Situated close to the cyclone centre of the western North Pacific, Vietnam is one of the most cyclone-prone countries in the Mekong region. This thesis aims to estimate the benefits to households of an improved tropical cyclone warning service, which can be considered as a 'no or low-regrets' option for adaptation to the adverse effects of climate change in Vietnam. A review of the literature indicates that a choice experiment (CE), a stated-preference technique, is appropriate to estimate the economic value of improvements in cyclone warning services.
The CE exercise reported in this thesis involved several rounds of design and testing. The design began with reviewing attributes of an improved cyclone warning service, which have been applied in previous studies. From a list of attributes used in previous studies, surveys were conducted in coastal areas of Vietnam to select three suitable attributes for this research: forecasting accuracy, frequency of update and mobile phone based warnings. An orthogonal choice task design was then used, resulting in twenty-four choice tasks. In order to reduce the cognitive burden on respondents, the twenty-four choice tasks were divided into four blocks of six choice tasks; each respondent randomly answered a block of six choice tasks. Two pilot surveys established the respondents' acceptance of the survey instruments. A face-to-face survey with a sample of 1014 households was subsequently undertaken at four sites, which represent both urban and rural coastal areas in the Northern and Central regions of Vietnam, to inform the estimation of willingness-to-pay (WTP) in this thesis.
The results of the WTP estimates for the attributes show that the WTP estimates for the accuracy attribute increase as the improvement levels are enhanced; thus, the WTP for accuracy improvement level 1 (USD1 - 1.3) is smallest when compared with improvement level 2 (USD1.6 - 3.1) and level 3 (USD3.5 - 5.4). Within the proposed improvement levels for frequency of update of 8 - 16 times per day, the WTP estimates for a unit change in this attribute are about USD0.14 - 0.20. A Vietnamese household is estimated to be willing to pay approximately USD2.2 - 4.2 for the mobile phone message based warning service. In relation to the total WTP estimates, the benefit estimates of the maximal improvement scenario, in which all cyclone warning attributes are set at the highest levels of proposed improvements, are approximately USD7.1 - 8.1 per household. The total WTP estimates are similar to the values of a cyclone warning service reported in the literature. An investigation of factors affecting preferences for improvements in cyclone warning services found that age, household size, fishing job, Central region residency, income, education and risk aversion influence the WTP.
In an ideal CE exercise, respondents are assumed to consider all presented attributes, have stable preferences when making choices in a sequence of choice tasks and choose alternatives that truthfully represent their preferences. In the real world, however, there is a possibility that some respondents would not behave in the ways assumed by researchers. To provide a better understanding of the estimated values, this thesis investigates a number of issues, including attribute non-attendance (ANA), effects of choice task order in a sequence of choice tasks and strategic behaviour of some respondents. In relation to the ANA phenomenon which asserts that some respondents may use only a subset of the presented attributes to make their choices, this thesis found no significant differences between the WTP estimates that accounted for ANA and the WTP estimated from the model not accounting for ANA. With regard to the ordering effects, an examination of this CE application revealed that estimates of total WTP estimates in the first position are larger than the equivalent values estimated in positions 2-6. The investigation into the ordering effects is extended with an analysis of strategic behaviour. Respondents who could behave strategically were detected using two follow-up questions on their beliefs about the consequentiality of the survey and of the payment obligation. After progressing through the first three choice tasks, the strategic respondents appeared to become more aware of the strategic opportunities in the series of repeated questions; and to behave strategically in the last three choice tasks. Their strategic behaviour results in significantly different total WTP from the equivalent WTP of non-strategic respondents over the last three choice questions.
This thesis also demonstrates that respondents hold not only a self-interest motivated value (termed use value), but also altruistic and bequest values with respect to the interests of others and future generations. To examine the relative importance of the values, respondents were requested to allocate 10 points between the different types of values. The more influential the value, the higher the point is scored. Use value was scored the highest mean point of 4.1 out of 10. Altruistic and bequest values were awarded similar points, approximately 2.9 and 3.0, respectively. Given the importance of non-use values, economic assessments focusing on only use value would underestimate benefits of an improved cyclone warning service to society.