial assets in smoothing consumption against

economy-wide shocks needs to be examined.

Loans are less available when the local economy is subject to a common shock, such as a

late monsoon (Rosenzweig, 1988). Thus weather-induced profit variability may be far

less insurable than idiosyncratic or household specific profit variability necessitating ex

ante risk reduction through altering of portfolio of investments that differ in their


68

sensitivity to weather outcomes. Investments would then be predominantly responsive to

weather risk.

Informal credit institutions are very important in LDCs and there is a huge diversity of

them in Asia as reported by Ghate (1992). Among the main sources of informal credit

are: loans from friends, relatives and community members; rotating savings and credit

association; moneylenders and informal banks; tied credit and pawning.

Udry’s (1990, 1994) studies of Northern Nigeria are based largely on loans from friends,

relatives and community members. Udry focuses on the risk-sharing function of loans,

with repayments being indexed to the borrowers’ and lenders’ economic circumstances.

Educational loans between family members may also be important with repayment

appearing as urban-to-rural remittances. The main enforcement mechanisms for such

loans tend to be informal social sanctions.

2.4.3 Credit, insurance and long-run development and growth

The relation between credit, insurance and long-run development and growth

works via the role of intermediaries in providing finance for industrialization and,

consequently, much lending of this sort occurs in an environment that is, arguably, not

very special to developing countries. For any given aggregate level of savings, the quality

of financial inter-mediation is a crucial determinant of the efficiency of investment

choices, i.e., in ensuring that savings find their way into the most productive

opportunities. Insurance may also be important, especially in relation to incentives to

adopt new, riskier technologies.

An important theme in the relationship between credit markets and long-run development

countries today, and many now developed countries historically, was a lack of institutions

for funds to flow to where capital could be most productively used. The evolution of

financial institutions can be understood in large part as trying to overcome this, leading to

a more efficient allocation of capital throughout the economy. This view is based on


69

realizing gains from trade from differences in technologies. This discussion also seems

relevant to modern day developing countries where, as we remarked above, market

segmentation is significant. Improvements in infrastructure and communications, more

generally, also play a central role in providing market integration. Temporary disruption

of these facilities after an earthquake may result in breaking up the integration that

directly affects the development of the affected region.

Many of the inefficiencies that arise because insurance possibilities are lacking may take

the form of a failure to adopt new technologies and appropriate investments. In a LDC

context Eswaran and Kotwal (1989) have discussed how access to credit may affect

technology adoption decisions. Townsend’s (1992) study of northern Thai villages

attributes the non-adoption of new rice varieties to the absence of insurance possibilities.

Insurance is a key loss-sharing strategy in the developed countries. Like disaster aid, it is

a redistributive method but in this case people at risk join forces with a large financial

organization to spread the costs more widely. Most insurance takes place when an

individual perceives a hazard and purchases a policy from a commercial company, which

guarantees that any specified losses will be reimbursed. Hence, the policyholder spreads

the possibly crippling cash burden from one catastrophe over a number of years through

the repayment of an annual premium.

For the insurance company, risk spreading starts with the underwriting of property, such

as buildings or crops, against natural hazards. Policy underwriters try to ensure that the

property they insure is spread over diverse geographical areas so that only a small

fraction of the total value at risk could be destroyed by a single event. By this means,

payments to those policyholders suffering loss are spread over all policyholders.

Assuming the premiums are set at an appropriate rate, the money received from

policyholders can be used to compensate those suffering loss.


70

2.5 Aid and recovery

Disaster aid is the inevitable outcome of humanitarian concern following a

catastrophic event, usually involving the loss of life. Though necessary for immediate

post-event relief aid can never fully alleviate the profound economic and social

disparities around the world, which are responsible for so much hazard vulnerability. Aid

is not a good long-term solution for disaster reduction since victims come to rely on such

external support. They might be indirectly encouraged to settle in high-risk areas driven

by the expectation of being compensated after a catastrophic event. Data from past

catastrophes indicates that more of government’s allocation of its expenditures to some

form of aid is associated with lager catastrophic losses as a percent of GDP (Fig. 2.21).

Fig. 2.21 Larger aid is associated with larger losses

R2 = 0.2069

0.0%

0.1%

1.0%

10.0%

100.0%

1000.0%

(2.0) (1.5) (1.0) (0.5) - 0.5 1.0 1.5 2.0 2.5

Econ

omic

loss

as

a %

of G

DP

Aid (% of central government expenditures)


71

2.5.1 Disaster aid at the Macro Level

At a macro level, disaster aid flows to victims via governments and charitable

non-governmental organizations (NGOs), such as the International Federation of Red

Cross and Red Crescent Societies (IFRCRCS), Oxfam, Save the Children Fund and the

religious agencies. As many governments in the developed countries are less willing to

take welfare responsibility for the poor and the vulnerable, and governments in the LDCs

are less able to do so, the role of NGOs in disaster relief increases. Since the 1970s there

has been a substantial increase in the proportion of aid channeled through the NGOs. For

example, the European Community, which is the world’s largest single provider of

disaster aid, raised the proportion of its funding through NGOs from zero in 1976 to 40

percent by the mid-1980s (IFRCRCS, 1993, 1994). Other sources of International aid

include UN’s Disaster Relief Organization (UNDRO), Department of Humanitarian

Affairs (DHA), and USA’s Office of Foreign Disaster Assistance (OFDA).

Despite all these efforts to organize disaster relief, the results are often disappointing.

International aid flowing from developed to developing countries can be unreliable and

may well not reflect the true need. For example, severe earthquakes and tropical

cyclones, which invariably result in a high ratio of casualties to survivors, usually

generate a large donor response irrespective of need. Alternatively, droughts and floods

tend to produce comparatively low responses, despite the large numbers of survivors who

will be adversely affected and in need of support. Not only are some disasters apparently

more fashionable than others are but aid is often highly political and may even be used as

a weapon by the powerful donor nations.

The political relations of the affected country with the donor countries often affect the

flow of development and disaster aid. Development aid, in particular, may be tied to trade

agreements rather than targeted at the countries in most need. In Europe a great deal of

disaster aid is raised for former colonies in Africa and Asia whilst the USA most actively

supports friendly Latin American countries within its sphere of influence. Very abrupt


72

changes in policy may occur. For example, the 1988 earthquake in Armenia, formerly

USSR, generated the largest initial donation (5 million pounds sterling) from the British

government following any natural disaster, despite the fact that the government had never

given any disaster aid to the former Soviet Union. This happened in 1976 when

Guatemala rejected earthquake assistance from Britain because the two countries were

then engaged in a territorial dispute over Belize in Central America.

Examples of badly managed aid relief abound. After the hurricane Gilbert aid consisted

of simply dumping surplus commodities such as fur coats, high heel shoes and heavy

winter clothing for victims in the tropics. Consideration should be given to food

requirements and the development needs of the recipient country since over-generous

donations can lower market prices and disrupt the local agricultural economy in some

developing countries. Excessive food aid may induce the government of the affected

region to lower its priority for self-sufficiency in agricultural base.

There is little hard evidence that disaster aid results in net benefit. Chang (1984) applied

an economic model of disaster recovery to coastal Alabama, USA, following a damaging

hurricane in 1979, and concluded that outside assistance from federal agencies and

insurance companies was not sufficient to replace lost assets. The inflows of capital into

an area immediately after a disaster and the improvement of some facilities do not

amount to an overall net gain accruing from a catastrophe.

Getting assistance after the catastrophe may be arduous because of the following reasons:

Getting to the correct location to file an application may be difficult because public

transportation is not available because of massive environmental destruction. Most street-

signs as well as prominent landmarks will also be damaged making the location of

assistance centers problematic.

Successfully negotiating the aid process – getting all of the assistance for which a

household is qualified – typically takes a great deal of time, energy, and skill in dealing

with bureaucracies. Poor people lack these assets. In case of Hurricane Andrew, it usually


73

took several trips to the FEMA Disaster Assistance Centers (DACs) or other centers to

complete various, rarely integrated application processes.

The destruction of entire communities places extraordinary demands on insurance

companies, government agencies, material suppliers, and skilled labor. Many

homeowners lack the necessary resources for repairing their homes, usually because they

are uninsured, underinsured, or their insurance companies folded or paid too little.

2.5.2 Disaster aid at the Micro Level

There are a host of reasons why there may be economic links across households,

and these linkages manifest themselves in a variety of ways, including transfers of money

goods and time. First, households may be altruistic. Second, transfers may be motivated

by exchange through an implicit contract or strategic behavior (Berheim, Scheifler and

Summers 1985; Cox 1987). A particular form of exchange, insurance against income

shocks, may be a key motive for inter-household links in a developing country context

(Rosenzweig 1988a, Rosenzweig and Stark 1989). If so, then this suggests the pool of

potential donors and recipients may be very large.

Peacock, Killian, and Bates (1987) found that households residing in peripheral isolated

rural villages recovered more slowly than households residing in more economically and

politically complex central cities. Households within larger communities had better

access to external resources and were able to take advantage of post-disaster funding and

resource opportunities (Bates and Peacock 1993). The implication is that a community’s

position within a regional stratification system and exchange network has important

consequences for disaster-related recovery processes.

For example the 1987 drought in Pakistan’s Thar Region found many villagers living

with their relatives in other villages in their district. But risk sharing at the provincial

level is less likely because at that level distances may be too great to enact transactions


74

and kinship may be weaker. Risk sharing has implications for the timing of transfers

(Rosenzweig 1988). The direction of net transfers should depend on which party has

faced positive or negative shocks and is unrelated to the life cycle. This hypothesis is

very hard to test without a (long) time series of data on both sides of the contract (and

possibly on all households in the pool.

Households may choose to mitigate risk through spatial diversification in living

arrangements (Rosenzweig 1988). Rosenzweig and Stark (1987) argue that in south India

where women, rather than men, marry and then to move to new households, families will

seek to locate daughters in different places if risk is a concern. The authors find almost

complete diversification, and that the extent of diversification is greatest for the least

wealthy, who are, presumably, those at the greatest risk in the face of a weather shock.

Furthermore, daughters tend to marry kin, who it is argued, are more likely to be

concerned about the origin family after marriage. Rosenzweig (1988) reports that

conditional on wealth, variability of agricultural profits positively affects the number of

co-resident daughters-in-law, which he interprets as a measure of intergenerational and

spatial extension. Though the effect is not precisely estimated, it is interesting to note the

various ways in which risk is managed in developing countries.

2.6 Policy issues

This section discusses some specific issues of policy towards orientating the

development process taking catastrophes into account. There are broadly two normative

criteria that can be applied to motivate policy. The first is based on concerns about

equity. Poor people are most likely to be excluded from trade in formal financial markets.

Some reasons for this is that poor people lack reliable forms of collateral, are less likely

to be literate, numerate, may face higher transactions costs and lack the influence needed

to gain subsidized loans. This suggests that interventions that genuinely broaden the

scope of financial inter-mediation may have a major impact on the poor in terms of

raising their self-reliance in times adverse situations that may result from a catastrophe.


75

While this is a useful measure the question of whether intervention in financial markets

alone is an appropriate policy response is debatable. This raises the large subject of what

are appropriate policy interventions are required to reduce vulnerability against natural

hazards.

Recent discussions of disaster-relief policy have emphasized the synergy with longer-

term development goals (Anderson and Woodrow 1989). Catastrophes triggered by

natural events can have a significant effect on development, at all levels from that

individual households and local communities through to national level. Moreover,

development and its consequences are key factors in determining whether a natural

hazard is transformed into a catastrophe. Management of anticipated catastrophic events

should be integrated into a region’s development plans and no longer left as viewed in

terms of immediate post-event crisis management strategies of relief teams.

For disaster reduction and development to occur together, a reliance on local knowledge

rather than imported technology is required. In rural areas, for example, successful

development means “bottom up” strategies that start at community level, such as the

establishment of cooperatives to provide seed banks, crop insurance and credit for tools

and other assets lost in a disaster. Such measures would help to stabilize the rural base

and halt the migration to unsafe urban environments. Conversely, technical aid,

particularly beyond the emergency relief phases, is perceived as increasing vulnerability

by making a short-term problem semi-permanent through additional dependency.

It is difficult to disentangle disaster and development problems in the developing

countries and to make reliable assessments of disaster aid as an adjustment to an

environmental hazard. But, wherever possible, disaster aid should be minimized. Given

the fact that some emergency response will always be necessary in certain cases, attention

should be given to optimizing this form of relief. More training of local aid workers

would help, especially if continuity could be maintained by re-training core staff from

event to event. Aid needs to be carefully targeted in order to improve the situation of the

most vulnerable people.


76

2.7 Summary

Hazards of similar intensities result in only a few people being affected in the

developed countries but result in thousands being affected in the developing regions of

the world. Why does this happen? In other words, what transforms a hazard into a

catastrophe? The answer to this complex question can be summarized in one word –

vulnerability. Hazards are converted into catastrophes by vulnerabilities in the existing

socioeconomic fabric of a community or a nation. What are the factors that determine

vulnerability?

Using literature on development and growth economics, and sociology of

disasters as pointers, the determinants of vulnerability (measured by the loss-GDP ratio)

to natural hazards were identified. Data on catastrophes demonstrates statistically

significant associations between the various socioeconomic indicators and the loss-GDP

ratios. These associations map the complex relationships between ongoing

socioeconomic and development processes and the determinants of vulnerability to

catastrophes.

Corrupt and inefficient governments and bureaucracies, poor physical

infrastructure facilities, excessive dependence on imports, poor health infrastructure,

large uncertainties in the macroeconomic environment, and low levels of literacy are all

factors contributing towards the vulnerability. It is not surprising to note that these factors

also determine the per capita income of a nation, and it has been shown (Fig. 2.2) that

they are inter-related. However, physical and human capital losses also depend on hazard

intensity.

It is apparent that reducing disasters is possible not only by modifying the hazard, but

also by reducing vulnerability. However, most of the efforts of those concerned with

disasters are focused either on reducing the impact of the hazard itself (sometimes in

expensive and inappropriate ways), or on reducing one aspect of vulnerability – social


77

protection through certain forms of technological preparedness. The major determinants

that make people vulnerable (i.e. the social, economic and political factors, which

determine the level of resilience of people’s livelihoods, and their ability to withstand and

prepare for hazards) are rarely tackled. Building storm shelters on the cyclone-prone

regions of eastern India is a solution, which can manage the needs of a community

immediately after a cyclone. It does not address the wider problems associated with

growing vulnerability of certain sections of the population during “normal” times.

Mitigation of hazards is normally associated with attempts to reduce the intensity of a

hazard or to make some other modification, which is supposed to lessen its impact. It is

often a hazard-centered rather than a people centered-approach. Large-scale engineering

works to counter river floods and expensive satellite early warning systems for tropical

cyclones are two examples of the “technocratic” solutions to the problem. Other

approaches rely on the state and through local groups or NGO activities. But state

intervention is often unreliable. Though well intentioned, the state is usually a party to the

very same economic and social processes that lead people to be unable to protect

themselves in the first place.

More generally, national development plans and budgeting exercises should take

potential losses from disasters into account and make assessments of the likely

investments necessary to recover from future natural hazards, which may transform into

catastrophes.

Chapter Three: Theoretical Models

78

Chapter Three

Short-Run Analysis of the Economic Consequences of Catastrophes _____________________________________________________________

3.1 Introduction

This chapter investigates the dynamic effects of a catastrophic event that destroys

substantial capital stock of an economy. Three models are presented to address various

aspects of the problem. For the three models, a catastrophe, due to the occurrence of an

earthquake or a hurricane, is modeled by a discontinuous change in the capital stock.

Post-event reconstruction and behavior is modeled using perturbations in productivity

and the inflow of investment exogenously.

A major strength of the framework on which the models are built is that they are founded

in standard micro-economic principles. The study examines how an economy initially in

a steady state responds to unanticipated and large change in the capital stock followed by

an arbitrarily complex change in the affected region’s productivity. Laplace transforms

are used to study the perturbations of the resulting system of ordinary differential

equations.

The results presented below indicate the initial impact on investment, consumption, and

production due to occurrence of a catastrophe and the resultant changes in productivity.

For the purposes of the study we examine the dynamic effects two or three years after the

event. Hence the term ‘short run’ is used in the title of this chapter.

The first model is based on Ramsey’s growth model. Perturbations of the model to

simulate the effect of a catastrophe reveal an initial drop in consumption consequent to

the loss in capital. After an event there is a discontinuous drop in the output of region,

followed by a rapid growth rate. The growth rate is changed by the magnitude of change

in the productivity of the affected that occurs after the event. Immediately after the event,

the affected region’s productivity decreases due to non-availability of infrastructure,


79

factory shutdowns, and increase in prices of various commodities. However, when new

capital is constructed, the productivity increases, because old and least maintained capital

stock is replaced by more productive capital. It may be noted that it is the old and least

maintained capital stock that is most vulnerable to natural hazards. These changes in

productivity are modeled by appropriate perturbations of the production functions.

The second model is an extension of the Ramsey’s growth model, which includes two

different types of capital – the maturing capital and the productive capital. The maturing

capital is present in the economy due to various ongoing construction processes. The

productive capital is the capital that has been commissioned and is being used to produce

goods or services. When a catastrophe strikes a region, the model simulates the impact of

efficiency of the reconstruction process. The results suggest that the output of the

economy drops discontinuously and starts to grow only after a period of time when its

productivity increases. This result is different form the previous model, wherein

economy grows, without delay, after the event. The model is able to simulate a more

realistic behavior of the economy as will be clear from the econometric evidence

presented in a Chapter 4. The reasoning behind it is that in real-life, reconstruction of

damaged capital takes time. It is impossible to fully utilize the capital that is being

reconstructed immediately, as a result of unavoidable start-up problems.

The third model examines the consequences of a catastrophe on a region that is

interacting with another region. This model is again an extension of Ramsey’s model.

The model simulates how resultant changes in productivity of the affected region are

propagated to the other region. The model simulates behavior of county level economies

interacting with state-level economies.

The chapter is organized as follows. Section 3.2 examines the first model. Effect on

initial consumption and investment are derived in Section 3.2.3. Results and discussion

based on a numerical simulation are presented in Section 3.2.4. The next section (3.3)

presents the second model. Results based on a numerical simulation are examined to

bring out salient features of modeling the efficiency of the post-event reconstruction


80

process in Section 3.3.2. Section 3.4 presents the third model. Section 3.4.1 presents

numerical simulation results of the consequences of a catastrophe on interacting

economic regions. Section 3.5 summarizes the chapter’s main points.

