Report No. 143 March 2003 the John A

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 439 Panama Mall, Bldg. 02-540 Stanford CA 94305

(650) 723-4150 (650) 725-9755 (fax) [email protected] http://blume.stanford.edu

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

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

i 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.

ii Abstract 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.

iii Acknowledgments

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 iv

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 v

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 vi

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 vii

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

Table of Contents 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 4.8 Specifications and regression analysis describing the effect of catastrophes on average economic growth 4.9 Specifications and regression analysis describing the effect of catastrophes on external debt 4.10 Specifications and regression analysis describing the effect of catastrophes on resource balance 4.11 Specifications and regression analysis describing the effect of catastrophes on 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

viii List of Figures 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

ix List of Figures 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

x List of Figures 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 capital to productive capital using extended model 3.3e Change in consumption with various rates of conversion from maturing capital to productive capital using extended model 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 capital to productive capital using extended model 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

xi List of Figures 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

xii List of Figures 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 using regional model 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 using regional model 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)

xiii List of Figures 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 5.12 Effect on gross regional product – probable scenarios with ant without aid assuming 20% capital loss 5.13 Effect on gross regional product – probable scenarios with ant without aid assuming 28% capital loss 5.14 Effect on gross regional product – probable scenarios with ant without aid assuming 35% 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

xiv List of Figures 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

xv List of Figures 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

1 Chapter One: Introduction 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.

2 Chapter One: Introduction 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

3 Chapter One: Introduction 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

4 Chapter One: Introduction 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 regionwise 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

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% $100 $1,000 $10,000 $100,000 GDP per capita (USD current)

5 Chapter One: Introduction

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

6 Chapter One: Introduction sections, a summary of the results, which will be discussed in detail in the subsequent chapters of this dissertation, will be presented.

Fig. 1.2 Regionwise distribution of the loss to GDP ratios

100.0%

10.0%

1.0%

0.1% Lossa proportion as ofGDP

0.0% Category of events that occurred in a year (1970-'98) in a nation

Fig. 1.3 Number of events causing more than US$ 0.5 billion economic loss are increasing R2 = 0.53 16 14 12 10 8 6 4 Number of events 2 0 1970 1975 1980 1985 1990 1995 2000 Year

7 Chapter One: Introduction

Fig 1.4 Loss ratios do not exhibit any trend with time

0.5 0.0 -0.5 -1.0 -1.5 -2.0 -2.5 -3.0

Loss ratio (log scale) (log ratio Loss -3.5 -4.0 1970 1975 1980 1985 1990 1995 2000 Year

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

0.30

0.25

0.20

0.15

0.10

0.05 total population log scale) 0.00 Population affected (as a ratio of 1970 1975 1980 1985 1990 1995 2000 Year

8 Chapter One: Introduction Fig. 1.6 Smaller economic loss is associated with smaller percentage of population affected R2 = 0.30, N= 118 100.000% 10.000% 1.000% 0.100% 0.010% 0.001% 0.000%

affected(log log scale) 0.000% Percentage of population population of Percentage 0.000% 0.0% 0.1% 1.0% 10.0% 100.0% 1000.0% Economic loss as a percentage of GDP

9 Chapter One: Introduction

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

10 Chapter One: Introduction 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

11 Chapter One: Introduction 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.

12 Chapter One: Introduction Table 1.1 Data Description and Sources

Variables Description Source Primary Loss Current US dollars CRED Number of Includes persons dead, injured, homeless CRED people affected or otherwise affected Type of disaster Earthquake, floods, storms, hurricanes, CRED and its year of cyclones, drought, forest fires, and occurrence avalanches Control Growth of GDP Average annual growth of real GDP per 2001 WDI, capita World Bank Institutions Bureaucratic quality 0-6 index PRS; ICRG Dates : 1982, 1990 Government Rule of law 0-6 index Easterly Dates : 1982, 1990 Levine (1997) Institutions Freedom from Corruption 1-7 index Easterly Dates : 1982, 1990 Levine (1997) Government Repression of Civil Liberties 1-7 index Gastil (1990), Dates : 1980, 1990 Gastil (1987). Literacy Percent of population literate Dates : Banks (1984) 1980 Illitercy Percentage of "no schooling" in Barro and Lee population (1993) Enrollment Gross enrollment ratio for higher Barro and Lee education (1993) Enrollment Gross enrollment ratio for secondary Barro and Lee education (1993) Enrollment Gross enrollment ratio for primary Barro and Lee education (1993) Life expectancy Life expectancy at age zero Barro and Lee at age zero (1993) Health Daily calorie intake. Could be used as a World Bank's measure of poverty. BESD database Health Daily protein intake (grams). World Bank's BESD database Health services Number of hospital beds per thousand World Bank's availability inhabitants Dates : 1980, 1990 BESD database Roadways Paved Roads/Highways 2001 WDI, Dates : 1980, 1990 World Bank Railroad Mileage per square mile 2001 WDI,

13 Chapter One: Introduction Variables Description Source Dates : 1980, 1990 World Bank Average Income Gini coefficient for income that can Deininger and Inequality range from a low of 0 to a high of 100 Squire (1996) Poorest Bottom quintile in income distribution Deininger and 1980, 1990; Average Squire (1996) Richest top quintile in income distribution Deininger and 1980, 1990; Average Squire (1996) Gender Female to male average schooling years, Barro and Lee Inequality age 26+ 1980, 1990; (1993) Balance of Balance of Payments as a percentage of 2001 WDI, Payments GDP World Bank 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 Derived from supply during the last five years minus 2001 WDI, the potential growth rate of real GDP World Bank Inflation Standard deviation of the annual inflation Derived from rate during the last five years 2001 WDI, World Bank Consumption Household, government, as a percentage 2001 WDI, of GDP and their growth World Bank Genuine savings Savings as a percentage of GDP 2001 WDI, World Bank Interest Rates Real, nominal, interest rate spreads 2001 WDI, World Bank Government (Government consumption/GDP) 2001 WDI, Size World Bank Takings Transfers and subsidies as a percent of Gwartney and GDP Lawson, 1997 International Difference between the official exchange Gwartney and exchange rate and the black market rate Lawson, 1997 International Actual size of the trade sector compared Gwartney and exchange to the expected size Lawson, 1997

14 Chapter One: Introduction Fig. 1.7 Layout of the thesis and interdependencies among the chapters

1 Introduction

2.1 Catastrophe Definition Perspectives

2.2 2.3 3 and 5 2.4 How do development How do catastrophes Catastrophes and Methods of processes determine affect development interaction between Coping vulnerability? processes? regional economies Aid and Recovery

3 4 Theoretical models Empirical Data 3.3 5-5.2 Economic growth, output Macro-economic Model 3 Simulation and consumption Indicators Interacting Regions results from a regional model

3.2 3.3 4.2-6 Model 1 Model 2 Effect of direct loss on 5.3-7 Ramsey's growth Model including economic indicators Case Studies model efficiency of reconstruction including growth Loma Prieta, Hurricane Andrew Northridge Earthquake

6 Summary Conclusions Future work

15 Chapter One: Introduction

16 Chapter One: Introduction

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.

17 Chapter Two: Catastrophes and Development

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.

18 Chapter Two: Catastrophes and Development

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

19 Chapter Two: Catastrophes and Development

“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-

20 Chapter Two: Catastrophes and Development

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.

21 Chapter Two: Catastrophes and Development

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.

22 Chapter Two: Catastrophes and Development

23 Chapter Two: Catastrophes and Development

Fig 2.1 Determinants of Macro- and Micro- Vulnerability

Overall Vulnerability

Openess to Poverty and Urbanization Large scale Phase of world economy population Forces poor development development growth to occupy vulnerable projects regions

Vagaries of Health, Poor construction quality Poor Fragile international Nutrition, and infrastructure social markets Education facilities networks

Low levels Inadequate of disaster labor supply Sloppy Lack of Lack of awareness after event residential disaster insurance, credit and preparedness construction recovery facilities Inadequate savings

Low levels of heath Household Susceptible to Vulnerability disease after event

24 Chapter Two: Catastrophes and Development

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

25 Chapter Two: Catastrophes and Development

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.

26 Chapter Two: Catastrophes and Development

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 2 1000.00% R = 0.4143

100.00%

10.00% GDP 1.00%

0.10%

Annual economicloss a % as of 0.01% 100 1,000 10,000 100,000 Per capita income (current US$)

Fig. 2.3 Greater per capita income is associated with smaller population affected R2 = 0.20, N= 117 100.0000%

10.0000%

1.0000%

0.1000%

0.0100%

0.0010%

affected (loglogscale) 0.0001% Percentage of population of population Percentage 0.0000% 100 1,000 10,000 100,000 Per capita income (current US$)

27 Chapter Two: Catastrophes and Development

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.

28 Chapter Two: Catastrophes and Development

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

29 Chapter Two: Catastrophes and Development

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.

30 Chapter Two: Catastrophes and Development

Fig. 2.4 Larger negative external balance on goods and services is associated with larger losses R2 = 0.2201 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0% (35.0) (30.0) (25.0) (20.0) (15.0) (10.0) (5.0) - 5.0 10.0

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 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0% - 2.0 4.0 6.0 8.0 10.0 12.0

Government repudiation of contracts

31 Chapter Two: Catastrophes and Development

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

32 Chapter Two: Catastrophes and Development

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.

33 Chapter Two: Catastrophes and Development

Fig. 2.6 Higher population growth is associated with larger losses

R2 = 0.3359 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss%as a of GDP

0.0% - 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Average population grow th

Fig. 2.7 Larger urban population growth is associated with larger losse s R2 = 0.3471 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss%as a of GDP

0.0% 012345678 Urban population grow th (annual %)

34 Chapter Two: Catastrophes and Development

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.

35 Chapter Two: Catastrophes and Development

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.

36 Chapter Two: Catastrophes and Development

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 losses R2 = 0.1595 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0% - 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

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

37 Chapter Two: Catastrophes and Development

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.

38 Chapter Two: Catastrophes and Development

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 losses R2 = 0.4922 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0% - 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Electric pow er consumption (kw h per capita)

39 Chapter Two: Catastrophes and Development

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 losses R2 = 0.2931 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0% - 1234567 Bureaucratic quality (1-poor, 6-best)

40 Chapter Two: Catastrophes and Development

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 losses R2 = 0.271 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss%as a of GDP

0.0% - 1234567 Rule of law (1- Poor, 6-Good)

Fig. 2.12 Higher prevalence of corruption is associated with larger losse s R2 = 0.1743 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0% - 1234567 Corruption (1-High, 6-Low )

41 Chapter Two: Catastrophes and Development

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

42 Chapter Two: Catastrophes and Development

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.

43 Chapter Two: Catastrophes and Development

Fig. 2.13 Greater uncertainty in economic growth is associated with larger losses R2 = 0.1324 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0% 1.0 10.0 Standard deviation of annual economic grow th

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

R2 = 0.115 1000.00%

100.00%

10.00%

1.00%

0.10% Economic loss%as a of GDP

0.01% (0.5) - 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Log of Pre-event standard deviation of inflation

44 Chapter Two: Catastrophes and Development

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 45 Chapter Two: Catastrophes and Development

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.

46 Chapter Two: Catastrophes and Development

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

47 Chapter Two: Catastrophes and Development

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 losses R2 = 0.4001 1000.0%

100.0%

10.0%

1.0%

0.1% Economicloss a % as of GDP

0.0% - 0.5 1.0 1.5 2.0 2.5 Pre-event secondary school enrollment

Fig. 2.16 Availability of physicians is associated with a decrease in the losses R2 = 0.3035 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0% (2.5) (2.0) (1.5) (1.0) (0.5) - 0.5 1.0

Pre-event physicians (per 1,000 people, log) 48 Chapter Two: Catastrophes and Development

Fig. 2.17 Higher infant mortality rate is associated with greater losses

R2 = 0.4048 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss%as a of GDP

0.0% - 0.5 1.0 1.5 2.0 2.5 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 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss%as a of GDP 0.0% 1 10 100 1,000 10,000 # of hospital beds per thousand

49 Chapter Two: Catastrophes and Development

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

50 Chapter Two: Catastrophes and Development

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 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0%

1,000Daily calorie intake 10,000

Fig. 2.20 Greater daily protein intake is associated with smaller losses R2 = 0.304 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss as a % of GDP

0.0% -20406080100120 daily protein intake (grams)

51 Chapter Two: Catastrophes and Development

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

52 Chapter Two: Catastrophes and Development

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,

53 Chapter Two: Catastrophes and Development

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.

54 Chapter Two: Catastrophes and Development

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.

55 Chapter Two: Catastrophes and Development

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

56 Chapter Two: Catastrophes and Development

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 57 Chapter Two: Catastrophes and Development

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.

58 Chapter Two: Catastrophes and Development

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).

59 Chapter Two: Catastrophes and Development

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

60 Chapter Two: Catastrophes and Development

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

61 Chapter Two: Catastrophes and Development

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

62 Chapter Two: Catastrophes and Development

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

63 Chapter Two: Catastrophes and Development

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.

64 Chapter Two: Catastrophes and Development

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.

65 Chapter Two: Catastrophes and Development

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.

66 Chapter Two: Catastrophes and Development

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

67 Chapter Two: Catastrophes and Development

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

68 Chapter Two: Catastrophes and Development

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.

69 Chapter Two: Catastrophes and Development

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 1000.0%

100.0%

10.0%

1.0%

0.1% Economic loss%as a of GDP

0.0% (2.0) (1.5) (1.0) (0.5) - 0.5 1.0 1.5 2.0 2.5

Aid (% of central government expenditures)

70 Chapter Two: Catastrophes and Development

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

71 Chapter Two: Catastrophes and Development

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

72 Chapter Two: Catastrophes and Development

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

73 Chapter Two: Catastrophes and Development

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.

74 Chapter Two: Catastrophes and Development

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.

75 Chapter Two: Catastrophes and Development

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 76 Chapter Two: Catastrophes and Development

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.

77 Chapter Two: Catastrophes and Development

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,

78 Chapter Three: Theoretical Models

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

79 Chapter Three: Theoretical Models

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:

∞ U = ∫ e −ρt u(c)dt (3.2.1) 0 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: c1+γ u(c) = (3.2.2) 1+ γ 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:

∞ −ρt V (k0 ) = max e u(c)dt (3.2.3) c(t) ∫ 0 s.t. k& = f (k) − c − βk (3.2.4)

* k0 = k (3.2.5)

Eq. 3.2.4 represents the fact that the change in capital stock, k& , occurs due to output, f (k) , 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: f (k) = Ak σ (3.2.6)

80 Chapter Three: Theoretical Models

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: c& = c /γ (ρ + β − f ′(k)) (3.2.7) k& = f (k) − c − βk (3.2.8)

At equilibrium, c& = k& = 0 . This implies that the equilibrium values of c and k, denoted by c* and k*, respectively, are given by:

1  ρ + β  σ −1 k * =   (3.2.9)  Aσ  c* = f (k * ) − βk * (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:

σ → σ (1+ εh1 (t)) (3.2.11a)

A → A(1+ εh2 (t)) (3.2.11b)

β → β + εh3 (t) (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:

σ (1+εh1 (t))−1 c& = c /γ [ρ + β + εh3 (t) − Aσ (1+ εh2 (t))(1+ εh1 (t))k (3.2.12)

σ (1+εh1 (t)) k& = [A(1+ εh2 (t))k − c − (β + εh3 (t))k] (3.2.13) h3 (t) = εa3δ (t −τ ) (3.2.14)

81 Chapter Three: Theoretical Models

δ(t-τ) is the Dirac Delta function. To make the dependence on ε explicit the equations are written as:

σ (1+εh1 (t))−1 ct (t,ε) = c(t,ε) /γ [ρ + β + εh3 (t) − Aσ (1+ εh2 (t))(1+ εh1 (t))k (3.2.15)

σ (1+εh1 (t)) kt (t,ε) = [A(1+ εh2 (t))k(t,ε) − c(t,ε) − (β + εh3 (t))k(t,ε)] (3.2.16) k(0,ε) = k * (3.2.17a) limt→∞ k(t) < ∞ (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: c (t,0) = −c* /γAσ (σ −1)(k * )σ −2 k (t,0) + εt ε (3.2.18) * * σ −1 2 * c /γ [h3 (t) − Aσ (k ) {}h1 (t) + h2 (t) + h1 (t)σLog(k ) ] k (t,0) = −c (t,0) + k (t,0){Aσ (k*)σ −1 − β} + εt ε ε (3.2.19) k*{−h (t) − β + Aσ (k*)σ −1h (t)Log(k*)}+ h (t)A(k*)σ 3 1 2 In matrix form:

* * c&ε (t,0)  0 − c /γf ′′(k )cε (t,0)   =   +    *   k&ε (t,0) −1 f ′(k ) − β kε (t,0) (3.2.20)  c* /γ [h (t) − Aσ (k * )σ −1 h (t) + h (t) + h 2 (t)σLog(k * ) ]   3 {}1 2 1   * * σ −1 * * σ  k {−h3 (t) − β + Aσ (k ) h1 (t)Log(k )}+ h2 (t)A(k )  Using Laplace transform

 C (s)   C (s)  s ε  = J ε  +  K (s)  K (s)  ε   ε  (3.2.21) c (0,0) + c* / γ [H (s) − Aσ (k * )σ −1 H (s) + H (s) + H 2 (s)σLog(k * ) ]  ε 3 {}1 2 1   * * σ −1 * * σ   kε (0,0) + k {−H 3 (s) − β + Aσ (k ) H1 (s)Log(k )} + H 2 (s)A(k )  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

82 Chapter Three: Theoretical Models

* 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:

* * σ −1 2 *  Cε (s) c (0,0) + c / γ [H (s) − Aσ (k ) {H (s) + H (s) + H (s)σLog(k )}]   = (sI − J) −1  ε 3 1 2 1     * * σ −1 * * σ   K ε (s)  k {−H 3 (s) − β + Aσ (k ) H1 (s)Log(k )} + H 2 (s)A(k )  (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: µ = (ρ + ρ 2 − 4c ss /γf ′′(k ss ))/ 2 (3.2.23) since at steady state, ρ = f ′(k ss ) − β 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

1  s − J 22 J12    (3.2.24) (s − µ)(s −ξ )  J 21 s − J11 

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:

83 Chapter Three: Theoretical Models

 s − J 22 J12     J 21 s − J11 

c (0,0) + c* /γ [H (s) − Aσ (k * )σ −1{}H (s) + H (s) + H 2 (s)σLog(k * ) ] 0  ε 3 1 2 1  =    * * σ −1 * * σ     k {−H 3 (s) − β + Aσ (k ) H1 (s)Log(k )}+ H 2 (s)A(k )  0 (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: c (0,0) = −(µ − J )[k *{−H (s) − β + Aσ (k * )σ −1 H (s)Log(k * )}+ H (s)A(k * )σ ]/ J ε 11 3 1 2 12 * * σ −1 2 * − c /γ [H 3 (s) − Aσ (k ) {}H1 (s) + H 2 (s) + H1 (s)σLog(k ) ] (3.2.26)

Substituting the values for J11 (=0) and J21 (=-1), the following equation results: c (0,0) = µ[k *{−H (s) − β + Aσ (k * )σ −1 H (s)Log(k * )}+ H (s)A(k * )σ ] ε 3 1 2 (3.2.27) * * σ −1 2 * − c /γ [H 3 (s) − Aσ (k ) {}H1 (s) + H 2 (s) + H1 (s)σLog(k ) ] -τµ Simplifying the above expression and substituting H3(µ) = e , c (0,0) = H (µ) f ′(k * )[c* /γ (1+ H (µ)σLog(k * )) + µk * Log(k * )] ε 1 1 (3.2.28) * * * −τµ * * * + H 2 (µ)[µf (k ) + c /γf ′(k )] − e [c /γ + µk ] − βk 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 c* /γ is the measure of coefficient of absolute constant risk aversion. The term f ′(k * ) is the marginal product of capital at steady state. This measures the price of capital at steady state. µ µ measures the after-event change in productivity, that is, H 1( ) 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 84 Chapter Three: Theoretical Models

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 1 K (s) = ε (s − µ)(s −ξ ) * * σ −1 2 * (3.2.29) − ()cε (0,0) + c /γ [H 3 (µ) − Aσ (k ) {}H1 (µ) + H 2 (µ) + H1 (µ)σLog(k )   * * σ −1 * * σ  + s()k {−H 3 (µ) − β / µ + Aσ (k ) H1 (µ)Log(k )}+ H 2 (µ)A(k )  Eq.3.2.29 can be written as:

2 (s − s(µ + ξ ) + µξ)K ε (s) = * * σ −1 2 * − ()cε (0,0) + c /γ [H 3 (µ) − Aσ (k ) {}H1 (s) + H 2 (s) + H1 (s)σLog(k ) (3.2.30) * * σ −1 * * σ + s()k {−H 3 (s) − β / s + Aσ (k ) H1 (s)Log(k )} + H 2 (s)A(k ) Taking limit of s → ∞ and using the relations:

2 limit(s Kε (s) − skε (0,0)) = k&ε (0) (3.2.31) s→∞ limit sKε (s) = kε (0) (3.2.32) s→∞ the following relation for initial investment can be obtained:

* * * * Iε (0) = −cε (0,0) + k (−h3 (0) − β + f ′(k )Log(k )h1 (0)) + f (k )h2 (0) (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.

