The willingness to pay for flood insurance in mekong river delta

Based on the estimate result, WTP for attributes and the WTP for specific insurance package are calculated carefully, we also evaluate the probability of levels of WTP for flood insuranc

UNIVERSITY OF ECONOMICS ERASMUS UNVERSITY ROTTERDAM

HO CHI MINH CITY INSTITUTE OF SOCIAL STUDIES

VIETNAM – THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

WILLINGNESS TO PAY FOR

FLOOD INSURANCE IN THE MEKONG RIVER DELTA

BY

NGUYEN NGOC QUE ANH

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY, SEPTEMBER 2016

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

WILLINGNESS TO PAY FOR

FLOOD INSURANCE IN THE MEKONG RIVER DELTA

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

NGUYEN NGOC QUE ANH

Academic Supervisor:

TRUONG DANG THUY

HO CHI MINH CITY, September 2016

DECLARATION

In order to fulfill the requirements for the degree of Master of Art in Development Economics

to Vietnam – The Netherlands Programme (VNP), this thesis entitled “Willingness to pay for Flood Insurance in the Mekong River Delta” is submitted

This declaration certify that this thesis constitutes on my original work only All materials used

in this thesis have been acknowledged and cited properly following the Programme’s standards

NGUYEN NGOC QUE ANH

ACKNOWLEDGMENTS

Doing thesis is an amazing adventure but it is also a tough path, without support and encouragement of my family, my teachers, my friends and Vietnam Netherland Program, I might not complete it

I would like to send my sincerest thanks and gratitude to my supervisor - Dr Truong Dang Thuy who always gives me invaluable advice and instructs me wholeheartedly From the initial ideas to finish, he is very patient, listens to my opinions and helps me to correct mistakes delicately

I am deeply grateful to Mr Phung Thanh Binh Many thanks for giving me the precious opportunity to join in this research, inspiring me and allowing me to employ the data, so that I can pursue this topic Memories and experience from this research will be unforgettable

I greatly appreciate the enthusiasm and kindness of lecturers and staffs of Vietnam Netherland Program I would like to thank my teachers, the founders and staffs of VNP for be willing to help and give me invaluable knowledge

Last but not least, I am very thankful that my family and my friends are always with me, love

me and support me wholeheartedly From the bottom of my heart, I would send my sincerest gratitude to my parents Without their love, this thesis would be not accomplished

NGUYEN NGOC QUE ANH

ABSTRACT

In the circumstance of climate change and series of dams built in Mekong River Basin, risk of flood damages and productivity loss in Mekong River Delta tend to be ambiguous and unpredictable Choice experiment is applied to assess the stated preference and the willingness to pay for flood insurance of local farmers Based on the data obtained from survey in Mekong River Delta, we consider the impacts of attributes on the utility of insurance buyer and the willingness

to pay of them As a result, flood insurance attributes except deductible have impact on utility of farmers and the willingness to pay for flood insurance of them Especially, buying the flood insurance provided by corporation will raise the utility and willingness to pay of local farmer Since the deductible have no impact on the utility of flood insurance buyers, local farmers are willing to share the burden with providers When the effects of challenges for flood insurance development are controlled, only provider and policy types have strong positive impact on utility

of insurance

Based on the estimate result, WTP for attributes and the WTP for specific insurance package are calculated carefully, we also evaluate the probability of levels of WTP for flood insurance with different levels of important attributes The potential development of insurance companies are very bright, especially corporation The profitability will be large, if the provider provide insurance packages with favorite policy When premium of the most preferred flood insurance vary from 5,000 (VND/1000𝑚2/ Farming season) to 200,000 (VND/1000𝑚2/ Farming season), 90 percent to

79 percent of farmer willing to pay it

JEL Classification: Q11, Q12, Q14

Keywords: Willingness to pay, Flood insurance, Mekong River Delta, Random Utility Model

TABLE OF CONTENT

ABSTRACT i

ABBREVIATIONS vi

LIST OF FIGURES vii

LIST OF TABLES ix

CHAPTER 1 INTRODUCTION 1

1.1 Problem statement 1

1.2 Research objectives 6

1.3 Scope of the study 6

1.4 Structure of this thesis 7

CHAPTER 2 LITERATURE REVIEW 8

2.1 Previous studies without using RUM 8

2.1.1 Researches on agricultural insurance using secondary data or combine with primary data 8

2.1.2 Researches not applying RUM on agricultural insurance using primary data 9

2.2 Random utility model (RUM) and applications 11

2.2.1 Random utility model (RUM) 11

2.2.2 Researches applying RUM on agricultural insurance 13

2.2.3 Review of flood insurance demand research using RUM 18

2.3 Challenges of disaster insurance market 21

2.3.1 The ambiguity 21

2.3.2 Adverse selection 23

2.3.3 Charity hazard 25

CHAPTER 3 RESEARCH METHODOLOGY 28

3.1 Demand for flood insurance 28

3.2 The advantages of Choice Experiment compared to Contingent Value Method 28

3.3 General model 29

3.4 Estimation 30

3.4.1 Exogenous sample 30

3.4.2 Estimation on Subset of Alternatives 33

3.5 Description of variables 35

3.5.1 Description of all attributes and levels 35

3.5.2 Description of variables used to capture challenges for flood insurance market development 37

3.6 Empirical models 42

3.6.1 Empirical model with only attribute variables 42

3.6.2 Empirical model with attribute variables and their interaction with non-attribute variables 43

3.7 Calculation of Willingness-to-Pay (WTP) for specific insurance packages, and probability of buying specify insurance packages with the changes in premium 44

3.7.1 Calculation of Willingness to Pay (WTP) for each attribute and for specific insurance packages 44

3.7.2 Probability of buying specify insurance packages with changes in premium levels 47

3.8 Data collection 47

CHAPTER 4

RESEARCH RESULTS 50

4.1 Descriptive statistics 50

4.2 Bivariate analysis 56

4.2.1 No selection without consideration 56

4.2.2 Bivariate analysis about the effects of personal perspectives and externalities on flood insurance purchasing decision 57

4.3 Empirical results 59

4.3.1 Estimation results 59

4.3.2 The willingness to pay (WTP) 67

4.3.3 The probability of willingness to pay of most preferred insurance packages 73

CHAPTER 5 CONCLUSION 80

5.1 Conclusion remark 80

5.2 Policy implications 82

5.3 Limitations 82

REFERENCES 84

APPENDIX 92

APPENDIX A: Questions are used from the survey 92

APPENDIX B: Conceptual Framework 94

Appendix C: The statistic results about impacts of challenges 95

Appendix D: The variation of WTP for flood insurance probability, with difference levels of cover 100

Appendix E: The variation of WTP for flood insurance probability, with difference levels of deductible rate 101 Appendix F: the regression result of models controlling the impacts of challenges and

household characteristics 102 Appendix G: The regression result in Stata of models 104 Appendix H: The regression result from applying Nested Logit Model in Stata 108