3.2 Modeling a catastrophe

Assume that an economy consists of a large fixed number of identical and

infinitely lived economic agents. The common utility functional is assumed to be

additively separable in time with a constant pure rate of time preference, ρ:

∫∞

−=0

)( dtcueU tρ (3.2.1)

where c(t) is consumption of a single item at time t and u is the instantaneous utility

function. We assume the following form for the utility:

γ

γ

+=

+

1)(

1ccu (3.2.2)

γ measures the coefficient of risk aversion. The representative agent will chose a

consumption path, c(t), capital accumulation, k(t) such that he maximizes his utility:

∫∞

−=0

)(0 )(max)( dtcuekV t

tc

ρ (3.2.3)

s.t.

kckfk β−−= )(& (3.2.4) *

0 kk = (3.2.5)

Eq. 3.2.4 represents the fact that the change in capital stock, k& , occurs due to

output, )(kf , of the economy and decreases due to diversion of output towards

consumption, c, and depreciation of existing capital stock, k, at rate β. In other words,

investment, which measures the change in capital, is output less consumption and

depreciation. At t=0, the capital stock is assumed to be in equilibrium, k*. The following

form of the production function is assumed – the Cobb-Douglas form: σAkkf =)( (3.2.6)


81

It is easily recognized that A is a measure of productivity of the economy and σ is a

measure of the capital share in the output. The resulting equilibrium solves the system of

differential equations:

))((/ kfcc ′−+= βργ& (3.2.7)

kckfk β−−= )(& (3.2.8)

At equilibrium, 0== kc && . This implies that the equilibrium values of c and k, denoted by

c* and k*, respectively, are given by:

11

* −

+

=σ

σβρ

Ak (3.2.9)

*** )( kkfc β−= (3.2.10)

A catastrophe is modeled by a discontinuous change in capital, for one period after the

occurrence of an event at time τ. This is accompanied by a change in productivity

associated with post-event reconstruction due to the inflow of aid. To model these

changes due to occurrence of a catastrophe the following perturbation equations are used:

))(1( 1 thεσσ +→ (3.2.11a)

))(1( 2 thAA ε+→ (3.2.11b)

)(3 thεββ +→ (3.2.11c)

Eq. (3.2.11a) simulates the changes in capital share,σ, in output after the occurrence of an

event using a perturbation ε, and a time varying function h1(t). In the rest of the in this

chapter, ε, will denote the perturbation. Eq.(3.2.11b) models the change in total factor

productivity, A, due to construction of capital. The discontinuous reduction in capital

stock after an event is modeled by assuming a Dirac Delta function for h3(t) in Eq.

3.2.11c.

Applying the perturbation in Eq.3.2.11, to Eq.3.2.7 and 3.2.8, the perturbed system of

differential equations can be written as follows: 1))(1(

1231))(1))((1()([/ −+++−++= thkththAthcc εσεεσεβργ& (3.2.12)

]))(())(1([ 3))(1(

21 kthckthAk th εβε εσ +−−+= +& (3.2.13)

)()( 33 τδε −= tath (3.2.14)


82

δ(t-τ) is the Dirac Delta function. To make the dependence on ε explicit the equations are

written as: 1))(1(

1231))(1))((1()([/),(),( −+++−++= th

t kththAthtctc εσεεσεβργεε (3.2.15)

)],())((),(),())(1([),( 3))(1(

21 εεβεεεε εσ tkthtctkthAtk th

t +−−+= + (3.2.16)

*),0( kk =ε (3.2.17a)

∞<∞→ )(lim tkt (3.2.17b)

Eq. 3.2.17a states the fact that the system’s initial condition coincides with its

equilibrium. Since the economy is initially at the ε=0 steady state, the occurrence of a

catastrophic event essentially implies that ε has changed. The impact of this change in ε

on the critical variables at future times will be studied herein. Differentiating Eqs.3.2.15

and 3.2.16 with respect to ε and evaluating at the steady state, the following equations

result:

{ }])()()()()()([/

)0,())(1(/)0,(*2

1211*

3*

2**

kLogthththkAthc

tkkActc t

σσγ

σσγσ

εσ

ε

++−

+−−=−

−

(3.2.18)

σσσβ

βσσεεε

)*()(2)}*()(11)*()(3{*

}1)*(){0,()0,()0,(

kAthkLogthkAthk

kAtktcttk

+−+−−

+−−+−= (3.2.19)

In matrix form:

{ }

++−−++−

+

−′−

′′−=

−

−

σσ

σ

ε

ε

ε

ε

σβσσγ

βγ

)()()}()()()({])()()()()()([/

)0,()0,(

)(1)(/0

)0,()0,(

*2

*1

1*3

*

*2121

1*3

*

*

**

kAthkLogthkAthkkLogthththkAthc

tktc

kfkfc

tktc

&&

(3.2.20)

Using Laplace transform

{ }

++−−+++−+

+

=

−

−

σσε

σε

ε

ε

ε

ε

σβσσγ

)()()}()()()({)0,0(])()()()()()([/)0,0(

)()(

)()(

*2

*1

1*3

*

*2121

1*3

*

kAsHkLogsHkAsHkkkLogsHsHsHkAsHcc

sKsC

sKsC

s J (3.2.21)

where Hi(s) is the Laplace transform of hi(t). It should be noted here that cε(0,0) is the

change in c at t = 0 induced by ε. cε(0,0) is an unknown at this point, but the initial value


83

of k is fixed at k*. This fact yields an initial condition kε(0,0)= k* = 0. Rewriting Eq.

3.2.21, the following equation results:

{ }

++−−++−+

−=

−

−−

σσ

σε

ε

ε

σβσσγ

)()()}()()()({])()()()()()([/)0,0(

)()()(

*2

*1

1*3

*

*2121

1*3

*1

kAsHkLogsHkAsHkkLogsHsHsHkAsHcc

ssKsC

JI

(3.2.22)

Eq. 2.22 is used to obtain the solution for Cε(s) and Kε(s) in terms of cε(0,0). To ensure

the uniqueness of cε(0,0) it is assumed that the eigenvalues of the linearized system, J,

are distinct, real and have opposite signs. That such is the case for Ramsey’s growth

model is explained in Barro and Sala-i-Martin (1995). Let µ and ξ be the two eigenvalues

of the Jacobian J and furthermore let µ be the positive eigenvalue and ξ be the negative.

Then, the positive eigenvalue is given by:

( ) 2/)(/42 ssss kfc ′′−+= γρρµ (3.2.23)

since at steady state,

βρ −′= )( sskf

The eigenvalues will be used in studying the impact on consumption and investment. 3.2.1 Impact on Consumption and Investment

We assume that the initial change in capital due to a perturbation ε, kε(t,0), is

bounded, which implies that a catastrophe causes a finite and bounded change in the

capital. In this case Kε(s) must be finite for all s>0, since Kε(s) is the Laplace transform of

kε(t,0). This again implies that when s = µ, Kε(µ) must be finite for any positive

eigenvalue, µ. However, when Eq. 3.2.22 is evaluated at s = µ, a singularity exists since

by definition of µ being an eigenvalue, µI – J, is a singular matrix. The matrix (sI – J)-1

can be written as

−

−−− 1121

1222

))((1

JsJJJs

ss ξµ (3.2.24)

In particular, the denominator is zero when s = µ. Therefore the only way for Kε(µ) to be

finite is that the following should be satisfied:


84

{ }

=

++−−++−+

−

−

−

−

00

)()()}()()()({])()()()()()([/)0,0(

*2

*1

1*3

*

*2121

1*3

*

1121

1222

σσ

σε

σβσσγ

kAsHkLogsHkAsHkkLogsHsHsHkAsHcc

JsJJJs

(3.2.25)

This implies two conditions for cε(0,0); however, since µ is an eigenvalue, these

conditions are not independent thus giving a unique value of cε(0,0). Therefore:

{ }])()()()()()([/

/])()()}()()()({)[()0,0(*2

1211*

3*

12*

2*

11*

3*

11

kLogsHsHsHkAsHc

JkAsHkLogsHkAsHkJc

σσγ

σβµσ

σσε

++−−

++−−−−=−

−

(3.2.26)

Substituting the values for J11 (=0) and J21 (=-1), the following equation results:

{ }])()()()()()([/

])()()}()()()({[)0,0(*2

1211*

3*

*2

*1

1*3

*

kLogsHsHsHkAsHc

kAsHkLogsHkAsHkc

σσγ

σβµσ

σσε

++−−

++−−=−

−

(3.2.27)

Simplifying the above expression and substituting H3(µ) = e-τµ,

******2

***1

**1

]/[)](/)()[(

)]())()(1(/)[()()0,0(

kkcekfckfH

kLogkkLogHckfHc

βµγγµµ

µσµγµτµ

ε

−+−′++

++′=−

(3.2.28)

The Eq.3.2.28 gives the initial impact on consumption due to a sudden change in the

capital stock accompanied by changes in productivity. The expression tells us that

consumption decrease exponentially after the event (the third term), which is

accompanied by the normal depreciation of capital (the fourth term). This is in part offset

by increases in productivity during reconstruction represented by the first two terms. The

term γ/*c is the measure of coefficient of absolute constant risk aversion. The term

)( *kf ′ is the marginal product of capital at steady state. This measures the price of

capital at steady state. µµ)(H 1 measures the after-event change in productivity, that is,

discount the change in productivity at rate µ and multiply the result by µ. The expression

for µ, Eq.3.2.23, implies that it is greater than the pure rate of time preference, ρ, for

realistic values of crucial parameters. Since µ>ρ, µH1(µ) puts more weight on changes in

productivity immediately after the event relative to distant future changes than does

ρH(ρ). This implies that productivity changes decay rapidly relative to the utility

discount rate as the economy evolves away from the event date. This in turn implies that


85

the initial effects of a catastrophe on the consumption, lasts for short spans after the

event. Numerical experiments, reported in a later section, indicate that the impact on

consumption lasts for a time when the changes in productivity have stabilized.

The initial change in capital kε(0,0), which is the impact on investment, can be

determined as shown below. Since

{ }( )( )

++−−+

++−+−

−−=

−

−

σσ

σε

ε

µµσµβµ

σµµµσµγ

ξµ

)()()}()()(/)({

)()()()()()([/)0,0(

))((1)(

*2

*1

1*3

*

*2121

1*3

*

kAHkLogHkAHks

kLogHHHkAHcc

sssK

(3.2.29)

Eq.3.2.29 can be written as:

{ }( )( )σσ

σε

ε

σβ

σσµγ

µξξµ

)()()}()()(/)({

)()()()()()([/)0,0(

)())((

*2

*1

1*3

*

*2121

1*3

*

2

kAsHkLogsHkAssHks

kLogsHsHsHkAHcc

sKss

++−−+

++−+−

=++−

−

− (3.2.30)

Taking limit of ∞→s and using the relations:

)0())0,0()((limit 2

s εεε ksksKs &=−∞→

(3.2.31)

)0()(limits εε kssK =

∞→ (3.2.32)

the following relation for initial investment can be obtained:

)0()())0()()()0(()0,0()0( 2*

1**

3* hkfhkLogkfhkcI +′+−−+−= βεε (3.2.33)

Since the assumed perturbations occur after the catastrophe (that occurs at time τ), the

functions, h1(0), h2(0), and h3(0) are zero. Eq.3.2.33 implies that initial investment occurs

to compensate for the changes in consumption.

The system of equations (Eq. 3.2.20) becomes a non-autonomous linear initial value

problem, which can be solved to yield a solution for kε(t,0) and cε(t,0). The procedure of

solving the initial value problem (IVP) is shown below.

( )

−−−

−−=

−−

=− −

skfcs

ssssadjs

1)(/

))((1

)det()( *''*

1 γρξµAI

AIAI (3.2.34)

Taking the inverse Laplace transform, the following results:

( )

−−−

−−=−= −−−

skfcs

ssLsLe

sst

1)(/

))((1 ''*

111 γρξµ

AIA (3.2.35)


86

[ ] ( ) [ ]

( ) [ ] [ ]

++−−

−

−−

−−+−−=

tttt

tttt

t

eeee

eekfeee

ξµξµ

ξµξµ

ξξµµξµ

ξµγξρρµ

ξµ

)(11

)()()()(

1 *''

A (3.2.36)

The solution can be written as follows:

{ }[ ]{ } ds

kAthkLogthkAthkkLogthththkAthc

ee

kc

etktc

tst

t

∫

++−−++−

+

=

−

−

0*

1*

11*

3*

*2121

1*3

*

)()()()()()()()()()()()(/

)0()0(

)()(

σσ

σ

ε

ε

ε

ε

σβσσγAA

A

(3.2.37)

Eq. 3.2.37 gives the evolution of the consumption and capital stock when the economic

system 3.2.3-4 is perturbed by a catastrophic event. Numerical experiments on the

solution 3.2.37 are presented in Section 3.2.4.

3.2.2 Impact on Welfare

The overall welfare function can be written as:

∫∞

−=0

)),(()( dttcueU t εε ρ (3.2.38)

U(ε) represents the present value of overall welfare associated with different ε. One

measure of the impact of the catastrophe is dU/dε, which the change in the overall

welfare due to perturbations induced by catastrophe. This is calculated in the

neighborhood of the ε=0 paths. The change in U due to an infinitesimal change in ε is

∫∞

−=0

* )0,( dttcecddU t

ερ

ε (3.2.39)

( ) [ ]))()()()(())((

))(()(''

21**

3*

*1*

ρρσβρξρµργε

γ

HHkLogkfHkkfcddU

+++−−−

=+

(3.2.40)

The above expression (Eq. 3.2.40) can be used as a measure of the secondary effects of a

catastrophe. Since f’’(k)<0, ρ<µ, and γ<0, the first term of Eq.3.2.40 is positive. The

welfare change is negatively affected by the capital loss: - k*(H3(ρ)+β). This loss in

welfare can be compensated by appropriate reconstruction measures that boost the


87

productivity. The changes in productivity are given by: f(k*)(σLog(k*)H1(ρ)+H2(ρ)) that

depends crucially on the steady state production level, f(k*). Steady state production level

means pre-event per capita output. H1(ρ) is the discounted (at rate ρ) change in the capital

share, σ, after the event. H2(ρ)is the discounted (at rate ρ) change in the technology, A,

after the event. Depending on the nature of these changes in productivity, the net overall

change in welfare may be controlled.

The following expressions (based on Taylor series expansion around the steady state)

could be used to determine the unit impulse response functions due to a catastrophe:

)0,()()1,( ** tckkftc εβ +−≅ (3.2.41)

)0,()1,( * tkktk ε+≅ (3.2.42)

3.2.4 Numerical Experiments

The Eq. 3.2.20 is solved numerically. The Mathematica© program is presented in

electronic form in Appendix C. The following Cobb-Douglas form of production function

is chosen:

σσρ kkf /)( = (3.2.43)

It is assumed that σ =0.25, and ρ = 0.04. ρ measures the time rate of preference or the

real interest rate, hence 4% is a reasonable value. σ measures the share of capital per unit

labor of 25% in the output, which is a reasonable assumption. It is also assumed that γ =

-0.5 and that time at which the catastrophe occurs, τ = 0.5. Productivity changes due to

the catastrophe are handled by the two functions. For changes in the capital share in the

production function, σ, the form of the function is

h1(t) = [(a4+b1)/(2-τ)t - (2a4+b1τ)/(last-τ)][UnitStep(t-τ) -UnitStep(t-τ)]

+ b1UnitStep(t-2) (3.2.44)

where, a4 = 0.05. The model is simulated for various values of b1. The plot for various

assumed functions simulating the changes in the capital share in the production function

is shown in Fig. 3.1a (b1 = 0.0, 0.05, 0.1, 0.15, 0.2). This type of change in productivity

may occur when the reconstructed capital replaces pre-event old and ill-maintained

capital stock. The reconstructed capital incorporates latest technology and hence causes


88

Note: In all subsequent figures in this chapter time is measured in years, and the catastrophic

event occurs 0.5 units of time (years) after the system starts to evolve from its steady state. A bar

denotes the occurrence of an event.

Fig. 3.1a Changes in capital share in the production function

-10%

-5%

0%

5%

10%

15%

20%

25%

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5Time (in years)

Cap

ital s

hare

0% Change in capital share5%10%15%20%

Time of occurrence of the event

Fig. 3.1b Effect of loss intensity on initial consumption

-0.014-0.012-0.01

-0.008-0.006-0.004-0.002

00.0020.004

0 0.05 0.1 0.15 0.2 0.25

loss ratio

Initi

al c

onsu

mpt

ion

chan

ge


89

permanent shifts in the capital share in the production function. The total factor

productivity (TFP) change is modeled using:

[ ])3())5.0(()sin()( 12 −−+−= tUnitSteptUnitSteptath τ (3.2.45)

The plot for the TFP change is shown in Fig. 3.2b. Capital reduction at time τ due to the

occurrence of a catastrophe is modeled using:

)()( 33 τδ −= tath (3.2.46)

It is assumed that 10 percent of the capital stock is destroyed or a3 = 0.1. The results are

shown in Fig. 3.1b to 1f.

Fig. 3.1b plots the initial consumption change required for a stable solution (Eq.3.2.28)

for various values of b1. The curve implies that as the after event capital share in the

production function increases, smaller initial consumption levels can ensure stability.

This implies that an economy with a lower pre-event consumption, for a given percentage

of capital loss, requires higher post-event capital share in its production function to

achieve stability. Conversely, a higher pre-event consumption level demands lesser

increase in post-event capital share for stability.

Fig. 3.1c shows the discontinuous drop in consumption when the event occurs. The

unanticipated change in income (output, Fig. 3.1e) due to sudden loss of capital causes a

drop in consumption and the empirical evidence presented in Chapter 4 supports this.

There is a drop of around 40% in consumption due to a 22% percent loss in capital stock.

It takes approximately 3 units of time to reach to its post-event stable consumption level,

which is 25%, less than its pre-event consumption level. From the graph (Fig. 3.1c) it is

clear that increase in post-event share of capital in production function increases the post-

event level of consumption.

Fig. 3.1d plots the evolution of capital after 22% destruction due to a catastrophic event.

It takes approximately 3 units of time to reach to its pre-event level. We see that the loss

of capital is compensated for the decrease in consumption level. The evolution of the

output of the economy is shown in Fig3.1e. The output clearly portrays the changes that

occur in the capital share. Increase in the TFP cause the output to rise at t=1. The output

traces this increase till it lasts (t=3). Greater the rise in TFP, greater is the output.


90

Fig 3.1c Evolution of consumption

0%

1%

2%

3%

4%

5%

6%

7%

8%

9%

0 0.5 1 1.5 2 2.5 3 3.5

Time

Con

sum

ptio

n


Fig 3.1d Changes in capital assets

0.10

0.12

0.14

0.16

0.18

0.20

0.22

0 0.5 1 1.5 2 2.5 3 3.5

Time

Prod

uctiv

e ca

pita

l



91

Fig 3.1e Evolution of output

0.6

0.61

0.62

0.63

0.64

0.65

0.66

0.67

0.68

0.69

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time

Out

put


3.1f Changes in economic growth over time

-2.0%-1.5%-1.0%

-0.5%0.0%0.5%1.0%1.5%2.0%

2.5%3.0%3.5%

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time

Econ

omic

gro

wth



92

Growth (Fig. 3.1f) increases immediately after the loss in capital, consistent with the

standard growth model prediction that lesser the output, higher is the growth. The growth

then drops till t=1, where the rise in total TFP boosts growth rate. The growth rate then

continues to fall till t=3, when the changes in TFP cease. The kink at t=2, is the point

where the changes in capital share of production stabilize. After all the changes in the

production function has stabilized, at t=3, greater increases in productivity result in

greater post event growth rates.

After studying the impact of changes in productivity on the post-event behavior of an

economy, the impact of changes in the capital loss is studied. It is assumed that

a2=b1=0.1 (Fig.3.2a and 3.2b) but a3 varies from 0.0 to 0.2 simulating various levels of

capital loss. Greater the loss, lower the level of consumption (Fig. 3.2c) and capital (Fig.