* '' * −1 adj(sI − A) 1 s − ρ − c /γf (k ) ()sI − A = =   (3.2.34) det(sI − A) (s − µ)(s −ξ )  −1 s  Taking the inverse Laplace transform, the following results:

 1 s − ρ − c* /γf '' (k ss ) etA = L−1 sI − A −1 = L−1   (3.2.35) ()     (s − µ)(s −ξ )  −1 s 

85 Chapter Three: Theoretical Models

'' *  1 µt ξt f (k ) µt ξt   [](µ − ρ)e + (ρ −ξ )e − []e − e  etA = (µ −ξ ) γ ()µ −ξ  (3.2.36) 1 1  − []e µt − eξt [](µ + ξ )e µt + ξeξt   ()µ −ξ µ  The solution can be written as follows:

cε (t) tA cε (0)   = e   + kε (t) kε (0) (3.2.37) t  c* /γ []h (t) − Aσ (k * )σ −1{}h (t) + h (t) + h 2 (t)σLog(k * )  etA e sA  3 1 2 1 ds ∫  * * σ −1 * * σ  0 k {}− h3 (t) − β + Aσ (k ) h1 (t)Log(k ) + h1 (t)A(k )  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:

∞ U (ε) = ∫ e −ρt u(c(t,ε))dt (3.2.38) 0 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

dU ∞ = c* e −ρt c (t,0)dt (3.2.39) ∫ ε dε 0

* γ +1 * dU ()c f ''(k ) * * * = []− k (H (ρ) + β ) + f (k )(σLog(k )H (ρ) + H (ρ)) dε γ (ρ − µ)(ρ −ξ ) 3 1 2 (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

86 Chapter Three: Theoretical Models

* * 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:

* * c(t,1) ≅ f (k ) − βk + cε (t,0) (3.2.41)

* k(t,1) ≅ k + kε (t,0) (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: f (k) = ρ /σk σ (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 87 Chapter Three: Theoretical Models

Fig. 3.1a Changes in capital share in the production function 0% Change in capital share 5% 25% 10% 15% 20% 20% 15% 10% 5% Capital share 0% -5% Time of occurrence of the event -10% 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Time (in years)

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. 3.1b Effect of loss intensity on initial consumption

0.004 0.002 0 -0.002 -0.004 -0.006 -0.008 -0.01 -0.012

Initial consumption change -0.014 0 0.05 0.1 0.15 0.2 0.25 loss ratio

88 Chapter Three: Theoretical Models

permanent shifts in the capital share in the production function. The total factor productivity (TFP) change is modeled using: h2 (t) = a1 sin(t)[UnitStep(t − (τ + 0.5)) −UnitStep(t − 3)] (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: h3 (t) = a3δ (t −τ ) (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. 89 Chapter Three: Theoretical Models

Fig 3.1c Evolution of consumption

9% 8% 7% 6% 5% 4% 0% Change in capital share 3%

Consumption 5% 2% 10% 15% 1% 20% 0% 00.511.522.533.5 Time

Fig 3.1d Changes in capital assets

0.22

0.20

0.18

0.16

0.14 0% Change in capital share 5%

Productive capital 10% 0.12 15% 20%

0.10 0 0.5 1 1.5 2 2.5 3 3.5 Time

90 Chapter Three: Theoretical Models

Fig 3.1e Evolution of output

0.69 0.68 0.67 0.66 0.65 0.64 Output 0.63 0% Change in capital share 5% 0.62 10% 0.61 15% 20% 0.6 00.511.522.533.544.5 Time

3.1f Changes in economic growth over time

3.5% 0% Change in capital share 3.0% 5% 10% 2.5% 15% 2.0% 20% 1.5% 1.0% 0.5% 0.0%

Economic growth Economic -0.5% 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 -1.0% -1.5% -2.0% Time

91 Chapter Three: Theoretical Models

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.

92 Chapter Three: Theoretical Models

Fig. 3.2a Changes in capital share in the production function

10%

0% Capital share 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 -2%

-4%

-6% Time

Fig 3.2b Changes in total factor productivity with time

0.12

0.1

0.08

0.06

0.04

0.02

0 00.511.522.533.544.5

93 Chapter Three: Theoretical Models

Fig 3.2c Evolution of consumption

0.09 0.08 0.07 0.06 0.05 0.04 Loss ratio = 0% 0.03

Consumption 5% 0.02 10% 15% 0.01 20% 0.00 00.511.522.533.5 Time

Fig 3.2d Changes in capital assets

0.24

0.22

0.20

0.18

0.16 Loss ratio = 0% 5% 0.14 10% Productive capital 15% 0.12 20%

0.10 0 0.5 1 1.5 2 2.5 3 3.5 Time

94 Chapter Three: Theoretical Models

Fig 3.2e Evolution of output

0.72

0.7

0.68

0.66 Output 0.64 Loss ratio = 0% 5% 10% 0.62 15% 20% 0.6 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Time

3.2f Changes in economic growth over time

4% Loss ratio = 0% 5% 3% 10% 15% 2% 20%

0% 00.511.522.533.544.5

Economic growth Economic -1%

-2%

-3% Time

95 Chapter Three: Theoretical Models

Fig. 3.2g How does consumption vary with changes in capital? 0.09

Loss ratio = 0% 0.08 5% 10% 15% 20% 0.07

0.06

Consumption 0.05

0.04

0.03 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 Capital

Fig3.2h Changes in welfare

0 0.2 0.4 0.6 0.8 1 -0.3 -0.32 -0.34 -0.36 -0.38 -0.4 -0.42 -0.44 -0.46 Present value of utility -0.48 -0.5 Loss ratios

96 Chapter Three: Theoretical Models

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:

∞ −ρt V (k0 ) = max e u(c)dt (3.3.1) c(t) ∫ 0 such that:

k&1 = f (k2 ) − c −αk1 (3.3.2)

k&2 = αk1 − βk2 (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

97 Chapter Three: Theoretical Models

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:

σ f (k) = sk 2 (3.3.4)

The current value Hamiltonian for Eqs.3.3.1-3 is given by

H = u(c) + λ1 ()f (k2 ) − c −αk1 + λ2 (αk1 − βk2 ) (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:

λ&1 = (ρ +α)λ1 −αλ2 (3.3.6)

λ&2 = − f '(k2 )λ1 + (ρ + β )λ2 (3.3.7)

H c = 0 ⇒ u'(c) − λ1 = 0 (3.3.8) Assuming the utility function as in Eq.2.2 the following relationship is obtained:

1/ γ c = λ1 (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:

α → α(1+ εh1 (t)) (3.3.10a) external aid → εh2 (t) (3.3.10b)

β → β + εh3 (t) (3.3.10c)

σ → σ (1+ εh4 (t)) (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

98 Chapter Three: Theoretical Models

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:

σ (1+εh4 (t)) 1/ γ k&1 (t,ε) = sk2 (t,ε) − λ1 (t,ε) − (α + εh1 (t))k1 (t,ε) + εh2 (t) (3.3.11) k&2 (t,ε) = (α + εh1 (t))k1 (t,ε) − (β + εh3 (t))k2 (t,ε) (3.3.12)

λ&1 (t,ε) = (ρ +α(1+ εh1 (t)))λ1 (t,ε) − (α + εh1 (t))λ2 (t,ε) (3.3.13)

σ (1+εh4 (t))−1 λ&2 (t,ε) = −sσ (1+ εh4 (t))k2 (t,ε) λ1 (t,ε) + (ρ − β 2 − εδ (t −τ ))λ2 (t,ε) (3.3.14) Differentiating the above system with respect to ε, and evaluating at ε=0, the following system of equations result:

* * 1/ γ −1  k&1 (t,0)  −α f '(k2 ) −1/γ (λ1 ) 0  k1ε (t,0)       k (t,0)  k&2 (t,0)  α − β 0 0  2ε    = + λ& (t,0)  0 0 (ρ +α) −α  λ (t,0)   1    1ε    ss * * λ (t,0) λ&2 (t,0)  0 f "(k2 )λ1 f '(k2 ) (ρ + β ) 2ε  (3.3.15)  σf (k * )h (t) −αk *h (t) + h (t)   2 4 1 1 2  * *  αk1 h1 (t) − h3 (t)k2   αh (t)(λ* − λ* )   1 1 2   * * * *  h3 (t)λ2 − f '(k2 )h4 (t)(1+ σLog(k2 ))λ1  Performing the Laplace transform:

 Κ& (s)  k (0,0) + σf (k * )h (t) −αk *h (t) + h (t)   1ε   1ε 2 4 1 1 2  * * Κ& 2ε (s) −1  k2ε (0,0) +αk1 h1 (t) − h3 (t)k2  = (sI − J)  * *  (3.3.16)  Λ& (s)  λ (0,0) +αh (t)(λ − λ )  1ε   1ε 1 1 2     * * * *   Λ& 2ε (s) λ2ε (0,0) + h3 (t)λ2 − f '(k2 )h4 (t)(1+ σLog(k2 ))λ1 

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:

99 Chapter Three: Theoretical Models

 σf (k * )Log(k * )H (µ) −αk * H (µ) + H (µ)  0  2 2 4 1 1 2     αk * H (µ) − H (µ)k *  0 adj(sI − J) 1 1 3 2 = (3.3.17)  λ (0,0) +αH (µ)(λ* − λ* )  0  1ε 1 1 2     * * * *    λ2ε (0,0) + H 3 (µ)λ2 − f '(k2 )H 4 (µ)(1+ σLog(k2 ))λ1  0

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.

3.3.2 Numerical Experiments

The following Cobb-Douglas from of production function is chosen: f (k) = sk σ (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): h2 (t) = a2 sin(t)[]UnitStep(t − (τ + 0.5)) −UnitStep(t − 3) (3.3.19) A similar function is used to model the changes in rate at which maturing capital is converted to productive factor: h1 (t) = a1 sin(t)[]UnitStep(t − (τ + 0.5)) −UnitStep(t − 3) (3.3.20) 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.

100 Chapter Three: Theoretical Models

Fig. 3.3a Changes in total factor productivity

25% 0% change in TFP 20% 5% 10% 15% 15% 20%

10% Capital share

0% 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Time

Fig. 3.3b Changes in aid and capital share in the production function 3% 2% 1% 0% -1% -2% -3%

Capital share aid -4% capital share -5% -6% 00.511.522.533.544.5 Time

101 Chapter Three: Theoretical Models

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.

102 Chapter Three: Theoretical Models

Fig 3.3c Changes in maturing capital

0.25 Conversion rate 0% 5% 0.20 10% 15% 20% 0.15

0.10 Maturing capital 0.05

0.00 0 0.5 1 1.5 2 2.5 3 3.5 Time

Fig 3.3d Changes in productive capital

0.05

0.00

-0.05

Conversion rate 0% -0.10 5% 10%

Productive capital 15% 20% -0.15

-0.20 00.511.522.533.5 Time

103 Chapter Three: Theoretical Models

Fig 3.3e Evolution of consumption

0.00

-0.10

-0.20

-0.30

-0.40

Consumption Conversion rate 0% -0.50 5% 10% -0.60 15% 20% -0.70 0123456 Time

Fig 3.3f Evolution of output

1.18

1.17 Conversion rate 0% 1.16 5% 10% 1.15 15% 20% 1.14

1.13 Output

1.12

1.11

1.1

1.09 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Time

104 Chapter Three: Theoretical Models

3.3g Changes in economic growth over time

4.5% Conversion rate 0% 4.0% 5% 10% 3.5% 15% 20% 3.0%

2.5%

2.0%

1.5% Economic growth Economic 1.0%

0.5%

0.0% 0123456 Time

105 Chapter Three: Theoretical Models

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 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.

106 Chapter Three: Theoretical Models

Fig. 3.4a Changes in capital share in the production function

3.0%

2.5%

2.0%

1.5%

1.0% Capital share

0.5%

0.0% 012345 Time

Fig. 3.4b Changes in aid and total factor productivity

12% 10% 8% 6% 4% 2%

Capital share 0% -2% 012345 -4% -6% Time

107 Chapter Three: Theoretical Models

Fig 3.4c Changes in maturing capital

0.80 Loss ratio = 0.0 0.60 5% 10% 15% 0.40 20%

0.20

0.00 00.511.522.533.5

Maturing capital -0.20

-0.40

-0.60 Time

Fig 3.4d Changes in productive capital

0.50 0.40 Loss ratio = 0.0% 5% 0.30 10% 15% 0.20 20% 0.10 0.00 -0.10 00.511.522.533.5

Productive capital -0.20 -0.30 -0.40 Time

Fig 3.4e Evolution of consumption

1.00

0.50

0.00 00.511.522.533.5 -0.50

Consumption Loss ratio = 0.0% 5% -1.00 10% 15% 20% -1.50 Time 108 Chapter Three: Theoretical Models

Fig 3.4f Evolution of output

1.18 1.16 Loss ratio = 0.0% 1.14 5% 10% 1.12 15% 20% 1.1 Output 1.08 1.06 1.04 00.511.522.53 Time

3.4g Changes in economic growth over time

5% Loss ratio = 0.0% 5% 4% 10% 15% 3% 20%

0% Economic growth Economic 00.511.522.53 -1%

-2% Time

Fig. 3.4h Effect of loss intensity on initial consumption and welfare 0.1 0.05 0 -0.05 0 0.05 0.1 0.15 0.2 0.25 -0.1 -0.15 Welfare -0.2 InitialConsumption -0.25 -0.3 -0.35 loss ratio 109 Chapter Three: Theoretical Models

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:

∞ −ρt V (k0 ) = max e (u(c1 ) + u(c2 ))dt (3.4.1) c (t),c (t) ∫ 1 2 0 subject to the following conditions: k&1 = a11 f (k1 ) + a12 f (k1 ) − c1 − βk1 (3.4.2) k&2 = a21 f (k1 ) + a22 f (k1 ) − c2 − βk2 (3.4.3) The current value Hamiltonian is: H = u(c ) + u(c ) + λ (a f (k ) + a f (k ) − c − βk ) 1 2 1 11 1 12 1 1 1 (3.4.4) + λ2 (a21 f (k1 ) + a22 f (k1 ) − c2 − βk2 )

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,

110 Chapter Three: Theoretical Models

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: H = 0 ⇒ u'(c ) − λ = 0 and H = 0 ⇒ u'(c ) − λ = 0 (3.4.5) c1 1 1 c2 2 2 Differentiating the relations in Eq.3.4.5 with respect to time we have:

λ&1 = u"(c1 )c&1 and λ&2 = u"(c2 )c&2 (3.4.6) The co-state equations are:

' λ&1 = ρλ1 − λ1 (a11 f (k1 ) − β ) − λ2 a21 f '(k1 ) (3.4.7)

' λ&2 = ρλ2 − λ1a12 f (k2 ) − λ2 (a22 f '(k2 ) − β ) (3.4.8) Using the relations in Eq.3.4.6 in Eqs. 3.4.7 and 3.4.8, the following equations result:

γ c1 ' c2 c&1 = [ρ + β − a11 f (k1 )] − γ −1 a21 f '(k1 ) (3.4.9) γ γc1

γ c2 ' c1 c&2 = [ρ + β − a22 f (k2 )] − γ −1 a12 f '(k2 ) (3.4.10) γ γc2 The following perturbations are introduced to model the behavior of the economies after a catastrophic event in Region 1.

σ → σ (1+ εh1 (t)) (3.4.11a) a12 → a12 (1+ εh2 (t)) (3.4.11b) a21 → a21 (1− εh2 (t)) (3.4.11c)

β → β + εh3 (t) (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:

111 Chapter Three: Theoretical Models

σ (1+εh1 (t)) k&1 (t,ε) = a11sk1 (t,ε) + a12 (1+ εh1 (t)) f (k2 (t,ε)) − c1 (t,ε) − (β + εh2 (t))k1 (t,ε) (3.4.12)

σ (1+εh1 (t)) k&2 (t,ε) = a21sk1 (t,ε) + a22 (1− εh1 (t)) f (k2 (t,ε)) − c2 (t,ε) − βk2 (t,ε) (3.4.13) c (t,ε) 1 σ (1+εh1 (t))−1 c&1 (t,ε) = [ρ + β + εh3 (t) − a11sσ (1+ εh1 (t))k1 (t,ε) ] γ (3.4.14) c (t,ε)γ 2 σ (1+εh1 (t))−1 + γ −1 a21sσ (1+ εh1 (t))k1 (t,ε) γc1 (t,ε)

c2 (t,ε) σ −1 c&2 (t,ε) = [ρ + β − a22 sσ (1− εh2 (t))k2 (t,ε) ] γ γ (3.4.15) c1 (t,ε) σ −1 + γ −1 a12 (1+ εh2 (t))sσk2 (t,ε) γc2 (t,ε) Differentiating the above system (Eqs.3.4.12-15) with respect to the perturbation ε around the steady state, the following equations result:  c (t,0)   c (t,0)   &1ε   1ε  c&2ε (t,0) c2ε (t,0) = J + d (3.4.16)  k& (t,0)  k (t,0)  1ε   1ε      k&2ε (t,0) k2ε (t,0) where the elements of the matrix J and vector d are given below:

* 1 * c1 γ J11 = [ρ + β − f ′(k1 )(a11 + (γ −1)a21 ( * ) )] (3.4.17a) γ c2

* * c1 γ J12 = −a21 f ′(k1 )( * ) (3.4.17b) c2

* ′′ * f (k1 ) * * c2 γ −1 J13 = − [a11c1 + a21c2 ( * ) ] (3.4.17c) γ c1

J14 = 0 (3.4.17d)

* * c1 γ J 21 = −a12 f ′(k2 )( * ) (3.4.18a) c2

* 1 * c1 γ J 22 = [ρ + β − f ′(k2 )(a22 + (γ −1)a12 ( * ) )] (3.4.18b) γ c2

J 23 = 0 (3.4.18c)

* ′′ * f (k2 ) * * c1 γ −1 J 24 = − [a22c2 + a12c1 ( * ) ] (3.4.18d) γ c2 112 Chapter Three: Theoretical Models

J 31 = −1 (3.4.19a)

J 32 = 0 (3.4.19b)

J = a f ′(k * ) − β (3.4.19c) 33 11 1 J = a f ′(k * ) (3.4.19d) 34 12 2

J 41 = 0 (3.4.20a)

J 42 = −1 (3.4.20b) J = a f ′(k * ) (3.4.20c) 43 21 2 J = a f ′(k * ) − β (3.4.20d) 44 22 2 The components of the vector d (Eq.3.4.16) are shown below.