ABBREVIATIONS

BDT Bangladeshi Taka

CE Choice Experiment CLL Conditional Log Likelihood Function CVM Contingent Value Method

GIS Geographic Information System IFRC International Federation Red Cross IPCC The Intergovernmental Panel on Climate Change

LL Log Likelihood Function MRD Mekong River Delta MRC Mekong River Commission OLS Ordinary Least Squares RUM Random Utility Model VND The Vietnamese Dong USD The United States Dollar WTP Willingness to Pay

LIST OF FIGURES

Figure 1.1 Flood damages in the Mekong River Delta from 1990s to 2000s 2

Figure 1.2 Disbursement process of IFRC funding contribution for Floods in MRD 3

Figure 3.1 One choice card are used in the survey 48

Figure 4.1 Statistical result of percentage of policy types chosen 51

Figure 4.2 Statistical result of percentage of providers chosen 52

Figure 4.3 Statistical result of percentage of deductibles rates chosen 53

Figure 4.4 Statistical result of percentage of cover level chosen 54

Figure 4.5 Statistical result of percentage of premium levels chosen 55

Figure 4.6 Statistic results of the most interested attribute of farmers 56

Figure 4.7 Percentages of observations choosing to purchase flood insurance 57

Figure 4.8 Amount of observations divided according to respondents’ perception 57

Figure 4.9 Probability of WTP for flood insurance covering triple disaster damages 74

Figure 4.10 Probability of WTP after controlling impacts of challenges 74

Figure 4.11 Probability of buying flood insurance covering flood & inundation damages 75

Figure 4.12 Probability of WTP for flood insurance covering flood and windstorm damages 76

Figure 4.13 Probability of WTP for flood insurance packages provided by corporation 77

Figure 4.14 Probability of WTP after controlling impacts of challenges 78

Figure 4.15 Probability of WTP for flood insurance provided by foreign company 79

Figure 4.16 Probability of WTP for flood insurance provided by private company 79

LIST OF TABLES

Table 2.1 Attributes in the insurance package in Nganje’s study 15

Table 2.2 Attributes of flood insurance package and their levels in previous study 19

Table 3.1 Attributes of flood insurance packages 35

Table 3.2 Variables used to capture impacts of challenges for flood insurance development 39

Table 3.3 Interactions between perceptive about flood insurance challenges and attributes 40

Table 3.4 Insurance packages are used to calculate the WTP 45

Table 4.1 Amount of observations choosing alternatives in each order of choice cards 56

Table 4.2 Estimate result of models 60

Table 4.3 WTP (VND/1000 𝑚2/ Farming season) for each levels of attributes 67

Table 4.4 WTP (VND/1000 𝑚2/Farming season) for specific insurance packages 70

CHAPTER 1

INTRODUCTION 1.1 Problem statement

Natural disasters have been causing many losses of human life and property including direct damage and indirect damage The frequency of disasters is increasing together with climate change In the announcement of IPCC in 2007, since 1900 climate change has caused the increases

in global temperatures (roughly 0.76𝑜𝐶 ) and sea level (approximately 20 centimeters) Along with climate change, the damages caused by natural disasters are increasing gradually over time and tend to be more serious in regions that are prone to be affected by calamities (Pielke et al., 2005)

Mekong River Delta, located in downstream the Mekong River Basin, have been experiencing seasonal floods due to great flow rates in the wet season (more than 65,000 𝑚3/s) and low terrain According to the Vietnam Academy for Water Resources, from 1991 to 2009, annual floods have caused damaged areas to increase from 1.6 million to 2 million hectares (To & Tang, 2011) Annually, before the flood season, the local residents have carefully preparation to reduce the flood damage According Vietnam Mekong River Delta Project for enhancing resistance to flood for poor households in 2011 with the support of United Nation, local government and residents reinforced houses, had careful examinations of flood protection infrastructures, practiced aid and provided necessary medicine for prioritized targets based on prudent plans Despite of these careful preparations, the 2011 flood caused more than 1,000 billion VND in property damage, 27,000 hectares area of rice and vegetables are damaged in which 10,000 hectares damaged 100%, nearly 12.000 ha of fruit area also flooded

The statistical data of flood damage in the Mekong River Delta from 1990s to 2000s shows that flood damages in downstream Mekong River Delta become abnormal

Figure 1.1 Flood damages in the Mekong River Delta from 1990s to 2000s

Source: Collect from Nguyen, 2006; Dao & Bui, 2009; MRC, 2011; MRC, 2012

Recovery after floods in Mekong River Delta residents takes a long time This is partly due to the dependence on the aids from government and humanitarian organizations In reality, the source

of aids is unstable, and people should not completely rely on them In 2011, the severe flood happened, local residents had to wait at least one year for entire of the aids from International Federation Red Cross

Figure 1.2 Disbursement process of IFRC funding contribution for Floods in MRD

Source: Report of International Federation Red Cross, 2013

In agricultural production, the post-flood relief cannot fully compensate for the costs of agriculture, because farmers in the Mekong Delta invested heavily in their farms with an average

of 15 million dong per hectare Thus, even if the government doubled the level of post-disaster subsidy for local farmers to 5millions dong per hectare paddy after the severe flood in 2011, the new subsidy level only offset against 30% of agricultural cost invested (Ngoc Anh, 2011) Furthermore, the burdens of agricultural cost and default risk will increase more after floods or other kinds of natural disaster, the farmers might have to face with threat of double liabilities In reality, because of financial constraints, many farmers have to purchase inputs on credit and pay after harvesting at higher prices In each season, many farmers and agents agreed to debit the purchase agreement for four months, if these farmers cannot pay off the loan on the maturity, the remaining debt will continue to be charged at the rate of 3-4% per month (Ngo, 2013)

Although farmers can access to bank loan, the threat of debt piling up is still there While farmers growing rice can only borrow from banks around 1 million VND per 1000𝑚 2, the average costs for cultivation and harvesting of rice vary from 2.2 million VND to 2.5 million VND per 1000𝑚 2, and transferring value is about 40-50 million VND (Ngo, 2013) Therefore, occurrence of flood might induce farmer to fall into debt piling up

In 2000s, Mekong River Delta has contributed more than 48% of food production of Viet Nam and 85% rice export volume (To & Tang, 2011) According to information of The Flood Prevention Agency in HCMC about the prediction for 2030, approximately 45% area of Mekong River Delta will be affected by salinity and damaged by severe floods and inundations, the potential losses would be 17 billion USD (2008) In addition, a series of dams constructed and operated in the Mekong River Basin has caused Vietnam to be passive in the flood discharge and flood prevention, the consequences are unpredictable, downstream area of Mekong river basin may be drought or flooded severely For example, when a certain dam discharge due to experiencing unusually heavy rains, it will create a domino effect for the whole system of 12 dams and the damage is enormous This hydropower system threats the future of the Mekong River

Delta and the whole country (Huynh & Phan, 2015 )