3.2d). It takes around 3 units of time for the economy to attain stable solution (which is

the middle path in the five paths shown). Fig. 3.2e shows the evolution of output. Greater

the losses, lower are the output. The output grows after the event. At t=1 there is a

noticeable kink due to increase in the TFP. The output grows till it stabilizes around t=3,

when all the changes in productivity have stabilized.

Growth is plotted in Fig. 3.2f. The results show that greater the loss higher the growth

rate. This does not change even after all the changes in productivity have stabilized.

Empirical evidence from post event growth rates indicate that they are negatively related

to losses, i.e. greater the loss lower is the post event growth rate. The model is not able to

explain this empirical observation. The second model presented in the next section helps

us to better understand this empirical observation.

Fig. 3.2g gives a phase space portrait of the evolution of consumption and capital after a

catastrophe. Fig. 3.2h plots the changes in the welfare due to the occurrence of a

catastrophe. The plot indicates that a greater loss in capital results in greater loss in

welfare. This change in welfare due to occurrence of a catastrophe can be used as a

measure of the secondary impact.


93


-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time

Cap

ital s

hare

Fig 3.2b Changes in total factor productivity with time

0

0.02

0.04

0.06

0.08

0.1

0.12

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5


94

Fig 3.2c Evolution of consumption

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0 0.5 1 1.5 2 2.5 3 3.5

Time

Con

sum

ptio

n

Loss ratio = 0%5%10%15%20%

Fig 3.2d Changes in capital assets

0.10

0.12

0.14

0.16

0.18

0.20

0.22

0.24

0 0.5 1 1.5 2 2.5 3 3.5

Time

Prod

uctiv

e ca

pita

l

Loss ratio = 0%5%10%15%20%


95

Fig 3.2e Evolution of output

0.6

0.62

0.64

0.66

0.68

0.7

0.72

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time

Out

put

Loss ratio = 0%5%10%15%20%

3.2f Changes in economic growth over time

-3%

-2%

-1%

0%

1%

2%

3%

4%

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time

Econ

omic

gro

wth

Loss ratio = 0%5%10%15%20%


96

Fig. 3.2g How does consumption vary with changes in capital?

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24Capital

Con

sum

ptio

n

Loss ratio = 0%5%10%15%20%

Fig3.2h Changes in welfare

-0.5-0.48-0.46

-0.44-0.42-0.4

-0.38-0.36

-0.34-0.32-0.3

0 0.2 0.4 0.6 0.8 1

Loss ratios

Pres

ent v

alue

of u

tility


97

3.3 Model including the effects of efficiency of post-event reconstruction

In an economy, construction activity of some kind is almost always going on.

Capital in an economy can therefore be classified into two categories – the maturing

capital and the productive capital. The capital that is being constructed does not become

immediately productive. For example, it takes time to build a bridge. It is crucial to

model the effects of speed with which the maturing capital becomes productive capital.

The model described in the previous section does not model the effects of post event

construction. Typically, after a catastrophic earthquake, productive capital including

buildings and infrastructure will be damaged or destroyed. Reconstruction activities start

some time after an event. With the reconstruction efforts gaining momentum, the

conversion of maturing capital to productive capital may temporarily exceed its normal

values for a period of time depending on the inflow of investment in the affected region.

The model described below tries to simulate this behavior of an economy.

3.3.1 Model

The representative agent will choose a consumption path, c(t), capital

accumulation, k2(t) such that he maximizes his utility:

∫∞

−=0

)(0 )(max)( dtcuekV t

tc

ρ (3.3.1)

such that:

121 )( kckfk α−−=& (3.3.2)

212 kkk βα −=& (3.3.3)

where k1 is the maturing capital and k2 is the productive capital. Eq.3.3.2 states that the

maturing capital grows from investment (f(k2) ) but is partly offset by the consumption (c)

and partly by conversion into productive capital (αk1). Eq. 3.3.3 states that the productive

capital grows depending on a portion of the maturing capital (αk1) but the growth is

negatively affected by the depreciation of the productive capital (βk2). To study the

consequences of a catastrophe on an economy described in Eqs. 3.3.1-3, it is assumed


98

that the economy initially is at equilibrium with the steady state values of capital given by

k*1 and k*

2 and consumption given by c*.

The following form of the production function is assumed – the Cobb-Douglas form:

σ2)( skkf = (3.3.4)

The current value Hamiltonian for Eqs.3.3.1-3 is given by

( ) ( )212121 )()( kkkckfcuH βαλαλ −+−−+= (3.3.5)

λi(t) (i=1,2) can be interpreted in simple economic terms as shadow prices. λ1(t) is the

change in the maximum attainable value of the objective function discounted to t, when

the economy has to acquire a single unit of maturing capital at time t. Similarly, λ2(t) is

the change in the maximum attainable value of the objective function discounted to t,

when the economy has to acquire a single unit of productive capital at time t.

The resulting equilibrium solves the system of differential equations:

211 )( αλλαρλ −+=& (3.3.6)

2122 )()(' λβρλλ ++−= kf& (3.3.7)

0)('0 1 =−⇒= λcuH c (3.3.8)

Assuming the utility function as in Eq.2.2 the following relationship is obtained: γλ /1

1=c (3.3.9)

A catastrophe is modeled by a discontinuous change in capital, for one period after the

occurrence of an event at time τ. This is accompanied by a change in productivity

associated with post-event reconstruction. To model these changes due to occurrence of a

catastrophe the following perturbation equations are used:

))(1( 1 thεαα +→ (3.3.10a)

)(2 thaidexternal ε→ (3.3.10b)

)(3 thεββ +→ (3.3.10c)

))(1( 4 thεσσ +→ (3.3.10d)

Eq. 3.3.10a simulates the changes in the rate at which maturing capital is converted to

productive capital. Eq.3.3.10b models the flow of external aid. The discontinuous


99

reduction in capital stock after an event is modeled by assuming a Dirac Delta function

for h3(t) in Eq. 3.3.10c. Eq.3.3.10d models the productivity due to a change to the capital

share in the production function.

The perturbed system of differential equations can be written as follows:

)(),())((),(),(),( 211/1

1))(1(

214 thtkthttsktk th εεεαελεε γεσ ++−−= +& (3.3.11)

),())((),())((),( 23112 εεβεεαε tkthtkthtk +−+=& (3.3.12)

),())((),()))(1((),( 21111 ελεαελεαρελ tthttht +−++=& (3.3.13)

),())((),(),())(1(),( 2211))(1(

2424 ελτεδβρελεεσελ εσ ttttkthst th −−−++−= −+& (3.3.14)

Differentiating the above system with respect to ε, and evaluating at ε=0, the following

system of equations result:

+−−

−+−

+

+−+

−−−

=

−

*1

*24

*2

*23

*2

*11

*231

*1

21*14

*2

2

1

2

1

*2

*12

1/1*1

*2

2

1

2

1

))(1)(()(')())((

)()()()()()(

)0,()0,()0,()0,(

)()(')("0)(00

000)(/1)('

)0,()0,()0,()0,(

λσλλλα

αασ

λλ

βρλααρ

βαλγα

λλ

ε

ε

ε

εγ

kLogthkfthth

kththkththkthkf

tttktk

kfkf

kf

tttktk

ss&

&

&

&

(3.3.15)

Performing the Laplace transform:

+−+−+

−++−+

−=

ΛΛΚΚ

−

*1

*24

*2

*232

*2

*111

*231

*12

21*14

*21

1

2

1

2

1

))(1)(()(')()0,0())(()0,0(

)()()0,0()()()()()0,0(

)(

)()()()(

λσλλλλαλ

αασ

ε

ε

ε

ε

ε

ε

ε

ε

kLogthkfthth

kththkkththkthkfk

s

ssss

JI

&

&

&

&

(3.3.16)

Let the positive eigenvalue of J be µ. Then the conditions that k1(t), k2(t) < ∞ as t → ∞,

k1ε(0,0) = k2ε(0,0) = 0, and the singularity of the matrix (sI-J) at s = µ imply that:


100

=

+−+−+

−+−

−

0000

))(1)(()(')()0,0())(()0,0(

)()()()()()()(

)(

*1

*24

*2

*232

*2

*111

*231

*1

21*14

*2

*2

λσµλµλλλµαλ

µµαµµαµσ

ε

ε

kLogHkfHH

kHHkHHkHkLogkf

sadj JI (3.3.17)

Eqs.3.3.17 can be solved to obtain the initial conditions λ1ε(0,0) and λ2ε(0,0). The

Eqs.3.3.15 can then be solved using standard procedures (Appendix ). The solutions so

derived will be used to explain the relevance of conversion of maturing capital into

productive capital in the post-event dynamics.


The following Cobb-Douglas from of production function is chosen:

σskkf =)( (3.3.18)

It is assumed that σ =0.25, and ρ = 0.04. It is also assumed that γ = -0.5 and that time at

which the catastrophe occurs, τ = 0.5. Productivity changes due to the catastrophe are

handled by the function in Eq. 3.3.10d for changes in the capital share in the production

function, σ. The flow of external aid is modeled using (Fig 3b):


A similar function is used to model the changes in rate at which maturing capital is

converted to productive factor:


The Mathematica© program for simulating these equations is presented in electronic

form in Appendix D. The plot for various assumed functions simulating the changes in

the capital share in the production function is shown in Fig. 3.3a (a1 = 0.0, 0.05, 0.1, 0.15,

0.2). This change in conversion from maturing capital to productive capital occurs based

on the efficiency of the reconstruction process. More efficient the reconstruction, higher

the rate at which reconstructed capital becomes productive. Capital reduction at time τ

due to the occurrence of a catastrophe is modeled using Eq.3.3.10c, where it is assumed

that 10 percent of the capital stock is destroyed. Changes in capital share are modeled

using Eq.3.3.10d (Fig.3b). The results are shown in Figs. 3.3c to 3g.


101

Fig. 3.3a Changes in total factor productivity

0%

5%

10%

15%

20%

25%

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time

Cap

ital s

hare

0% change in TFP5%10%15%20%

Fig. 3.3b Changes in aid and capital share in the production function

-6%

-5%

-4%-3%

-2%

-1%

0%1%

2%

3%

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time

Cap

ital s

hare

aid

capital share


102

Fig. 3.3c plots evolution of maturing capital after the event. The maturing capital grows

slower after the catastrophe, until there is a change in the conversion factor α at t= 0.5.

Greater the change in conversion factor lesser is the level of maturing capital. Fig. 3.3d

plots the evolution of the productive capital. The productive capital is higher if the

conversion factor is higher as can be seen in Fig. 3.3d.

Fig. 3.3e shows the drop in consumption after the event. But unlike Figs. 3.1c or 3.2c,

where there is a discontinuous drop in consumption, Fig. 3.3e shows that the

consumption level drops and remains constant after the event until there is a change in

the conversion factor. Lower the conversion factor lower is the consumption. It takes

approximately 3 units of time to reach to its post-event stable consumption level, which is

7% point less than its pre-event consumption level.

Fig. 3.3d plots the evolution of productive capital after 15% destruction due to a

catastrophic event. It takes approximately 3 units of time to reach its stable path. There is

a permanent increase in capital by 7%. Thus, increase in capital is compensated by the

decrease in consumption level. The evolution of the output of the economy is shown in

Fig. 3.3f. The output clearly portrays the changes that occur in the capital. Greater the

rise in conversion factor, lesser is the output beyond t=3, when changes in productivity

stabilize.

Growth (Fig. 3.3g) increases immediately after the loss in capital, consistent with the

standard growth model prediction that lesser the output, higher is the growth. The growth

increases till t=1, where the rise in conversion factor boosts growth rate. Greater the

conversion factor higher is the growth rate. When the change in conversion factor stops

(t=2), the growth rate falls. It then stabilizes at t=3. After all the changes in the

production function has stabilized, at t=3, greater increases in conversion factor result in

greater post event growth rates.


103

Fig 3.3c Changes in maturing capital

0.00

0.05

0.10

0.15

0.20

0.25

0 0.5 1 1.5 2 2.5 3 3.5

Time

Mat

urin

g ca

pita

l

Conversion rate 0%5%10%15%20%

Fig 3.3d Changes in productive capital

-0.20

-0.15

-0.10

-0.05

0.00

0.05

0 0.5 1 1.5 2 2.5 3 3.5

Time

Prod

uctiv

e ca

pita

l



104

Fig 3.3e Evolution of consumption

-0.70

-0.60

-0.50

-0.40

-0.30

-0.20

-0.10

0.00

0 1 2 3 4 5 6

Time

Con

sum

ptio

n


Fig 3.3f Evolution of output

1.09

1.1

1.11

1.12

1.13

1.14

1.15

1.16

1.17

1.18

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time

Out

put



105

3.3g Changes in economic growth over time

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

3.5%

4.0%

4.5%

0 1 2 3 4 5 6

Time

Econ

omic

gro

wth



106


economy, the impact of changes in the capital loss is studied. It is assumed that a1=0.1

(Fig. 3.4a). Changes in the capital share and the external aid are shown in Fig. 3.4b. a3

varies from 0.0 to 0.2 simulating various levels of capital loss. Greater the loss, lower is

the consumption (Fig. 3.4e) and capital (Fig. 3.4c and d). It takes around 3 units of time

for the economy to attain stable solution (which is the middle path in the five paths

shown). Fig. 3.4f shows the evolution of output. Greater the losses, lower is the output.

The output grows after the event. The output grows till it stabilizes around t=3, when all

the changes in productivity have stabilized.

Growth is plotted in Fig. 3.4g. The results show that greater the loss, higher the growth

rate, immediately after the event. However this changes soon after the changes in the

conversion factor. After t= 1.2, greater loss result in lesser growth rates. This behavior

continues till all the changes in productivity have stabilized and the conversion rate has

returned to its normal level. This concurs with the empirical evidence of post event

growth rates. The evidence indicates that losses are negatively associated with post event

growth rates, i.e. greater the loss, lower is the post event growth rate. Unlike the previous

model, the second model able to explain this empirical observation. The results point to

the importance of modeling two types of capital in an economy – the maturing and the

productive capital and the changes in the conversion process that typically follow after a

catastrophic event.

Fig. 3.4h plots the initial changes in consumption required for stable post event behavior.

The curve implies that the greater the loss, smaller should be the initial consumption

levels to ensure stability. Fig. 3.4h plots the changes the welfare due to the occurrence of

a catastrophe. The plot indicates that a greater loss in capital results in greater loss in

welfare. This change in welfare due to occurrence of a catastrophe can be used as a

measure of the secondary impact and is consistent with the results of the previous model.


107


0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

0 1 2 3 4 5

Time

Cap

ital s

hare

Fig. 3.4b Changes in aid and total factor productivity

-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

12%

0 1 2 3 4 5

Time

Cap

ital s

hare


108

Fig 3.4c Changes in maturing capital

-0.60

-0.40

-0.20

0.00

0.20

0.40

0.60

0.80

0 0.5 1 1.5 2 2.5 3 3.5

Time

Mat

urin

g ca

pita

lLoss ratio = 0.05%10%15%20%

Fig 3.4d Changes in productive capital

-0.40

-0.30

-0.20

-0.10

0.00

0.10

0.20

0.30

0.40

0.50

0 0.5 1 1.5 2 2.5 3 3.5

Time

Prod

uctiv

e ca

pita

l

Loss ratio = 0.0%5%10%15%20%

Fig 3.4e Evolution of consumption

-1.50

-1.00

-0.50

0.00

0.50

1.00

0 0.5 1 1.5 2 2.5 3 3.5

Time

Con

sum

ptio

n

Loss ratio = 0.0%5%10%15%20%


109

Fig 3.4f Evolution of output

1.04

1.06

1.08

1.1

1.12

1.14

1.16

1.18

0 0.5 1 1.5 2 2.5 3

Time

Out

put

Loss ratio = 0.0%5%10%15%20%

3.4g Changes in economic growth over time

-2%

-1%

0%

1%

2%

3%

4%

5%

0 0.5 1 1.5 2 2.5 3

Time

Econ

omic

gro

wth

Loss ratio = 0.0%5%10%15%20%

Fig. 3.4h Effect of loss intensity on initial consumption and welfare

-0.35

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0 0.05 0.1 0.15 0.2 0.25

loss ratio

WelfareInitialConsumption


110

3.4 Interacting Regions

The model in this section simulates the behavior of two interacting regions when

a catastrophe strikes one of the regions. Region 1 is affected by a catastrophe. Region 1

interacts with Region 2 by exporting and importing goods. After a catastrophe strikes

Region 1, its productive capacity decreases. Region 2 tries to mitigate the situation in

Region 1 by diverting some of its output for relief and reconstruction to Region 1. After a

period of time, due to the construction of new capital in Region 1, its productive capacity

may increase. As a result, Region 1 is able to export more to Region 2. This situation

describes some of the dynamics when the Northridge Earthquake struck Los Angeles

County. As will be clear from the data and simulation presented in a later chapter, the

economy of Los Angeles did better after the Northridge Earthquake. In cases where there

is no appreciable change in the productive capacity, the economy may reach its pre-event

output levels. The model presented herein explains why and under what condition

economies revive and often do better after a catastrophe.

In the following the subscript 1 will denote variables belonging to Region 1 and subscript

2 denotes variables belonging to Region 2. The representative agents in both the regions

will choose consumption paths c1(t) and c2(t) such that they maximize their utility:

∫∞

− +=0

21)(),(0 ))()((max)(21

dtcucuekV t

tctc

ρ (3.4.1)

subject to the following conditions:

111121111 )()( kckfakfak β−−+=& (3.4.2)

221221212 )()( kckfakfak β−−+=& (3.4.3)

The current value Hamiltonian is:

))()(())()(()()(

221221212

11112111121

kckfakfakckfakfacucuHβλβλ

−−++−−+++=

(3.4.4)

λ1(t) is the change in the maximum attainable value of the objective function discounted

to t, when Region 1’s economy acquires a single unit of its capital at time t. Similarly,


111

λ2(t) is the change in the maximum attainable value of the objective function discounted

to t, when Region 2’s economy acquires a single unit of its capital at time t.

Differentiating the Hamiltonian with respect to c1 and c2, the following first order

conditions result:

0)('00)('0 2211 21=−⇒==−⇒= λλ cuHandcuH cc (3.4.5)

Differentiating the relations in Eq.3.4.5 with respect to time we have:

222111 )(")(" ccuandccu &&&& == λλ (3.4.6)

The co-state equations are:

)('))(( 12121'

11111 kfakfa λβλρλλ −−−=& (3.4.7)

))('()( 22222'

12122 βλλρλλ −−−= kfakfa& (3.4.8)

Using the relations in Eq.3.4.6 in Eqs. 3.4.7 and 3.4.8, the following equations result:

)(')]([ 12111

21

'11

11 kfa

cckfacc −−−+= γ

γ

γβρ

γ& (3.4.9)

)(')]([ 21212

12

'22

22 kfa

cckfacc −−−+= γ

γ

γβρ

γ& (3.4.10)

The following perturbations are introduced to model the behavior of the economies after

a catastrophic event in Region 1.