* * ′ * c1 f (k1 ) * * * c2 γ −1 d1 = h1 (t) − (1+ σLog(k1 ))(a11c1 + a21c2 ( * ) ) (3.4.21a) γ γ c1

* ′ * f (k2 ) * * c1 γ −1 d 2 = h1 (t)(a22c2 + a12c1 ( * ) ) (3.4.21b) γ c2

* * * * d3 = a12 f (k2 )h2 (t) − h3 (t)k1 + a11σf (k1 )h1 (t)Log(k1 ) (3.4.21c)

* * * d 4 = a21σf (k1 )h1 (t)Log(k1 ) − a22 f (k2 )h2 (t) (3.4.21d)

Performing the Laplace transform:  C& (s)   c (0,0) + d   1ε   1ε 1   C& 2ε (s)  −1 c2ε (0,0) + d 2    = (sI − J) (3.4.22) Κ& (s)  k (0,0) + d   1ε   1ε 3      Κ& 2ε (s) k2ε (0,0) + d 4 

3.4.1 Numerical Experiments

The following Cobb-Douglas from of production function is chosen: f (k) = sk σ (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

113 Chapter Three: Theoretical Models

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: h2 (t) = a2 sin(t)[]UnitStep(t − (τ + 0.5)) −UnitStep(t − 3) (3.4.24) A similar function is used to model the changes in rate at which maturing capital is converted to productive factor: h1 (t) = a1 sin(t)[]UnitStep(t − (τ + 0.5)) −UnitStep(t − 3) (3.4.25) 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

114 Chapter Three: Theoretical Models

(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

1.00% Increase in TFP of affected region: 0% 5% 0.00% 10% 15% 20% -1.00%

-2.00%

-3.00%

-4.00% Consumption indisaster region -5.00% 012345678 Time

Fig 3.5b Evolution of consumption in the adjacent region

3.00% Increase in TFP of affected region: 0% 2.00% 5% 10% 1.00% 15% 20% 0.00% -1.00% -2.00% -3.00% -4.00% -5.00%

Consumptionin adjacent region -6.00% 012345678 Time

115 Chapter Three: Theoretical Models

Fig 3.5c Changes in capital in the affected region

0.10

0.08

0.06

0.04

0.02

0.00 Increase in TFP of affected region: 0% 5% 10% -0.02 15% 20%

Capitalin the affectedregion -0.04

-0.06 012345678 Time

Fig 3.5d Changes in capital of the unaffected region

0.20 Increase in TFP of affected region: 0% 5% 0.15 10% 15% 0.10 20%

0.05

0.00

-0.05

-0.10

-0.15 Capital in the unaffectecd region -0.20 012345678 Time

116 Chapter Three: Theoretical Models

Fig. 3.5e How does consumption vary with changes in capital of the affected region? 0.01

0 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1

-0.01

-0.02

-0.03

Increase in TFP of affected region: 0%

Consumption indisaster region 5% -0.04 10% 15% 20%

-0.05 Capital in disaster region

Fig. 3.5f How does consumption vary with changes in capital of the unaffected region? 0.03 Increase in TFP of affected region: 0% 5% 0.02 10% 15% 20% 0.01

0 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 -0.01

-0.02

-0.03

-0.04 Consumption in unaffected region unaffected in Consumption -0.05

-0.06 Capital in unaffected region

117 Chapter Three: Theoretical Models

Fig 3.5g Output in the disaster region

0.85

0.84

0.83

0.82

0.81 Increase in TFP of affected region: 0% 0.80 5% 10% 15%

Output in disaster region 0.79 20%

0.78 012345678 Time

Fig 3.5h Economic growth in the disaster region

3.50% Increase in TFP of affected region: 0% 3.00% 5% 10% 2.50% 15% 20% 2.00%

1.50% 1.00%

0.50% 0.00%

Growth in disaster region -0.50% -1.00% 012345678 Time

118 Chapter Three: Theoretical Models

Fig 3.5i Output in the adjacent region

1.28 Increase in TFP of affected region: 0% 1.27 5% 10% 1.27 15% 1.26 20% 1.26 1.25 1.25 1.24 1.24

Output in the adjacent region 1.23 012345678 Time

Fig 3.5j Economic growth in the adjacent region

1.00%

0.50%

0.00%

-0.50%

-1.00% Increase in TFP of affected region: 0% -1.50% 5% 10% -2.00% 15% 20% -2.50% Growth inthe adjacent region -3.00% 012345678 Time

119 Chapter Three: Theoretical Models

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.

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 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.

120 Chapter Three: Theoretical Models

Fig 3.6a Evolution of consumption in the disaster region

0.04

0.02

0.00

-0.02

-0.04

-0.06 Loss ratio = 0.0 5% -0.08 10%

Consumption indisaster region 15% 20% -0.10 012345678 Time

Fig 3.6b Evolution of consumption in the adjacent region

0.03 Loss ratio = 0.0 0.02 5% 10% 0.01 15% 20% 0.00

-0.01

-0.02

-0.03

-0.04 Consumptionin adjacent region -0.05 012345678 Time

121 Chapter Three: Theoretical Models

Fig 3.6c Changes in capital in the affected region

0.30 Loss ratio = 0.0 0.25 5% 10% 0.20 15% 20% 0.15 0.10

0.05 0.00

-0.05

Capital in the affected region affected the in Capital -0.10

-0.15 012345678 Time

Fig 3.6d Changes in capital of the unaffected region

4.0% Loss ratio = 0.0 3.0% 5% 10% 2.0% 15% 20% 1.0% 0.0% -1.0% -2.0% -3.0% -4.0% -5.0% Capital in the unaffectecd region -6.0% 012345678 Time

122 Chapter Three: Theoretical Models

Fig. 3.5e How does consumption vary with changes in capital of the affected region? 0.04 Loss ratio = 0.0 5% 0.02 10% 15% 20% 0 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3

-0.02

-0.04

-0.06 Consumption indisaster region -0.08

-0.1 Capital in disaster region

Fig. 3.5f How does consumption vary with changes in capital of the unaffected region? 0.03

Loss ratio = 0.0 0.02 5% 10% 15% 0.01 20%

0 -0.06 -0.04 -0.02 0 0.02 0.04 -0.01

-0.02

-0.03 Consumption in unaffected region unaffected in Consumption -0.04

-0.05 Capital in unaffected region

123 Chapter Three: Theoretical Models

Fig 3.6g Output in the disaster region

0.95 Loss ratio = 0.0 5% 0.90 10% 15% 20%

0.85

0.80

0.75 Output indisaster region

0.70 012345678 Time

Fig 3.6h Economic growth in the disaster region

7.0% 6.0% Loss ratio = 0.0 5% 5.0% 10% 15% 4.0% 20% 3.0% 2.0% 1.0% 0.0% -1.0% -2.0% Growth in disaster region -3.0% -4.0% 012345678 Time

124 Chapter Three: Theoretical Models

Fig 3.6i Output in the adjacent region

1.26 Loss ratio = 0.0 5% 1.26 10% 15% 1.26 20%

1.25

1.25 Outputin the adjacentregion 1.25 012345678 Time

Fig 3.6j Economic growth in the adjacent region

0.8% 0.6% 0.4% 0.2% 0.0% -0.2% -0.4% Loss ratio = 0.0 -0.6% 5% 10% -0.8% 15% 20% -1.0%

Growth in the adjacent region -1.2% -1.4% 012345678 Time

125 Chapter Three: Theoretical Models

Fig3.6k Changes in welfare

0 0.05 0.1 0.15 0.2 0.25 0 Welfare in disaster region -0.1 Welfare in adjacent region -0.2

-0.3

-0.4

-0.5

-0.6

-0.7 Present value of utility -0.8

-0.9 Loss ratios

126 Chapter Three: Theoretical Models

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

127 Chapter Three: Theoretical Models

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.

128 Chapter Three: Theoretical Models

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, 129 Chapter Four: Empirical Studies

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.

130 Chapter Four: Empirical Studies

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.

131 Chapter Four: Empirical Studies

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 Average growth rate GDP 1980-88 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

132 Chapter Four: Empirical Studies

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.

133 Chapter Four: Empirical Studies

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.

134 Chapter Four: Empirical Studies

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.

135 Chapter Four: Empirical Studies

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.

136 Chapter Four: Empirical Studies

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). 137 Chapter Four: Empirical Studies

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.

138 Chapter Four: Empirical Studies

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

139 Chapter Four: Empirical Studies

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. 140 Chapter Four: Empirical Studies

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.

141 Chapter Four: Empirical Studies

Table 4.2 Summary statistics for short-term growth Pre event Post event Mean 3.96 3.29 Standard Error 0.30 0.33 Median 3.80 3.18 Standard Deviation 3.70 4.06 Sample Variance 13.67 16.46 Kurtosis 2.01 2.61 Skewness -0.16 -0.61 Range 23.36 26.60 Minimum -9.10 -12.57 Maximum 14.27 14.03 Sum 605.25 503.99 Count 153 153 Confidence Level(95.0%) 0.59 0.65

Fig 4.1 Event year growth is clearly lower than the pre-event year growth

100% 90% 80% 70% 60% 1yrBefore 50% Eve n t Ye ar 40% Percentile 30% 20% 10% 0% 0246810 GDP annual growth (%)

Fig 4.2 Average pre- and post-event growths

100% 90% 80% 70% Avg3yrsAfter 60% Avg3yrsBefore 50% 40%

Percentile 30% 20% 10% 0% 0246810 GDP annual growth (%)

142 Chapter Four: Empirical Studies

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 Average 3 years before years after Mean 3.83 3.55 Standard Error 0.27 0.24 Median 3.48 3.61 Standard Deviation 3.32 3.00 Sample Variance 11.03 9.02 Kurtosis 1.89 3.35 Skewness 0.15 -0.44 Range 22.96 22.66 Minimum -9.42 -10.20 Maximum 13.54 12.46 Sum 586.68 542.90 Count 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.

143 Chapter Four: Empirical Studies

Fig. 4.3 Effect of catastrophes on consumption

100

Post event consumption (% of GDP) 40 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Annual loss as a % of GDP

Fig. 4.4 Greater losses are associated with larger amount of government spending

45 40 35 30 25 20 of GDP 15 10 5 0 Post-event investment as a % 0.0% 0.1% 1.0% 10.0% 100.0% 1000.0% Annual loss as a % of GDP

Fig. 4.5 Greater losses are associated with larger amount of government spending 40 35 30 25 20 15 10 spending as a % of GDP Post event government 5 0 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Annual loss as a % of GDP 144 Chapter Four: Empirical Studies

Fig. 4.6 Greater losses are associated with higher openness

140

120

100

GDP 60

Post event openness as a % of 0 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Annual loss as a % of GDP

Fig. 4.7 Larger losses are associated with smaller post event savings 50 40 30 20 10

GDP 0 -10 -20 -30 Post event savings%as a of -40 0.0% 0.1% 1.0% 10.0% 100.0% 1000.0% Annual loss as a % of GDP

Fig. 4.8 Greater losses are associated with lower post event GDP per capita

100,000

10,000

1,000 equivalent adult Post-event GDP real per 100 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Annual loss as a % of GDP 145 Chapter Four: Empirical Studies

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

1000.0

100.0

10.0

Post Event Inflation 1.0

0.1 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Log(Loss/GDP)

146 Chapter Four: Empirical Studies

Fig. 4.10 Greater loss ratios are associated with higher post- event real interest rates

-5

Post event interest real rate -10 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Annual loss as a % of GDP

147 Chapter Four: Empirical Studies

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

148 Chapter Four: Empirical Studies 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.

149 Chapter Four: Empirical Studies

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

150 Chapter Four: Empirical Studies

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). 151 Chapter Four: Empirical Studies

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

152 Chapter Four: Empirical Studies

(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

153 Chapter Four: Empirical Studies 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

154 Chapter Four: Empirical Studies

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

155 Chapter Four: Empirical Studies

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

156 Chapter Four: Empirical Studies

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

157 Chapter Four: Empirical Studies

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.

158 Chapter Four: Empirical Studies

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

159 Chapter Four: Empirical Studies

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 160 Chapter Four: Empirical Studies

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. 161 Chapter Four: Empirical Studies

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 162 Chapter Four: Empirical Studies

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

163 Chapter Four: Empirical Studies 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.

Table 4.3 Result of a typical regression used for testing the negative association of economic loss to post-event growth Summary |R| 0.543 R2 0.295 R2 adjusted 0.271 Standard Error 3.486 # Points 151 PRESS 1963.96 R2 for Prediction 0.214 Durbin-Watson d 1.787 First Order Autocorrelation 0.106 Collinearity 0.334 Coefficient of Variation 105.286

ANOVA Source SS SS% MS F F Signif df Regression 736.81 29 147.36 12.12 8.030e-10 5 Residual 1762.30 71 12.154 145 Total 2499.1 100 150

Gr1yrAfter = b0 + b1*Log10(Loss/GDP) + b2*Log(1+TotoAff/Pop) + b3*Log10AvgGDP_per_capita + b4*COV_AnnualGrowth + b5*Log10Std_Dev_Inflation P value Std Error -95% 95% t Stat VIF b0 8.73 0.000 1.604 5.56 11.90 5.4 b1 -2.37 0.000 0.500 -3.36 -1.38 -4.7 2 b2 60.47 0.003 19.796 21.34 99.60 3.1 1 b3 -3.10 0.000 0.511 -4.11 -2.09 -6.1 2 b4 0.42 0.001 0.128 0.16 0.67 3.2 1 b5 -1.32 0.006 0.475 -2.25 -0.38 -2.8 1

164 Chapter Four: Empirical Studies

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

165 Chapter Four: Empirical Studies

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.

166 Chapter Four: Empirical Studies

Table 4.4 Result of a typical regression used for testing the negative association of economic loss to post-event average growth Summary |R| 0.633 R2 0.401 R2 adjusted 0.376 Standard Error 2.276 # Points 151 PRESS 862.82 R2 for Prediction 0.307 Durbin-Watson d 1.382 First Order Autocorrelation 0.308 Collinearity 0.311 Coefficient of Variation 60.682

ANOVA Source SS SS% MS F F Signif df Regression 499.10 40 83.18 16.05 4.387e-14 6 Residual 746.27 60 5.182 144 Total 1245.4 100 150

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 P value Std Error -95% 95% t Stat VIF b0 3.54 0.217 2.854 -2.10 9.18 1.2 b1 -1.68 0.000 0.327 -2.33 -1.04 -5.1 2 b2 38.22 0.005 13.264 12.01 64.44 2.9 1 b3 -2.72 0.000 0.336 -3.39 -2.06 -8.1 2 b4 -0.73 0.021 0.311 -1.34 -0.11 -2.3 1 b5 4.17 0.024 1.836 0.54 7.80 2.3 1 b6 0.31 0.000 0.084 0.14 0.47 3.6 1

167 Chapter Four: Empirical Studies

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.

Indicators of the monetary health of an economy, namely the inflation variability, the 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 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 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.

168 Chapter Four: Empirical Studies

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. 169 Chapter Four: Empirical Studies

Table 4.5 Effect of catastrophes on change in consumption Summary |R| 0.24 R2 0.06 R2 adjusted 0.05 Standard Error 0.04 # Points 150.00 PRESS 0.20 R2 for Prediction 0.03 Durbin-Watson d 1.87 First Order Autocorrelation 0.05 Collinearity 1.00 Coefficient of Variation 3.61

ANOVA Source SS SS% MS F F Signif df Regression 0.01234 6 0.01234 9.396 0.00259 1 Residual 0.194 94 0.00131 148 Total 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 1.046 121 b1 0.011 0.003 0.004 0.004 0.019 3.065

Table 4.6 Effect of catastrophes on investments Summary |R| 0.915 R2 0.837 R2 adjusted 0.835 Standard Error 2.477 # Points 149 PRESS 951.04 R2 for Prediction 0.827 Durbin-Watson d 1.508 First Order Autocorrelation 0.243 Collinearity 0.814 Coefficient of Variation 12.814

ANOVA Source SS SS% MS F F Signif df Regression 4592.7 84 2296.4 374.14 3.469e-58 2 Residual 896.11 16 6.138 146 Total 5488.8 100 148

AvgAfter = b0 + b1*Log10(Loss/GDP) + b2*AvgBefore P value Std Error -95% 95% t Stat VIF b0 1.998 0.0056 0.71 0.60 3.40 2.81 b1 -0.896 0.0020 0.28 -1.46 -0.33 -3.15 1.23 b2 0.791 0.0000 0.03 0.72 0.86 23.16 1.23 170 Chapter Four: Empirical Studies

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.

171 Chapter Four: Empirical Studies

Fig. 4.11 There is perceptible increase in real interest rates after a catastrophic loss

100% 90% 80% AvgBefore 70% AvgAfter 60% 50% 40% 30% 20% 10% 0% -5.0 0.0 5.0 10.0 15.0 20.0 Real Interest rates

Table 4.7 Effect of catastrophes on the real interest rates Summary |R| 0.581 R2 0.337 R2 adjusted 0.306 Standard Error 3.424 # Points 112 PRESS 1626.91 R2 for Prediction 0.132 Durbin-Watson d 1.311 First Order Autocorrelation 0.343 Collinearity 0.585 Coefficient of Variation 62.590

ANOVA Source SS SS% MS F F Signif df Regression 632.53 34 126.51 10.79 2.097e-08 5 Residual 1242.5 66 11.72 106 Total 1875.0 100 111

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 -0.09 b1 9.130 0.003 3.051 3.082 15.18 2.99 1.2 b2 3.526 0.000 0.756 2.028 5.025 4.67 1.2 b3 1.245 0.031 0.568 0.119 2.372 2.19 1.5 b4 -7.396 0.042 3.587 -14.51 -0.285 -2.06 1.2 b5 -0.316 0.004 0.106 -0.527 -0.105 -2.97 1.2 172 Chapter Four: Empirical Studies

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 173 Chapter Four: Empirical Studies

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,

174 Chapter Four: Empirical Studies

and in particular lags of income should not help predict consumption in period t. The expression for changes in consumption (Deaton, 1992:83) is:

∞ -k ∆ct = r/(1+r) Σ k=0 (1+r) (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.

175 Chapter Four: Empirical Studies

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 176 Chapter Four: Empirical Studies

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

177 Chapter Four: Empirical Studies

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 t R2 F N Savings ∆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 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 178 Chapter Four: Empirical Studies

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 t R2 F N Savings ∆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 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 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-

179 Chapter Four: Empirical Studies

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.

180 Chapter Four: Empirical Studies

• 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 catastrophes change ex ante saving behavior at least for two years after the event.

181 Chapter Four: Empirical Studies

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.

182 Chapter Four: Empirical Studies

Chapter Four

An Empirical Study of the Macro-economic Effects of Catastrophes Triggered by Natural Events

4. INTRODUCTION

130 Chapter Four: Empirical Studies

4.1 PREVIOUS STUDIES

131 Chapter Four: Empirical Studies

Taken together these studies produce ambiguous conclusions regarding the effect of catastrophes on ongoing economic processes.

132 Chapter Four: Empirical Studies

133 Chapter Four: Empirical Studies

4.2 CATASTROPHES AND ONGOING DEVELOPMENT PROCESSES

134 Chapter Four: Empirical Studies

135 Chapter Four: Empirical Studies

4.2.1 Change in Indicators Due to Catastrophes

136 Chapter Four: Empirical Studies

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.