In this circumstance, adaptation or “Living with floods” is the optimal solution to exploit the benefits of flood and maintain the Vietnamese rice granary Besides flood prevention infrastructures, disaster insurance is a useful tools in adaptation strategy For developing countries, supporting development of catastrophe insurance is compared as a judicious investment It will facilitate damage reduction and repel disaster-induced poverty trap (Barnett et al., 2008), as suitable designed insurance does not only have characteristics of useful instruments

in deployment of adaptation process but also contributes effectively in risks management and recovery after adverse events (Botzen & Van den Bergh, 2008) Besides, flood insurance is also able to remove financial burdens of government in recovery after natural disaster, this instrument helps society to get back to routine faster In the cases that natural disasters consequences are severe, and households do not have resilience, insurance companies could spread the risk by utilizing the premiums collected from other households to pay for devastation

Market principles would promote private insurance companies to work more effectively in implement risk-reduce processes than public ones (Priest, 1996) Devastation could be reduced

by rewarding for design climate-adaptable constructions or premium discount programs Through insurance policies, insurance companies could encourage households to participate in risk-reduction activities In the cases of seasonally flooded areas, in order to encourage residents to use tile floors or flood-adapted building, insurance contracts might not pay for damages of wooden floor, or they could introduce flood-adapted materials and buildings (Thieken et al., 2006) After

the 2002 flood, a survey conducted in Germany indicates that insurance buyers pay more attention

to flood mitigation than who does not (Thieken et al., 2006) As a result, recovery costs and negative effects of natural disasters could be moderated

However, flood insurance and other kinds of agricultural insurance are new products to Vietnamese farmers According to the Ministry of Finance, revenue from agricultural insurance accounts for a very small proportion (0.015%) in total insurance revenue, and the implementation

of agricultural insurance has been in the pilot stage since 2011 The reason is that authorities have yet to identify the objects, the risks to be insured, while the scope of target clients and geographical deployment are quite wide; disaster and disease occur diversely, each locality has a different situation; technical facilities, information technology systems of provinces and insurance company are limited (Pham, 2015) In addition, many other obstacles such as ambiguity, adverse selection, moral hazard, correlated risks make the penetration of private insurance market become unappealing (Botzen & Van den Bergh, 2008) Therefore, conducting a research to capture the demand for flood insurance and other kinds of agricultural insurance in disaster-prone areas is necessary for flood insurance development projects and insurance companies

The demand for flood insurance is not a new topic with developed countries, most of them used the insurance statistics data to investigate the flood insurance demand (Kunreuther et al., 2009; Michel-Kerjan & Kousky, 2010) However, it cannot be applied in the case that disaster insurance market has yet to be formed such as Vietnam Thus, employing primary data tends to be supported (Aliagha et al., 2015; Brouwer & Akter, 2010; Brouwer et al., 2013, Reynaud et al., 2012) Some

of these studies tried to use Choice Experiment Model to capture potential markets of flood insurance, but besides worth learning points, many mistakes still remain

In Vietnam, since flood insurance has been a new product with farmers, using choice experiment

is suitable Only two studies was conducted capture the demand for flood insurance in Central Vietnam, the authors employed Choice Experiment Model as method for their studies But they still did not correct the mistakes of the previous studies and neglected the effect of local residents’ perspective their flood insurance demand Furthermore, there is no study conducted to capture the demand of flood insurance in Mekong River Delta

Overall, facilitating the disaster insurance sector in developing countries like Vietnam is necessary The information about the willingness-to-pay (WTP) of Vietnamese in disaster-prone

areas as well as the impacts of obstacles mentioned above on WTP are very helpful in this circumstance

1.2 Research objectives

In this study, Choice Experiments was employed to achieve the main objective in this study that

is the evaluation of WTP for flood insurance of farmers in the Mekong River Delta

Firstly, we estimate the impacts of flood insurance attributes on the utility of farmers in Mekong River Delta

Secondly, we consider how the effects of challenges for flood insurance market development and local irrigation services such as ambiguity, adverse selection, charity hazard, improvement of irrigation and accessibility to pumping station on the impacts of these attributes Based on findings

of previous studies, the effects of these challenges are captured by using perception of local farmers about their vulnerability, fear of flood, government responsibility, and local irrigation services

Thirdly, based on the estimate results above we will evaluate willingness-to-pay of local farmers for each level of attributes and willingness-to-pay for specific flood insurance packages with the different combinations of policy types and providers

Fourthly, we determine the variation of probability of WTP of Mekong River Delta farmers who are willing to pay for flood insurance with the changes in premium level

Finally, we would like to present some appropriate suggestions to insurance companies and policymakers

1.3 Scope of the study

In order to conduct this study, the data of this study was collected in three districts in Mekong River Delta including Gao Giong, Phu Loc, Tan Cong Chi in October, 2015 These districts have been affected by flood and other natural disaster in recent years

1.4 Structure of this thesis

This thesis consists of five chapters Theories about choice experiment and challenges of flood insurance will be presented in Chapter 2 Besides, many empirical studies about the demand for flood insurance and other similar types insurance are also reviewed In Chapter 3, the data collection and methodology used for this study will be mentioned Interpretations and discussion about statistics and empirical results will be presented in Chapter 4 Based on the results of this study, the conclusions and recommendations for flood insurance company will be presented in the Chapter 5

CHAPTER 2

LITERATURE REVIEW

This chapter firstly introduce different approaches of studying insurance demand After that, the advantages and disadvantages of approaches will be discussed Then, theories about Random Utility Model will be presented, so that we could know the advantages of this method and how it was applied to study about the demand for flood insurance and other similar types of insurance Some definitions and studies about challenges of development of flood insurance are also presented in this chapter

2.1 Previous studies without using RUM

2.1.1 Researches on agricultural insurance using secondary data or combine with primary data

The empirical attention for the demand of crop insurances is still very small compared to its development capabilities, though agricultural insurances could be considered as useful tools in government burden reduction such as damage reduction facilitation and repelling disaster-induced poverty trap for developing, financial difficulties mitigation due to crop failures or devaluation (Barnett et al., 2008) Most of empirical studied focus on North American market or other developed countries, and these studies tend to employ secondary data (Atreya et al, 2015; Dumm

et al., 2012) or combining with raw data were conducted (Enjolral et al, 2012; Sherrick et al., 2003)