))(1( 1 thεσσ +→ (3.4.11a)

))(1( 21212 thaa ε+→ (3.4.11b)

))(1( 22121 thaa ε−→ (3.4.11c)

)(3 thεββ +→ (3.4.11d)

Eq.3.4.11d models the fact that when a catastrophic event occurs in Region 1, there is a

change in its capital. This is followed by an increase in the amount that Region 1 imports

from Region 2, which is modeled by Eqs.3.4.11b-c. Reconstruction in Region 1 changes

the capital share coefficient of Region 1’s production function, which is modeled in Eq.

3.4.11a. Substituting the relations in Eqs.3.4.2-3 and Eqs.3.4.9-10, the following

perturbed system of equations result:


112

),())((),()),(())(1(),(),( 1212112))(1(

11111 εεβεεεεε εσ tkthtctkfthatskatk th +−−++= +&

(3.4.12)

),(),()),(())(1(),(),( 222122))(1(

12121 εβεεεεε εσ tktctkfthatskatk th −−−+= +& (3.4.13)

1))(1(11211

1

2

1))(1(11113

11

1

1

),())(1(),(),(

]),())(1()([),(),(

−+−

−+

++

+−++=

th

th

tkthsatctc

tkthsathtctc

εσγ

γ

εσ

εεσεγε

εεσεβργε

ε&

(3.4.14)

122121

2

1

12222

22

),())(1(),(),(

]),())(1([),(),(

−−

−

++

−−+=

σγ

γ

σ

εσεεγε

εεσβργε

ε

tksthatctc

tkthsatctc& (3.4.15)

Differentiating the above system (Eqs.3.4.12-15) with respect to the perturbation ε

around the steady state, the following equations result:

dJ +

=

)0,()0,()0,()0,(

)0,()0,()0,()0,(

2

1

2

1

2

1

2

1

tktktctc

tktktctc

ε

ε

ε

ε

ε

ε

ε

ε

&

&&

&

(3.4.16)

where the elements of the matrix J and vector d are given below:

)])()1()(([1*2

*1

2111*111

γγβργ c

caakfJ −+′−+= (3.4.17a)

γ))(( *2

*1*

12112 cckfaJ ′−= (3.4.17b)

])([)( 1*221

*111

*1

13 *1

*2 −+

′′−= γ

γ cccacakfJ (3.4.17c)

014 =J (3.4.17d)

γ))(( *2

*1*

21221 cckfaJ ′−= (3.4.18a)

)])()1()(([1*2

*1

1222*222

γγβργ c

caakfJ −+′−+= (3.4.18b)

023 =J (3.4.18c)

])([)( 1*112

*222

*2

24 *2

*1 −+

′′−= γ

γ cccacakfJ (3.4.18d)


113

131 −=J (3.4.19a)

032 =J (3.4.19b)

β−′= )( *1133 1

kfaJ (3.4.19c)

)( *1234 2

kfaJ ′= (3.4.19d)

041 =J (3.4.20a)

142 −=J (3.4.20b)

)( *2143 2

kfaJ ′= (3.4.20c)

β−′= )( *2244 2

kfaJ (3.4.20d)

The components of the vector d (Eq.3.4.16) are shown below.

))())((1()()( 1*221

*111

*1

*1

1

*1

1 *1

*2 −++

′−= γσ

γγ cccacakLogkfthcd (3.4.21a)

))()(()( 1*

112*2221

*2

2 *2

*1 −+

′= γ

γ cccacath

kfd (3.4.21b)

)()()()()()( *11

*111

*132

*2123 kLogthkfakththkfad σ+−= (3.4.21c)

)()()()()( 2*222

*11

*1214 thkfakLogthkfad −= σ (3.4.21d)

Performing the Laplace transform:

++++

−=

ΚΚ

−

42

31

22

11

1

2

1

2

1

)0,0()0,0()0,0()0,0(

)(

)()()()(

dkdkdcdc

s

sssCsC

ε

ε

ε

ε

ε

ε

ε

ε

JI

&

&

&

&

(3.4.22)


The following Cobb-Douglas from of production function is chosen: σskkf =)( (3.4.23)

where it is assumed that σ =0.25, ρ = 0.04. It is also assumed that γ = -0.5 and that time at

which the catastrophe occurs, τ = 0.5. Productivity changes due to the catastrophe are


114

handled by the function in Eq. 3.2.44 for changes in the capital share in the production

function, σ. The flow of external aid is modeled using:


A similar function is used to model the changes in rate at which maturing capital is

converted to productive factor:


Capital reduction at time τ due to the occurrence of a catastrophe is modeled using

Eq.2.46, where it is assumed that 5 percent of the capital stock is destroyed. The

simulation is done for various values of a1 = 0.0, 0.05, 0.1, 0,15, 0.2. The Mathematica©

program for simulating these equations is presented in electronic form in Appendix E.

The results are shown in Figs. 3.5a to 5j. Fig. 3.5a shows the discontinuous drop in consumption for Region 1 when the event

occurs. There is a drop of around 5% in consumption due to a 5% percent loss in capital

stock. It takes approximately 3 units of time to reach to its post-event stable consumption

level, which is 3% less than its pre-event consumption level. Fig.5b plots the evolution of

consumption in Region 2 after a catastrophic event has occurred in Region 1.

Consumption levels keep rising in Region 2 until it starts giving some of its output to

Region 1 for reconstruction. The consumption in Region 2 falls for the period when it

diverts its output (t=2) after which consumption rises again to reach its pre-event level at

t=3.

Fig. 3.5c plots the evolution of capital in Region 1 after a catastrophic event. It takes

approximately 3 units of time to reach to a stable level. The post event stable level of

capital is higher than the pre-event level. The increase in capital is compensated for by

the decrease in consumption level. The capital in Region 2 (Fig. 3.5d) reaches it pre-

event level at t= 4 after some changes. The phase portraits of the evolution of capital and

consumption in Regions 1 and 2 are shown in Figs. 3.5g and 5f, respectively. The

evolution of the output of the economy of Regions 1 and 2 are shown in Fig.5g and 5i,

respectively. The output in Region 1 stabilizes at t=3 above its prevent level. Growth rate


115

(Fig. 3.5h) of Region 1 rises sharply after the event and the drops only to be raised by

increase of flow of funds from Region 2. Greater the inflow, greater is the growth.

Fig 3.5a Evolution of consumption in the disaster region

-5.00%

-4.00%

-3.00%

-2.00%

-1.00%

0.00%

1.00%

0 1 2 3 4 5 6 7 8

Time

Con

sum

ptio

n in

dis

aste

r reg

ion

Increase in TFP of affected region: 0%5%10%15%20%

Fig 3.5b Evolution of consumption in the adjacent region

-6.00%

-5.00%

-4.00%

-3.00%

-2.00%

-1.00%

0.00%

1.00%

2.00%

3.00%

0 1 2 3 4 5 6 7 8

Time

Con

sum

ptio

n in

adj

acen

t reg

ion Increase in TFP of affected region: 0%

5%10%15%20%


116

Fig 3.5c Changes in capital in the affected region

-0.06

-0.04

-0.02

0.00

0.02

0.04

0.06

0.08

0.10

0 1 2 3 4 5 6 7 8

Time

Cap

ital i

n th

e af

fect

ed re

gion


Fig 3.5d Changes in capital of the unaffected region

-0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

0 1 2 3 4 5 6 7 8

Time

Cap

ital i

n th

e un

affe

ctec

d re

gion



117

Fig. 3.5e How does consumption vary with changes in capital of the affected region?

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

-0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1

Capital in disaster region

Con

sum

ptio

n in

dis

aste

r reg

ion


Fig. 3.5f How does consumption vary with changes in capital of the unaffected region?

-0.06

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

Capital in unaffected region

Con

sum

ptio

n in

una

ffec

ted

regi

on



118

Fig 3.5g Output in the disaster region

0.78

0.79

0.80

0.81

0.82

0.83

0.84

0.85

0 1 2 3 4 5 6 7 8

Time

Out

put i

n di

sast

er re

gion


Fig 3.5h Economic growth in the disaster region

-1.00%

-0.50%

0.00%

0.50%

1.00%

1.50%

2.00%

2.50%

3.00%

3.50%

0 1 2 3 4 5 6 7 8

Time

Gro

wth

in d

isas

ter r

egio

n



119

Fig 3.5i Output in the adjacent region

1.23

1.241.24

1.251.25

1.261.26

1.271.27

1.28

0 1 2 3 4 5 6 7 8

Time

Out

put i

n th

e ad

jace

nt re

gion Increase in TFP of affected region: 0%

5%10%15%20%

Fig 3.5j Economic growth in the adjacent region

-3.00%

-2.50%

-2.00%

-1.50%

-1.00%

-0.50%

0.00%

0.50%

1.00%

0 1 2 3 4 5 6 7 8

Time

Gro

wth

in th

e ad

jace

nt re

gion



120

Growth of Region 2’s economy (Fig. 3.5j) increases after the event but drops after

Region 2 starts diverting its funds to Region 1. There is a sharp change in growth rate

once Region 2 stops diverting its funds to Region 1 at t=2. The growth then falls down

and reaches a lower value than its pre-event level after t=3.


economy, the impact of changes in the capital loss is studied. It is assumed that there are

no changes to productivity and that the aid from region 2 is constant. Various levels of

capital loss are modeled by varying a3 from 0.0 to 0.2. The levels of consumption in

Region 1 (Fig. 3.6a) and capital (Fig. 3.6c) are lower greater the loss. It takes around 3

units of time for the economy to attain stable solution (which is the middle path in the

five paths shown). The levels of consumption in Region 2 (Fig. 3.6b) and capital (Fig.

3.6d) are higher, greater the loss in Region 1. It takes around 3 units of time for the

economy to attain stable solution (which is the middle path in the five paths shown). The

phase portraits of the evolution of capital and consumption in Regions 1 and 2 are shown

in Figs. 3.6e and 6f, respectively. The evolution of the output of the economy of Regions

1 and 2 are shown in Fig.6g and 6i, respectively. The output in Region 1 stabilizes at t=3

above its pre-event level. Growth rate (Fig. 3.6h) of Region 1 rises sharply after the event

and the drops only to be raised by increase of flow of funds from Region 2. Greater the

loss, greater is the growth. This does not change even after all the changes in productivity

have stabilized. Growth of Region 2’s economy (Fig. 3.6j) increases after the event but

drops after Region 2 starts diverting its funds to Region 1. There is a sharp change in

growth rate once Region 2 stops diverting its funds to Region 1 at t=2. The growth then

falls down and reaches a lower value than its pre-event level after t=3. Greater the loss in

Region 1 greater is the post-event growth rate. This is due to the fact that unlike model 2,

the effects of the efficiency of the reconstruction process has not been modeled for the

affected Region 1.

Fig. 3.6k plots the changes the welfare due to the occurrence of a catastrophe for Regions

1 and 2 respectively. The plot indicates that a greater loss in capital results in greater loss

in welfare for both the regions.


121

Fig 3.6a Evolution of consumption in the disaster region

-0.10

-0.08

-0.06

-0.04

-0.02

0.00

0.02

0.04

0 1 2 3 4 5 6 7 8

Time

Con

sum

ptio

n in

dis

aste

r reg

ion

Loss ratio = 0.05%10%15%20%

Fig 3.6b Evolution of consumption in the adjacent region

-0.05

-0.04

-0.03

-0.02

-0.01

0.00

0.01

0.02

0.03

0 1 2 3 4 5 6 7 8

Time

Con

sum

ptio

n in

adj

acen

t reg

ion Loss ratio = 0.0

5%10%15%20%


122

Fig 3.6c Changes in capital in the affected region

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0 1 2 3 4 5 6 7 8

Time

Cap

ital i

n th

e af

fect

ed re

gion

Loss ratio = 0.05%10%15%20%

Fig 3.6d Changes in capital of the unaffected region

-6.0%

-5.0%

-4.0%

-3.0%

-2.0%

-1.0%

0.0%

1.0%

2.0%

3.0%

4.0%

0 1 2 3 4 5 6 7 8

Time

Cap

ital i

n th

e un

affe

ctec

d re

gion

Loss ratio = 0.05%10%15%20%


123

Fig. 3.5e How does consumption vary with changes in capital of the affected region?

-0.1

-0.08

-0.06

-0.04

-0.02

0

0.02

0.04

-0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3

Capital in disaster region

Con

sum

ptio

n in

dis

aste

r reg

ion

Loss ratio = 0.05%10%15%20%

Fig. 3.5f How does consumption vary with changes in capital of the unaffected region?

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

-0.06 -0.04 -0.02 0 0.02 0.04

Capital in unaffected region

Con

sum

ptio

n in

una

ffec

ted

regi

on

Loss ratio = 0.05%10%15%20%


124

Fig 3.6g Output in the disaster region

0.70

0.75

0.80

0.85

0.90

0.95

0 1 2 3 4 5 6 7 8

Time

Out

put i

n di

sast

er re

gion

Loss ratio = 0.05%10%15%20%

Fig 3.6h Economic growth in the disaster region

-4.0%-3.0%-2.0%-1.0%0.0%1.0%2.0%3.0%4.0%5.0%6.0%7.0%

0 1 2 3 4 5 6 7 8

Time

Gro

wth

in d

isas

ter r

egio

n

Loss ratio = 0.05%10%15%20%


125

Fig 3.6i Output in the adjacent region

1.25

1.25

1.25

1.25

1.25

1.26

1.26

1.26

0 1 2 3 4 5 6 7 8

Time

Out

put i

n th

e ad

jace

nt re

gion

Loss ratio = 0.05%10%15%20%

Fig 3.6j Economic growth in the adjacent region

-1.4%-1.2%-1.0%-0.8%-0.6%-0.4%-0.2%0.0%0.2%0.4%0.6%0.8%

0 1 2 3 4 5 6 7 8

Time

Gro

wth

in th

e ad

jace

nt re

gion

Loss ratio = 0.05%10%15%20%


126

Fig3.6k Changes in welfare

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

00 0.05 0.1 0.15 0.2 0.25

Loss ratios

Pres

ent v

alue

of u

tility

Welfare in disaster regionWelfare in adjacent region


127

3.5 Conclusions

In the preceding sections three models simulating the behavior of an economy

after the occurrence of a catastrophic event were studied. Changes in capital due to a

catastrophe and the subsequent changes in productivity are modeled by perturbing a

dynamic model of the economy. The simulation results point to the importance of

modeling the efficiency of the reconstruction processes after an event. Classical models

of growth (as demonstrated in the first model) suggest that lower levels of capital result

in higher growth rates. This would imply that if a catastrophe strikes a region and

destroys substantial capital stock it would grow at a faster rate than a region that loses

smaller amounts of capital. However, empirical evidence, based on data from 43

countries in which catastrophes have occurred, strongly suggest that greater loss is

associated with smaller post-event growth rates.

The main contribution of this chapter is to show that, unless the process whereby

the maturing capital is converted to productive capital is modeled, the fact that post-event

growth rate is negatively correlated with the magnitude of loss cannot be explained. Pre-

event conditions are important in post-event recovery. Pre-event conditions determine the

efficiency of the processes whereby newly reconstructed capital becomes fully

productive. In particular, pre-event conditions determine the changes in productivity.

The third model simulates the interaction between two regions one of which is struck

by a catastrophe. Model results suggest that greater the loss more is the effect felt on the

unaffected region. This is supported by examining the evidence from three catastrophic

events – 1989 Loma Prieta Earthquake, 1992 Hurricane Andrew, and 1994 Northridge

Earthquake. Higher loss ratio in Dade County made its effect felt at the state level,

whereas the lower loss ratios in Los Angeles and the Loma Prieta affected counties

resulted in localized effects. Evidence regarding the personal income suggests that

growth rate after the events are typically higher. This is also borne out in the model

behavior results. Though the models do not go into simulating sector specific effects such


128

as impact on housing prices or government expenditures, by modeling consumption a

general equilibrium approach has been adopted that allows for qualitatively verifying the

behavior.

The models consistently demonstrate the importance of investment in the

disaster region. Coupled with efficient reconstruction, the models show that investment in

the disaster region can make the region more productive than before the event. Disasters

offer an opportunity to rebuild and convert vulnerable communities into robust ones.

Chapter Four: Empirical Studies

129

Chapter Four

An Empirical Study of the Macro-economic Effects of Catastrophes Triggered by Natural Events

4. INTRODUCTION

In this chapter we re-examine our understanding of the effects of catastrophes on

the economy based on empirical evidence. Questions addressed include the change or

absence thereof, in economic growth, consumption, saving, inflation, and real interest

rates. Data on these economic indicators are compiled for various countries for periods

immediately preceding and following the occurrence of a catastrophe. Data regarding

catastrophes such as the estimates of direct losses is also compiled. The regression

analysis employed suggests that catastrophes are negatively associated with all the

aforementioned economic indicators.

In order to study the effect of a catastrophe on an economy the factors that describe

socio-economic conditions prior to occurrence of the hazard event have to be identified.

The vulnerability of a society to natural hazards is the result of various on-going

economic, social, and political processes, as has been discussed in Chapter 2. For large

segments of the world's underdeveloped population, occurrence of a natural hazard may

worsen an already deteriorating or fragile situation. In such regions even a moderate

hazard, such as the 1985 Mexico earthquake, could trigger a catastrophe. Oliver-Smith

(1994) brings this out clearly in his analysis of the 1970 Peru Earthquake. He points out

that Peru's catastrophe was some 500 years in the making, rooted in the complex of

economic and political forces that structured development and the human-environment

relations. The earthquake and subsequent landslides was a trigger for a catastrophe

grounded in poverty, political oppression, and the subversion of previously sustainable

indigenous practices (Bolin and Stanford, 1998).

Socioeconomic conditions in a region are mainly as a result of the developmental

processes. The effect of a major catastrophe on the developmental process is complex,


130

especially for developing regions. Globally, economies are evolving ‘complex’ systems.

This complexity in the economic systems is the result of the historical geography, the

political economy, the increased interdependencies among various sectors and regions of

an economy facilitated by the quantum leaps in the communication technology, and the

rapid globalisation of trade. In order to study the effect of a catastrophe on an economy

the factors that describe socio-economic conditions prior to occurrence of the hazard

event have to be identified. General statements regarding the economic consequences of a

catastrophe can be made only when these complexities are appropriately modeled.

An overview of some studies, which partly address these questions, is presented in the

next section. In section 4.2 connections between the occurrence of a catastrophe and

ongoing development processes of an affected region are made. Section 4.3 describes the

data used for the present study. The general framework and the particular econometric

model used to estimate the effect of catastrophes are presented in the next section.

Various factors that affect the growth rate are then presented. Section 4.5 presents a

discussion of various factors that may be important in determining the post event

economic indicators. Results of regression analysis are discussed in Section 4.6.


131

4.1 PREVIOUS STUDIES

Studies on the effects of natural hazards on an economy have discussed direct and

indirect losses that result from such events (Development Technologies, 1992). Direct

losses are usually associated with direct physical damage and secondary effects, such as

damage caused by fire following an earthquake. Indirect damages relate to the effect on

flows of goods that will not be produced and services that will not be provided after a

catastrophe. They are measured in monetary terms. The impact of the catastrophe on

overall economic behavior, which has sometimes been termed as secondary effects, is

measured by changes in macro-economic variables. The work reported in this dissertation

focuses on secondary effects.