138 Chapter Four: Empirical Studies

4.3 GENERAL FRAMEWORK AND ECONOMETRIC MODEL

* growthwith hazard = f(damage, productivity-changes; y ) (4.1)

Damage, in general, depends upon the intensity of the hazard and the vulnerability. Vulnerability is the susceptibility of the exposed constructed facilities, economic and

139 Chapter Four: Empirical Studies

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)

not correlated with the error term ε1. The reduced form given in Eq.4.3 is estimated.

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

141 Chapter Four: Empirical Studies

Fig 4.1 Event year growth is clearly lower than the pre-event year growth

100% 90% 80% 70% 60% 1yrBefore 50% Eve n t Ye ar 40% Percentile 30% 20% 10% 0% 0246810 GDP annual growth (%)

Fig 4.2 Average pre- and post-event growths

100% 90% 80% 70% Avg3yrsAfter 60% Avg3yrsBefore 50% 40%

Percentile 30% 20% 10% 0% 0246810 GDP annual growth (%)

142 Chapter Four: Empirical Studies

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.

4.4.2.2 Effect on consumption, investment, government expenditure, net exports and income

143 Chapter Four: Empirical Studies

Fig. 4.3 Effect of catastrophes on consumption

100

Post event consumption (% of GDP) 40 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Annual loss as a % of GDP

Fig. 4.4 Greater losses are associated with larger amount of government spending

45 40 35 30 25 20 of GDP 15 10 5 0 Post-event investment as a % 0.0% 0.1% 1.0% 10.0% 100.0% 1000.0% Annual loss as a % of GDP

Fig. 4.6 Greater losses are associated with higher openness

140

120

100

GDP 60

Post event openness as a % of 0 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Annual loss as a % of GDP

Fig. 4.7 Larger losses are associated with smaller post event savings 50 40 30 20 10

GDP 0 -10 -20 -30 Post event savings%as a of -40 0.0% 0.1% 1.0% 10.0% 100.0% 1000.0% Annual loss as a % of GDP

Fig. 4.8 Greater losses are associated with lower post event GDP per capita

100,000

10,000

1,000 equivalent adult Post-event GDP real per 100 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Annual loss as a % of GDP 145 Chapter Four: Empirical Studies

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.

Fig. 4.9 Larger losses are associated with higher inflation

1000.0

100.0

10.0

Post Event Inflation 1.0

0.1 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Log(Loss/GDP)

146 Chapter Four: Empirical Studies

Fig. 4.10 Greater loss ratios are associated with higher post- event real interest rates

-5

Post event interest real rate -10 0.01% 0.10% 1.00% 10.00% 100.00% 1000.00% Annual loss as a % of GDP

147 Chapter Four: Empirical Studies

4.4 EFFECT ON THE ECONOMIC GROWTH

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

148 Chapter Four: Empirical Studies community's assets. The direct loss is one measure of the change in community assets after a catastrophe.

4.4.1.2 Percentage affected

149 Chapter Four: Empirical Studies

4.4.1.3 Type of hazard

Location and climate have large effects on income levels and income growth, through their effects on transport costs, disease burdens, and agricultural productivity, among

150 Chapter Four: Empirical Studies

4.4.2 Control variables

4.4.2.1 Pre-existing Economic Conditions

4.4.2.2 Health

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

152 Chapter Four: Empirical Studies

4.4.2.4 Government, Bureaucracy, and Institutions

154 Chapter Four: Empirical Studies

155 Chapter Four: Empirical Studies

4.4.2.5 Infrastructure

156 Chapter Four: Empirical Studies

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

157 Chapter Four: Empirical Studies

The percentage of "no schooling" in the population is used as an indicator for education.

4.4.2.7 Trade

158 Chapter Four: Empirical Studies

4.5 INTRODUCTION TO ECONOMETRIC ISSUES

is therefore more reasonable to estimate yt at a given level of loss variables, as on average

159 Chapter Four: Empirical Studies

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

Algebraically, the stochastic nature of the relationship for the post event growth rate is represented as

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.

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

4.8 RESULTS FROM REGRESSION ANALYSIS

4.8.1 Growth rates – Short term

particular specification, the coefficient for the loss term (b1) is –2.37 and highly

ANOVA Source SS SS% MS F F Signif df Regression 736.81 29 147.36 12.12 8.030e-10 5 Residual 1762.30 71 12.154 145 Total 2499.1 100 150

164 Chapter Four: Empirical Studies

165 Chapter Four: Empirical Studies

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

166 Chapter Four: Empirical Studies

ANOVA Source SS SS% MS F F Signif df Regression 499.10 40 83.18 16.05 4.387e-14 6 Residual 746.27 60 5.182 144 Total 1245.4 100 150

167 Chapter Four: Empirical Studies

168 Chapter Four: Empirical Studies

4.9 EFFECT ON MAJOR ECONOMIC INDICATORS

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)

ANOVA Source SS SS% MS F F Signif df Regression 0.01234 6 0.01234 9.396 0.00259 1 Residual 0.194 94 0.00131 148 Total 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 1.046 121 b1 0.011 0.003 0.004 0.004 0.019 3.065

ANOVA Source SS SS% MS F F Signif df Regression 4592.7 84 2296.4 374.14 3.469e-58 2 Residual 896.11 16 6.138 146 Total 5488.8 100 148

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)

171 Chapter Four: Empirical Studies

Fig. 4.11 There is perceptible increase in real interest rates after a catastrophic loss

100% 90% 80% AvgBefore 70% AvgAfter 60% 50% 40% 30% 20% 10% 0% -5.0 0.0 5.0 10.0 15.0 20.0 Real Interest rates

ANOVA Source SS SS% MS F F Signif df Regression 632.53 34 126.51 10.79 2.097e-08 5 Residual 1242.5 66 11.72 106 Total 1875.0 100 111

4.10 Catastrophes and consumption smoothing

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

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

174 Chapter Four: Empirical Studies

and in particular lags of income should not help predict consumption in period t. The expression for changes in consumption (Deaton, 1992:83) is:

∞ -k ∆ct = r/(1+r) Σ k=0 (1+r) (Et+1 − Et)yt+k

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

175 Chapter Four: Empirical Studies

The instruments in the IV are ∆yt and ∆ct lagged twice. t-values are shown in brackets.

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

177 Chapter Four: Empirical Studies

If we use income lagged twice as an instrument in the regression on consumption change on lagged saving we essentially get the same results.

4.12 Conclusions, Extensions, and Limitations

179 Chapter Four: Empirical Studies

180 Chapter Four: Empirical Studies

181 Chapter Four: Empirical Studies

182 Chapter Four: Empirical Studies

Chapter Five

Regional Impact of Catastrophes

5. Introduction

The main purpose of this chapter is to study the regional impact of three catastrophes – 1989 Loma Prieta earthquake, 1992 Hurricane Andrew, and 1994 Northridge earthquake. Results from simulating a standard regional economic model are compared to the results obtained from the theoretical model presented in Section 3.5. The regional economic model is used for studying the effect of probable earthquake scenarios in the Bay Area.

An important question, in the field of disaster research, is what are the effects of catastrophic events such as earthquakes and hurricanes on a regional economy. A catastrophic event inflicts heavy damage to the capital assets of the affected community. This damage has consequences that can be measured at three levels: direct damages, indirect damages, and secondary effects (ECLAC, 1991). Direct losses are the damages to fixed assets (including property), capital and inventories of finished and semi-finished goods, raw materials and spare parts that occur as a direct consequence of the natural phenomenon triggering a catastrophe. Indirect damages relate to the effect on flows of goods that will whose supply and demand will be affected. They are measured in monetary, rather than physical terms. Secondary effects refer to the impact on overall economic performance as measured through important macro-economic variables. As such, they cannot be added mathematically to the sum total of direct and indirect damages. Relevant variables may include gross state product (GSP), net earnings of a county, the employment and unemployment levels, inflation, and the state of public finances.

Studies relating to the direct and indirect effects of catastrophes are extensive (FEMA, 1994), though much needs to be done. Methodologies have been developed to evaluate the direct and indirect effects. In particular, HAZUS (1999) is a software that brings

183 Chapter Five: Regional Impact of Catastrophes together state-of-the-art techniques from engineering and economics to estimate losses from earthquakes in US. It is being extended to include other hazards as well.

Several methodologies have been used to study the indirect effects of catastrophes. Roberts, Milliman and Ellson (1982) use a macro-econometric model for predicting the effects of an earthquake. In a macro-econometric model of a region, a baseline forecast of economic activity is generated. Shocking the exogenous variables of the model to yield a post-disaster forecast then simulates the earthquake and the impacts of the disaster are found by taking the difference between the baseline and post-disaster forecasts. However, the post-disaster impacts are based on coefficients, which were derived under factor supply conditions that are no longer relevant after a catastrophic event. Roberts, Milliman and Ellson (1982) circumvented this problem by joining engineering process models with an econometric modeling approach to produce an impact analysis that took into account the changes in the stocks of capital caused by the earthquake. This technique is extremely data-intensive.

To estimate direct and indirect losses from lifeline damage, Boisvert (1992) adopted a national input-output model to develop improved estimates of the losses. Input-output models are a description of the inter-industry flows in an economy. The critical assumption is that the money value of goods and services delivered by an industry to other producing sectors is a linear and homogenous function of the output level of the purchasing sectors. It is also assumed that the system is in equilibrium at given prices, constant returns to scale, and there is no substitution between inputs. Boisvert (1992) argues that for every billion dollars of direct damages to lifelines direct business losses were about $1.8 billion. Indirect business losses increase by three-quarters of a billion for every billion dollars’ increase in direct business losses. In most cases, the combined economic losses are less than one percent of the regional economies. Boisvert (1992) concludes that from a national perspective, it is unlikely that such losses will seriously disrupt markets.

184 Chapter Five: Regional Impact of Catastrophes Brookshire and McKee (1992) introduce Computable General Equilibrium models (CGE) for indirect loss measurement in a region. CGE models extend the framework of input- output models to include multiple households and to include substitution (not fixed) input possibilities in the production side and on the consumption side. The CGE models determine welfare changes that occur as a result of exogenous changes and shocks to the economic system being examined.

The following is the outline of this chapter. The next section gives an overview of the methodologies for studying the regional economic impacts. Section 5.2 mentions some difficulties in assessing the regional impact of catastrophes. Section 5.3 gives a description of the events considered – 1989 Loma Prieta earthquake, 1992 Hurricane Andrew, and 1994 Northridge earthquake. A comparative study of the impacts of these events is presented in Section 5.4. A standard regional economic model is used to simulate the economic effects of the three events and comparison of the results with observed values personal income are made in Section 5.5. Using this validated model, the behavior of the San Francisco Bay Area and Silicon Valley economies to scenario earthquakes is studied in Section 5.6 and 5.7. Section 5.8 concludes this chapter.

5.1 Methodologies used to study regional impacts

Cochrane (1992) using the income identity explains macro-economic effects of a catastrophe: Y = C + I + G + (EX - IM) (5.1) Y is the gross regional product, C is the spending on consumption, I is the spending on investment, G is the government spending, EX is the exports, and IM the imports in a region. The occurrence of a catastrophic earthquake would cause a loss of income for individuals who are laid-off or lose their jobs because of business interruptions and failures. The physical destruction of private and public facilities however would create new jobs in the construction industry as destroyed facilities are repaired and rebuilt. Losses in income for some individuals would therefore be offset by gains in income for others. Cochrane (1992) presents qualitative reasons for the effect of an earthquake on each of the variables. His conclusions are:

185 Chapter Five: Regional Impact of Catastrophes • The net effect on the aggregate income of a region is ambiguous. • Investments in other areas may be more expensive and consequently may decrease. • After reinvestment takes place, local incomes may be higher than before the event. • The way in which individual losses and gains come together to determine aggregate losses is a complex issue and unique to each set of local economic and physical conditions, not to mention the severity of the event that occurs.

The estimation of macro-economic effects can also be based on a comparison between the economic performance anticipated by public and private sector organizations and academic or consultant analysts (if they are available), and a modified projected performance estimated after the direct and indirect damages are assessed and valued. (ECLAC, 1991). This involves a with-without rather than before-after analysis of economic performance.

In assessing the overall macro-economic and social impact of a catastrophe consideration must be given to: i) the time frame in which the disaster occurs, such as its timing relative to agricultural cycle and, more broadly, to the economic cycle or any short or medium trends ii) dis-aggregation of the sectoral impacts, and iii) structure of the economy

Analytical techniques used to evaluate disaster losses and reconstruction gains have become increasingly sophisticated. As West and Lenze (1994) point out, the techniques have advanced from descriptive case studies (Haas, Kates, and Bowden 1977) to formal implementation of regression and time series techniques (Friesema et al. 1979; Chang 1983) to analysis based on regional econometric models (Ellson, Milliman, Poberts 1984; Guimaraes, Hefner, and Woodward 1993), input-output models (Cochrane 1992a; Boisvert 1992; Gordon and Richardson 1992), and computable general equilibrium models (Brookshire and McKee 1992).

186 Chapter Five: Regional Impact of Catastrophes 5.2 Modeling Problems

Modeling the impact of a catastrophe on a regional economy presents many problems. Estimates of the direct losses can only be given in probabilistic terms. This is in marked contrast to most impact studies that start with some firm numerical input on expenditure, employment, income, or tax rate changes. In addition, a catastrophe may affect many sectors of the economy simultaneously unlike an event like the inauguration of a theme park, which may affect only a specific sector(s). Extensive damages to physical facilities including infrastructure and buildings may cause direct supply interruptions to the affected region. Reconstruction immediately after the event increases the demand, which may strain regional capacity. Another problem is modeling household reactions to unanticipated destruction of homes, personal property, and neighborhoods. It is difficult to identify conditions under which households will resort to energetic rebuilding as opposed to migration from the impacted area. Changes of consumption patterns after a catastrophe destroys most of the wealth are not well documented.

A regional economic model originally developed by Treyz (1993) will be used in studying the regional impacts of catastrophes such as earthquakes and hurricanes. A separate Appendix J details the model and the associated MATLAB program. The model is validated by studying the effects of three recent catastrophes – 1989 Loma Prieta earthquake, 1992 Hurricane Andrew, and 1994 Northridge earthquake. Studying these historical events helps us to understand the complexity of behavior that a catastrophe can induce. The model, thus calibrated, is then used for studying the effect of probable earthquake scenarios in the Bay Area.

The model has five basic building blocks, namely, output, labor and capital, population and labor supply, wages, prices, and profits, and market shares.

Comprehensive modeling of regional impact of a catastrophe requires data at considerable levels of detail. There is a lack of data on direct exogenous and endogenous variable impacts as well as changes in ‘normal’ linkages between the output, labor and

187 Chapter Five: Regional Impact of Catastrophes capital, population and labor supply, wages, prices, and profits, and market shares modules. Ex post analysis of major events are to a certain extent amenable to analysis, since we have observed values of personal income to calibrate the model by changing the parameters. Ex ante analysis of probable events requires careful study by changing the parameters to establish reasonable bounds on the behavior of a regional economy.

5.4 Description of Events

5.4.1 Loma Prieta Earthquake

On October 17, 1989 at 5:04 p.m., an earthquake of a 7.1 magnitude struck the San Francisco Bay Area and its environs. The earthquake caused over $6 billion in direct property damage and disrupted transportation, communications and utilities. Brady and Perkins (1991) observed that workers were affected by layoffs for a maximum period of four months. The total number of workers affected by layoffs was 7100 (out of a total of 3 million jobs = 0.237 percent). This resulted in direct potential loss of wages and salaries of about $54 million, resulting in a minimum potential loss in gross output (including wages and salaries) of about $110 million during this period. The total economic disruption resulted in an estimated maximum potential Gross Regional Product (GRP) loss ranging from $735 million in one month to $2.9 billion over a maximum of two months following the event. However, at least 80 percent of that loss was recovered during the 1st and 2nd quarters of 1990. This implies maximum GRP lost ranges from $181 million to $725 million only. The 1989 GRP for the Bay area was $174 billion. The losses, when compared to the total size of the regional economy, can be viewed as isolated.

San Francisco (SF) experienced the greatest loss in retail activity for the 4th quarter. The damage and disruption to the Bay Bridge connecting SF to East Bay is a good indicator of how a major transportation network disruption could affect economic activity. Data analysis indicates the loss of approximately $73 million in taxable sales due to the closure of the Bay Bridge for several weeks. The damage of the Cypress Freeway was minimal because of alternative routes. However, the Bay Area Economic Forum and Metropolitan

188 Chapter Five: Regional Impact of Catastrophes transportation Commission, Oakland, CA, have documented economic impacts of approximately $20 million annually. This implies that major failure of infrastructure in a future quake will result in severe regional impacts. Santa Cruz County experienced 85 percent increase in unemployment insurance claims.

Table 5.1 Loma Prieta Earthquake - Damage observed (Brady and Perkins 1991): Homes Damaged 24347 Homes Destroyed 1119 Businesses damaged 4316 Businesses destroyed 382 Road Damage ($million) 833 Public utilities damage ($million) 43 PG&E losses ($million) 74

5.4.2 Hurricane Andrew

On August 24, Hurricane Andrew hit South Florida, resulting in the destruction of 85,000 dwelling units and buildings - nearly $23 billion (West and Lenze, 1994) physical damages -- leaving hundreds of thousands of people homeless. Major areas of impact were the residential suburbs south of downtown Miami. The following counties in South Florida were affected - Broward, Collier, Dade, and Monroe. The Dade County bore the brunt of the damages. Hurricane Andrew has been classified as one of the costliest weather related catastrophe to have hit US in the recent years. The impact on measured income was substantial. The hurricane reduced 1992 real income growth statewide in Florida from 1.8 percent to 1.0 percent, turning stable per capita income into real per capita income decline.

5.4.3 Northridge Earthquake

The Northridge earthquake of January 17, 1994 killed 57 people and injured an estimated 10,000. The counties affected were Los Angeles, Ventura, and Orange. The costs of repairing earthquake damage and providing relief to victims probably exceeded

189 Chapter Five: Regional Impact of Catastrophes $30 billion, including $12- 15 billion in insured losses, making that event the most costly disaster in U. S. history. The number of households and businesses that suffered losses in the Northridge earthquake far exceeded the size of the victim population in other recent major disasters in the U. S., including Hurricane Hugo in 1989 and Hurricane Andrew in 1992. The assistance effort launched after the earthquake was the largest ever undertaken for an U.S. disaster. Applications to the Federal Emergency Management Agency for various forms of housing assistance totaled well over half a million. In the year following the earthquake, over 50,000 businesses applied to the U.S. Small Business Administration for disaster loans, and over $1.3 billion in loans had been paid out.

5.5 A comparison of the impacts of the events

The gross state product of California in 1989 was $591 billion. The personal income of the worst affected county i.e. San Francisco – San Jose was $ 149 billion. The Loma Prieta earthquake occurred in one of the most technology advanced counties of California. Fortunately the disruption was minimal since the highway and rail network was not substantially affected. Businesses in downtown Santa Cruz were able to relocate to other areas of Santa Cruz County (Brady and Perkins, 1991). The direct losses were relatively minimal. The loss to county personal income ratio was around 4 per cent. The loss to gross state product ratio was around 1.1 per cent (Table 5.2).

The gross state product of California in 1994 was $722 billion. The personal income of the worst affected county i.e. Los Angeles was $205 billion. The Northridge earthquake occurred in one of the most prosperous counties of California. Perhaps, for this reason the direct economic losses were very high given the fact that the earthquake itself was not of a severe magnitude. The loss to county personal income ratio was around 12 per cent. The loss to gross state product ratio was around 4 per cent (Table 5.2).