By employing the data of 135 counties in Georgia from 1978 to 2010, the determinants of flood insurance purchasing were investigated (Atreya, 2015) They found that economic variables, demographic variables have influence on flood insurance demand but flood mitigation assistance does not In another study, similar above method and the impact of representative heuristic on distorting the resident behavior in risk assessment (Volkman-Wise, 2012) were combined to evaluate the flood insurance demand (Dumm et al., 2012) As a result, the damage caused by natural disaster in recent years have the positive effect on the disaster insurance demand, but this effect will decease with time, and predictions before disaster happen will be underweighted Even

though using secondary data may be economical, it cannot be applied in the case that disaster insurance market has yet to be formed Furthermore, if the authors only base on the secondary data, it will be hard explicit the factors influencing the insurance purchase decision of farmers Many studies tried to combine the available secondary data with primary data In the context of innovating insurance program, many attributes would be changed and new agricultural insurances were introduced, a survey was conducted to learn how attributes influence farmers’ preference (Sherrick et al., 2003) In order to get an insight into effects of economic and weather condition

on buying insurance decision, secondary data were employed and combined with primary data in this study Interestingly, the methodology applied in Sherrick’s study is conjoint analysis which support to form a specific product through statistics about attributes evaluation of respondents Sherrick and his partners found that flexibility in choosing the type of insurance and coverage level have strong influence in insurance preference of Montana farmers However, conjoint analysis also make insurance design seem to be more complex, and this method also require the large sample size and meticulous attribute levels to achieve effectiveness In other study, the determinants such as farm size, irrigation and some individual indicators have effects on buying crop insurance (Enjolral et al, 2012) Researches which can access to a sufficient data and perform

a two-stage analysis in expected utility framework including measurement of crop insurance demand elasticity and determination of purchasing crop insurance decision factors such as Enjolral’s study (2012) is rare, because using secondary data or combining of secondary data and primary data only cannot be applied, if the market of that agricultural has yet to be formed or inaccessible

2.1.2 Researches not applying RUM on agricultural insurance using primary data

In some developing countries, since many kinds of agricultural insurance is still a new product, secondary date is not sufficient, conducting a survey to capture potential development capacity of crop insurance or disaster insurance is preferred In order to define the determinants of flood insurance demand in Malaysia, demographics, aspects of exposure level, the ability to resilience and adapt, residents’ perception about vulnerability level were considered (Aliagha et al., 2015) But this study is quite simple and the result does not explicit any prominent issue in insurance market or any particular flood insurance package should be launched

In insurance market, due to information asymmetries, market failure such as adverse selection and moral hazard can occur Therefore, studies applying Expected Utility Maximization in theoretical framework were expanded besides controlling heterogeneity They considered the effects of adverse selection by testing the ability to reflect individual risk perception of premium (Smith & Baquet, 1996) Based on the survey in Montana, Smith and Baquet (1996) applied two stages estimation procedures of Heckman including participation in any agricultural insurance decision (probit model) and level of coverage decision (OLS) Many determinants affected the participation decisions of local farmer Due to adverse selection, the coverage-levels decisions of farmer having positive expected return is different from farmers having negative expected return, and premium cannot promote as a useful tool in reducing loss ratio

A few year later, Expected Utility Maximization and Random Utility Theorem in theoretical framework was also applied to study the influences of insurance contract perceptions on participation decision and multi-peril insurance demand (Ye at al., 2015) Expected Utility Maximization framework is used to learn participation decision of farmers which depends on premium, insurance perception and indirect utility The effect of insurance perception on participation is evaluated by Framework of Random Utility Theorem, insurance participation and perception have simultaneous correlation which is expressed by simultaneous equations But the econometric analysis based on the data obtained from a survey in Hubei, China indicates that local farmers have a low perception about insurance contract, despite of long-time government subsidiaries May be due to this context, the strong learning-by-doing evidence of insurance market cannot be found (Ye at al., 2015)

The demand for flood insurance demand of farmers would increase after experiencing a severe flood, the result would be more precise if risk behaviors, mitigation and adaptation activities, financial constraints, psychology, and information gaps were controlled (Turner et al., 2014) Since risk behaviors of farmers is hard to capture, a lottery experiment were conducted to evaluate risk attitude of respondents In the case of Pakistan, Expected Utility Maximization framework, exhibited in Probit Model with binary insurance choice experiment as depend variable, was employed to investigate the determinants of flood insurance participation (Turner et al., 2014) Although using Binary Probit Models in Expected Utility Maximization framework is quite effective in finding the determinant of flood or crop insurance participation, it does not consider

the effects of attributes on insurance participation decision Potential clients might envision unreasonable insurance packages, and when real flood insurance packages were introduced, they would be likely to have bad perceptions about insurance or be disappointed In order to overcome this issue, some studies applied Contingent Value Method (CVM) to capture the demand for

agricultural insurances

Contingent Value Method was used to obtain the WTP for breeding-sow insurance (Wan, 2014)

In the survey, the respondents were provided a payment card method containing table of calculated premium, coverage and the premium per coverage ratio and asked to choose the most suitable combination If the respondent cannot found the most suitable one, some open-end question will be asked to support The Tobit Model were applied in this case because premium, coverage willingness to pay (depend variable) is always positive After running models with premium and coverage as depend variable, average WTP for premium and average coverage level found is quite larger than the current level (Wan, 2014) However, one of the noticeable limitations of CVM in evaluating WTP is determining the maximum amount of WTP right from one question, this might cause a significant bias Besides, restriction in presentation of circumstances or period and limitations on considering changes in insurance contracts simultaneously have urged researchers to apply better approaches

2.2 Random utility model (RUM) and applications

2.2.1 Random utility model (RUM)

According to classical economic theory, the self-interest of consumers will be tried to maximize

by themselves In the heterogeneous preferences theories, the preferences of consumer in maximizing their utility were expressed as a utility function U(x) of vector x, and x stands for the level of goods consumed under the budget constrain The budget constrain can be represented by

px ≤ a, in which p is the price vector, a is income The demand is x = d (a, p) +𝜀, where 𝜀 is the disturbance due to errors in x measurement These disturbance in consumer behavior observation might come from random factors existing in economic agent objectives or their constraints (Griliches, 1975)

In 1927, a law of comparative judgment of Thurstone suggests that individuals will response to stimulus, and during the making choice process, people will choose the alternative which provides

the highest stimulus level The alternative j have objective level 𝑉𝑗 and random error term 𝜀𝑗 In

1960, the stimulus level in Thurstone’s study were applied in economics and presented as utility

in Random Utility Maximization Model by considering random factors effects in the choice and multiple choice probability of utility maximization (Marschak, 1960)

binary-Marschak expressed as utility level will be 𝑈𝑗 = 𝑉𝑗 +𝜀𝑗 = 𝑋𝑗𝛽 +𝜀𝑗, where 𝑉𝑗 and 𝜀𝑗 also were considered as systematic component and random component respectively

In 1966, the appearance of the “new theory of consumer demand” of Lancaster was the modification in theory of demand in standard microeconomic This theory argue that the consumption purpose is not acquisition of goods themselves, but obtain characteristics that goods contain This theory help researchers in considering the combination of desired characteristics of consumers and estimating the demand curve for new products McFadden developed his study by exploiting this “new theory of consumer demand” of Lancaster, and the systematic component does not only mention about quantity of good but also about the characteristics of goods