There are few studies on the macro-economic effects of catastrophes. They are based on

small data sets. Moreover, the conclusions are seemingly contradictory. Albala-Bertrand

(1993:163) argues "GDP normally does not fall after a disaster impact and if anything

tends to improve at least for a couple of post-disaster years." Albala-Bertrand's study

(1993) is based on a sample of catastrophes that occurred in the 1970's in mostly

developing countries. He uses three criteria for examining the effect of catastrophes on

economic growth, investment and sector outputs, public finance, and balance of

payments. The three criteria include: i) examining the change in the indicators according

to sign (positive meaning 'growth') and direction of change (up meaning 'acceleration'),

ii) the figures are averaged in per country terms for each period, and iii) comparison

between pre- and post-disaster averages. Limited by sample size, no other statistical

inferential procedures are used. The hypothesis he proposes is not validated since there

could be many factors that explain post-event economic behavior. For example, a country

might have experienced increased growth after an event because of reasons totally

unrelated to the occurrence of a catastrophe or due to efficient reconstruction policies.

However, this does not imply that a similar economy will sustain economic growth in the

absence of efficient reconstruction. Inferences from cross-country data are general only if

they are ‘normalized’ using control and environmental variables.


132

The World Disasters Report (1997) expresses an apparently opposite viewpoint. The

report states, “Caribbean disasters can be costly, especially as a proportion of GDP. The

impact on national economies has been significant: hurricanes between 1980 and 1988

effectively reversed the growth rates.” This statement is again based on a simple

comparison of average growth for the affected countries between 1980-88 and 1989-91

(Table 4.1). All the five countries are small islands, which makes it difficult to generalize

the result.

Taken together these studies produce ambiguous conclusions regarding the effect of

catastrophes on ongoing economic processes.

Friesema et al. (1979) is an early study to analyze the effect of disasters on the long-term

growth patterns of four cities - Conway, Galveston, Topeka, and Yuba City. Their null

hypothesis is that disasters had no significant effect on employment, small business

activity (number of gas stations and restaurants), retail sales, and public finance. They

examine a time series of the indicators for a time period ten years before and after an

event. They conclude that local economic behavior patterns, barring slight disruptions,

were scarcely interrupted by the disaster events considered. They also mention that their

results are not surprising since in all the four cases the basic capital stock remained, and

the production process continued. This makes their sample unrepresentative of post

catastrophic economic behavior.

Table 4.1: Disasters in the Caribbean can have a significant impact on GDP and growth (World Disasters Report, 1997)

Country Average growth rate GDP 1980-88

Average growth rate GDP 1989-91

Dominica 4.9 4.3 Montserrat 3.7 -4.4 St.Kitts/Nevis 6.0 4.9 Antigua/Barbuda 6.8 2.2 Jamaica 5.0 0.8


133

Wright et al. (1979) examine data for over 3100 counties in the US for effects of disasters

on growth trends of population and housing. Damage inflicted by the typical disaster in

their sample affected only a small proportion of structures, enterprises, and households of

typical counties. Based on regression studies they conclude that there are no significant

effects on growth trends in population and housing. However, these findings have been

questioned by the research of Yezer and Rabin (1987), who distinguish between

anticipated and unanticipated disasters. Their hypothesis is that “expected” disasters,

those occurring at a rate predicted by historical experience in a region, have no impact on

migration – such expectations have already been reflected in trend rate of migration. In

contrast, “unexpected disasters”, a spate in excess of those predicted by historical

experience, discourage migration. Empirical testing that explicitly distinguishes

“anticipated” from “unanticipated” supports the hypothesis.

The inferences from these studies cannot be generalized to effects of catastrophe in a

developing economy for several reasons. Firstly, the studies concentrate on regional

localized effects in a developed country. Secondly, the direct loss reported in the studies

is relatively small compared to the overall capital stock of the affected region. Finally,

they only examine changes in a subset of indicators that describe the social and economic

conditions of a region.


134

4.2 CATASTROPHES AND ONGOING DEVELOPMENT PROCESSES

Losses from a catastrophe may be readily absorbed by a developed economy. To

cite an example, the Northridge earthquake occurred in a state with a Gross Regional

Product ranked 6th largest in the world. A US $30 billion direct loss due to the earthquake

manifested itself as a minor perturbation. This contrasts with the devastating Third World

disasters such as the 1976 Guatemala earthquake or the 1985 Mexico City earthquake. In

both cases, the catastrophes produced national crises with effects well beyond the

immediate physical impacts.

For a developing economy, like Bangladesh, direct losses from a catastrophe, which are

comparable to the Gross Domestic Product (GDP) might divert scarce resources from

development plans to reconstruction. Almost half of the 1988/89 Bangladesh's national

development budget was diverted to pay for ad-hoc relief and rehabilitation programs

(Brammer, 1990) after the 1988 flood. Development plans may include improving health

care, education, food supply, and institutions for crisis management. As Bates and

Peacock (1993) point out catastrophes "intervene in the development process as it

pertains to other important adaptive problems, and they redirect, deflect, retard, and on

rare occasions accelerate the development process."

The deep indebtedness of many Third World countries has made the cost of

reconstruction and the transition from rehabilitation to development unattainable. To see

how foreign debt burden can adversely affect the loss that a country suffers, take the case

of Jamaica struck by Hurricane Gilbert in 1988 (Blaikie, et al. 1994). Prior to the

Hurricane Gilbert, part of Jamaica's debt burden was in part due loans used to pay for

damages from previous hurricane. Jamaica introduced a structural adjustment program

that typically involved cuts in public spending. Services such as education, health, and

sanitation were reduced. Government programs to introduce preparedness or mitigation

measures were also cut as result of economic constraints. These decisions greatly reduced

the ability of the community to recover from the effects of a major hazard like Hurricane

Gilbert.


135

Foreign debt also forced the government to intervene in the financial sector that resulted

in an increase of interest rates to over 20% and home mortgage rates ran between 14-

25%. Government forced rent control and import duty on construction materials. This

resulted in a rapid decline in new construction and other maintenance activity. The

quality of new construction also declined, since contractors tried to maximize profit by

using unsafe practices. This may have been partly responsible for the huge magnitude of

losses observed.

Delica (1993) brings out the relation between disasters and economic growth based on

her study of the natural hazards affecting Philippines. She argues that disasters have

practically negated the real economic growth achieved during the administration of

Carazon Aquino. From 1986 to 1991, damage to infrastructure, property, agriculture, and

industry from disasters were enormous, averaging about 2% of the GNP. Using simple

arithmetic, she argues that with an annual population growth of 2.3%, the economy needs

greater than 4.3% annual growth simply to maintain per capita income levels. But the

economy had only about 4% average annual growth, with the result vulnerability to

disasters has increased rather than decreased. This is because Philippines’ foreign debt

obligations have increased, from $26 billion in 1985 to $29 billion in 1992. The

government's spending on relief and rehabilitation has been tightly controlled and

increasingly dependent on external sources. Government's development strategy puts a

premium on export-orientation and attraction of foreign investment. This is at the

expense of ecological sustainability and environmental protection. Out of the 54% forest

cover required for a stable ecosystem only 20% remains as a result of deforestation. This

in turn increases the severity of floods and landslides.

Many poor countries try to solve their debt problems by adopting national policies

favoring raw material export. This typically results in land degradation since new land is

cleared for ranching and commercial cropping. Land degradation increases vulnerability,

which in turn increases the potential for catastrophic losses.


136

Long-term development projects may be adversely affected by diversion of resources to

help an affected community rebuild. Twigg (1998) reports that the World Bank diverted

some $2 billion of existing loans between the 1987 and 1988 financial years to fund

reconstruction and rehabilitation after catastrophes triggered by natural events.

Catastrophes reveal the robustness or vulnerability of a country's socioeconomic

conditions. Various indicators can be used to quantitatively measure robustness or

vulnerability. The importance of these parameters, which are perhaps ignored in less

turbulent times, is revealed, tragically, only after a catastrophic event. A catastrophe can

unmask social and economic inequalities that come to the fore in the distribution of relief

aid. Catastrophes usually result in worsening the pre-event economic inequalities. It is

important to identify the factors that can be associated with vulnerability that explains the

wide variety of post event economic behavior. For example, we could examine the role of

infrastructure in changes in post-event economic growth. As has been pointed out by

Hewitt (1983), catastrophes are shaped and structured by economic, social, political and

cultural 'practices' and processes that existed prior to the occurrence of a physical event.

Indicators that describe some of these initial practices and processes need to be identified.

Whether there exists empirical evidence to support the hypothesis that ongoing

socioeconomic processes determine the post event economic behavior will be examined

in this chapter.

4.2.1 Change in Indicators Due to Catastrophes

In the past decade there has been an explosion of empirical studies of growth and

development. Efforts have been made to account for differences in growth rates between

various countries using indicators of education, health, infrastructure, institutions, and

political freedom. Results from these studies will be used to identify variables that can

make cross-country comparisons of changes in macro-economic indicators possible. The

parameters will act to control some of the variability across countries. Any effect due to

catastrophes on macro-economy can be detected only after the control variables

explained variability from other sources.


137

As has been mentioned previously, there is an intimate relation between ongoing

development processes and the occurrence of a catastrophe. The various parameters,

which are associated with development such as education, infrastructure, and health, are

hypothesized as measures of a community's robustness (or pessimistically, vulnerability)

to a catastrophe. A combination of these parameters can be used to assess a community's

robustness. It is reasonable to expect that a robust community's development or growth

should not be adversely affected by occurrence of a catastrophe. The ongoing dynamics

of the developmental processes are capable of absorbing the effects of catastrophe.

Conversely, a society is weak if its development process is adversely affected by the

occurrence of catastrophe triggered by a natural event. The present work relies on

previous studies on the determinants of growth for choosing parameters that are

associated with the development process.

One important indicator of development of a country is its economic growth rate. The

following is a summary of some of the parameters that have been shown to be

determinants of growth. A percentage point in economic growth is associated with the

following:

• Increase of 1.2 years in average schooling of labor force

• An increase in secondary enrollment of 40 percentage points

• A reduction of 28 percentage points in the share of central bank in total credit

• An increase of 50 percentage points in financial depth (M2/GDP)

• An increase of 1.7% of GDP in public investment in transport and communication

• A fall in inflation of 26 percentage points

• A reduction in the government budget deficit of 4.3 percentage points of GDP

• An increase in (exports + imports)/GDP of 40 percentage points

• A fall in government consumption/GDP of 8 percentage points

• An increase in foreign direct investment/GDP of 1.25 percentage points.

(Barro 1991, Barro and Lee 1993, King and Levine 1993, Easterly and Rebelo 1993,

Fisher 1993, Easterly and Levine 1997, Easterly, Loayza, and Monteil 1997, Borensztein,

De Gregorio, and Lee 1994).


138

These inferences are used to identify the variables that can be controlled when making a

cross-country comparison of post-event behavior of the economic growth.


139

4.3 GENERAL FRAMEWORK AND ECONOMETRIC MODEL

The general framework to be used in empirical studies reported here will be

developed in this section. In Chapter 3, theoretical models simulated the occurrence of a

catastrophe as a perturbation of the ‘normal’ economic processes. A catastrophe was

modeled as a reduction of capital and subsequent changes in productivity of the affected

region. The economy was assumed to be initially in its steady state. Inspection of

Eq.3.3.15 reveals that growth of capital due to the catastrophe depends on the steady state

(the Jacobian term) and the perturbations. Growth of the economy in turn depends on the

changes in capital stock. Hence, the following relation is used to estimate the effect of a

catastrophe on the post-event growth rates:

growthwith hazard = f(damage, productivity-changes; y*) (4.1)

y* is the long-run steady-state level of per capita output and depends on the steady state

levels of capital stock, as shown in Eq. 3.2.9. y* depends on an array of choice and

environmental variables. The private sector’s choices include saving rates, labor supply,

and fertility rates, each of which depends on preferences and costs. The government’s

choices involve spending in various categories, tax rates, the extent of distortions of

markets and business decisions, maintenance of rule of law and property rights, and the

degree of political freedom. Also relevant for an open economy is the terms-of-trade,

typically given to a small country by external conditions. A cross-country empirical

analysis requires conditioning on the determinants of the steady states. Also, the pre-

event conditions to a large extent determine the post event productivity. These

determinants or the country specific factors, along with their relation to catastrophes, are

presented in Section 4.4. It is assumed that the country specific factors are invariant over

the period of interest – five years. Data for these factors are typically available as

constants over five- to ten-year periods.

Damage, in general, depends upon the intensity of the hazard and the vulnerability.

Vulnerability is the susceptibility of the exposed constructed facilities, economic and


140

social structures of a region to be affected given a specified level of hazard. As discussed

in Chapter 2, vulnerability is intimately related to ongoing socio-economic processes.

Damage may be expressed as:

damage = h(hazard, vulnerability) (4.2)

It should be mentioned here the relation Eq.4.2 is expected to be highly non-linear. Even

for relatively simple structures such as single-family dwellings, the damage curves –

which relate the hazard intensity to the damage level (RMS, 1996) – are non-linear. Data

regarding the loss of capital and the changes in productivity are hard to come by. Hence

the loss of capital is modeled by the direct losses recorded after the event.

4.3.1 Approximation

The first step to estimate the model expressed in the relations (Eqs.4.1-2) above is

to use an approximate linear relation. Consequently, the relation in Eq.4.1 is

approximated by:

growthwith hazard = α1 + β1E + β2Damage + β3Hazard_type + ε1 …(4.3)

ε1 is an unobserved disturbance term. The indicators for Damage are the direct-loss to

GDP ratio and the percentage of population affected. E is a vector of time-invariant

country specific indicators of the economy that are considered as determinants of

economic growth. The vector E contains indicators from each of the following categories

of determinants of growth - Economic conditions, Individual Rights and Institutions,

Education, Health, Transport and Communications, Inequality across income and gender.

In particular the following indicators are used: Inflation variability, Average pre event

decade growth, SD of pre event decade growth, annual money growth, black market

premium, political rights, civil liberties, bureaucratic quality, government enterprises,

percent “no schooling” in population, daily protein or calorie intake, life expectancy at

age zero, radios per capita, and TVs per capita. Hazard-type is a dummy variable to

account for the type of hazard – earthquake, hurricane, or drought.


141

It should be mentioned here that Damage as such would depend on factors in the vector

E. It is implicitly assumed that the indicators for damage are not correlated with the

factors in E. This may be a strong assumption if the measure of loss is in terms of

destroyed productive capital stock and E includes factors such as capital stock per

worker. Indicators chosen in E are such that they are only indirectly related to direct loss

term. Therefore, the assumption that E and damage are not significantly correlated is

reasonable. It is also assumed that the errors in measurement/estimation of damage are

not correlated with the error term ε1. The reduced form given in Eq.4.3 is estimated.

The results presented in Fig. 3.4h indicate that loss is negatively correlated with the post-

event growth rate. The hypothesis to be tested is that the coefficients β2 in Eq.4.3 are

statistically significant and negative.

Similar models for other economic indicators are estimated where the dependent variable

is chosen to be the post event budget deficit, external debt, resource balance, inflation,

interest rates, or consumer price index. Again, the hypotheses to be tested are that the

coefficients β’2 in Eq. 4.3 are statistically significant. 4.3.2 Summary Statistics and Discussion of the Sample

4.3.2.1 Economic growth

As a first step, the growth rates between two adjacent years are compared, that is,

the growth rate during the event year is compared with the growth rate immediately

preceding year. Both mean and median of the pre-event annual percentage growth are

greater than their post-event counterparts (Table 4.2). Presumably catastrophic events

also induce greater variance for the growth, as evidenced by comparing the pre- and post-

event variances in the growth (Table 4.2). Distribution of the pre- and post- event growth

rates are shown in Fig.4.1.


142

Table 4.2 Summary statistics for short-term growth Pre event Post event

Mean 3.96 3.29Standard Error 0.30 0.33Median 3.80 3.18Standard Deviation 3.70 4.06Sample Variance 13.67 16.46Kurtosis 2.01 2.61Skewness -0.16 -0.61Range 23.36 26.60Minimum -9.10 -12.57Maximum 14.27 14.03Sum 605.25 503.99Count 153 153Confidence Level(95.0%) 0.59 0.65

Fig 4.1 Event year growth is clearly lower than the pre-event year growth

0%10%20%30%40%50%60%70%80%90%

100%

0 2 4 6 8 10

GDP annual growth (%)

Perc

entil

e 1yrBeforeEventYear

Fig 4.2 Average pre- and post-event growths

0%10%20%30%40%50%60%70%80%90%

100%

0 2 4 6 8 10


Perc

entil

e Avg3yrsAfterAvg3yrsBefore


143

It is apparent from Fig. 4.1 that the distribution for growth in the event year is shifted to

the left relative to growth one year before the event.

Table 4.3 summarizes the statistics for pre- and post- event average growth. Here again

average post-event growth rate is smaller than pre-event growth rate. But sample variance

of the average post-event growth is smaller than average pre-event growth, indicating

perhaps that the effects of the events are reducing.

Table 4.3 Summary statistics for average growth Average 3 years before

Average 3 years after

Mean 3.83 3.55Standard Error 0.27 0.24Median 3.48 3.61Standard Deviation 3.32 3.00Sample Variance 11.03 9.02Kurtosis 1.89 3.35Skewness 0.15 -0.44Range 22.96 22.66Minimum -9.42 -10.20Maximum 13.54 12.46Sum 586.68 542.90Count 153 153

It is apparent from Fig. 4.2 that the average post event growth is shifted to the left relative

to average pre event growth rates, though in this case the effect is not as pronounced for

growths less than 5%.

4.4.2.2 Effect on consumption, investment, government expenditure, net exports and income

The main components of the GDP are the consumption, investment, government

expenditure and net exports. Using the latest Penn World Table data (2002), the effect of

catastrophes on each of these macroeconomic indicators is investigated. As a first step,

each of these variables is graphed with the loss-GDP ratios. These graphs are shown in

Fig. 4.3 to 4.8.


144

Fig. 4.3 Effect of catastrophes on consumption

40

50

60

70

80

90

100

0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%

Annual loss as a % of GDP

Post

eve

nt c

onsu

mpt

ion

(% o

f GD

P)

Fig. 4.4 Greater losses are associated with larger amount of government spending

05

1015202530354045

0.0% 0.1% 1.0% 10.0% 100.0% 1000.0%


Post

-eve

nt in

vest

men

t as

a %

of

GD

P


0

5

10

15

20

25

30

35

40

0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%


Post

eve

nt g

over

nmen

t sp

endi

ng a

s a

% o

f GD

P


145

Fig. 4.6 Greater losses are associated with higher openness

0

20

40

60

80

100

120

140

0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%


Post

eve

nt o

penn

ess

as a

% o

f G

DP

Fig. 4.7 Larger losses are associated with smaller post event savings

-40

-30-20

-100

1020

3040

50

0.0% 0.1% 1.0% 10.0% 100.0% 1000.0%


Post

eve

nt s

avin

gs a

s a

% o

f G

DP

Fig. 4.8 Greater losses are associated with lower post event GDP per capita

100

1,000

10,000

100,000

0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%


Post

-eve

nt re

al G

DP

per

equi

vale

nt a

dult


146

These (Fig. 4.3 to 4.8) depict important observed regularities between magnitude of

losses and post-event macroeconomic variables. For example, Fig. 4.5 depicts the

observation that higher losses are associated with higher post-event governmental

spending as a fraction of the GDP. Fig. 4.8 establishes a clear negative association

between loss magnitude and post-event GDP per capita. Regressions in the later sections

are performed to determine the robustness of these associations by accounting for country

specific factors.