190 Chapter Five: Regional Impact of Catastrophes

Table 5.2 Observations on the effects of the Loma Priets and Northridge earthquakes and Hurricane Andrew

Loma Prieta Earthquake Northridge Earthquake Hurricane Andrew Place of occurrence (state) California California Florida Date of occurrence 17-Oct-89 17-Jan-94 24-Aug-92 Estimated direct losses (upper bound) in dollars 7,000,000,000 30,000,000,000 31,000,000,000 Estimated direct losses (lower bound) in dollars 6,000,000,000 25,000,000,000 26,500,000,000 Gross state product in the year of the event 590,961,519,000 722,223,500,000 270,820,652,000

Personal earnings of the worst affected San Francisco-San Jose county in the year of the event CMSA Los Angeles Dade 148,574,598,000 204,872,592,000 34,303,921,000 Ratio of loss to gross state product 1.1% 3.8% 10.6% Percentage loss to county personal income (assuming 90% loss occurred in worst affected county) 3.9% 12.1% 75.4%

191 Chapter Five: Regional Impact of Catastrophes We can contrast Northridge earthquake to the occurrence of the Hurricane Andrew that occurred in the state of Florida. The gross state product of Florida, in 1992, was $271 billion, almost one-third of California. Yet, Hurricane Andrew resulted in approximately the same amount of dollar damages as Northridge Earthquake. Dade, the southern Florida County worst affected by Hurricane Andrew, had personal earnings of around $34 billion, almost a sixth of the net earnings of Los Angeles. The loss to county personal income ratio was around 75 percent. The loss to gross state product ratio was around 10 percent (Table 5.2).

San Francisco – San Jose CMSA contribution to the California's gross product remained almost unchanged if we compare the pre- and post - disaster trends. It was 25.31 percent in 1988, 25.14 percent in 1989, and 25.16 percent in 1990 (Table 5.3). Los Angeles contribution to the California's gross product remained almost unchanged if we compare the pre- and post - disaster trends. It was 28.66 percent in 1993, 28.37 percent in 1994, and 28.3 percent in 1995. The slight drop is the long-term trend, rather than the effect of Northridge earthquake (Table 5.3). Dade's contribution to Florida's gross product showed the effects of Hurricane Andrew. Dade's contribution was 13.81 percent in 1991, dropped to 12.7 percent in 1992, but rebounded to 13.5 percent in 1993 (Table 5.3).

The difference in the loss-to-Gross Regional Product ratios is partly responsible for the different ways the macro-economic variables such as net earnings are impacted. It can be readily inferred that a high loss-to-Gross Regional Product ratio results in a greater regional impact. Thus, Loma Prieta and Northridge earthquakes had relatively minimal impacts on the regional economies, whereas Hurricane Andrew’s was felt statewide.

192 Chapter Five: Regional Impact of Catastrophes

Table 5.3 Effect on county's components of personal income

Loma Prieta Earthquake Hurricane Andrew Northridge Earthquake 1988 1989 1990 1991 1992 1993 1993 1994 1995 Effect on county's contribution to GSP 25.3% 25.1% 25.2% 13.8% 12.7% 13.5% 28.7% 28.4% 28.3% (Net earnings by place of work)/Personal income State 70.4% 69.7% 69.3% 55.8% 57.1% 56.5% 67.1% 66.7% 65.8% Affected county 70.2% 69.8% 69.2% 61.5% 67.8% 63.1% 67.7% 67.2% 65.9% (Dividends, interest, and rent)/Personal income State 18.2% 18.7% 18.7% 27.0% 24.3% 25.0% 17.2% 17.5% 18.6% Affected county 19.6% 19.9% 20.2% 22.4% 12.7% 19.0% 16.6% 16.8% 18.3% (Transfer Payments)/Personal income State 12.7% 12.7% 13.1% 17.2% 18.7% 18.6% 15.7% 15.8% 15.6% Affected county 10.6% 10.8% 11.0% 16.8% 20.2% 18.6% 15.6% 16.0% 15.8%

193 Chapter Five: Regional Impact of Catastrophes

5.5.1 Effects on the components of personal income

The personal income of an area (BEA, 1997), is defined as the income that is received by, or on behalf of, all the individuals who live in the area. It is calculated as the sum of: the net earnings by place of work, the personal dividend income and the personal rental income, and the transfer payments. To delineate the effects of events like the Northridge earthquake from general economic trends, we compare the values of these components at county level to the state level. If we compare the trends of the California state to that of Los Angeles, they are almost indistinguishable (Table 5.3). For example, California earned 67.11 percent of its personal income from net earnings by place of work in 1993, 66.67 percent in 1994 and 65.82 percent in 1995. The trend in the affected county, in this case Los Angeles, is almost same - 67.71 percent in 1993, 67.22 percent in 1994, 65.92 per cent in 1995. Notice, however, that the transfer payments have slightly increased for Los Angeles in 1994 and 1995 (Table 5.3). This should be apparent since federal aid is accounted for under transfer payments. Similar trends can also be observed for the Loma Prieta earthquake.

In the case of Hurricane Andrew, the effects were more pronounced and different. There was a drop of about 10 percentage points in Dade's dividend, interest and rent component of the personal income though there was no appreciable change of the same component in Florida's personal income (Table 5.3). Florida's net earnings by place of work component of the personal income increased in 1992, but Dade's increase was higher. Details of this will become apparent if we examine the various components that make up the net earnings by place of work.

5.5.2 Effects on the components of net earnings by place of work

The net earnings by place of work have two components - farm and non-farm. In this paper, we do not consider the effects of the catastrophic events on farm earnings

194 Chapter Five: Regional Impact of Catastrophes since they form a small part of the San Francisco – San Jose CMSA, Los Angeles and Dade's earnings.

The components of non-farm earnings include earnings from construction, manufacturing, transportation, wholesale trade, retail trade, finance insurance and rental income, services and government.

The methodology used herein to study the effects on each of the sectors is by tracking the changes in the growth rates. For example, the growth rates for construction earnings for a period two years prior (pre-event) to the occurrence of the catastrophe and three years after (post-event) the occurrence of the catastrophe (including the year of occurrence) are calculated (Table 5.4). The mean of the pre-event growth rates is compared to the post event means.

From Table 5.4 it is clear that excluding the Loma Prieta earthquake, the events tended to cause similar changes in direction of the growth rates for all sectors except finance, insurance, and rental income. But the magnitude of change of growth rates was more intense in the case of Hurricane Andrew in the following sectors – construction, manufacturing, transportation, retail trade, governmental spending. In the case of Loma Prieta the trends were opposite, except for the transportation sector. Partial explanation for Bay Area’s opposite behavior is that U.S. economy was in a recession during the 1990-1991 period.

If we exclude the Loma Prieta case, all the sectors received a boost to their growth rates except the government and finance, insurance, rental income of Los Angeles. For example, the pre-event construction growth rate jumped from a negative 5.5 percent to a positive 6.4 percent.

But to answer the question as to whether these changes in growth rates were propagated to the state level we have to analyze taking into account the changes in the net earnings by place of work of the state.

195 Chapter Five: Regional Impact of Catastrophes

Table 5.4 Growth rates of the components of net earnings Pre- Post- Event event event t t Year t t t mean mean Direction -2 -1 0 1 2 Construction Los Angeles -9.5% -7.1% 12.7% 2.2% -1.8% -8.3% 4.4% increase Dade -0.4% -10.8% 3.8% 21.7% -6.1% -5.6% 6.5% increase SF-San Jose 5.6% 8.4% 4.2% 2.2% -8.6% 7.0% -0.7% decrease Manufacturing Los Angeles -3.5% -3.9% -2.7% -3.1% 2.9% -3.7% -1.0% increase Dade 0.4% 1.1% 4.0% 0.8% 1.8% 0.8% 2.2% increase SF-San Jose 5.3% 8.2% 6.9% 4.7% 4.7% 6.7% 5.4% decrease Transportation Los Angeles 3.5% 4.2% 4.7% 4.5% 3.0% 3.9% 4.1% increase Dade 8.9% -0.1% 0.4% 13.6% 5.7% 4.4% 6.6% increase SF-San Jose 5.0% 2.5% 4.7% 8.4% 3.9% 3.7% 5.7% increase Wholesale Los Angeles 2.6% -5.1% 2.9% 4.7% 1.5% -1.2% 3.0% increase Dade 7.1% 1.4% 7.0% 3.2% 3.5% 4.2% 4.6% increase SF-San Jose 5.9% 13.1% 8.0% 6.2% 1.3% 9.5% 5.2% decrease Retail Los Angeles 1.0% -1.9% 1.8% 3.7% 2.7% -0.4% 2.7% increase Dade 2.9% -3.1% 3.8% 10.6% 3.0% -0.1% 5.8% increase SF-San Jose 2.4% 7.0% 6.9% 4.8% 1.9% 4.7% 4.5% decrease Finance,insurance, and real estate Los Angeles 10.7% 8.4% -3.7% 2.1% 4.8% 9.6% 1.1% decrease Dade 1.5% -2.7% 14.1% 6.9% 3.8% -0.6% 8.3% increase SF-San Jose 9.1% 8.2% -1.7% 7.0% 3.6% 8.7% 3.0% decrease Services Los Angeles 4.5% 1.5% 1.6% 6.2% 6.8% 3.0% 4.9% increase Dade 6.6% 2.9% 7.9% 7.9% 4.8% 4.7% 6.9% increase SF-San Jose 10.9% 14.6% 9.5% 11.2% 4.9% 12.8% 8.5% decrease Government Los Angeles 4.3% -0.4% 1.2% 2.3% 1.0% 1.9% 1.5% decrease Dade 9.6% 6.4% -1.6% 7.1% 5.6% 8.0% 3.7% decrease SF-San Jose 7.8% 6.0% 6.8% 9.0% 5.7% 6.9% 7.2% increase

196 Chapter Five: Regional Impact of Catastrophes 5.5.3 Dampening out effect

The methodology adopted for this section relies on comparing the changes in the components of earnings at the state and county levels for the years immediately before and after the event (Table 5.5).

Hurricane Andrew affected the components that make up Dade’s net earnings by place of work and these changes were reflected in Florida’s earnings. For example, the construction component of earnings for Dade changes from 4.7% in 1992 to 5.3% in 1993, an increase of 12.5% while Florida’s increase was 3.8% (Table 5.5). An important observation in the case of Hurricane Andrew is that all the components of net earnings of Dade and Florida showed similar trends. One possible inference from this observation is that the effect of Hurricane Andrew was propagated to the state level, though these effects were considerably dampened at the state level.

The same cannot be said for the Northridge earthquake. Though the damage was of the same magnitude as that of hurricane Andrew, the robustness of Los Angeles economy was partly responsible for dampening out the effects at the county level itself. This can be inferred from the behavior of each of the components of net earnings (Table 5.5). For example, while the wholesale earnings of California dropped from 2.3% in 1994 to 1% in 1995, wholesale earnings in Los Angeles dropped by only 0.6 percentage points in the same period (Table 5.4). In the case of the Loma Prieta earthquake, the direct losses were relatively lower. The changes in the components were not propagated to the state level as is clear from Table 5.5.

197 Chapter Five: Regional Impact of Catastrophes Table 5.5 Effect on county's components of net earnings by place of work

Loma Prieta Earthquake Hurricane Andrew Northridge Earthquake Construction/Net earnings No dampening effect Dampened out Dampened out* State 1.0% -5.6% -6.6% 3.9% 5.6% -0.7% Affected county -2.2% -5.0% -2.4% 12.5% 11.6% -1.1% Manufacturing/Net earnings No dampening effect Dampened out Dampened out State -1.3% -4.4% -2.9% -4.9% -1.1% -2.0% Affected county 0.4% -2.7% -2.3% -6.8% -3.6% -6.2% Transportation/Net earningsNo dampening effect Dampened out Dampened out State 0.29% -0.28% -0.22% 2.19% 0.78% -0.90% Affected county -1.69% 0.76% -5.59% 5.04% 3.72% 1.17% Wholesale/Net earnings Dampened out Dampened out No Dampening observed# State 1.22% -0.17% 0.13% -3.95% 2.34% 0.96% Affected county 1.34% -1.26% 0.60% -4.56% 1.92% 1.29% Retail/net earnings Dampened out Dampened out No Dampening observed State -0.1% -2.8% -1.8% -0.5% 1.1% -1.6% Affected county 0.4% -2.6% -2.4% 2.3% 0.9% 0.3% Fin.,ins.,etc./net earnings No dampening effect Dampened out No Dampening observed State -9.3% -1.7% 8.1% 3.6% -4.6% -0.2% Affected county -7.7% -0.5% 7.2% -1.1% -4.6% -1.2% Services/net earnings No dampening effect Dampened out No Dampening observed State 2.7% 4.4% 1.1% 0.6% 0.1% 1.8% Affected county 2.8% 3.4% 1.5% -0.2% 0.7% 2.8% Government/net earnings No dampening effect Dampened out No Dampening observed State 0.9% 2.0% -3.4% -2.1% -0.1% -2.2% Affected county 0.2% 1.3% -6.4% -0.8% 0.2% -0.8%

Notes: *Dampened out here means that the state trend follows the affected county's trend but to a lesser degree # No Dampening observed implies that the state trend is different from the affected county's trend Components of earnings = Component earnings/total earnings

198 Chapter Five: Regional Impact of Catastrophes Hurricane Andrew and Northridge earthquake caused almost equal amounts of direct losses. But the effects were both similar and different. The effects were similar to the extent that almost all sectors of the economy received a boost to their growth rates, probably due to flow of external aid and reconstruction. The effects were different in the sense that Northridge earthquake occurred in a county with a robust economy as compared to Hurricane Andrew. The effects of Loma Prieta and Northridge earthquakes were localized, whereas effects of Hurricane Andrew were felt at the state level. Data presented herein suggests that shock to a robust economy, such as California's from a hazardous event of magnitude comparable to the Northridge earthquake, results in localization of the effects. A higher magnitude of shock may propagate the effects to a greater extent, but will be nevertheless localized. Hurricane Andrew was a greater shock in terms of the loss-GSP ratio. But the propagation of the shock was felt at the state level, though it dampened out rapidly both spatially and temporally.

5.6 Simulation of the effects with the regional model

In this section we describe the application of this model to simulate the impacts of three events on the regional economies - Loma Prieta Earthquake, Hurricane Andrew, and Northridge Earthquake. Appendix J lists a computer program in MATLAB© that simulates this model. Personal income, population, and employment in the affected regions before the occurrence of the catastrophe are shown in Table 5.6. Table 5.6 Main economic indicators before the events Loma Prieta Hurricane Northridge Earthquake Andrew Earthquake Personal income (billions of dollars) 138.8 35.8 204.8 Population (million number of persons) 6.1 2.0 9.1 Earnings by place of work (billions) 105.3 26.7 163 Dividends, interest, and rent (billions) 27.0 7.8 32.5 Transfer payments (billions) 14.7 6.0 31.5 Total full- and part-time employment 3.8 1.1 4.9 (million) Government and government 0.52 0.13 0.56 employment Average earnings per job (dollars) 27446 25208 32982 Construction Employment (million) 0.19 0.04 0.16

199 Chapter Five: Regional Impact of Catastrophes 5.6.1 Loma Prieta Earthquake

The baseline prediction without an earthquake is shown as the curve labeled as ‘baseline’ in the Fig. 5.1. Assuming the earthquake caused a direct loss of $6.2 billion and 7100 jobs were lost, the model forecasts the curve labeled ‘earthquake’ in the Fig. 5.1. It was also assumed that the regional capacity to satisfy regional demand fell by 2.5% (since the direct capital loss was around 3% of the pre-disaster level of capital). Subsequently the regional capacity to satisfy regional demand is assumed to reach its pre- disaster levels over a period of five years after the event. This curve assumes that no aid was given to the affected area.

The model generates a forecast, which is labeled ‘aid’ in Fig. 5.1, based on two assumptions. The first assumption is that $1.0 billion is used as transfer payments in the first two periods after the earthquake. The second assumption is that the regional capacity to satisfy regional demands falls by 2% (instead of 2.5% in no aid case) and recovers to its pre-disaster values within a period of years as a result of reconstruction efforts. If we compare the ‘aid’ curve with the actual observed values given by Bureau of Economic Analysis (BEA) the mean absolute percentage error is 2.59 (Table 5.7).

Table 5.7 Comparison of model predictions with observed values – Loma Prieta Earthquake Observed With Eq. Without Eq. With aid Personal % error % error % error Income ($ Billions) 1989 149 -5.75 -1.56 -3.80 1990 161 -6.23 -2.35 -4.28 1991 166 -5.47 -2.11 -4.54 1992 176 -5.02 -2.24 -4.06 1993 181 -2.20 -0.01 -1.16 1994 188 0.37 1.83 1.46 1995 200 1.47 2.49 2.19 1996 216 1.10 1.62 1.39 1997 233 -0.67 -0.26 -0.43

Mean 3.14 1.61 2.59 absolute % error

200 Chapter Five: Regional Impact of Catastrophes Fig. 5.2a shows plots of percentage changes of the gross regional product with respect to no-event scenarios. The curve labeled ‘without aid’ corresponds to difference in behavior with respect to no-event scenario of the gross regional product when no external aid is given. The curve labeled ‘with aid’ is a similar curve where aid in the form of transfer payments is assumed. In subsequent discussions, the comparison with a baseline no-event scenario is illustrated via similar plots. From Fig. 5.2a it is clear that without aid the GRP would have been lower by 5.39 percent during 1989. Because of transfer payments and improved regional capacity due to reconstruction, the GRP was lower by 3.78, a gain of 1.64 percent. With aid and increased regional capacity, the region recovers within 5 years, whereas without aid and slower increases in regional capacity, the region would have recovered in 7 years. Consumption shows similar trends (Fig. 5.2b). The consumption in no-aid case is lower by 4.18 percent whereas with aid it is lower by 2.2 percent. The theoretical model presented in Section 3.5 also shows similar trends in the change output and consumption. Immediately following the event, consumption and output fall. The theoretical models also predict that greater aid results in more rapid recovery. Increases in regional capacity are modeled by changes in productivity.

The reasons for lower GRP than the no-event scenario can be readily discerned when one plots the capital and employment as in Figs. 5.2c and 5.2d, respectively. Capital (Fig. 5.2c) is lower by about 4 percentage points three years after the event in the no-aid scenario, whereas it is lower by 3.5 percent with aid. The capital steadily converges to the no-event scenario. Employment (Fig. 5.2d) is lower by 5.37 percentage points in the no- aid scenario and is somewhat ameliorated with aid (is lower by 3.76 points).

Prices are modeled as consumption deflator (Fig. 5.2e). Lower the deflator higher is the real price. Without aid the deflator is lower by 0.08 to 0.14 percentage points during the first three years. With aid, prices are nearer the no-event case. With aid the deflator is lower by only 0.06 to 0.1 percentage points, implying that prices are lower than the case of no-aid, but higher than no-event case.