In McFadden’s study (1978), systematic component is also the function of attributes of alternative j: 𝑉𝑗 = 𝛽0𝑗+ 𝛽1𝑥𝑗1+ 𝛽2𝑥𝑗2+ ⋯ + 𝛽𝐾𝑥𝑗𝑘, in which 𝑥𝑗𝑘 is attribute k level of alternative j, 𝛽𝑘 is the marginal utility of attribute k and 𝛽0𝑗 is the alternative specific constant which indicates the preference of alternative without attributes The error term is the random variable with Gumbel distribution: 𝜀~G(𝜂, 𝜇), in which 𝜂 and 𝜇 are location parameter and scale parameter respectively There is no doubt that the alternative having highest utility level will be chosen

In the binary choice, the utility function U(x) of vector x is random utility indicator when:

Pr (𝑈𝑥 ≥ 𝑈𝑦) = 𝑃𝑥𝑦

In the multiple-choice, the utility function U(x) of vector x is random utility indicator when:

Pr (𝑈𝑥 ≥ 𝑈𝑦, all y in 𝑋∗) = 𝑃𝑥(𝑋∗) with 𝑋∗ ∈ 𝑋0

To sum up, the probability of choosing alternative j:

𝑃𝑗 = 𝑒

𝑉𝑗

∑𝐽 𝑒𝑉 𝑙

𝑙=1The coefficients of the utility functions are estimated by maximizing the log-likelihood function:

Besides, the random factor in Random Utility Model caused by unobserved heterogeneity such as experience, tastes and information source about attributes These random factor distributions generate the choice probabilities model in terms of parameters Since the unobserved components

in consumer characteristics are assumed to correlate with observed components and participate in subjective perception constitution, there exist an index of continuous random field between unobserved and observed characteristics

2.2.2 Researches applying RUM on agricultural insurance

The potential bias and other disadvantages of CVM can be overcome by RUM, because this approach can evaluate the maximum WTP through considering the choses between different drafts of insurance contract Furthermore, providing drafts of reasonable insurance package will

help potential clients to have a clear visualization Due to many advantages, Random Utility Model is very suitable to study about the demand of agricultural insurances

Like flood insurance, holistic insurance is a very new product and quite complicated in operation,

so RUM were applied to capture the demand for this product (Nganje et al., 2008) Based on the Choice Experiment process, a survey were conducted after testing the levels of attributes and using D-optimal main effect to choose alternative This is a good point of Nganje’s study If this step is not performed careful, the result found might be affected the result negatively, and false expectations of potential customers for flood insurance could be created Many studies did not test the scale of level or consider about the income of local farmers, they just based on the average level obtain from the past (Mercadé, 2009) Some studies set inappropriate and unreasonable levels of attributes (Brouwer et al., 2013, Reynaud et al, 2012) These mistakes are usually found

in attributes such as premium, coverage

Questionnaires used in survey usually include demographic information and CE questions After received choice card containing three kinds of insurance package, respondents were asked to choose the package insurance that they think is most suitable or they could choose “none of them option” (Nganje et al., 2008) In some studies, the authors want to reduce the confusion and help respondents to make decision based on rational consideration, each choice card in study contains only two insurance packages and “none of them” option (Mercadé, 2009; Brouwer & Akter, 2010; Brouwer et al., 2013, Reynaud et al, 2012; Opiyo et al., 2014; Liesivaara & Myyra, 2014) Since holistic insurance is the combination between crop insurance and health insurance with the difference in level of subjects insured, attributes were build based on four scenarios, that farmers might face, include farm disaster and health good; farm disaster and health bad; farm good and health bad; farm good and health good (Nganje et al., 2008) Each insurance package usually has five attributes

Table 2.1 Attributes in the insurance package in Nganje’s study

Type of insurance RAH: revenue assurance, farm selects target revenue health and dental

AGRH: adjusted gross revenue guarantees farm historical revenue health, dental and vision

MPCIH: multi-perial crop insurance coverage for yield-related losses and health

Source: Nganje et al, 2008

Many attributes in Nganje’s study (2008) can be found in other studies, though some authors used the different names, the definition is the same

Types of insurance mainly mention about the combination between crop insurance and health insurance with the difference in level of subjects insured Some author named this attribute

“policy” or “risk covered” In Nganje’s study (2008), the types of insurance is suitable, they reflect the situation of each household and tendencies in their demand For example, households having the medical history may choose AGRH, and risk averse families might think MPCIH is the most suitable one Furthermore, types of insurance in this attribute did not violate the rule that alternatives must be mutually exclusive of RUM However, many other studies about flood insurance demand violate this basic rule of RUM They used mutually inclusive policies such as property damage insurance, crop damage insurance, health damage insurance, unemployment income insurance (Brouwer & Akter, 2010, Reynaud et al., 2012) or yield insurance and rainfall insurance (Jorgensen & Termansen, 2015)

Some studies called implementing agencies “provider” Options in implementing agencies attributes should be consistent, and do not make potential clients confuse Thus, the author should choose to present consistent levels based on proportion of business owners such as government

company, private company (Brouwer et al., 2013, Reynaud et al., 2012; Opiyo et al., 2014) or product provided by insurance company, micro-credit company Using inconsistent levels in providers attribute will affect negatively on study result (Brouwer & Akter, 2010)

“Coverage level” refers to percentage of damage insured (Nganje et al., 2008), but percentage of damage insured should be called “indemnity insured level” rather than coverage level However,

if we use percentage of damage insured as an attribute, the purpose of encouraging insurance buyer to joint in mitigation activities might not be emphasize Thus, this attribute should be replaced by “deductible” which refers as the part of loss would not be compensated and buyers have to bear this part (Liesivaara & Myyra, 2014) Sharing the burden of crop failure can be used

as motivation for farmers to participate in mitigation activities Through bringing “deductibles” into insurance package is a good contribution, but this have yet to be applied in flood insurance studies

In other flood insurance demand studies, cover is defined as the largest amount of money that farmers can received if productivity is lower than threshold in contract due to disaster (Brouwer

& Akter, 2010; Brouwer et al., 2013) However, many mistakes are found in the scale level of cover attribute In the study of Brouwer (2010), the level of insurance cover is quite low, since the highest flood damage in these districts is 738USD per household and the lowest damage is 224USD per household, cover levels cannot insure for damages at medium levels Some studies used unnumbered levels such as “low” and “high”, levels are unclear, but the authors did not mention about any specific regulations relating to cover levels (Opiyo et al., 2014)

Subsidy switch is a feature of holistic insurance, flood insurances and other single insurances do not have this attribute Subsidy switch is the health coverage levels that insurance buyers could switch to crop coverage level, and vice versa (Nganje et al., 2008) This attribute make holistic insurance package to be flexible and provide convenience for insured However, calculating the switch cost is quite complicated for farmers

Since holistic insurance is complicated, its complication also exists in calculating the premium Due to characteristic of holistic insurance, insurance payment were forced to express in percentages based on farm’s wealth, subsidy switching and historical regional payments Using percentages or unspecified amount of money makes understanding the insurance packages clearly

more difficult especially when buyers and respondents are required to calculate the insurance premium by themselves Thus, farmers who are inherently sensitive to the indemnity payment can feel that insurance costs are not always consistent with coverage levels