Other variables are also examined. In particular the effect of losses on inflation and real

interest rates are presented in Fig. 4.9 and 4.10, respectively.

The next section discusses the primary and control variables that are used in the

estimation. An overview of linear regression analysis for the estimation of Eq.4.2 along

with the model adequacy checking is presented in Section 4.6. Following that,

econometric evidence associating changes in economic indicators with magnitude of loss,

the percentage affected, and the type of catastrophe is presented.

Fig. 4.9 Larger losses are associated with higher inflation

0.1

1.0

10.0

100.0

1000.0

0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%

Log(Loss/GDP)

Post

Eve

nt In

flatio

n


147

Fig. 4.10 Greater loss ratios are associated with higher post-event real interest rates

-10

-5

0

5

10

15

20

0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%


Post

eve

nt re

al in

tere

st ra

te


148

4.4 EFFECT ON THE ECONOMIC GROWTH

Relating the magnitude of a catastrophe with a change in the growth of an

economy is very complex since there are many factors that determine the economic

growth (Barro, 1997). Recent research in the determinants of cross-country economic

growth has revealed much regularity. Investment in physical capital, educational

attainment of the population, stable macro-economic policies, open trade regimes, better

developed financial markets are important factors exerting positive effect on growth

(Barro and Sala-i-Martin, 1995). There are several other factors that retard growth -

population growth, political instability, budget deficits, shocks resulting from terms of

trade changes, internal strife, and wars (Rodrik, 1998), policy distortions, government

consumption, and low bureaucratic quality (Commander, et al, 1997). In the following

sections we describe some factors that may explain the variety of observed changes in

ongoing economic processes after a catastrophe. This discussion is similar to the

discussion in Chapter 2 regarding the factors that determine the vulnerability to natural

hazards. The important difference here is that these factors are explained here as factors

that may contribute towards the recovery of a community after a catastrophic event.

4.4.1 Primary variables

Three primary variables are used as indicators of the catastrophe. They are: i) the direct

physical loss, ii) the percentage of population affected, and iii) the type of natural hazard.

4.4.1.1 Direct physical loss

One of the important variables that characterize a catastrophe is the resulting

direct loss. Direct damages include all damage to fixed assets (including property),

capital and inventories of finished and semi-finished goods, and business interruption

resulting from a catastrophe (HAZUS, 1997). Estimation of the macro-economic effects

involves a comparison of economic behavior with and without the change in a


149

community's assets. The direct loss is one measure of the change in community assets

after a catastrophe.

Comparing direct loss across countries necessitates an approach based on purchasing

power parity (PPP). Converting the losses into a common currency, for e.g. the US dollar,

through the use of official exchange rates often misleads cross-country comparisons of

the losses. These nominal exchange rates do not reflect the relative purchasing power of

different currencies, and thus errors are introduced into the comparisons. Using PPP is

one way to obtain a correct measure of losses. In countries where the domestic prices are

low, the losses based on PPP will be higher than that obtained from official exchange

rates. For the purposes of this study we use ratio of loss (in current US dollars) to the

GDP (in current US dollars) as a measure of direct loss. Using a ratio makes comparison

of loss across countries valid, since PPP or exchange rates that appear both in numerator

and denominator of the ratio cancel out. As mentioned in the introduction, the loss to

GDP ratio does not exhibit any trend over the time period of the sample and hence is a

good indicator of catastrophes. This is important since the present study is based on

events during the last three decades. Comparison is only possible by using the annual

economic loss as a proportion of the total income (GDP).

4.4.1.2 Percentage affected

In a developing economy, where the majority are poor the number of people

affected is often a better indicator of the severity of a catastrophe than direct loss. The

number of people affected depends on the vulnerabilities of various groups that are

resident in the affected area. The vulnerability of groups in turn depends on the manner in

which assets and income are distributed between different social groups. Post event

recovery depends on the way resources are allocated and here too discrimination may

occur based on pre-existing conditions of inequality based on gender, ethnicity, and race.

It is these vulnerable sections of society that suffer most from catastrophes affecting their

lives, their settlements, and their livelihoods.


150

Blaikie et al. (1994) point out that in many parts of the world each household's bundle of

property and assets and economic connections with others may be lost, enhanced,

disrupted, or reinforced in a number of ways due to hazards. The impact of the hazards

operate under the influence of rules and structures derived from existing social and

economic system, but are modified by the distinct characteristics of a particular hazard

and patterns of vulnerability.

4.4.1.3 Type of hazard

Different types of disaster have varying direct and therefore indirect and secondary

impacts. Given a vulnerable habitat, the damage pattern depends on type and intensity of

the physical event. For example, droughts ruin crops and forests but cause relatively little

damage to infrastructure. As a result productivity may remain the same after the event. In

the case of droughts, if the country has surplus of domestic food production, drought can

be managed. For example, one year after the 1982 Australian drought the country's

economy was back to 'normal'. But in countries with little surplus, the effects are more

tangible. Countries whose GDP is mainly represented by the rural economies are

especially vulnerable to droughts. Droughts cause major production losses. If the net farm

income falls during a drought in a farm based economy, it my cause a decline in the

overall output.

In contrast, earthquakes cause relatively little damage to standing crops, other than

localized losses resulting from landslides. But an earthquake can damage buildings and

underground infrastructure. A hurricane may cause extensive crop damage as well as

damage to structures. Reconstruction may result in changes to the productivity due to the

destruction and subsequent construction of new capital. Such changes in productivity

were modeled in Chapter 3. It is important to find out whether the type of disaster affects

the post-event growth rates.

Location and climate have large effects on income levels and income growth, through

their effects on transport costs, disease burdens, and agricultural productivity, among


151

other channels. Major natural hazards that occur frequently in some parts of the globe

have definite effects on income levels and growth. Some countries may therefore be at a

geographical disadvantage due to being situated in hazard prone area.

4.4.2 Control variables

Previous studies do not explicitly spell out the explanatory variables that may be

related to the post-event economic growth rate. Furthermore, there is a lack of theoretical

analytical models describing the phenomena, which has been addressed in Chapter 3.

Theoretical models and simulations presented in Chapter 3 point to the importance for

modeling the post-event productivity changes. Changes in the productivity are reflected

in the post-event evolution of consumption, output, and growth. Based on a wealth of

studies conducted in the field of economic growth (mentioned in Section 4.2.1), variables

that may be important in determining the post-event productivity are discussed in the

following sections. These include indicators for describing pre-event economic

conditions, health, poverty and inequality, government, infrastructure, education, and

trade.

4.4.2.1 Pre-existing Economic Conditions

If a nation has a stable macro-economy with a steady growth, it would be

relatively easier to detect any fluctuations resulting from a catastrophe. Pre-event decade

mean and standard deviation of the annual percentage growth rates are included as

control variables, as indicators of past performance of a nation's macro-economy. Barro

(1995) finds that higher inflation variability goes along with a lower rate of economic

growth. Monetary institutions and policies that lead to substantial variations in the

general level of prices create uncertainty and undermine the efficacy of money. In the

event of a catastrophe, it is more likely in nations with high inflationary susceptibility

that the prices will go out of control. Inflationary pressures will have a negative effect on

the productivity. An indicator for standard deviation of the annual inflation rate during

the last five years is included as a control variable (Gwartney and Lawson, 1997).


152

Another indicator of monetary stability that is included is the average annual growth rate

of the money supply during the last five years minus the potential growth rate of the GDP

(Gwartney and Lawson, 1997).

4.4.2.2 Health

Health problems are particularly highlighted in studies of floods on the West

Coast of South America brought about by El Nino in 1982-83. Blaikie et al. (1994),

quoting from a study of government health centers in north Peru, report that there was an

almost two-fold increase in number of deaths as result of disease and illness due to

epidemics following floods. People's basic health and nutritional status relates strongly to

their ability to survive disruptions of their livelihood systems. This status is important for

their resilience in the face of external shock. For most people living on a subsistence diet

and without proper access to health care, even a mild epidemic after a catastrophe may

prove fatal. The pre-event socioeconomic processes, to a large extent, determine the pre-

event health conditions of the community which in-turn determines the percentage of

people affected by a catastrophe. The post event reconstruction depends on an adequate

supply of labor immediately after the event. If the majority of population is affected by a

catastrophe for health reasons, there may be inadequate supply of labor resulting in

adverse changes in post-event productivity.

Various indicators are used to summarize the 'health' of a community. These include:

i) Life expectancy at age zero,

ii) Number of hospital beds per thousand, indicating the accessibility of health services

after a catastrophe, and

iii) The daily calorie and protein intakes.

4.4.2.3 Poverty and Inequality

The burden of poverty is spread unevenly - among the regions of the developing

world, among countries within those regions, and among localities within those countries


153

(Meier, 1995, Ray, 1998). Alexander (1998) cites the example of Philippines and

compares it with Japan. Both the countries have similar risk profiles as far as occurrence

of types physical hazards are concerned. But Philippines has a GNP that is 2.75% of

Japanese and 49% of Philippines population lives below poverty line. This necessitates

Philippines to bear a heavier burden from losses it experiences from calamitous events.

Within regions and countries, the poor are often concentrated in vulnerable places: in

rural areas with high population densities, such as the Indo-Gangetic plain and the Island

of Java, Indonesia. Often the problems of poverty, population, and the environment are

intertwined: earlier patterns of development and pressure of rapidly expanding

populations mean that many of the poor are forced to live in highly vulnerable regions.

As Blaikie et al. (1994) point out, in Manila (Philippines) the inhabitants of squatter

settlements constitute 35% of the population vulnerable to coastal flooding, and Bogota

(Colombia) has 60% of population living on landslide prone steep slopes. Even in urban

areas, if there are no adequate measures to systematically maintain buildings, potential

losses may be high. For example in the 1985 Mexico earthquake, the decaying inner city

tenements were severely affected.

Rural-urban migration leads to the erosion of local knowledge and institutions required

for coping in the aftermath of a disaster. The loss of younger people, especially working

age males and those with skills which are marketable in the cities may alter the type of

building structures that can be constructed to something less safe than previously.

Obviously this results in greater number of people being affected by the catastrophe.

Certain groups within a community are more vulnerable. Women, children, elderly,

ethnic groups, and minorities suffer disproportionately as a result of catastrophe as has

been reported by Peacock et al. (1997) after Hurricane Andrew. Inequality is a crucial

factor in the ability of an affected community to recover after the occurrence of a

catastrophe. A more unequal society will result in a more unequal distribution of effects -

the poorest in the affected society bearing the brunt of the catastrophe. An inefficient

bureaucracy will allow the inequality to deepen by concentrating the relief in the already


154

affluent people of the community. It has already been demonstrated by various macro-

economists (Barro 1995, Easterly, 1997) that higher the inequality slower is the economic

growth. Thus one of the effects of a catastrophe, given an inefficient bureaucracy, is to

indirectly retard growth by deepening inequalities. On the other hand the government can

view the occurrence of a catastrophe as an opportunity for initiating various programs to

boost economic growth. More efficient and modernized infrastructure may be constructed

replacing damaged structures increasing productivity, which acts as a catalyst for

economic growth of the affected region.

Indicators used to summarize the 'poverty' include the percentage of people living on less

than $1 a day (PPP 1981-95) (World Bank, 1997). The daily calorie intake is also an

indicator of poverty, though controversial. The decade average for Gini coefficient is

used as an indicator for inequality (Easterly and Levine, 1997). The ratio of the share of

the top twenty percent in the income distribution to the first quintile is also used as an

indicator of inequality (Easterly and Levine, 1997). Gender bias is represented using the

ratio of female to male average schooling years.

4.4.2.4 Government, Bureaucracy, and Institutions

Whether a poor country recovers quickly from a catastrophe depends, among

other factors, on its government. If the government has effectively implemented the

policies that make the country's development potential realizable, then a catastrophe will

be absorbed without much negative impact. But in many poor countries, the political

foundations for developmental efforts are not yet firm. Political instability,

undifferentiated and diffuse political structures, and inefficient governments are still too

prevalent (World Bank, 1997).

Commander et al. (1997) look at factors explaining the size of government and the

consequences of government for income growth and other measures of well-being, such

as infant mortality and life expectancy. They present partial evidence for the view that

governments use consumption to buffer external risk, particularly in low-income


155

countries. With respect to the consequences for growth, they find a robust negative

association with government consumption and with an index of policy distortions and a

positive relationship with quality of bureaucracy. They also report that social sector

spending can exert a positive influence on infant mortality and life expectancy.

Primarily its bureaucracy (Knack and Keefer 1995 and Mauro 1993) gives an explicit

evaluation of the quality of government. This evaluation is put together from a set of

responses by foreign investors that focus on the extent of red tape involved in any

transaction, the regulatory environment and the degree of autonomy from political

pressure. These responses provide us with a composite index of the quality of

government bureaucracy or its capability. Mauro (1993) finds a strong relationship

between per capita income and average indices of red tape, inefficient judiciary, and

corruption. Clague, Keefer, Knack, and Olson (1996) likewise establish a relationship

between high per capita income and high quality institutions - freedom from

expropriation, freedom from contract repudiation, freedom from corruption, and rule of

law. After a catastrophe has occurred, it is the efficiency of the government bureaucracy,

which partly determines the efficiency of the processes that determine the post event

productivity. As Oliver-Smith (1994) points out, the assistance after the 1970 Peru

earthquake never reached the survivors because of the 'Byzantine bureaucratic design and

a bewildering division of responsibilities' of the principal agency in charge of relief and

reconstruction. Keefer and Knack (1997) find a strong association between per capita

income and trust between individuals in a society. Trust is important for post event

behavior.

Rodrik (1999) presents econometric evidence from countries that experienced the

sharpest drops in growth after 1975 were those with divided societies and with weak

institutions of conflict management. He contends that 'social conflicts and their

management - whether successful or not - played a key role in transmitting the external

shocks on to economic performance.' The strength of crisis management institutions

determines the recovery process of an affected community. Studies at community level

(e.g. Oliver-Smith 1990, and Bolin 1982) highlight the major impediments to the


156

community recovery process even when they have received aid. Aid is not effective for

the following reasons: i) local disaster management staff are unprepared to deal with aid

recipients, ii) aid does not meet the needs of the poor, iii) outside donor programs exclude

local involvement, and iv) poorly coordinated and conflicting demands from national

government agencies. Many national governments have begun to initiate programs that

assist their local jurisdictions to prepare recovery and development plans (Kreimer and

Munasinghe, 1991).

Political will and respect for human rights are important factors for the successful

implementation of such plans. Without strong political will and freedom of expression in

a country, methodologies devised by a vulnerable community for coping after a

catastrophic event will not receive necessary impetus. As a result development processes

may suffer. On the other hand, if the most vulnerable in the society are empowered

through grass-roots organizations and non-governmental organizations there would be

greater equity in distribution of relief aid thus sustaining the development process. In

Central America, the importance of local organization has been demonstrated in sudden

onset disasters (earthquakes) and in terms of events with a longer preparatory phase

(hurricanes and flooding).

An indicator of government size is the general government consumption as a percentage

of GDP. Human rights index, bureaucratic quality, rule of law, freedom from corruption,

and repression of civil liberties are all indicators for individual rights and democracy

(Easterly and Levine1997).

4.4.2.5 Infrastructure

Typically, a catastrophe results in major disruption of infrastructure facilities.

Given the fact that infrastructure facilities are poorly maintained in developing countries,

the extent of damage is severe even for a moderate hazard. Moreover, the emphasis shifts

to re-building the community in the immediate aftermath of a catastrophe. In its report on

Infrastructure and Development (1994), the World Bank points out that 'when times are


157

hard, capital spending on infrastructure is the first item to go and operations and

maintenance are often close behind. Despite the long-term economic costs of slashing

infrastructure spending, governments find it less politically costly than reducing public

employment or wages.' After a catastrophe, emphasis shifts to providing immediate relief

to the victims and as a result capital expenditures are cut with infrastructure capital

spending often taking the biggest reduction. The 1970 Peru earthquake completely

incapacitated the fragile infrastructure of roads, railways, airports and communications

(Oliver-Smith, 1994). The rails of one railway system that had been twisted beyond

repair have not been replaced 20 years after the event. Destruction of roadways or other

infrastructure may cause impediments to relief and rescue operations and lead to business

interruption. After the 1995 Kobe earthquake, the Kobe port was closed down for a long

time due to extensive damage. As a result ships started using other ports with a result that

there was a permanent damage to regional income. The degree to which key

infrastructure can function after a catastrophe determines the rate at which materials and

men can be moved to the affected region. This in turn affects the post-event productivity

of the region.

The indicators used for infrastructure are:

i) Number of television sets per capita, and

ii) Number of radios per capita

4.4.2.6 Education

Education is vital in creating awareness and facilitates communicating catastrophe

mitigation ideas. A society will be more aware of the risk of occurrence of natural hazard

event if specific measures are taken to incorporate an adequate knowledge of the

vulnerabilities of different zones or regions in the school curricula. An educated

community can devise and implement more effective self-management strategies in case

of a catastrophic event and rapidly recover thus restoring its pre-event productive

capacity or sometimes even exceeding it. But education is intimately related to ongoing

socioeconomic processes. Barro and Martin (1995) present evidence to show that


158

average-years of male secondary and higher schooling and average years of female

secondary and higher schooling tend to be significantly related to subsequent growth.

Thus the level of education of a community is an important control variable used to

compare post-event economic behavior across various countries.

The percentage of "no schooling" in the population is used as an indicator for education.

4.4.2.7 Trade

Trade forms an important component of a country's output in the modern global

economy. Manufactures form a major portion of imports of a developing country while

exports are usually non-manufactures such as cash crops. If a catastrophe severely

destroys the main export merchandise then the country may be forced into a balance of

payments crisis thus dampening its post-event reconstruction efforts. Productivity after

the event may be forced to be below its pre-event levels. Also, prices, for agricultural and

mineral exports on which Third World has traditionally had to depend, are falling. On the

other hand, prices of imported energy and technology have increased. This worsens

balance of payments crises. Hurricane Gilbert deprived Jamaica of more than US $27

million in foreign exports in 1988-89. This may partly explain the decline in its growth

rate (Table 4.1). Foreign debt amounted to 60 per cent of GDP for Latin America during

1985 (Branford and Kucinski 1988).

Indicators which summarize a governments policy stance on trade over time is:

i) The Economic Freedom Index (Gwartney and Lawson, 1997), and

ii) The degree to which a country's exchange rate has been over-valued - as measured by

the black market premium on the exchange rate.