201 Chapter Five: Regional Impact of Catastrophes Fig. 5.1 Effect of Loma Prieta Earthquake on Personal Income of San Francisco - San Jose CMSA

240 230 Bas e line Earthquake Aid Actual(BEA) 220 210 200 190 180 170 160 150 140 Personal income (billions, nominal) 1989 1990 1991 1992 1993 1994 1995 1996 1997

202 Chapter Five: Regional Impact of Catastrophes Fig. 5.2a Effect on Gross Regional Product (Loma Prieta) 1

-1

-2

-3 without aid, Regional capacity = 0.975,0.98,0.985,0.99,0.995 -4 with aid, Regional capacity=0.98,0.985,0.99,0.995

-5

-6 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 year

Fig. 5.2b Effect on Consumption (Loma Prieta)

0 -0.5 -1 -1.5 -2 witho ut aid with aid -2.5 -3 -3.5 -4 -4.5 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 year

Fig. 5.2c Effect on Capital stock (Loma Prieta) 0 -0.5 -1 -1.5 -2 -2.5 -3 -3.5 Capital stock - without aid Capital stock - with aid -4 -4.5 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 year

203 Chapter Five: Regional Impact of Catastrophes Fig. 5.2d Effect on Employment (Loma Prieta) 1

-1

-2

witho ut aid with aid -3

-4

-5

-6 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 year

Fig. 5.2e Effect on Consumption Deflator (Loma Prieta)

-0.02

-0.04

-0.06

-0.08 witho ut aid with aid -0.1

-0.12

-0.14

-0.16 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 year

Fig. 5.2f Effect on Government spending (Loma Prieta)

-0.5

-1

witho ut aid with aid -1.5

-2

-2.5

-3 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 year

204 Chapter Five: Regional Impact of Catastrophes 5.6.2 Hurricane Andrew

Assuming the hurricane caused a direct loss of $26 billion and 10500 jobs were lost, the model forecasts the curve labeled ‘hurricane’ in the Fig. 5.3. This curve assumes that no aid was given to the affected area. It also assumes that the regional capacity to satisfy regional demand was down by 15% (as compared to the no-event scenario) during 1992-1993. Hurricane Andrew destroyed about 65% of the capital stock. But the region was operating at approximately 50% of its capacity during the pre-event period (West and Lenze, 1994). It was therefore assumed that without aid the region has the capacity to supply 85% of its local demand. This capacity is assumed to steadily rise to its full no- event value within a period of 9 years, without aid. With aid, the model generates a forecast, which is labeled ‘aid’ in Fig. 5.3. It is assumed that transfer payments of $0.268 billion in 1992 and $0.403 billion are made in the years 1992 and 1993 (West and Lenze, 1994). It is also assumed that reconstruction expenditures are as follows: $2.377 billion in 1992, $7.939 billion in 1993, $4.282 billion in 1994, $2.670 billion in 1995, and $0.045 billion in 1996 (West and Lenze, 1994). This is translated in jobs (thousands) as 3.846 in 1992, 28.472 in 1993, 27.904 in 1994, 22.288 in 1995, and 0.396 in 1996 (West and Lenze, 1994). The ‘aid’ curve compares well with the actual observed values given by Bureau of Economic Analysis (BEA): the mean absolute percentage error is 2.70 percent (Table 5.8). Table 5.8 Comparison of model predictions with observed values - Hurricane Andrew Personal With Without With aid Income Hurricane hurricane % error (BEA $ % error % error billions) 1992 34.3 -6.12 16.26 -2.55 1993 39.2 -11.82 6.51 -2.00 1994 40.5 -9.37 7.33 1.61 1995 42.5 -7.39 6.87 1.85 1996 44.7 -3.06 8.55 2.35 1997 46.2 0.53 10.90 5.83

Mean 6.38 9.40 2.70 Absolute Errors

From Fig. 5.4a it is clear that without aid the GRP would have been lower by 25 percent during 1992. Because of reconstruction spending, transfer payments and improved

205 Chapter Five: Regional Impact of Catastrophes regional capacity due to reconstruction, the GRP was lower by 20 percent in 1992 (a gain of 5 points as compared to no-aid case) and jumped to 8.1 percent in 1993 (a gain of 11.5 points). Consumption shows similar trends (Fig 5.4b). The consumption in no-aid case is lower by 18.9 percent whereas with aid it is lower by 15.9 percent.

Capital (Fig. 5.4c) is lower by about 8 percent in 1992 and falls steadily to 12.5 percent in 1995 after which it shows an upward trend in the no-aid scenario. With aid, however it is 6.8 percent lower in 1992 and 7.2 percent lower in 1993. The capital steadily converges to the no-event scenario. Employment (Fig. 5.4d) is lower by 25 percentage points in the no-aid scenario and aid helps in rapidly improving the situation. Employment is lower by 21 percent in 1992 but is only 8 percent in 1993.

Without aid the deflator is lower by 0.46 to 0.78 percentage points during the first three years. With aid the deflator is lower by 0.4 percentage points during first three years and converges to the no event case.

Fig 5.3 Effect of Hurricane Andrew on Personal Income of Dade county

30 Personal income (billions, nominal) 1992 1993 1994 1995 1996 1997

Bas e line Hurricane Aid Actual(BEA)

206 Chapter Five: Regional Impact of Catastrophes Fig. 5.4a Effect on Gross Regional Product (Andrew) 0

-5

-10 GRP GRP(with aid)

-15

-20

-25

-30 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 year

Fig. 5.4b Effect on Consumption (Andrew) 0 -2 -4 -6 -8 -10 Consumption Co ns.(with aid) -12 -14 -16 -18 -20 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 year

Fig. 5.4c Effect on Capital Stock (Andrew) 0

-2

-4

-6

-8

-10 Capital stock capital stock(with aid)

-12

-14 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 year

207 Chapter Five: Regional Impact of Catastrophes Fig. 5.4d Effect on Employment (Andrew)

-5

-10

-15 Employment Employment (with aid)

-20

-25

-30 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 year

Fig. 5.4e Effect on Price Index (CP) (Andrew)

0 -0.1 -0.2 -0.3 -0.4 Price index Price index(with aid) -0.5 -0.6 -0.7 -0.8 -0.9 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 year

Fig. 5.4f Effect on Government spending (CP) (Andrew)

-2

-4

-6

-8

-10

-12 Government spending Government spending(with aid) -14 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 year 208 Chapter Five: Regional Impact of Catastrophes Fig. 5.4g Effect on Investment (Andrew)

-10

-20 Witho ut aid with aid

-30

-40

-50 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 year

209 Chapter Five: Regional Impact of Catastrophes

5.6.3 Northridge Earthquake

Northridge earthquake caused a direct loss of $25 billion and 25,800 jobs were lost in the Los Angeles County. The model forecasts the curve labeled ‘earthquake’ in the Fig. 5.5. This curve assumes that no aid was given to the affected area. It also assumes that the regional capacity to satisfy regional demand was down by 5 percent (as compared to the no-event scenario) during 1992-1993. Hurricane Andrew destroyed about 7 percent of the capital stock. Assuming the region had idle capacity of 2% of the capital stock it can be assumed that without aid the regional capacity to satisfy regional demand was lower by 5 percent. This capacity is assumed to steadily rise to its full no-event value within a period of 6 years, without aid. With aid, the model generates a forecast, which is labeled ‘aid’ in Fig. 5.5. It is assumed that transfer-payments of $0.5 billion each in 1994 and 1995. It is also assumed that reconstruction expenditures resulted in job (in thousands) impacts as follows: 6 in 1994, 50 in 1995, 45 in 1996, 40 in 1997, and 20 in 1998. The ‘aid’ curve compares well with the actual observed values given by Bureau of Economic Analysis (BEA): the mean absolute percentage error is 1.3 percent (Table 5.9).

Table 5.9 Comparison of model predictions with observed values – Northridge Earthquake Percent Errors Personal With Eq. Without Eq. With aid Income (BEA $ billions) 1994 204 -7.52 0.87 -2.91 1995 214 -3.45 0.73 -0.77 1996 224 -2.72 0.39 -0.72 1997 234 -2.82 -0.50 -0.76

Mean 4.12 0.62 1.29 Absolute Errors

From Fig. 5.6a it is clear that without aid the GRP would have been lower by 11 percent during 1994. Because of reconstruction spending, transfer payments and improved regional capacity due to reconstruction, the GRP was lower by 5 percent in 1994 (a gain

210 Chapter Five: Regional Impact of Catastrophes of 6 points as compared to no-aid case). Consumption shows similar trends (Fig 5.6b). The consumption in no-aid case is lower by 8 percent whereas with aid it is lower by 4 percent.

Capital (Fig. 5.6c) is lower by about –1.8 percent in 1994 and falls steadily to –2.4 percent in 1996 after which it shows an upward trend in the no-aid scenario. With aid, however it is 0.8 percent lower in 1994 and 1.0 percent lower in 1995-96. The capital steadily converges to the no-event scenario. Employment (Fig. 5.6d) is lower by 10.7 percentage points in the no-aid scenario and aid helps in rapidly improving the situation. Employment is lower by 5 percent in 1994 but is only 1.7 percent in 1995.

Without aid the deflator is lower by 0.15 to 0.17 percentage points during the first three years. With aid the deflator is lower by 0.07 percentage points during first three years and converges to the no event case.

Having discussed historical events, the next section presents simulation results of the impacts of probable earthquake scenarios in the San Francisco – San Jose region (Bay Area).

Fig. 5.5 Effect of Northridge Earthquake on Personal Income of Los Angeles County

235 230 225 220 215 210 205 200 195 190 185 Personal income (billions, nominal) 1994 1995 1996 1997

Bas e line Earthquake Actual(BEA) With Aid

211 Chapter Five: Regional Impact of Catastrophes Fig. 5.6a Effect on Gross Regional Product (Northridge) 2

-2

-4 witho ut aid with aid

-6

-8

-10

-12 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 year

Fig. 5.6b Effect on Consumption (Northridge) 0 -1 -2

-3 witho ut aid with aid -4 -5 -6 -7 -8 -9 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 year

Fig. 5.6c Effect on Capital Stock (Northridge)

-0.5

-1

-1.5 without aid with aid

-2

-2.5 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 year

212 Chapter Five: Regional Impact of Catastrophes Fig. 5.6d Effect on Employment (Northridge)

-2

-4 witho ut aid with aid

-6

-8

-10

-12 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 year

Fig. 5.6e Effect on Consumer price CP (Northridge)

0 -0.02 -0.04

-0.06 witho ut aid with aid -0.08 -0.1 -0.12 -0.14 -0.16 -0.18 -0.2 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 year

Fig. 5.6f Effect on Government Spending (Northridge)

-0.5

-1

-1.5 witho ut aid with aid -2

-2.5

-3

-3.5

-4 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 year

213 Chapter Five: Regional Impact of Catastrophes

5.7 Simulation of impact of probable earthquake scenarios in the Bay Area

In this section results of simulation studies for various earthquake scenarios in the San Francisco Bay Area are presented. The model that was used to study the Loma Prieta earthquake is extended to include an scenario earthquake that occurs in the year 2000. A catastrophe abruptly increases the difference between optimal and actual capital levels. The influx of aid and insurance payments speeds up the adjustment process. But this additional spending should be carefully modeled so as to avoid double counting by including only the direct endogenous spending increase from reconstruction. Loss of capital stock available to the production process reduces regional output, employment, wage income, and proprietors’ income. Unless these losses are accounted for as exogenous changes, the model will reflect only reconstruction gains from a catastrophe. As restoration and reconstruction evolves, the regional output, employment, wage income, and proprietors’ income return to their ‘normal values’ and this has to be suitably modeled.

There are two possible ways in which this increased reconstruction investment can be modeled. If the purpose of the model is to examine an event ex post and we have data regarding the investment spending for reconstruction, then the job and output losses can be restored in conjunction with the time path of capital reconstruction.

In the absence of data regarding reconstruction expenditure, we assume that only a portion of the investment is used for building new structures and the rest is used for reconstruction. The recovery of a region depends crucially on the pattern of the reconstruction expenditure and the fractions apportioned to new buildings and restoration of damaged structures. The uncertainty surrounding these fractions can be studied by suitably changing some parameters. In the model proposed, suitable changes were made to α to reflect the true dynamics of the recovery process.

One important effect of a catastrophe is to cause a change in the relationships between local demands and demand for locally produced output. Regional patterns change

214 Chapter Five: Regional Impact of Catastrophes abruptly as the regional share in satisfying local demand decreases abruptly. As a result purchases made from outside the region increase. The decrease in local potential to satisfy local needs may prolong the deleterious effect of a catastrophe if region has no in- built excess capacity or receives no external aid or receives external aid that is not suitably used for reconstruction. The earthquake scenarios are generated based on assumptions of direct losses to capital stock and the number of jobs lost. These scenarios are presented in Table 5.10. Four possible scenarios are simulated. For each of these scenarios assumptions regarding regional capacity are made. For example, we assume regional capacity is lower by 7% (Table 9, row 4, col. 2) when 10% of the capital has been lost in the no aid case. When the affected region gets external aid, its capacity is augmented, and the regional capacity is lower by 3% only in the year 2000 (Table 9, row 4, col. 3). These scenarios are simulated and the results are presented in Figs. 5.7 to 5.11. The results show the importance to regional capacity in dampening out the effects of a catastrophe. For example (Fig. 5.8), in the 10 percent capital loss scenario, a regional capacity increase of 4 percent (due to aid, rapid reconstruction, or inherent pre-event excess capacity) results in absorbing 7 percent (14.4 – 7.1) of loss in the gross regional product during the year 2000. This motivates the study of the model by varying other crucial parameters that may change after a catastrophe. Such studies will be discussed in the next section.

Table 5.10 Earthquake scenarios and assumptions about regional capacity Loss Scenarios Regional 10% of capital 20% of capital 28% of capital 35% of capital ($ capacities ($ 30 billion) ($ 60 billion) ($ 80 billion) 100 billion) and and 25,000 jobs and 50,000 jobs and 75,000 jobs 100,000 jobs Year No aid aid No aid aid No aid aid No aid aid 2000 7 3 17 10 25 18 32 25 2001 5 1 15 5 20 12 27 10 2002 3 - 10 3 15 7 22 15 2003 1 - 5 1 10 3 17 10 2004 - - 3 - 5 1 12 5 2005 - - 1 - 3 - 7 1 2006 - - - - 1 - 3 - 2007 ------1 - Fig. No. 5.7 5.8 5.7 5.9 5.7 5.10 5.7 5.11

215 Chapter Five: Regional Impact of Catastrophes

Fig. 5.7 Effect on Gross Regional Product - Probable scenarios with no external aid 10

-10

-20

-30

-40

-50

-60 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year $30 billion and 25,000 jobs $60 billion and 50,000 jobs $80 billion and 75,000 jobs $100 billion and 100,000 jobs

Fig. 5.8 Effect on Gross Regional Product - Probable scenarios with and without aid (10% loss) 2 0 -2 -4 -6 -8 -10 -12 -14 -16 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year $30 billion and 25,000 jobs $30 billion and 25,000 jobs with aid

216 Chapter Five: Regional Impact of Catastrophes Fig. 5.9 Effect on Gross Regional Product - Probable scenarios with and without aid (loss = 20% of capital) 5 0 -5 -10 -15 -20 -25 -30 -35 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year $60 billion and 50,000 jobs $60 billion and 50,000jobs (with aid)

Fig. 5.10 Effect on Gross Regional Product - Probable scenarios with and without aid (loss = 28% of capital) 5 0 -5 -10 -15 -20 -25 -30 -35 -40 -45 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year $80 billion and 75,000 jobs $80 billion and 75,000jobs (with aid)

Fig. 5.11 Effect on Gross Regional Product - Probable scenarios with and without aid (loss = 35 % of capital) 0

-10

-20

-30

-40

-50

-60 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year

$100 billion and 100,000 jobs $100 billion and 100,000jobs (with aid)

217 Chapter Five: Regional Impact of Catastrophes

5.8 Model behavior when crucial parameters are varied

By varying the many parameters of the model in a controlled fashion one can get a better understanding of the underlying mechanism of the model. In this section the 2nd damage scenario (loss $60 billion and 50,000 jobs) is used.

5.8.1 Transfer payments effects (Fig. 5.12)

In personal income, transfer payments are income payments to persons for which no current services are performed. They are payments by government and business to individuals and non-profit institutions. Federal aid after a disaster is put in this category. The effect of making transfer payments of 10 percent of the capital loss during the years 2000, 2001, and 2002 is not very significant. Increases in the transfer payments improve the gross regional product to a small degree.

5.8.2 Consumer spending effects (Fig. 5.13)

Immediately after a catastrophe, propensity to consume increases, at least temporarily, in the affected region. Consumption spending as a proportion of the real disposable income increases. People rely on savings, credit, or insurance to finance their reconstruction spending. Increases in consumption spending result in an increase in the personal income for the region. Whether this reconstruction gain offsets the losses from reduction in regional output, employment, wage income, and proprietors’ income can be answered only by examining data from a specific event.

Consumer spending preferences are increased by 10 percent. Within one year of the event the economy does better than the baseline no earthquake scenario due to increased spending.

218 Chapter Five: Regional Impact of Catastrophes Fig. 5.12 Effect on gross re gional product due to 10% increase in transfer payments 5

-5

-10

-15

-20

-25 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year $60 billion and 50,000 job loss transfer payments 10% of loss

Fig. 5.13 Effect on gross re gional product due to 1% increase in consumer spending 10

-5

-10

-15

-20

-25 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year $60 billion and 50,000 jobs loss 1% increase in consumption spending

Fig. 5.14 Effect on gross re gional product due to 10% increase in government spending 5

-5

-10

-15

-20

-25 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year Govt. spending increases by 10% $60 billion and 50,000 job loss 219 Chapter Five: Regional Impact of Catastrophes

5.8.3 Government spending effects (Fig. 5.14)

The government spending preference usually increases temporarily for the affected region. This is modeled by increasing the value of government propensity to spend, for some periods (usually two to three years) after the event. As in most of the model parameters, almost no data is available for these temporary increases in spending preferences. One way to understand the behavior of the model in absence of data is to use reasonable bounds.

A 10 percent increase in governmental spending does not seem to affect the behavior of the regional economy, though small improvements can be discerned.

5.8.4 Labor Supply effects (Fig. 5.15)

A decrease in employment as a result of catastrophe reduces the exports from the local area. A decrease in employment decreases the change in the wage rate, which in turn decreases the wage rate as compared to the no change in occupational-employment- demand scenario. Decrease in the wage rate decreases the relative production costs and relative sales price for regional industries. Decrease in wage rate reduces the labor and proprietor’s income thus reducing the overall output, GRP.

When the occupational wage supply is decreased due to a catastrophe, the change in wage rate relative to no-event scenario is smaller. This reduces the wage rate relative to the no-event scenario and consequently reduces the labor and proprietors’ income. Reduced wages also imply that the optimal capital stock reduces which may lead to lower investment levels.

220 Chapter Five: Regional Impact of Catastrophes Lower wages implies that the relative labor intensity average will be higher than the baseline. As a result the labor-output ratio will increase prompting employment to increase.

Increase in the changes in wages via a 10% increase in employment does not seem to have much impact to the behavior.

Fig. 5.15 Effect on gross regional product due to 10% decrease in occupational employment 5

-5

-10

-15

-20

-25 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year 10 % decrease in occupational employment $60 billion and 50,000 jobs loss

Fig. 5.16 Effect on gross re gional product due to 1% increase in migration 5

-5

-10

-15

-20

-25 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year Increase in migration 1% of population $60 billion and 50,000 jobs loss

221 Chapter Five: Regional Impact of Catastrophes

5.8.5 Migration effects (Fig. 5.16)

Immediately after a catastrophe, in-migration to the affected region decreases partly because of lower relative employment opportunity (REO) in most sectors except construction. The relative wage ratio (RWR) may also be smaller for the affected region when compared to neighboring unaffected regions. Individuals are suddenly released from liquidity constraint of selling their current homes and as jobs in national firms are relocated outside the disaster region. This may induce out-migration. The effect of reduced migration is a decrease in population when compared to a no-event scenario. Smaller population implies lesser transfer payments, lesser dividends, interest, and rent, and lesser governmental spending. But lesser (relative to baseline no-event scenario) population results in lesser demand for housing, thus reducing relative housing price and also a smaller consumer price deflator. A lower consumer price deflator reduces the wage rate thus reducing the labor and proprietors’ income. Reduced wages also imply that the optimal capital stock reduces which may lead to lower investment levels. Lower wages implies that the relative labor intensity average will be higher than the baseline. As a result the labor-output ratio will increase prompting employment to increase.

If it is assumed that one-percent of population migrates after the occurrence of an earthquake, then there is a small decrease in gross regional product. This results from the fact that as more people migrate, the population of the region decreases, which in turn implies a lower income from dividends, interest, and rent, and government spending.

222 Chapter Five: Regional Impact of Catastrophes

5.8.6 Production or fuel costs (Fig. 5.17)

Damage to infrastructure, machinery, and buildings as a consequence of a catastrophe may temporarily increase the production costs. Equipment and raw materials may have to be brought from nearby unaffected regions. The effect of increase in production costs is similar to a decrease in factor productivity, explained previously.