Although, using percentages or unspecified amount of money makes calculating the premium to

be not simple, many studies of flood insurance or other single insurance still used percentage to express levels of premium (Reynaud et al, 2012; Nganje et al., 2008; Opiyo et al., 2014; Ranganathan et al., 2014; Jorgensen & Termansen, 2015) Furthermore, premium were usually set inappropriate levels For example, authors did not consider about the income of local farmers, just based on the average level obtain from previous study (Mercadé, 2009) In other studies, the levels of premium is quite suitable with average per capita income in this region (13USD/month), but they are unreasonable and too low to motivate insurance company to join in this market (Brouwer & Akter, 2010) Therefore, setting an appropriate scales of level is very hard, but it is very important to obtain an authentic result

Besides attributes, the econometric models also play important roles in capturing the flood insurance demand and WTP evaluation In studies focusing on the demand for particular agricultural insurance package, using Conditional Logit Model is appropriate, because the authors want to capture the impacts of insurance package characteristics (the characteristics of alternative)

on each individual’s decision Conditional Logit Model will support to build functions reflecting the relationship between characteristics of insurance packages (the characteristics of alternatives) and each insurance package's utility (each alternative’s utility) to individual (McFadden, 1973)

In order to know more about parameter of attribute, some studies use Mixed Logit Model (Opiyo

et al., 2014; Liesivaara & Myyra, 2014) They found that premium have triangle distribution, and distribution of deductible and indemnity is normal with the control of premium attributes effects (Liesivaara & Myyra, 2014)

In addition, if the authors considered that choosing no insurance product were one nest and choosing one of three holistic insurance product were one nest The first-level decision is the function reflecting the effects of social-demography variables on respondents’ decision The second-level decision will show the effects of attributes on respondent’s decision that is choice experiments elicitation They would applied the Nested Logit Model (Nganje et al., 2008)

Since using RUM to study about agricultural insurance especially flood insurance is quite new, many aspects such as attributes are needed contributing Researchers should try to include attributes can motive farmer to safe environment such as deductibles (Liesivaara & Myyra, 2014) and mulching required (Jorgensen & Termansen, 2015)

Using new attributes to evaluate the insurance demand better is good, but including some attributes such as “Coverage Start Date”, “Coverage End Date”, “Claim Settlement Time” (Ranganathan et al., 2014) only made the choice set more complicated and hard to understand, because allowing to choose these attributes will make insurance packages inhomogeneous Thus, issues relating to insurance period or compensation claims process should be set as regulations in insurance contracts

Furthermore, some attributes added are unrealistic They are not characteristics of actual flood insurance package could be provided such as flood return period, probability of fatality, length of social disruption (Brouwer et al., 2013) Actually, these attributes are status of surrounding environment which might affect decision to purchase flood insurance of local residents, and insurance providers cannot affect these statuses immediately

2.2.3 Review of flood insurance demand research using RUM

Using primary data to examine the WTP and existent of flood insurance market has not yet been popular There are around three studies applying RUM to explore the potential of flood insurance market (Brouwer & Akter, 2010, Brouwer et al., 2013, Reynaud et al., 2012) None of them were conducted in Mekong River Delta In addition, there are no study creates the motivation for farmers to participate in mitigation activities and many mistakes can be found in these studies The pioneering research in this field is the study of Brouwer and Akter in 2010 Based on the records of flood situation, surveys were conducted, the authors provided respondents different choice cards containing information about two alternatives and none of them option Alternatives usually have 4 or 5 attributes in choice experiment

Table 2.2 Attributes of flood insurance package and their levels in previous study

4 Insurance premium

(BDT/week)

5 – 10 – 25 – 50 (0.07USD – 0.15USD – 0.38USD – 0.77USD)

Source: Brouwer & Akter, 2010

In three studies mentioned above, there are many common mistakes mentioned above such as violating the mutually exclusive rule or using inconsistent levels in insurance provider or setting inappropriate levels or using incomprehensibility attributes In this part, these mistakes will be briefly discuss again

When applying RUM to capture existent of flood insurance market, the common mistake are violating mutually exclusive rule In the study of Brouwer and Akter (2010), four kinds of insurance policy such as property damage insurance, crop damage insurance, health damage insurance, and unemployment income insurance can be owned simultaneously, so mutually exclusive rule in RUM are violated The study of Reynaud and partners (2012) have similar mistake, but this mistake also led to complication in setting scale of levels, since different kinds

of insurance require difference scales for each attributes

In some studies, types of insurance providers attribute levels are not consistent (Brouwer & Akter, 2010), the author should choose to present levels based on proportion of business owners such as government company, private company (Brouwer et al., 2013, Reynaud et al., 2012) or product provided by insurance company, micro-credit company

Setting inappropriate levels could affect the result and create false expectations of potential customers for flood insurance The levels of insurance cover is quite low, since the highest flood damage in these districts is 738USD per household and the lowest damage is 224USD per

household, cover levels cannot insure for damages at medium levels Although the premium is suitable with average per capita income in this region (13USD/month), it is still low to motivate insurance company to join in this market This mistake also remains in other studies mentioned above For instance, “maximal annual insurance cover” attribute in Reynaud’s study (2012) has three scale, and the coverage levels are quite high, but the average flood damages in 5 years before the survey is much lower than the coverage level Thus, testing levels should be based on annual income of local resident, flood damage statistics, opinions of experts from the insurance company

As for the comprehensibility of attributes, using specific number as levels will help to clarify attributes and insurance packages rather than using percentage Percentage may be flexible but it may cause some confusions for respondents when choose alternatives This issue might have caused the low demand in insurance package in Reynaud’s study (2012) Therefore, besides using images to exhibit the attributes and its levels, the authors should try to not use percentage as the levels of attributes relating to premium, cover to reduce the confusion of the respondents about attributes

Moreover, redundant attributes whose results can be predicted before survey should not consider

as attributes to prevent complication of choice cards In Reynaud’s study (2012), the authors argued that monthly payment attribute is used to take into account the liquidity constraint of rural households in this study However, this attribute is unnecessary, as the authors did not mention about any special policy payment for buyer who pay premium monthly in paper, and even if the buyers have monthly payment opportunity, they still choose to pay premium yearly

To evaluate the insurance demand better, some authors might want to use new attributes This action should be encouraged, but some of new attributes are unrealistic Some attributes of alternative in the survey is not characteristics of actual flood insurance package could be provided

In Brouwer’s study (2013), attributes such as flood return period, probability of fatality, length of social disruption in are actually status of surrounding environment which might affect decision to purchase flood insurance of local residents, and insurance providers cannot affect these status immediately

To sum up, determining attributes for flood insurance plans and setting the suitable scales level for them are very important In order to help the survey to be successful, the authors should choose attributes which reflect necessary information about flood insurance packages but not the operation

process The attributes also should be easy to understand Thus, attributes of flood insurance plan should be insurance policy, provider, cover and premium Our flood insurance packages will be better if we could encourage farmers to join in mitigation, so including deductible attribute should

be considered Besides, the insurance plans must have consistent levels in insurance provider and appropriate levels in other attributes, they also have to obey the rules of Random Utility Model such as mutually exclusive rule