159

4.5 INTRODUCTION TO ECONOMETRIC ISSUES

An overview of the main concepts involved in linear regression models is

presented in this section. An algebraic model represents the real–world system by a

system of equations. These equations may be behavioral, such as the consumption

function: C = C(Y), an equilibrium condition, such as the national income equilibrium

condition: Y = C+Z. The variables in this model are consumption C, national income Y,

and exogenous expenditure Z. All of these fundamental or basic equations of the model

may be termed structural equations. The model determines values of certain variables,

called endogenous variables, the jointly dependent variables of the model, which are

determined simultaneously by relations of the model. In the case of the model specified

in Eq.3.7, post event economic growth rate is the endogenous variable, which is to be

explained or predicted. The model also contains other variables, called exogenous

variables, which are determined outside the system but which influence it by affecting the

values of the endogenous variables. They affect the system but are not in turn affected by

it. Direct loss from a catastrophe is an example of exogenous variable. The model also

contains certain parameters (α and β), which are generally estimated using econometric

techniques and relevant data. Another important characteristic of an econometric model is

the fact that it is stochastic rather than deterministic. A stochastic model includes random

variables, whereas a deterministic model does not. Let the economic growth rate after a

catastrophic event be given by:

yt = α0 + β1yt-1 + β2D + β3E (4.4)

This function specifies that given the pre-event economic conditions yt-1, a vector D

describing the disaster including magnitude of direct loss, the number of people affected,

and a vector E describing the country specific characteristics, post-event economic

growth rate yt is determined exactly as given by Eq.4.4. This is clearly not reasonable.

Many factors other than the direct loss and the population affected determine the post

event growth rate, such as inflation variability, quality of the bureaucracy, health, and

education. Furthermore, the relationship may not be as simple as that given by the linear

relation as explained in Chapter 3 and the loss variables may be measured inaccurately. It

is therefore more reasonable to estimate yt at a given level of loss variables, as on average


160

equal to the right hand side of Eq.4.4. In general, yt will fall within a certain confidence

interval, after controlling for country specific fixed effects, that is,

yt = α0 + β1yt-1 + β2D + β3E ± ∆, (4.5)

where, ∆ indicates the level above or below the average value such that with a high

degree of confidence, yt fall in the defined interval. The value of ∆ can be determined by

assuming that yt is itself a random variable with a particular density function. Because of

the central limit theorem the normal distribution is typically assumed. The term “on

average” generally refers to the mean or expected value, so the right hand side of Eq. 4.5

is the mean of yt. The ∆ can then be chosen, as illustrated, so that 90% of the distribution

is included in the confidence interval, where each of the tails of the distribution contains

5% of the distribution. In general, an econometric model uniquely specifies the

probability distribution of each endogenous variable, given the values taken by all

exogenous variables and given the values of all parameters of the model.

Algebraically, the stochastic nature of the relationship for the post event growth rate is

represented as

yt = α0 + β1yt-1 + β2D + β3E ± ε (4.6)

ε is an additive stochastic disturbance term that plays the role of chance mechanism. In

general, each equation of an econometric model, other than definitions, equilibrium

conditions, and identities, is assumed to contain an additive stochastic disturbance term.

If the stochastic disturbance term has a variance that is always identically zero, the model

reduces to a deterministic one. The other extreme case is where the model is purely

stochastic. The stochastic terms are unobservable random variables with certain assumed

properties (e.g. means, variances, and covariances). The values taken by these variables

of the model are not known with certainty; rather, they can be considered random

drawing from a probability distribution. Possible sources of such stochastic perturbations

could be relevant explanatory variables that have been omitted from the relationship

shown in the model or possibly the effects of measurement errors in the variables - in

particular errors related to reporting of direct losses. Other sources of such stochastic

disturbances could be mis-specified functional forms, such as assuming a linear

relationship when the true relationship is nonlinear, or errors of aggregation, which might


161

be introduced into a macro equation when not all individuals possess the same underlying

micro relationship. It might be noted that sources of these perturbations can be quite

important in practice; thus the treatment of measurement error will be, in general, quite

different, depending on whether the measurement error is found in the dependent variable

or in one or more of the explanatory variables. In any case, the inclusion of such

stochastic disturbance terms in the model is basic to the use of tools of statistical

inference to estimate parameters of the model.

4.6 PROBLEMS WITH THE DATA

Data related to catastrophes are typically non-experimental data. There are several

problems encountered with these data, which are presented in the following.

The first is the degrees-of-freedom problem – that the available data simply do not

include enough observations to allow an adequate estimate of the model. In the use of

non-experimental data it is impossible to replicate the conditions that gave rise to them,

so additional data points cannot be generated. Data regarding catastrophic losses may be

available, but data on explanatory variables may be missing. This was particularly true in

panel data compiled for this study. In some cases the available was inadequate for

estimating a particular model but adequate for estimating an alternative model, which

will be clear when various model specifications are presented. By including more than

155 major events, the panel had adequate degrees of freedom in spite of the missing data.

Second is the multi-collinearity problem - the tendency of the data to bunch or move

together rather than being “spread out”. For example, for a complex model describing

economic growth, the variables exhibit the same trends over a cross-section of countries.

With experimental data it may be possible to vary the conditions of the experiment to

obtain an adequate spread. With non-experimental data such control does not exist, and

the real-world system may involve very small variation in the data, in particular a high

degree of interdependence among certain variables. This problem was circumvented to a

certain extent by including data from countries belonging to wide range economic

development as measured by per capita GDP.


162

Third is the serial-correlation problem – the fact that when using data in two consequent

years (before and after the catastrophic event), underlying changes occur very slowly

over time. Thus conditions in time periods that are close together tend to be similar. To

the extent that the stochastic disturbance term represents conditions relevant to the model

but not accounted for in it explicitly, such as omitted variables, serial correlation itself in

a dependence of the stochastic disturbance term in one period on that in another period.

Various tests, to be described later, were performed to detect the serial-correlation and

suitably interpret the results of the specifications.

Fourth is the errors-in-measurement problem – that data are measured subject to various

inaccuracies and biases. In fact, data are sometimes revised because of later recognition

of these inaccuracies and biases. More fundamentally, potential inaccuracies result from a

lack of precision in conceptualization. For example, the GNP accounts are revised from

time to time on the basis of such changes in conceptualization (e.g. defining what is

included in consumption). Such changes in conceptualization necessitate refining the data

to make them comparable and consistent over time. Also, as has been mentioned

previously, the reporting of data related to catastrophes may be inaccurate and biased due

to several factors related to political, sociological, and anthropological factors.

To address these issues, the Extreme Bounds Analysis is used and this is described in the

following.

4.7 LIMITATIONS OF CROSS-COUNTRY REGRESSION STUDIES

There are substantial conceptual and statistical problems that plague cross-country

investigations (Levine and Renelt, 1992). Levine and Renelt (1992) point out that

statistically entries are sometimes measured inconsistently and inaccurately. Even putting

measurement difficulties aside, it is not clear whether we can include countries as diverse

as Bangladesh and Canada in the same regression. These countries operate in different

policy regimes and under different environments. A country may be at a particular stage

in a business cycle, or may be undergoing major policy changes, or experiencing political


163

disturbances. All these factors affect economic activity and consequently economic

growth. Researchers (Barro, 1991; Easterly, 1997) have found that many individual

indicators of monetary, fiscal, trade, exchange-rate, and financial policies are

significantly correlated with long-run growth in cross country growth regressions. How

could one evaluate the “believability” of cross-country regressions? Extreme bounds

analysis (EBA) based Edward Leamer’s (1983, 1985) work can be used for testing the

results of regressions relating the direct loss to the post event indicators of the economy.

The EBA employs a linear, ordinary-least-squares regression framework. The variables in

the vector E are chosen from a set of indicators, which are known to affect the long-run

economic growth rate. The EBA involves varying the E variables to determine whether

that coefficient on the damage indicator, D, is consistently significant and of the same

sign when the right-hand-side variables change. If β2 is consistently significant and have

the same sign the results are termed as “robust”; otherwise the results are “fragile.” The

EBA is used to test the robustness of the empirical associations between the loss-GDP

ratio and various economic indicators. The results of these regressions are presented next.

4.8 RESULTS FROM REGRESSION ANALYSIS

Details of the regression results for growth changes and the effects of catastrophes

on consumption, savings, government expenditure, inflation and real interest rates are

presented in Appendices F to I. For these regressions various specifications are reported.

4.8.1 Growth rates – Short term

Based on the specifications (21 in all) for examining the effect of a catastrophe on

the short-term growth rate of an economy that the direct loss term enters statistically

significantly. Complete details of the specifications and the regressions are presented in

Appendix F, in electronic form. Table 4.3 presents one such specification. For this

particular specification, the coefficient for the loss term (b1) is –2.37 and highly

significant. Other specifications reveal that the coefficient (b1) ranges from –3.9 to –1.7

with a mean of –2.9. The coefficient remains highly significant (<0.001) in all

specifications. The coefficient for percentage of population affected is also statistically


164

significant but is positive. Dummy for earthquakes indicate that they are associated

positively and significantly with post event growth rate, whereas droughts are negatively

associated with post event growth rate.

Summary|R| 0.543R2 0.295R2 adjusted 0.271Standard Error 3.486# Points 151PRESS 1963.96R2 for Prediction 0.214Durbin-Watson d 1.787First Order Autocorrelation 0.106Collinearity 0.334Coefficient of Variation 105.286

ANOVASource SS SS% MS F F Signif df

Regression 736.81 29 147.36 12.12 8.030e-10 5Residual 1762.30 71 12.154 145Total 2499.1 100 150

P value Std Error -95% 95% t Stat VIFb0 8.73 0.000 1.604 5.56 11.90 5.4b1 -2.37 0.000 0.500 -3.36 -1.38 -4.7 2b2 60.47 0.003 19.796 21.34 99.60 3.1 1b3 -3.10 0.000 0.511 -4.11 -2.09 -6.1 2b4 0.42 0.001 0.128 0.16 0.67 3.2 1b5 -1.32 0.006 0.475 -2.25 -0.38 -2.8 1

Table 4.3 Result of a typical regression used for testing the negative association of economic loss to post-event growth

Gr1yrAfter = b0 + b1*Log10(Loss/GDP) + b2*Log(1+TotoAff/Pop) + b3*Log10AvgGDP_per_capita + b4*COV_AnnualGrowth + b5*Log10Std_Dev_Inflation


165

Some of the specifications control for immediate (one year preceding) pre event

economic conditions with indicators such as the pre event growth rate, the pre event gross

domestic fixed investment growth, and the pre event government size. The coefficients of

pre-event growth rate and the pre event gross domestic fixed investment growth enter

positively and significantly in explaining the post-event growth rate, as expected. One

point of the pre event growth rate explains 0.39 to 0.79 point of the post event growth

rate. One point growth in the pre event gross domestic fixed investment is associated with

0.22 points of post event growth rate. Greater share of the government expenditures in the

GDP appears negatively associated with the post-event growth rate.

Control variables such as measures of civil liberties, bureaucratic quality, black market

exchange rates, percentage of population without schooling are averages over longer

periods of time, typically five to ten years around the event. The inherent assumption is

that these variables change at a much slower rate than other macro-variables like the

annual percentage growth rate. As has been discussed in chapter two, these factors

nevertheless determine the vulnerability of a country to natural hazards, which in turn

determines the post event economic behavior. The regressions present econometric

evidence for associations between indicators of ongoing social, economic, and political

processes and the post event behavior.

Indicators of the monetary health of an economy, namely the inflation variability, the

monetary growth (average annual growth rate of the money supply during the last five

years minus the potential growth rate of real GDP) is negatively associated with growth.

Better bureaucracies are associated with higher post event growth. Indicators for

government enterprises (higher ranks imply lower role and presence of government

owned enterprises) are positively associated with post event growth. Greater civil

liberties and political rights are also positively and significantly associated with post

event growth. Better health (as indicated by the daily protein/calorie intakes) is also

positively associated with growth. Lack of education is negatively and significantly


166

associated with post event growth. The signs and significance of the determinants of

growth appear as expected and in accordance with the discussion in Section 4.5.

4.8.2 Growth rates – Average

Table 4.4 presents the regression results for average growth rates. As has been

already mentioned, average growth rates refer to means of growth rates three years prior

to and after the event. It can be seen from Table 4.4 that the loss term appears negatively

and significantly in all the specifications. The coefficient ranges from –1.95 to –0.67 with

a mean of –1.28 over the specifications. This again implies that loss term is negatively

correlated with post event growth. Simulations based on Ramsey’s model indicate that

(Fig. 2f) loss is positively correlated with the post-event growth. Only if the effects of the

efficiency of the post event reconstruction are taken into account, as described in Section

3.3 using an extended Ramsey’s model, can the negative correlation between loss and the

post-event growth be explained. This brings out the importance of modeling the transient

processes immediately after the event.

Dummy for earthquakes indicate that they are associated positively and significantly with

post event growth rate. Droughts are negatively associated with post event growth rate.

Earthquakes typically result in capital being damaged or destroyed. Droughts do not

cause relatively damage to capital stock. After an earthquake, reconstruction activities

may have a positive effect on the region’s productivity. Such a change in productivity

was assumed in the theoretical models developed in Chapter 3. Numerical simulations

indicated that increases in productivity (Figs. 3.1f and 3.3h) result in increases in post

event growth rates. Empirical evidence that earthquake dummy is positively correlated

with the post-event growth rate lends support to the theoretical result that capital

regeneration after an earthquake increase the post event growth rate. This is further

reinforced by the fact that drought dummy is negatively correlated.


167




P value Std Error -95% 95% t Stat VIFb0 3.54 0.217 2.854 -2.10 9.18 1.2b1 -1.68 0.000 0.327 -2.33 -1.04 -5.1 2b2 38.22 0.005 13.264 12.01 64.44 2.9 1b3 -2.72 0.000 0.336 -3.39 -2.06 -8.1 2b4 -0.73 0.021 0.311 -1.34 -0.11 -2.3 1b5 4.17 0.024 1.836 0.54 7.80 2.3 1b6 0.31 0.000 0.084 0.14 0.47 3.6 1

AvgAfter = b0 + b1*Log10(Loss/GDP) + b2*Log(1+TotoAff/Pop) + b3*Log10AvgGDP_per_capita + b4*Log10Std_Dev_Inflation +

b5*Log10AvgGrossCapitalFormation_%GDP + b6*COV_AnnualGrowth

Table 4.4 Result of a typical regression used for testing the negative association of economic loss to post-event average growth


168

The coefficient of average pre-event per capita income enters negatively and significantly

in explaining the post-event growth rate, as expected. Greater is the pre-event per capita

income, smaller is the post-event growth.


standard deviation of growth over the past ten years is negatively associated with post

event growth. These indicators of the susceptibility of the economy to price volatility in

an economy enter negatively because price increases after the event result in lower

productivity.

Better bureaucracies are associated with higher post event growth. One possible reason is

that better bureaucracies will fuel the post event productivity and hence growth. Lack of

corruption enters positively and significantly in the post event growth rates. Indicators for









169

4.9 EFFECT ON MAJOR ECONOMIC INDICATORS

After examining the data on economic growth, the effects of catastrophes on

major economic indicators such as the consumption, investment, government

consumption, inflation, and the real interest rates are examined. Details of these

regressions are presented in electronic form in Appendix G.

4.9.1 Consumption

It is clear from Table 4.5 that direct loss is positively and significantly associated

with consumption. The coefficient of the loss term has a minimum value of 2.0 and a

maximum of 3.2 with a mean of 2.6 over the specifications. This implies that a direct loss

of 10% of GDP is associated with 3.21 point increase in consumption. This response to

changes in income from catastrophes is further explained in the subsequent sections.

4.9.2 Investment

An examination of the specification (Table 4.6) reveals that catastrophes result in

lowering the amount of investments by concentrating on increases in consumption

expenditures.

4.9.3 Government expenditure

The loss term enters positively and significantly in all the specifications with a

mean of 1.14. A typical relation is as follows:

GovernmentConsumptionafter = -4.33

(0.002)

+ 0.86*GovernmentConsumptionbefore+1.4*Log(Loss/GDP)

(0.000) (0.011)

+ 2.395*Eq. + 0.559*GovtEnterp

(0.004) (0.006)

N=100; R2= 0.818; F=38; DW = 1.925

This implies that a direct loss of 10% of GDP is associated with 1.4 point increase in

government expenditures.


170

Table 4.6 Effect of catastrophes on investmentsSummary

|R| 0.915R2 0.837R2 adjusted 0.835Standard Error 2.477# Points 149PRESS 951.04R2 for Prediction 0.827Durbin-Watson d 1.508First Order Autocorrelation 0.243Collinearity 0.814Coefficient of Variation 12.814



AvgAfter = b0 + b1*Log10(Loss/GDP) + b2*AvgBeforeP value Std Error -95% 95% t Stat VIF

b0 1.998 0.0056 0.71 0.60 3.40 2.81b1 -0.896 0.0020 0.28 -1.46 -0.33 -3.15 1.23b2 0.791 0.0000 0.03 0.72 0.86 23.16 1.23

Table 4.5 Effect of catastrophes on change in consumption Summary

|R| 0.24R2 0.06R2 adjusted 0.05Standard Error 0.04# Points 150.00PRESS 0.20R2 for Prediction 0.03Durbin-Watson d 1.87First Order Autocorrelation 0.05Collinearity 1.00Coefficient of Variation 3.61


Regression 0.01234 6 0.01234 9.396 0.00259 1Residual 0.194 94 0.00131 148Total 0.207 100 149

ChangeInRealConsumption = b0 + b1*Log10(Loss/GDP)P value Std Error -95% 95% t Stat

b0 1.029 0.000 0.009 1.012 1.046 121b1 0.011 0.003 0.004 0.004 0.019 3.065


171

4.9.4 Inflation, and Interest rates

The regression associating the loss term and the inflation is as follows:

Log(Inflation)after = 0.18 + 0.89 *Log(Inflation)before+ 0.07*Log(Loss/GDP) (0.121) (0.000) (0.057)

N=114; R2= 0.686; F=204; DW = 1.68 This implies that a direct loss of 10% of GDP is associated with 0.07 point increase in log(inflation)

Fig. 4.11 clearly illustrates the fact that the post-event real interest rates are higher

than the pre-event counterparts. Details on the regressions for examining the effect

on interest rates is presented in electronic form in Appendix H. The impact of

catastrophes on real interest rates is presented in Table 4.7. It is clear from the

table that catastrophes, as measured by the number of people affected, are

positively associated with post event real interest rates. Pre-event levels of

inflation and per capita income are positively associated with the post event real

interest rates. Higher the pre event gross investment in fixed capital, lower is the

post event interest rate. And if household spend more money, smaller will be the

real interest rate. An important conclusion of this section is that catastrophes and

financial markets may not be totally uncorrelated after all. More research is

required to unravel the connections between the catastrophes and financial

markets.

Summarizing the results of this section, greater loss-to-GDP ratios are positively

and significantly associated with increases in consumption, government

expenditure, and inflation, and real interest rates and are negatively and

significantly associated with investments. The effect of catastrophes on a host of

other economic indicators is presented in electronic form in Appendix I. These are

not discussed explicitly herein, but they indicate negative effects of a catastrophe.