If the catastrophe causes extensive damage to capital stock, the factor productivity is expected to drop, at least temporarily till the damaged machinery has been repaired and the damaged buildings have been reconstructed. This drop in factor productivity may result in increase of relative production costs. Increase in relative production costs reduces the relative profitability for national industries thus lowering the attractiveness of the region to new investment, at least temporarily. Drop in relative profitability reduces the region’s share in satisfying the region’s demand, thus lowering the overall output.

A 10% market surge in prices causes the economy causes a further drop of 5 percentage points in the GRP compared to no price-rise case. The market surge hurts the economy since it takes a longer time to converge to the no-event scenario.

5.8.7 Business Taxes and credits (Fig. 5.18)

The effect of a decrease in business taxes and credits is to decrease the relative production costs. The effect of decrease in production costs is similar to an increase in factor productivity. An increase in factor productivity will encourage investors since the relative profitability will increase. The region’s share in satisfying region’s demand will increase, thus increasing the overall output. Thus decreasing business taxes in the affected region may prove to be a good incentive to revive an economy. However, the exact amount of credits that has to be given will depend on the severity of economic loss,

223 Chapter Five: Regional Impact of Catastrophes the jobs lost, the extent to which public infrastructure in the affected region can support new businesses.

Credits given to businesses immediately after a catastrophe, for example the SBA loans prove very useful in bringing back the economy to the baseline. A 10 percent decrease in business taxes or equivalently, 10 percent increase in business credits helps in even surpass the baseline scenario within three years of the event.

5.8.8 Consumer Prices (Fig. 5.19)

It has been documented in literature (Brookshire, Thayer, Tshirhart, and Schulze, 1985) that houses sold for less in areas exposed to earthquake risk. Tobin and Montz (1988, 1994) present empirical evidence that house values (median selling prices) fell after the occurrence of floods. A fall in the housing price causes the relative price changes to be smaller, thus reducing the wage rate. A fall in the wage rate reduces the relative production costs and increases the relative profitability. The relative sales price for regional industries drop. A fall in the wage rate causes a drop in labor and proprietors’ income and thus lowers the gross output.

Immediately preceding a hurricane-type event where there is sufficient warning and after a catastrophic event costs of goods may increase temporarily anywhere from 10% to 30%. This decreases the consumer price deflator, which in turn increases the wage rate. Increase in the wage rate raises the labor and proprietors’ income.

A 10 % increase in consumer prices has effects similar to a market surge in production costs.

224 Chapter Five: Regional Impact of Catastrophes Fig. 5.17 Effect on gross re gional product due to 10% increase in relative production or fuel costs 5

-5

-10

-15

-20

-25

-30 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year Increase in Rel. production costs by 10% $60 billion and 50,000 jobs loss

Fig. 5.18 Effect on gross regional product due to decrease in business taxes or tax credits 5

-5

-10

-15

-20

-25 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year 10% decrease in business taxes or 10% increase in tax credits $60 billion and 50,000 jobs loss

Fig. 5.19 Effect of wage rate changes on gross regional product 5

-5

-10

-15

-20

-25 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 year 1% decrease in wage rate $60 billion and 50,000 jobs loss 1% increase in wage rate

225 Chapter Five: Regional Impact of Catastrophes

5.9 Summary and Conclusions

Hurricane Andrew and Northridge earthquake caused almost equal amounts of direct losses. But the effects were both similar and different. The effects were similar to the extent that almost all sectors of the economy received a boost to their economic growth, due to flow of external investment, reconstruction and changes in the underlying factors of productivity. Replacement of old and damaged capital with new and efficient capital enhances the productivity. This confirms the results of various numerical simulations reported in Chapter 3. The effects were different in the sense that Northridge earthquake occurred in a county with a robust economy as compared to Hurricane Andrew. The effects of Loma Prieta and Northridge earthquakes were localized, whereas effects of Hurricane Andrew were felt at the state level. Pre-event socioeconomic conditions determine the post-event behavior – to repeat an observation from Chapter 2.

Data presented herein suggests that shock to a robust economy, such as California's from a hazardous event of magnitude comparable to the Northridge earthquake, results in localization of the effects. A higher magnitude of shock may propagate the effects to a greater extent, but will be nevertheless localized. Hurricane Andrew was a greater shock in terms of the loss-GSP ratio. But the propagation of the shock was felt at the state level, though it dampened out rapidly both spatially and temporally.

A regional model was used to analyze the regional impacts of three major events in the recent past – Loma Prieta earthquake, Hurricane Andrew, and Northridge Earthquake. Model predictions for regional personal income matched with the actual observed values within acceptable levels of error. The changes in consumption and output as predicted from simulation of model used in this chapter concurred with the trends predicted by the theoretical model presented in Section 3.3. The model was then used to study the impacts of possible earthquake scenarios in the San Francisco – San Jose CMSA.

226 Chapter Five: Regional Impact of Catastrophes What are the factors that contribute towards dampening the propagation of localized shocks caused by intense earthquakes or hurricanes to adjacent regions? Values of the parameters of the model were varied exogenously to understand the working of the model. These studies help one to design suitable policies for early recovery of a regional economy. The main inferences from these numerical experiments were as follows: 1. The model predicts large gains if incentives are given to businesses in the affected region. 2. Large increases in consumer prices and production and fuel costs after an event delay recovery 3. Migration from the affected region delays its recovery 4. Governmental spending and transfer payments have a small positive effect on the region’s recovery. 5. Increased consumer spending is an important factor driving the region’s economy to rapid recovery.

The main conclusion that can be drawn from this study is that efficient recovery from a catastrophic event is possible if the government introduces measures that check the short term inflation of commodities, give incentives for the establishment and continuation of investments, and encourage consumer spending (via tax rebates) in the disaster region.

227 Chapter Five: Regional Impact of Catastrophes

Chapter Six

Conclusions and Future Work ______

6. Introduction

The final chapter concludes with some final summary thoughts on the contributions of this study. This chapter looks to the future through a discussion of some natural extensions of the models and data described in the previous chapters. In particular it proposes studies that link the results to financial and insurance markets. It proposes refinements that might make it possible to model the post event economic behavior due to man- made catastrophes.

6.1 Conclusions

The research presented in this dissertation: • connects socioeconomic indicators to determinants of vulnerability of a nation to natural hazards, • develops dynamic economic models for studying the economic behavior of economies affected by catastrophes, and • validates these models based on an empirical study at national and regional levels.

There were several contributions as a result of this study.

What are the determinants of the vulnerability of a region to natural hazards? 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 a quantitative relationship exists between the two. However, physical and human capital losses also depend on hazard intensity.

228 Chapter Six: Future Work and Conclusions

What trends do past data on catastrophes suggest and can theoretical models replicate these trends? Do catastrophes actually retard economic growth? Data based on past catastrophes suggest a negative correlation of the loss with the post event economic growth. A theoretical model was developed that explained this negative correlation between the loss and the post event growth rate. This was achieved by modeling the efficiency of post-event reconstruction. The observation that earthquakes were associated positively with the post event growth rates were explained by the fact that reconstruction of destroyed or damaged capital results in increases in productivity of the region which in turn spurs the post-event economic growth. Empirical data also suggest negative impacts on inflation, interest rates, and savings.

What quantity can be used as a measure of a catastrophe? How do we quantify the secondary effects? The annual economic loss as a percentage of GDP is extensively used in this study and provided a robust indicator to study the impact of catastrophes. Catastrophe results in loss of physical and human capital and this loss combined with the changes in the productivity of the affected economy results in overall welfare losses. A measure of the secondary effects of catastrophes based on these welfare losses was devised, which can be used to assess to impact of a catastrophe on an economy.

How will a regional economy behave after a catastrophic event? The theoretical models predicted the declines in income and consumption of the affected region. Examination of three catastrophes – 1989 Loma Prieta earthquake, 1992 Hurricane Andrew, and 1994 Northridge earthquake, revealed similar results. Also, it was shown that a low loss to output ratio results in localizing the effect of a catastrophe. Hurricane Andrew’s effect was propagated to the state level, whereas Loma Prieta and Northridge earthquakes were localized. The study revealed that a catastrophic earthquake in San Francisco’s Silicon Valley causes the personal income and consumption of the affected counties to drop. Depending on the external aid it receives, Silicon Valley could

229 Chapter Six: Future Work and Conclusions fare better after the event. Since Silicon Valley’s interaction with the other regions is high, economic effects of an earthquake will be felt in these regions. The overall welfare losses to other regions are directly proportional to the direct losses caused by a catastrophe.

What measures will best help the affected community to recover? Theoretical model simulations reveal the importance of aid (in the form of investment) the affected region receives from unaffected region. This is crucial for reconstructing lost capital, thus, reviving the economy. The importance of incentives to business investments in the affected regions is clearly demonstrated.

Why and under what conditions does the affected community fare better after an event? How important and how long lasting are the various effects likely to be? The affected region can easily recover within two years of the catastrophe, sometimes even to better economic conditions as compared to the pre-event levels. This was clearly demonstrated after the Northridge earthquake and Hurricane Andrew. The affected region can fare better after an event if reconstruction results in permanent increase in the capital share of the production function. This typically follows after productive capital in an economy is destroyed and is replaced by new capital stock. Empirical evidence that earthquakes are positively correlated with post-event growth rates clearly lends support to the economic resurgence due to new capital influx in the affected region.

How closely are catastrophes and developmental process related? Catastrophes reveal the most vulnerable sections of a socioeconomic fabric. Vulnerability of a region to catastrophes is intimately related to the on going socioeconomic processes including development. If the threat of occurrence of natural hazards is taken into account while designing the development program of a region, then it may result in building of robust engineering as well as social structures. Ex ante a catastrophe-threat induced preparedness programs could result in many positive externalities. A well-planned development strategy would not only result in raising the levels of livelihood of the people concerned but also make them resilient towards the onslaughts natural hazards.

230 Chapter Six: Future Work and Conclusions Ex post catastrophes can result in the building of a robust and less vulnerable region, if appropriate measure are taken. The models and empirical data presented in the previous chapters bring out the importance of pre-event conditions and efficiency of post-event reconstruction in determining the evolution of an economy after an event. The occurrence of a catastrophe gives the opportunity to invest, rebuild, and revitalize the economy of the affected community. If this opportunity is seized, the affected community could emerge better off than it was prior to the event.

6.2 Future work

Future work related to study of economic behavior after a catastrophe requires two broad initiatives: relating the results of this study to financial markets and insurance and to man-made catastrophes.

Since it has been shown in this dissertation that catastrophes have negative impacts on such financial indicators as inflation, real interest rates, and more generally economic growth, it will be interesting to examine the affect of catastrophes on financial markets. Are catastrophes triggered by natural hazards and the financial markets really not correlated? The results of such a study will be important in devising various financial instruments including insurance and catastrophe bonds that will help in mitigating the effects of catastrophes.

Man-made catastrophes are important especially after September 11, 2001 attacks. An obvious question to answer is how are the affects of man-made catastrophes different from catastrophes triggered by natural hazards? Such a study will be useful in identifying macroeconomic policies that need be enforced for efficient recovery.

231 Chapter Six: Future Work and Conclusions References ______

Albala-Bertrand, J. M., (1993a) Natural Disaster Situations and Growth: A Macroeconomic Model for Sudden Disaster Impacts, September 1993, World Development, v21, n9, 0305-750X, 1417-34. Albala-Bertrand, J. M., (1993b) Political economy of Large Natural Disasters - with special reference to developing countries, Clarendon Press, Oxford, 1993 Anderson, Mary B. (1990), Analyzing the costs and benefits of natural disaster responses in the context of development, The World Bank Policy Planning and Research Staff, Environment Department. Anderson, Mary B. and Woodrow, Peter J. (1989), Rising from the ashes: development strategies in times of disaster, Boulder: Westview Press; Paris: UNESCO Ayscue, J.K., (1996), Hurricane damage to residential structures: risk and mitigation, Natural Hazards Research Working Paper #94, Institute of Behavioral Science, University of Colorado, Web site: http://www.colorado.edu/hazards/wp/wp94/wp94.html Backus, David K., Kehoe Patrick J., and Kydland, Finn E. (1992) “International Real Business Cycles,” Journal of Political Economy, 100, 745-775. Barro, Robert J. (1990), Macroeconomics, 3rd ed. Imprint: New York : Wiley Barro, Robert J. (1991) “Economic Growth in a Cross Section of Countries.” Quarterly Journal of Economics 106:407 –43. Barro, Robert J. (1997), Determinants of economic growth : a cross-country empirical study Cambridge, Mass. : The MIT Press, c1997. Barro, Robert J. and Xavier Sala-i-Martin. (1995), Economic growth, New York : McGraw-Hill. Bates, F.L., and Peacock, W.G., (1989) “Long term recovery”, International Journal of Mass Emergencies and Disaster 7: 349-365. Bates, F.L., and Peacock, W.G., (1993) Living Conditions, Disasters and Development - An Approach to Cross-Cultural Comparisons, The University of Georgia Press, Athens.

232 References

Behrman, J.R. and Deolalikar, A.B. (1990) “The intra-household demand for nutrients in rural south India: Individual estimates, fixed effects and permanent income”, Journal of Human Resources, 25(4):665-696. Behrman, J.R. and Wolfe, B.L. (1984) “More evidence on nutrition demand; Income seems overrated and women’s schooling under-emphasized”, Journal of Development Economics, 14:105-128. Behrman, Jere and Srinivasan, T.N., eds. (1995) “Handbook of Development Economics.” Vol.3. Amsterdam: Elsevier, North Holland, 1995. Behrman, Jere R. and Deolalikar, Anil B. (1987) “Will Developing Country Nutrition Improve with Income?” Journal of Political Economy, June 1987, 95(3): 492-507. Bernheim, D., Shleifer, A. and Summers, L. (1985) “The strategic bequest motive”, Journal of Political Economy, 93(6):1045-1076. Besley, Timothy. (1995) “Savings, Credit and Insurance.” Handbook of Development Economics, 1995. Bevan, D., Collier, P. and Gunning, J.W. (with Bigsten, A. and Horsnell, P.) (1989) “Peasants and Government: An Economic Analysis.” Oxford: Clarendon Press. Blaikie, P., Cannon, T., Davis, I., and Wisner, B. (1994) “At Risk - Natural Hazards, Peoples Vulnerability, and Disasters,” Routledge, London. Boarnet, M. G. (1995) "Business Losses, Transportation Damage, and the Northridge Earthquake. Irvine, CA: Department of Urban and Regional Planning, School of Social Ecology, University of California, Irvine. Boisvert, R.N., Direct and Indirect Economic Losses from Lifeline Damage, FEMA, Indirect Economic Consequences of a Catastrophic Earthquake, A study Conducted by Development Technologies, Inc., National Earthquake Reduction Program, July 1992. Bolin, R. C. (1982) “Long Term Family Recovery from Disaster”, Boulder, University of Colorado. Bolin, R. with Stanford, L. (1998) “The Northridge Earthquake,” Routledge, London. Brady, R.J. and Perkins, J.B., Macro-economic effects of the Loma Prieta earthquake, Association of Bay Area Governments (ABAG), Nov. 1991.

233 References

Branford, S. and Kucinski, Bernardo, (1988), The debt squads : the US, the banks, and Latin America / Imprint: London ; Atlantic Highlands, N.J. : Zed Books. Brookshire, D.S., and McKee, M., (1992) Indirect Loss Measurement and Computable General Equilibrium, FEMA, Indirect Economic Consequences of a Catastrophic Earthquake, A study Conducted by Development Technologies, Inc., National Earthquake Reduction Program, July 1992. Brookshire, D.S., Chang S.E., Cochrane, H. Olson, R.A., Rose A., and Steenson, J., (1997) Direct and Indirect Economic Losses from Earthquake Damage, Earthquake Spectra, 13(4), Nov. 1997, pp.683- 701. Campbell, John Y. (1987) “Does saving anticipate declining labor income? An alternate test of the permanent income hypothesis,” Econometrica, 55, 1249-73. Cannon, T. (1994) “Vulnerability analysis and the explanation of “Natural” Disasters,” In Varley, A. (ed.), 1994, Disasters, Development and Development, John Wiley, Chichester. Chang, S. (1984) “Do disaster areas benefit from disasters?” Growth and Change Oct.:24-31. Chiappori, P.-A. (1992) “Collective labor supply and welfare”, Journal of Political Economy, 100(3):437-467. Coate, Stephen and Ravallion, Martin. (1993) “Reciprocity Without Commitment: Characterization and Performance of Informal Insurance Arrangements,” J. Devel. Econ., Feb. 1993, 40(1), pp1-24. Commander, Simon, Davoodi, Hamid R., and Lee Une J., (1997), The Causes of Government and the Consequences for Growth and Well-Being, World Bank Reading list, 2000/2001 World Development Report: Attacking Poverty. Clague, Christopher, Philip Keefer, Stephen Knack and Mancur Olson (1996). “Property and Contract Rights under Democracy and Dictatorship.” The Journal of Economic Growth 1:2, 243-276 (June). Cochrane, H., (1992) Is Northridge a model for future events? What are the economic consequences with and without insurance and federal assistance? The Northridge Earthquake: Analysing Economic Impacts and Recovery from Urban

234 References

Earthquakes: Issues for Policy makers, Proceedings of the Conference, Oct. 1996, EERI and FEMA. Cochrane, John H. (1991) “A Simple Test of Consumption Insurance”, Journal of Political Economy, Vol. 99, Issue 5 (Oct. 1991), 957-976. Coles, Jeffrey l. and Hammond, Peter J. (1995) “Walrasian Equilibrium Without Survival: Existence, Efficiency and Remedial Policy,” in Kaushik Basu, Prasanta Pattanaik, and Kotaro Suzumura, eds. 1995. Cornia, G.A., Jolly, R. and Stewart, F., eds. (1987) “Adjustment with a human face,” Vol. 1, Oxford: Clarendon Press. Cox, D. (1987) “Motives for private income transfers,” Journal of Political Economy, 95(3) 508-546. Dacy, D.C. and Kunreuther, H., The Economics of Natural Disasters - Implications for Federal Policy, The Free Press, 1969 Dando, W.A. (1980) “The geography of famine.” Edward Arnold, London. Davis, L. (1971) “Capital Mobility and American growth”, In R.W. Fogel and S.L. Engerman, (eds.), “The reinterpretation of American economic history.” New York: Harper and Row. Deaton, Angus. (1989) “Saving in Developing Countries: Theory and Review.” In “Proceedings of the World Bank annual conference on development economics.” Stanley Fischer and Dennis De Tray (Eds.), Washington, DC: World Bank, 1989, pp. 611-96. Deaton, Angus. (1992a) “Saving and Income smoothing in Cote d’Ivoire.” Journal of African Economies, 1(1): 1-24. Deaton, Angus. (1992b) “Household saving in LDCs: Credit markets, insurance and welfare” Scandinavian Journal of Economics, 94(2): 253-273. Deaton, Angus. (1992c) “Understanding Consumption,” Clarendon Press, Oxford. Deolalikar, A.B. (1992) “Does the impact of government health expenditure on the utilization of health services by children and on child health outcomes differ across expenditure classes?,” mimeo, Department of Economics, University of Washington. Quoted in Strauss and Thomas (1995).