2.3 Challenges of disaster insurance market

Besides no compulsory flood insurance participation in Mekong River Delta before, flood insurance companies also need to consider some challenges for their market expansion in this region Some typical challenges preventing the demand for flood insurance are ambiguity, adverse selection, charity hazard (Botzen, 2010)

2.3.1 The ambiguity

2.3.1.1 Definition

Ambiguity is the difficulty in assessing accurately the probabilities of flood and its potential damages, predicting these issues requires a lot investment and many technical methods (Botzen, 2010) Moreover, the combination of climate change phenomena and variation of socio-economic characteristics contributes to make prediction to be more complicated (IPCC, 2007)

In reality, simulation circumstances of climate change and socio-economic movements are used

to evaluate future risks To obtain more details about the magnitudes or the appearance trends of floods or other adverse events, historical data and statistical methods could be employed (Saunders and Lea, 2008; Schmidt et al., 2010) Yet the limitation of this methods is dependence

on historical data Indeed, even utilizing the advancement of computer-based techniques such as catastrophe models to cope with ambiguity of natural disaster consequences, predictions are not always precise (Grossi & Kunreuther, 2005) Thus ambiguity has not yet to be overcome completely, and difficulties in setting suitable price level for flood insurance have not to be solved

2.3.1.2 The impact of ambiguity on flood insurance demand

Empirical study found that the more ambiguous adverse events are, the more insurance premium would be charged (Kunreuther, 1996) The flood insurance companies have to set high premium

levels to cover for risks taken Since the severity and frequency of natural disaster are more uncertain and ambiguous, disaster insurance premium are usually more expensive than other kinds

of high-probability events insurance such as fire insurance (Botzen, 2010) The high premium might reduce the demand for flood insurance

Furthermore, the ambiguity and low-probability of flood might also cause underinsurance problem According to some risk perception studies, low precaution against low-probability of flood might cause people to neglect information about potential disaster and not purchasing insurance (Sunstein, 2002; Kunreuther et al, 2009) After being suffered from severe flood, many people may suppose that that severe flood will not come back for 100 years, so researchers used the phrase “100 years return” to reflect this low precaution of people Besides, ambiguity or low-probability of flood make some optic clients believe that purchasing flood insurance is an inefficient investment and they quit insurance contract (Kunreuther et al, 1978)

However, experiencing with great losses caused by low-probability risks, people might have good visualizations about severe flood Thus, the demand for safety and assurance of individuals might

be promoted, and the WTP for reducing risks or floods insurance would increase (Tversky and Kahneman, 1973) Setting a high premium for flood insurance may be suitable

2.3.1.3 Vulnerability perspective is a good method the capture the impact of ambiguity on flood insurance demand

The vulnerability perspective will reflect whether local farmers recognize the threat of flood in the circumstance that potential flood damages tend to be ambiguous and unpredictable If they do, ambiguity of flood will not prevent flood insurance market development

Popularity of media will raise the vulnerability perspective of people and affect their insurance purchasing decisions Nowadays, people are easy to recognize that developments of socio-economy such as population growth and prosperous lives might make them be more and more vulnerable, when adverse events happen (Miller et al., 2008) Households having many family members and properties might know that they are more vulnerable Besides, social development also awakes people about their vulnerability such as increase in greenhouse gases level and global warming phenomena Perspectives on higher uncertainty level of flood in their local area might lead to higher demand for flood insurance and higher WTP level for flood insurance

Based on a study conducted to learn about effects of risk perspective or vulnerability perspective

on WTP for flood insurance, the results show that vulnerability perspective has strong and positive effects on flood insurance demand (Botzen and Bergh, 2008) Thus, using vulnerability perspective to capture the impact of ambiguous risk on WTP for insurance might be more helpful for flood insurance providers than basing only historical data

In Mekong River Delta, information about weather, climate changes, and natural disasters such

as flood, inundation, and windstorm were provided through media In our sample, one hundred percent of respondents said that they usually used TV, radio to track the news of weather and flood Annually, before the flood season, the local residents have carefully preparation to reduce the flood damage, but losses are still large and unpredictable According to enhancing resistance

in flood for poor households in Vietnam Mekong River Delta Project in 2011 with support of United Nation, local government and residents reinforced houses, had strict examinations of flood protection infrastructures, practiced aid and provided necessary medicine for prioritized targets based on prudent plans, but the 2011 flood caused more than 1000 billion VND in property damage, 27,000 hectares area of rice and vegetables are damaged in which 10,000 hectares damaged 100%, nearly 12,000ha of fruit area also flooded (Ngoc Anh, 2011) Thus, the farmers

in Mekong River Delta must conceive quite clearly about their vulnerability Therefore, consideration about effects of vulnerability perspective of local farmers on flood insurance demand is necessary

Besides using vulnerability perspective to capture the impact of ambiguity, we control the irrigation improvement and accessibility to pumping station of respondent, as these factors also affect the demand for flood insurance

2.3.2 Adverse selection

2.3.2.1 Definition of adverse selection and its impact on flood insurance demand

Adverse selection could be express briefly as a phenomenon that insurance were only provided to high-risk individuals with a very high premium In detail, this issue coexists with an apparent problem that the insurance demand is related positively to the expected level of risks that people have to face, and the premium is expected to reflect accurately the risk expectation of insurance buyers However, information asymmetries usually exist between potential insurance buyers and

insurance company Since people think that they could anticipate their potential natural disaster consequences, while due to the ambiguous risk, it is very difficult for insurance companies to classify the customer types based on simple risk prediction results If the insurance premiums are based on the expected average loss, the insurance companies might have to face with insufficient capital status As a result, insurance companies would charge with high premium for only high-risk or risk averse individuals, and adverse selection occurs (Botzen & van den Bergh, 2008)

2.3.2.2 Reason for using risk perception or fear of individuals to determine adverse selection issue

In a study, even if insurance companies could evaluate the frequency and severity of floods for each customer groups, problems caused by information asymmetry might not be solved The natural risk expectation of individuals, called risk perception, tends to be based on intuition and

do not similar with insurance companies’ predict (Slovic, 1987; Slovic, 2000) According to one

of heuristics in the study of Kunreuther (2001), individual feelings such as fear will determine purchasing insurances instead of rational calculation Moreover, when control the government relief, people who fear of risk might choose to buy insurance, while people supposing to bear low-risk will rely on aids from government or charity organizations (Kim and Schlesinger, 2005) Therefore, risk perceptions of individual plays an important role in making decision about mitigation and choosing self-protective measures (Burn, 1999; Flynn et al, 1999)