172

Fig. 4.11 There is perceptible increase in real interest rates after a catastrophic loss

0%10%20%30%40%50%60%70%80%90%

100%

-5.0 0.0 5.0 10.0 15.0 20.0

Real Interest rates

AvgBeforeAvgAfter

Table 4.7 Effect of catastrophes on the real interest ratesSummary




RealIntRateEventYear = b0 + b1*Log10(1+TotAff/Pop) + b2*LogInflationCPI + b3*Log10AvgGDP_per_capita + b4*AvgGrossCapitalFormation_%GDP +

b5*Household_final_consumption_expenditure_(annual_%_growth)P value Std Error -95% 95% t Stat VIF

b0 -0.706 0.925 7.437 -15.45 14.04 -0.09b1 9.130 0.003 3.051 3.082 15.18 2.99 1.2b2 3.526 0.000 0.756 2.028 5.025 4.67 1.2b3 1.245 0.031 0.568 0.119 2.372 2.19 1.5b4 -7.396 0.042 3.587 -14.51 -0.285 -2.06 1.2b5 -0.316 0.004 0.106 -0.527 -0.105 -2.97 1.2


173

4.10 Catastrophes and consumption smoothing

The main purpose of this section is to test the validity or otherwise of the

permanent income hypothesis (PIH) during the years surrounding the occurrence of

catastrophe. Do catastrophes cause predictable shifts in consumption? If the nations are

not able to smooth consumption in these adverse circumstances, then policies need to be

devised which will help mitigate the effects of a catastrophe. If there are predictable

shifts in consumption after a catastrophe, then this information could be used to design

policies to smooth consumption after a catastrophe.

Whenever a catastrophe occurs it will cause rational agents to change the way in which

past incomes affect forecasts of future incomes. Consumption depends on expected future

incomes. Flavin’s (1981) result shows that changes in consumption are predictable by

lagged changes in income (the excess sensitivity hypothesis). The first part of the test will

establish whether the catastrophe results in predictable changes in income. Knowing

about the process generating income (in this case the date of occurrence of a catastrophe)

we could generate forecast for consumption change based on lagged values of income

and consumption. There is excess sensitivity if consumption responds to any previously

predictable component of income change (Deaton, 1992: 164). For the purposes of this section cross-sectional data was used. These include differences

in consumption and income three years preceding and following the event. The data was

not pooled since the main intention is to find out whether lagged changes in income can

predict consumption changes with and without the occurrence of a catastrophic event.

Table 4.8 shows the means and standard deviations of income, saving, and consumption

for five years with the catastrophic event year as the third year. The one noticeable

characteristic is the enormous variation in data. This is to be expected since we have

pooled data over a wide spectrum of countries. Another noticeable feature is that the

standard deviation of income and consumption become largest at t+1 and then drop to

their smallest value at t+2, t being the time of occurrence of the catastrophe. For saving

too the standard deviation becomes large at t+1 and drops at t+2. One plausible inference

is that the occurrence of a catastrophe, which results in direct losses comparable to the


174

GDP (typically greater than 1% of the GDP), may induce greater fluctuations in income,

savings, and consumption. Another plausible inference is that effects of catastrophe

attenuate two years after the event as evidenced by the relatively sharp drop in the

standard deviations for income, consumption and savings.

The data in Table 4.9 illustrate the fact that mean growth rates for both income and

consumption fall during the disaster year (t0 - t-1). The variability of the growth rates

increases for the income whereas for consumption the variability remains almost

constant.

Table 4.8 Summary statistics for income, consumption, and savings in the five years enveloping the disaster year Income Consumption Savings Mean s.d. Mean s..d. mean s.d. t-2 5168.5 7711.9 4039.0 5933.1 1187.7 2072.2 t-1 5198.8 7709.0 4100.0 5993.2 1157.0 1990.3 t 5267.8 7823.1 4170.5 6100.3 1153.8 1979.6 t+1 5399.8 8061.3 4228.8 6176.8 1185.2 2070.0 t+2 4707.9 6224.0 3751.0 5108.4 973.8 1238.2

Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of occurrence of the catastrophe and the previous year. Table 4.9 Summary statistics for percentage growth rates for income, consumption, and savings in the five years enveloping the disaster year

Income Consumption

Mean s.d. Mean s.d. (t-1- t-2) 0.64 1.60 0.69 2.24 (t0 - t-1) 0.38 2.34 0.45 2.12 (t+1- t0) 0.87 1.81 0.87 2.11 (t+2- t+1) 0.72 1.74 0.74 2.13

According to the permanent income hypothesis (PIH), changes in aggregate consumption

cannot be predicted by lags in income. Hall (1978) first proposed tests for the PIH by

adopting the technique of regressing changes in consumption using lagged income,

conditional on lagged consumption. For the PIH to hold variables lagged t-1 or earlier,


175

and in particular lags of income should not help predict consumption in period t. The

expression for changes in consumption (Deaton, 1992:83) is:

∆ct = r/(1+r) Σ∞k=0 (1+r)

-k (Et+1 − Et)yt+k

This implies that a change in consumption ought to be the amount warranted by

innovation in expectations about future labor income. Knowledge of the process

generating income will enable us to check this prediction too (Deaton, 1992:84).

Presumably innovations in income occur after a catastrophic event. Do these innovations

predict the change in consumption? What differences do we observe if we compare

consumption change after a catastrophe with consumption change without any

catastrophic event?

In an important paper, Flavin (1981) tested the null hypothesis the truth of the PIH as

expressed in Eq. 4.7, together with an auto regressive specification for the process

governing labor income. Flavin’s ‘excess sensitivity’ hypothesis allows consumption to

respond to current and lagged changes in income by more or less than is required by the

PIH. The measurement of excess sensitivity is the measurement of the extent to which

consumption responds to previously predictable changes in income.

By using Deaton’s (1992:94) specification of regressing the change on consumption on

the lagged change in income we could avoid the unit root problems that may affect the

income generating process:

∆ct = α + β∆yt-1

However time-averaging problems may induce spurious correlation for adjacent

observations of a series that has been first differenced. This implies that Eq. 4.8 may

yield inconsistent estimates because ∆yt-1 is spuriously correlated with ∆ct. To avoid such

problems Deaton (1992) suggests that variables lagged two periods may be used as

instruments. Instrumentation by variables lagged by variables enables us to account for

transitory consumption that may result from a catastrophe.


176

The results are shown in Table 4.10. The first row presents the results of regressing on

changes in consumption on lagged changes in income for the period t-1 (the period before

the event occurs). The second row presents a similar regression using the twice-lagged

changes in income and consumption as instruments. The third and fourth rows repeat the

same for the period when the catastrophe strikes. The rest of the table presents more

regressions for two following periods. From Table 4.10 it is clear that lagged changes in

income does not enter significantly in explaining the changes in consumption for the

periods: t-1, t+1, and t+2. This implies that there is evidence that the PIH is valid for the

above periods. For years immediately after the event (i.e. results for ∆ct), the results

indicate the change in consumption is the amount warranted by innovations in income.

Instrumentation of lagged changes in income with twice lagged changes in income and

consumption still leaves a significant (albeit reduced) coefficient for ∆yt-1. There is

therefore evidence of excess sensitivity once possible timing and transitory consumption

problems have been taken into account. These results can be taken to mean that

innovations in income due to the occurrence of the catastrophe result in predictable

changes in consumption.

Table 4.10 Estimates of consumption changes using lagged changes in income

Constant t Lagged ∆y t R2 F N ∆ct-1 (OLS) 57.66 (2.620) 0.29 (3.16) 0.16 9.95 48 ∆ct-1 (IV) 57.41 (2.458) 0.30 (1.49) - - 48 ∆ct (OLS) 54.00 (2.527) 0.51 (6.40) 0.46 40.89 48 ∆ct (IV) -8.43 (-0.191) 1.23 (3.75) - - 48 ∆ct+1 (OLS) 24.90 (0.821) 0.15 (1.64) 0.03 2.69 48 ∆ct+1 (IV) 50.63 (1.323) 0.18 (1.11) - - 48 ∆ct+2 (OLS) 70.52 (2.829) 0.19 (1.75) 0.04 3.06 48 ∆ct+2 (IV) 64.71 (2.001) 0.26 (0.90) - - 48

Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World

Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of

occurrence of the catastrophe and three years prior and following the event. Place and occurrence of

catastrophe are from Center for Research on Epidemiology of Disasters (Sapir and Misson, 1992 CRED).

The instruments in the IV are ∆yt and ∆ct lagged twice. t-values are shown in brackets.

In this section PIH and excess sensitivity were examined. There is evidence of excess

sensitivity once possible timing and transitory consumption problems have been taken


177

into account. These results can be taken to mean that innovations in income due to the

occurrence of the catastrophe result in predictable changes in consumption.

4.11 Consumption smoothing and savings behavior

A catastrophic loss in a country's income will lead to changes in consumption

only if the savings are not able to offset the income fluctuations. Incomes of LDCs are

both low and uncertain. Losses of incomes of LDCs due to catastrophes may seriously

undermine the ability to smooth consumption. When insurance markets are incomplete

(this is true for most LDCs), saving and credit transactions assume a special role by

allowing households to smooth their consumption streams in the face of random income

fluctuations.

If income can be treated as stationary then PIH implies that savings have a mean of zero.

Assets are built up in advance of expected declines in income, and are run down when

current income is lower than its expected future level. Under the PIH saving acts as a

sufficient statistic for the agent’s future income expectations. Saving behavior contains

information about what nations expect to happen to their incomes. Forecasts of income

conditional on saving help us deal with the fact that representative agents may possess

private more information about future income than does an observer. This helps us to

infer whether the catastrophic events considered are truly unanticipated. If the events

were anticipated, the changes in consumption could well be explained by expected values

of income, which in-turn would have been predicted by the lagged savings. This has

important consequences for policies designed for preparedness against catastrophes

triggered by natural events. If surprises in income due to occurrence of a catastrophe are

unanticipated then consumption will not be smooth even if we use the agent’s private

information. Nations that expect catastrophes (of the type and magnitude considered in

this study) to occur should devise policies for precautionary savings to smooth

consumption.

As pointed out by Campbell (1987), past savings is a predictor of how income will

change the next period. It is possible that countries anticipate the occurrence of a


178

catastrophe triggered by natural hazards. Though natural hazards occur with certain

regularity, their magnitude and point of occurrence remains uncertain. But a good

preparedness program in place would help nations to smooth their income. By regressing

the change income with lagged savings and comparing the no-disaster year with the

disaster year we could infer about the efficiency of the precautionary savings of the

countries to income shortfalls from a catastrophe triggered by a natural hazard. From

Table 4.11 it is clear that before the catastrophe occurs, lagged savings do not explain the

income change. This situation, however, changes one year after the event. The coefficient

for lagged savings becomes positive and significant in explaining income changes. The

value of the coefficient again drops two years after the catastrophe. This means that the

catastrophe changes the ex ante saving behavior at least for two years after the event.

Consumption change regressions (Table 4.12) show that it is positively related to lagged

values of savings. Before the catastrophic event the coefficient on savings has a lower

significance than two years after the event. One plausible inference is that the changes in

consumption due a catastrophe are only weakly anticipated for the collection of events

that have been considered. Since the events cover a large range of loss/GDP ratios (Table

4.13) it is plausible that the data set dampens out effects of unanticipated losses for

LDCs. Three years after the event the significance and magnitude fall to their pre-

disaster levels and lagged savings are not able to explain consumption changes in

accordance with the PIH.

If we use income lagged twice as an instrument in the regression on consumption change

on lagged saving we essentially get the same results.

Table 4.11 Estimates for income changes using lagged savings

Constant t Lagged Savings

t R2 F N

∆yt-1 37.32 (0.94) 0.0061 (0.36) 0.003 0.13 46 ∆yt 26.21 (0.67) 0.0520 (3.07) 0.17 9.45 46 ∆yt+1 -10.69 (-0.38) 0.1249 (10.22) 0.70 104.4 46 ∆yt+2 -22.71 (-0.64) 0.1031 (4.52) 0.31 20.43 45




179

occurrence of the catastrophe and two years prior and following the event. Place and occurrence of

catastrophe are from Center for Research on Epidemiology of Disasters (Sapir and Misson, 1992 CRED). t-

values are shown in brackets.

Table 4.12 Estimates for consumption changes using lagged savings Constant t Lagged

Savings t R2 F N

∆ct-1(OLS) 20.74 (0.66) 0.034 (2.57) 0.11 6.62 48

∆ct-1 (IV) 26.34 (0.82) 0.029 (2.11) - - 48

∆ct(OLS) 9.42 (0.45) 0.053 (5.83) 0.42 33.95 48

∆ct(IV) 6.67 (0.31) 0.055 (5.78) - - 48

∆ct+1(OLS) 20.02 (0.88) 0.067 (6.76) 0.49 45.71 48

∆ct+1(IV) 16.83 (0.73) 0.069 (6.71) - - 48

∆ct+2(OLS) 21.21 (0.66) 0.020 (1.51) 0.03 2.27 48

∆ct+2(IV) 11.30 (0.35) 0.028 (2.01) - - 48 Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World




The instrument in the IV is income lagged twice. t-values are shown in brackets.

In this section the efficiency of the precautionary savings of the countries to income

shortfalls from a catastrophe triggered by a natural hazard is examined. Results from

regressions of income change on lagged savings and comparison of the no-disaster year

with the disaster year are used for arriving at conclusions. Before the catastrophe occurs,

lagged savings do not explain the income change. But one year following the event,

lagged savings anticipate income changes. Evidence is presented to show that

catastrophes change ex ante saving behavior at least for two years after the event.

4.12 Conclusions, Extensions, and Limitations

The problem of finding empirical regularities in the ongoing socioeconomic

processes after the occurrence of a catastrophe was addressed in this chapter.

Connections between these statistical regularities and the results of the theoretical model

simulations presented in Chapter 3 were made. The results of the regression analysis

indicate that by studying disasters much can be learned about the way large-scale socio-


180

economic systems affect and are affected by the occurrence of catastrophes. By making a

cross-country study with countries from all income groups affected by different types of

natural hazards, the results are expected to be sufficiently general. Previous empirical

results from the literature on the determinants of economic growth and on economic

development helped in identifying the explanatory control variables.

The main results of this study can be summarized as follows:

Summarizing the regressions on growth the following statistical regularities are

discerned:

• The models indicate very significant negative coefficient for the direct loss variable in

regressions for short-term growth. The coefficient for the loss variable in the long-term

growth has a lower significance, but remains negative. The magnitude of the coefficient

in the average growth rate regression is less than the short-term regression. This implies

that the associations between the loss term and the economic growth rate become harder

to detect with the passage of time. These results corroborate the results obtained by

simulating the model presented in Section 3.3.

• The pre-event economic growth rate is positive and very significantly associated with the

post-event growth rate, in both the short-term and average regressions. This implies that,

other variables being constant, an economy with a sufficient growth rate can absorb the

effect of a catastrophe. Growth itself is an indicator of the robustness of ongoing

developmental processes. This brings out the importance of having a robust

developmental process in place in absorbing the effect of a catastrophe. The coefficient

for pre-event general government consumption is significant and negative. This agrees

with the known fact that heavy consumption by the government sector retards growth.

• The coefficient for the percentage of people affected is positive and significant in short-

term growth regressions. Though this seems odd, it is should be noted that a catastrophe

affects many people only in developing countries. The amount of aid is to a certain extent

decided by the figures regarding people affected. It is probably this external aid

associated with the percentage affected that spurs growth. As the models described in

Section 3.3 and 3.4, greater inflow of aid results in greater growth.


181

• The coefficient for daily protein/calorie intake appears positive in the short-term growth

regressions associating a healthier community with a more robust developmental process

• If the institutions of crisis management can be proxied by a combination of the size of the

government and the efficiency of the bureaucracy, then their coefficients are positively

and significantly associated with short- and long term (average) post event growth. This

brings out the importance governmental bureaucracy in mitigating the effects of a

catastrophe.

• The coefficient for inflation variability, which is a measure of the monetary robustness of

an economy, is associated negatively and significantly with the post event short- and

long-term growth. This once again ascertains the importance of the ongoing economic

processes in explaining the post-event economic behavior.

• Other factors including civil liberties, percentage of no schooling, economic freedom

index, freedom from corruption, and land-area had the expected signs.

The main results of examining the effects of catastrophes on consumption, investment,

government expenditure, net exports, inflation, interest rates gave the following results:

Large economic losses as a proportion of the GDP are associated with:

1. greater post-event consumption,

2. greater post-event government expenditure,

3. smaller post-event investments,

4. higher inflation, and

5. an increase in real interest rates.

Innovations in income due to the occurrence of the catastrophe result in predictable

changes in consumption. The efficiency of the precautionary savings of the countries to

income shortfalls from a catastrophe triggered by a natural hazard is examined. Results

from regressions of income change on lagged savings and comparison of the no-disaster

year with the disaster year are used for arriving at conclusions. Before the catastrophe

occurs, lagged savings do not explain the income change. But one year following the

event, lagged savings anticipate income changes. Evidence is presented to show that



182

There are limitations of the study, which are discussed in the following. The first is

regarding the heterogeneity and panel data that arise naturally in cross-country studies.

Omitted heterogeneity induces correlations between explanatory variables and the error

term in a way that has the same consequences as simultaneity bias. The factors that

appear on the right hand side of the specification (Eq.4.6) such as pre event growth may

have no general claim to exogeneity. The combination of genuine simultaneity and

heterogeneity has the further effect of ruling out the use of lags to remove the former.

These considerations would typically require further examination of the effect of

catastrophe on the economic indicators using alternative specifications based on first

differences. Another important limitation is the lack of appropriate instruments, which

are correlated with direct loss term but un-correlated with error term. These instruments

can be used to check whether the coefficients on the loss terms remain robust when they

are instrumented. If data on sectoral distribution of losses is available, this can be used to

instrument the direct loss variable. In other words, this requires details regarding losses in

the agriculture, industry, and service sectors. But such data is hard to obtain. It would be

ideal to develop a system of structural equations to explain the connections between all

the macro-economic variables affected by catastrophes. Lack of underlying theoretical

models forces us to use reduced form equations. These result in inference of statistical

regularities as opposed to full-fledged causal models. Increase of representation in the

sample of higher loss-GDP ratio events is required for the sake of generality.


129

Chapter Four

An Empirical Study of the Macro-economic Effects of Catastrophes Triggered by Natural Events

4. INTRODUCTION

In this chapter we re-examine our understanding of the effects of catastrophes on

the economy based on empirical evidence. Questions addressed include the change or

absence thereof, in economic growth, consumption, saving, inflation, and real interest

rates. Data on these economic indicators are compiled for various countries for periods

immediately preceding and following the occurrence of a catastrophe. Data regarding

catastrophes such as the estimates of direct losses is also compiled. The regression

analysis employed suggests that catastrophes are negatively associated with all the

aforementioned economic indicators.

In order to study the effect of a catastrophe on an economy the factors that describe

socio-economic conditions prior to occurrence of the hazard event have to be identified.

The vulnerability of a society to natural hazards is the result of various on-going

economic, social, and political processes, as has been discussed in Chapter 2. For large

segments of the world's underdeveloped population, occurrence of a natural hazard may

worsen an already deteriorating or fragile situation. In such regions even a moderate

hazard, such as the 1985 Mexico earthquake, could trigger a catastrophe. Oliver-Smith

(1994) brings this out clearly in his analysis of the 1970 Peru Earthquake. He points out

that Peru's catastrophe was some 500 years in the making, rooted in the complex of

economic and political forces that structured development and the human-environment

relations. The earthquake and subsequent landslides was a trigger for a catastrophe

grounded in poverty, political oppression, and the subversion of previously sustainable

indigenous practices (Bolin and Stanford, 1998).

Socioeconomic conditions in a region are mainly as a result of the developmental

processes. The effect of a major catastrophe on the developmental process is complex,