235 References

Deininger, K., Squire, L. (1996). Measuring Income Inequality: A New Data Set. World Bank Economic Review 10(3), 565-91 Development Technologies, Inc. (1992) Indirect Economic Consequences of a catastrophic earthquake, National Earthquake Hazards Reduction Program, NEHRP and FEMA. Devereux, Stephen. (1993) “Theories of famine.” New York: Harvester Wheatsheaf, 1993. Drabek, T.E., Key, W.H., Erickson, P.E. and Crowe, J.L. (1975) “The impact of disaster on kin relationships,” Journal of Marriage and the Family, 481-494. Dreze, Jean and Sen, Amartya, eds. “The Political economy of hunger” Vols. 1,2, and 3. Oxford: Oxford U. Press, 1990. Dreze, Jean and Sen, Amartya. (1989) “Hunger and public action.” Oxford: Oxford U. Press, 1989. Dufflo, Esther (1998) “Schooling and labor market consequences of school construction in Indonesia: Evidence from an unusual policy experiment,” mimeo, Massachusetts Institute of Technology, 1998. Dynes, R. Russell. (1998) “Coming to terms with community disaster,” In “What is a Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge, London, 1998 Eguchi, R.T., Goltz, J.T, Taylor, C.E., Chang, S.E., Flores, P.J., Johnson,L.A., Seligson, H.A., and Blais, N.C., The Northridge Earthquake as an Economic event: Direct Capital Losses, Paper No.1, The Northridge Earthquake: Analysing Economic Impacts and Recovery from Urban Earthquakes: Issues for Policy makers, Proc. of the Conference, Oct. 1996, EERI and FEMA. Elo, O. (1994) “Making disaster reduction a public policy.” STOP Disasters 17:14-15, 1994 EQE, Hurricanes Andrew and Iniki 1992 - An EQE Summary Report, 1992, Web site: http://www.eqe.com/publications/iniki/ Easterly, William, and Rebelo, Sergio. (1993) “Fiscal Policy and Economic Growth: an Empirical Investigation,” Journal of Monetary Economics December 1993, 32(3), pp. 417-57.

236 References

Easterly, William and Levine, Ross. (1997) “Africa’s Growth Tragedy: Policies and Ethnic Divisions,” Quarterly Journal of Economics, November 1997. Eswaran, M. and Kotwal, A. (1989) “Credit as insurance in agrarian economies,” Journal of Development Economics, 31(1): 37-53. Fafchamps, M. (1993) “Sequential labor decisions under uncertainty: An estimable household model of West African Farmers”, Econometrica, 61(5):1173-1198. Fischer, Stanley, (1993) “The Role of Macroeconomic Factors in Growth,” Journal of Monetary Economics, December, 32(3), pp. 485-511. Flavin, Marjorie, (1981). “The adjustment of consumption to changing expectations about future income,” Journal of Political Economy, 89, 974-1009. Foster, Andrew D. (1995) “Prices, Credit Markets and Child Growth in Low-income Rural Areas,” Economic Journal, May 1995, 105(430), pp. 551-70. Friesema, H. Paul, [et al.]. (eds.) (1979), Aftermath : communities after natural disasters, Beverly Hills : Sage Publications, c1979. Gastil,R.D.(1975 –90,various issues) Freedom in the World. New York: Freedom House. Gersovitz, Mark. (1983) “Savings and Nutrition at Low Incomes,” Journal of Political Economy, October 1983, 91(5), pp.841-55. Ghate, P. (1992) “Informal finance: Some findings from Asia.” Manila: Oxford University Press for the Asian Development Bank. Gilbert, Claude. (1998) “Studying disaster: Changes in the main conceptual tools,” In “What is a Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge, London, 1998 Gordon, P. and H. W. Richardson, with B. Davis, C. Steins, and A. Vasishth. 1995. "The Business Interruption Effects of the Northridge Earthquake." Los Angeles, CA: Lusk Center Research Institute, School of Urban and Regional Planning, University of Southern California. Gordon, Peter and Richardson, H.W., Business Interruption Effects of a Major Earthquake in the Newport/Inglewood Fault Zone, Submitted by the planning Institute, School of Urban and Regional Planning, University of Southern California, to the National Committee on Property Insurance, Boston, Massachusetts, July, 1992

237 References

Gwartney, J.; Lawson,R.; and Block,W. (1997) Economic Freedom of the World,1975 – 1995.Vancouver:The Fraser Institute. Haas, E.J., Kates, R.W., and Bowden, M.J., (1977) (eds.) Reconstruction following Disaster, MIT Press, Cambridge, Massachusetts Levine, R. and S. Zervos, 1998, "Stock Markets, Banks, and Economic Growth", American Economic Review, vol 88, no 3, 537-558. Hadfield, Peter. (1992) “Sixty Seconds that will change the World: The Coming Tokyo Earthquake.” Boston: Charles E. Tuttle Company. HAZUS (1997), Earthquake Loss Estimation Methodology, Technical Manuals (vols. I to III), National Institute of Building Sciences, FEMA Hewitt, Kenneth. (1998) “Excluded perspectives in the social construction of disaster,” In “What is a Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge, London, 1998 Hill, K. and Palloni, A. (1992) “Demographic responses to economic shocks: The case of Latin America”, mimeo, Department of Population Dynamics, School of Public Health, Johns Hopkins University. Quoted by Strauss and Thomas (1995). Hossain, S. (1989) “Effect of public programs on family size, child education and health”, Journal of Development Economics, 30(1):145-158. IFRCRCS (1993) World Disasters Report. Martinus Nijhoff, Dordrecht. IFRCRCS (1994) World Disasters Report. Martinus Nijhoff, Dordrecht. IFRCRCS (1999) World Disasters Report. Oxford University Press, Oxford. INCEDE, 1995, The first 55 hours - Great Hanshin earthquake, January 17, 1995, INCEDE newsletter Special Issue, January 1995, Institute of Industrial science, The University of Tokyo. Jacoby, Hannan and Skoufias, Emmanuel (1992) “Risk, Seasonality and School Attendance: Evidence from Rural India.” Mimeo, University of Rochester, 1992. Jones, E.L. (1987) “The European Miracle – Environments, economies and geopolitics in the history of Europe and Asia.” Second Edition. Cambridge University Press, Cambridge. Kathleen J. T., Business Impacts Of The Northridge Earthquake, Disaster Research Center, Department of Sociology and Criminal Justice, University of Delaware,

238 References

Newark, DE 19716 USA Paper prepared for Pan Pacific Hazards 96, Vancouver, British Columbia, July 29-Aug. 2, 1996. Kreimer, A. and Munasinghe, M. (eds.) (1991) “Managing natural Disasters and the environment,” Environment Department, The World Bank, Washington, DC. Knack, Stephen and Philip Keefer (1995). "Institutions and Economic Performance: Cross- Country Tests Using Alternative Institutional Measures." Economics and Politics 7:3, 207-228 (November). Knack, S. and Keefer, P., 1996. "Does social capital have an economic payoff?: a cross- country investigation ", The Quarterly Journal of Economics, 112(4): 1251. Kroll, C. A., J. D. Landis, Q. Shen, and S. Stryker. 1991. "Economic Impacts of the Loma Prieta Earthquake: A Focus on Small Business." Berkeley, CA: University of California--Berkeley, University of California Transportation Center and the Center for Real Estate and Economics. Working Paper No. 91-187. Leamer, E. Edward. (1983) Let ’s take the con out of econometrics.American Economic Review 73,1:31 –43 Leamer, E. Edward. (1985) “Sensitivity Analyses Would Help.” American Economic Review 75 (3):308 –13. Lewis, Karen (1996) “What can explain the apparent lack of international consumption risk sharing?” Journal of Political Economy, 104(2), 267-297. Levine, Ross and David Renelt (1992). “A Sensitivity Analysis of Cross-Country Growth Equations.” American Economic Review, September. Levine, R. And S. Zervos, 1998, "Stock Markets, Banks, and Economic Growth", American Economic Review, vol 88, no 3, 537-558. Lipton, Michael and Ravallion, Martin. (1995) “Poverty and Policy” in Jere Behrman and T.N. Srinivasan 1995, pp.2551-2657. Lipton, Michael. (1977) “Why Poor People Stay Poor” Temple Smith, London, 1977 Quoted by Devereux (1993). Mace, Barbara J. (1991) “Full Insurance in the Presence of Aggregate Uncertainty.” Journal of Political Economy, Vol. 99, Issue 5 (Oct. 1991), 928-956.

239 References

Maskrey, A. (1994) “Disaster Mitigation as a Crisis of Paradigms: Restructuring after the Alto Mayo Earthquake, Peru,” In Varley, A. (ed.), 1994, Disasters, Development and Development, John Wiley, Chichester. MATLAB compiler (2001): the language of technical computing: user's guide. Version 2. Natick, MA: MathWorks, Inc., 2001. Mauro, Paolo. (1995) “Corruption and Economic Growth,” Quarterly Journal of Economics, August, 110(3), pp. 681-712. McGregor, Pat. (1990) “A Model of Crisis in a Peasant Economy,” Oxford Economic Papers, Oct. 1990, 4, pp.793-811. Meier, Gerald M. (ed.) (1995) Leading issues in economic development, 6th ed., New York, Oxford University Press, 1995. Milliman, J.W. and Sanguinetty, J.A., Basic Concepts and Simple Analytic of Loss Measurement for Major Earthquakes, in FEMA, Indirect Economic Consequences of a Catastrophic Earthquake, A study Conducted by Development Technologies, Inc., National Earthquake Reduction Program, July 1992. Morduch, Jonathan (1994) “Poverty and Vulnerability” American Economic Review, May 1994, 84(2), pp. 221-25. Morduch, Jonathan. (1993) “Risk, Production and saving: Theory and Evidence from Indian Households” Mimeo, Harvard U., 1993. Morrow, B.H. (1997) “Stretching the Bonds: The Families of Andrew”, In Peacock, W.G., Morrow, B.H., and Gladwin, H. “Hurricane Andrew – Ethnicity, gender and the sociology of disasters”, Routledge, London, 141-169. Munich Re (2000), Topics 2000, Natural Catastrophes: The current position. Nigg, J.M. and Perry, R.W. (1988) “Influential first sources; Brief Statements with long- term effects”, International Journal of Mass Emergencies and Disasters, 6(3), 311- 343. NOAA, The costliest hurricanes in the United States - 1900-1996, Web site: http://www.nhc.noaa.gov/pastcost.html/ Oliver-Smith, A. (1994) “Peru’s Five Hundred-Year Earthquake: Vulnerability in Historical Context,” In “Disasters, Development and Environment,” A. Varley (ed.), John Wiley, Chichester.

240 References

Oliver-Smith, Anthony. (1998) “Global changes and the definition of disaster.” In “What is a Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge, London, 1998 Olsen, R. and Wolpin, K. (1983) “The impact of exogenous mortality on fertility; A waiting time regression with dynamic regressors”, Econometrica, 51(3):731-749. Palloni, A. and Hill, K. (1992) “The effects of structural adjustment on mortality by age and cause in Latin America”, mimeo, Center for Demography and Ecology, University of Wisconsin. Quoted by Strauss and Thomas (1995). Palloni, A. and Hill, K. and Pinto Aguirre, G. (1993) “Economic swings and demographic changes in the history of Latin America”, mimeo, Center for Demography and Ecology, University of Wisconsin. Quoted by Strauss and Thomas (1995). Paxson, Christina H. (1992) “Using Weather Variability to estimate the response of savings to Transitory Income in Thailand.” American Economic Review, March 1992, 82(1), pp.15-33. Peacock, W.G., Killian, C.D. and Bates, F.L. (1987) “The effects of disaster damage and housing aid on household recovery following the 1976 Guatemalan earthquake,” International Journal of Mass Emergencies and Disasters, 5, 63-88. Peacock, W.G., Morrow, B.H., Gladwin, H. (eds.), (1997) “Hurricane Andrew - Ethnicity, gender and the sociology of disasters,” Routledge, London. Peek, Joe and Rosengren Eric S. (1997) “The International Transmission of Financial Shocks: The Case of Japan” American Economic Review, Sep. 1997, pp. 495- 505. The Penn World Table (PWT ver 6.1), Extension of - The Penn World Table (Mark 5): An Expanded Set of International Comparisons, 1950- 1988," The Quarterly Journal of Economics, May, 1991. Pielke,Jr., R. A., and Landsea, C.W., Normalized hurricane damages in the United States: 1925-1995,  Copyright 1998 American Meteorological Society, Web site: http://www.dir.ucar.edu/esig/HP_roger/hurr_norm.html Pielke,Jr., R. A., Trends in Extreme Weather Impacts, Web site: http://ycymis.hq.nasa.gov/web_serv/files/southeast/

241 References

Porfiriev, B.N. (1992) “The environmental dimensions of national security: a test of systems analysis methods.” Environmental Management 16: 735-742. Quarantelli, E.L. (ed.) (1998) “What is a Disaster? Perspectives on the question,” Routledge, London, 1998 Ravallion, M. (1987) “Markets and famines,” Oxford: Clarendon Press. Ravallion, M. (1990) “Income effects on under-nutrition”, Economic Development and Cultural Change, 38(3): 489-515. Ravallion, Martin. (1997) “Famines and Economics”, Journal of Economic Literature, Vol. XXXV (Sep. 1997), pp.1205-1242. Razzaque, A., Alam, N., Wai, L. and Foster, A. (1990) “Sustained effects of the 1974-5 famine on infant and child mortality in a rural area of Bangladesh”, Population Studies, 44(1): 145-154. Ray, Debraj, (1998) Development economics, Princeton, N.J.: Princeton University Press, c1998. Red Cross, 1996, Hurricane Andrew results in one of the most comprehensive disaster relief efforts in Red Cross history, Web site: http://www.redcross.org/hec/1980- present/andrew.html/ RMS and FaAA, 1995, Japan - The great Hanshin Earthquake, Risk Management Solutions, Inc., (RMS) and Failure Analysis Associates, Inc. (FaAA), 1995 RMS, Northridge earthquake January 17, 1994 - Event report, Risk Management Software Inc., Menlo Park, California, 1994. RMS, What if the 1906 Earthquake Strikes Again? A San Francisco Bay Area Scenario, Topical issues series, Risk Management Solutions, Inc., CA, 1995. Roberts, Blaine R., Milliman, Jerome W., and Ellson, Richard W. (1982) Earthquakes and Earthquake Predictions: Simulating their Economic Effects, Technical Report, University of South Carolina, Columbia, SC 29208 Rodrik, D. (1999) Where did all the growth go? External shocks, social conflict, and growth collapses. Journal of Economic Growth 4(4), 385-412 Rosenthal, Uriel. (1998) “Future disasters, future definitions.” In “What is a Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge, London, 1998

242 References

Rosenzweig, Mark R. (1982) “Educational subsidy, agricultural development and fertility change”, Quarterly Journal of Economics, 97(1) 67-88. Rosenzweig, Mark R. (1988) “Risk, implicit contracts and the family in rural areas of low income countries”, Economic Journal, 98(4): 1148-170. Rosenzweig, Mark R. (1990) “Population growth and human capital investments: Theory and evidence”, Journal of Political Economy, 98(5): S38-S70. Rosenzweig, Mark R. and Binswanger, Hans P. (1993) “Wealth, Weather Risk and Profitability of Agricultural Investments,” Economic Journal, Jan. 1993, 103(416), pp.56-78. Rosenzweig, Mark R. and Evenson, R.E. (1977) “Fertility, schooling and the economic contribution of children in rural India: An Econometric analysis”, Econometrica, 45(5):1065-1079. Rosenzweig, Mark R. and Stark, O. (1989) “Consumption smoothing, migration and marriage: Evidence from rural India”, Journal of Political Economy, 97(4): 905- 927. Rosenzweig, Mark R. and Wolpin, Kenneth I. (1993) “Credit Market Constraints, Consumption Smoothing and Accumulation of Durable Production Assets in Low-Income Countries” Journal of Political Economy, April 1993, 101(2), pp.223-44. Sapir, D.G. and Misson, C. (1992) “The development of a database on disasters,” Disasters 16: 74-80. (Center for Research on the Epidemiology of Disasters – CRED) Sen, Amartya. (1981) “Poverty and famines: An essay on entitlement and deprivation.” Oxford: Oxford U. Press, 1981. Smith, K. (1996) “Environmental Hazards - Assessing risk and reducing disaster,” Second Ed., Routledge, London Spurr, G.B. (1983) “Nutritional status and physical work capacity”, Yearbook of Physical Anthropology, 26:1-35. Stallings, A. Robert. (1998) “Disaster and the theory of social order,” In “What is a Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge, London, 1998

243 References

Strauss, J. and Thomas D.(1990) “The shape of the calorie expenditure curve”, Economic Growth Center Discussion, Paper No. 595, Yale University. Strauss, J. and Thomas, D. (1995) “Human Resources: Empirical modeling of household and family decisions”, Chapter 34: In Behrman, Jere and Srinivasan, T.N., eds. Handbook of Development Economics. Vol.3. Amsterdam: Elsevier, North Holland, 1995. Subramanian, S. and Deaton, A. (1992) “The demand for food and calories: Further evidence from India”, mimeo, Woodrow Wilson School , Princeton university. Susman, p., O”Keefe, P. and Wisner, B. (1983) “Global disasters: a radical interpretation.” In Hewitt, K. (ed.) “Interpretations of Calamity,” pp. 263-283. Allen and Unwin, Boston and London. Thomas D., Lavy,V. and Strauss, J. (1992) “Public policy and anthropometric outcomes in the Cote d”Ivoire”, Living Standards Measurement Study Working Paper no. 89, World Bank, Washington D.C. Townsend, R. (1992) “Financial systems in Northern Thai Villages,” Typescript. Townsend, Robert M. (1994) “Risk and Insurance in Village India.” Econometrica, May 1994, 62(3), pp. 539-91. Treyz, George I., (1993) Regional economic modeling: a systematic approach to economic forecasting and policy analysis, Boston : Kluwer Academic. George I. Twigg, John. “Disasters, development and vulnerability.” In “Development at Risk? Natural Disasters and the Third World,” The UK National Coordination Committee for Natural Disaster Reduction (IDNDR), pp.2-7. Tyler, C. (1990) “Earthquake Zones – Sites of the World’s Largest cities.” Geographical Magazine 62, 3: 28-32. Udry, Christopher R. (1990) “Credit markets in Northern Nigeria: Credit as insurance in a rural economy”, World Bank Economic Review, 3:251-271. Udry, Christopher R. (1994) “Risk and Insurance in a Rural Credit Market: An Empirical Investigation in Northern Nigeria.” Review of Economic Studies, July 1994, 61(3), 495-526. Udry, Christopher R. (1995),“Risk and saving in Northern Nigeria.” American Economic Review, Dec. 1995, pp. 1287-1300.

244 References

Varley, A. (ed.), (1994), “Disasters, Development and Environment,” John Wiley, Chichester. West, C. and D. Lenze, 1994. Modeling the regional impact of natural disaster and recovery: A general framework and an application of Hurricane Andrew. International Regional Science Review, 17(2): 121-150. Whitman, R.V., Anagnos, T, Kircher, C.A., Lagorio, H.J., Lawson, S., and Schneider, P. Development of a National Loss Estimation Methodology, Earthquake Spectra, Vol.13, No.4, Nov. 1997, pp. 643-661. Wolfram, Stephen. (1999) The mathematica book, 4th ed. Champaign, IL : Wolfram Media ; New York : Cambridge University Press. Wolpin, K.I. (1982) “A new test of the permanent income hypothesis: The impact of weather on the income and consumption of farm households in India”, International Economic Review, 23(3): 583-594. World development report, 1994: Infrastructure and Development. New York ; Oxford : Oxford University Press, 1994. World Development Indicators, 2001: World Bank. World Disasters Report, (various years), International Federation of Red Cross and Red Crescent Societies, Geneva. Wright James D. [et al.]. (1979) After the clean-up: long range effects of natural disasters, Beverly Hills: Sage Publications. Yezer Anthony M. and Rubin Claire B. (1987) The local economic effects of natural disasters, Imprint: [Boulder, Colo.]: Institute of Behavioral Science, University of Colorado. Zupka, D. (1988) “Economic impact of disasters.” UNDRO News Jan./Feb.: 19-22.

245 References