Through Nash-equilibrium model,Rothschild and Stiglitz (1976) argue that the equilibrium of insurance market will be achieved, if we can classify two types of natural disaster insurance including one for high-risk individuals and one for low-risk individuals Thus, capturing individual risk perceptions or their fear will help to determine whether our potential client are high-risk and low-risk, so that problem of adverse selection be handled However, no many study using the fear to capture the effects of adverse selection on WTP for flood insurance were conducted, and topics of individual risk perceptions of flood also has not been exploited, and results obtained are mixed (Peacock et al, 2005)

As for flood risk perception studies, some previous researches studied this topic by evaluating the effects of socio-economic development, household characteristics, or the consequences of the previous disaster In flood risk perceptive estimations, the characteristics related to geography were also used as proximity of hazard The anxiety about flood risk of people have positive

relation with the occurrence of floods (Brilly & Polic, 2005) Using flood risk maps, Seigrist and Gutscher (2006) found that though individual fear of flood is significant and have positive direction with evaluation results of experts, deviations are still exist, depend on each area Moreover, the complicated methods using mental models were also applied Notably, from the combination of the two methods above, employing the Geographical Information System (GIS) allowed Botzen (2010) to exploit geographical characteristics in a new direction based on risk exposure level of objective such as potential flood levels, the levee system condition or distance

to the main river He also controlled characteristics of socio-economic development and floor experiences

However, this method is quite complicated, this study does not use GIS In order to capture the effect of adverse selection issue on WTP for flood insurance, the fear of flood is employed to examine whether it has impacts on flood insurance demand or marginal utility of flood insurance attributes or not If it does have impacts, the adverse selection might challenges the development

of flood insurance in Mekong River Delta In this study, dummy variables are used to distinguish between observations having great anxiety or fear about flood damages and the other ones Besides, we also control the irrigation improvement and accessibility to pumping station

2.3.3 Charity hazard

2.3.3.1 Definition of charity hazard

Charity hazard is the general status of developing countries This concept refers to circumstance that people refuse to cover by disaster insurance due to habit of relying on government and international aids Charity hazard could be considered as the special case of moral hazard, since this problem occurs when individuals hazard their recovery after flood on helping of relatives, friends or charity organizations (Browne and Hoyt, 2000)

Solving this issue is really important for government of developing countries, even if they usually receive compensation from international organizations after disaster Post-disaster assistance might cause financial burdens for governments In some cases, post-disaster assistance required government to mobilize financial resources from budget, raising tax bases, taking loan from domestic or international organizations Serious devastations after catastrophes in low-income countries might lead to budget deficit or decrease in international credit code (Mechler et al.,

2006) As for household, post-disaster assistance are not always trusted, when governments cannot compensate for resilience, the recipients have to bear the financial burden In some destitute situations, farmers or households in developing countries sometimes might be forced to sell or pawn valuable assets, despite having subsidies This may cause negative impacts on socio-economic development in future, especially in low-income countries (Linnerooth-Bayer et al, 2009)

2.3.3.2 The impact of charity hazard on flood insurance demand

With charity hazard, residents might refer that the government will and has responsibility to compensate for disaster damages (Botzen, 2010) Thus, they hazard or depend too much on government supports and do not concern about mitigations (Botzen, 2010) Supporting unconditionally for post-disaster recovery will reduce the motivation to mitigate the effects of natural disasters (Priest, 1996) Besides reducing the demand for flood insurance, charity hazard also affect the supply-side, as lack of customers might cause market to be unprofitable, the insurance companies have to set the higher premium to cover the cost, and higher cost will make fewer and fewer individuals want to insure for their potential risks (Raschky & Weck-Hannemann, 2007)

Similar to other developing countries, charity hazard could be a problem for Vietnam, the recoveries after natural disasters of Vietnam mainly base on post-disaster public assistance Farmers in Mekong River Delta might not support flood insurance purchasing Furthermore, households in poor regions such as Mekong River Delta could obtain financial supports from their relatives, money lenders, and aids from local government or international organizations (Warner

et al., 2008) The charity hazard might strongly challenge the development of flood insurance market in Mekong River Delta

2.3.3.3 The reason of using perception of government responsibility for post-flood recovery

to capture impact of charity hazard on flood insurance demand

Using the governmental relief or post-disaster subsidy statistics in the past as a factor to capture the impact of charity hazard might not enough, the results obtained are still mixed In some study, unconditional post-disaster relief in the past is found as the main cause of charity hazard and no conception of mitigations (Raschky & Weck-Hannemann, 2007) However, the result in the case

of US is adverse, the post-disaster assistance has the positive relationship with flood insurance demand (Browne and Hoyt, 2000) Although the author argued that natural disaster exposure will increase flood insurance demand and post-disaster assistant, mixed results still lead to the ideas

of using perception of government responsibility to capture the impact of charity hazard

In some other studies, they found that perception of government responsibility is the main cause

of low demand for flood insurance When charity organization and government relief work as the free disaster insurance, individuals recognize that not buying insurance will let them be the most beneficial ones As a result, individuals believe that government has responsibility for recovery

of society and economy after disaster and high-level of uninsured property, because government does not supply disaster insurance properly This perception is the main cause of low demand for flood insurance (Prettenthaler et al, 2004) The expectation about government relief also provides similar result (Botzen, 2010) Based on the theoretical framework, the expected government relief

of individuals will reduce the demand for flood insurance and mitigations (Kelly and Kleffner, 2003) Under expected post-disaster relief will cause rational people to not insure themselves (Lewis and Nickerson, 1989)

In this study, dummy variable was used to capture the effect of charity hazard in farmers’ perception on WTP for flood insurance We could expect that the charity hazard in perception of

farmers will have the negative relation with demand for flood insurance of residents

CHAPTER 3

RESEARCH METHODOLOGY

This chapter has four sections First, we will briefly present the current situation of flood insurance market in MRD and the advantages of Choice Experiment in analyzing the demand for flood insurance Second, general model and estimation methods will be present Third, we will introduce the details about empirical models and willingness-to-pay calculation Finally, data collection process will be described

3.1 Demand for flood insurance

In the circumstance that natural disasters occur annually in Mekong River Delta and cause many damages to agricultural production, local farmers have no solution to mitigate the burdens of cultivation costs but rely on subsidies or aids from humanitarian organizations Mekong River Delta has been recognized as the potential market of many insurance companies since 2013, but some agricultural insurances including flood insurance is still in pilot stage (Pham, 2015) Since flood insurance has been a new product with Mekong River Delta farmers, using choice experiment is suitable

In insurance sector, people usually used the insurance statistics data to investigate the flood insurance demand (Kunreuther et al., 2009; Michel-Kerjan & Kousky, 2010) In recent years, some studies using Choice Experiment Model to capture potential markets of flood insurance were conducted, but besides worth learning points, many mistakes still remain

3.2 The advantages of Choice Experiment compared to Contingent Value Method

Besides Choice experiment (CE), Contingent Value Method (CVM) is also a popular method applied by economists Contingent Value Method (CVM) is the non-market value method which

is used to determine the preference of individual for public goods, environment quality and other specific goods (Carson et al, 2005)