A computational framework to support government decision-making in regional hurricane risk management
Date
2018
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
This dissertation introduces a computational framework that can be used to identify hurricane disaster risk management policy solutions based on behavior of the system as a whole, including interactions among multiple types of stakeholders (homeowners, insurers, government, reinsurers) and strategies (insurance, retrofit, property acquisition), Specifically, it supports the following government decisions: (1) how much to spend on mitigation, (2) how to regulate the price of extreme event insurance, (3) how to allocate spending between homeowner retrofit grants and property acquisition, and (4) how to design retrofit grant and acquisition programs. The framework includes four interacting mathematical models—stochastic programming optimization models to represent government and insurer decisions, empirical discrete choice models of individual homeowner decisions, and a regional loss estimation model. It includes a description of how insurers and homeowners are expected to respond to government policies and what the outcomes will be for each. A full-scale application for eastern North Carolina suggests it is possible to identify system-wide win-win solutions that are better both for each stakeholder type individually and for society as a whole. For comparison, another version of the framework that uses a utility-based homeowner decision model is also presented. The comparison shows some similarities and differences between the two frameworks and in particular, suggests the utility-based framework is more sensitive to price changes. ☐ Within this extended framework, understanding the circumstances under which homeowners will purchase insurance is critical to creating an effective insurance market for hurricane wind and flood loss. This dissertation introduces empirical homeowner wind and flood insurance purchase decision models, which contribute to the empirical literature on the subject through an analysis of survey data for homeowners in North Carolina. Separate mixed logit models for flood insurance and wind insurance purchasing decisions are developed. The analysis uses stated preference data on the influence of premium and deductible to address some limitations of revealed preference data in which it is difficult to fully decouple effects of premium, deductible, risk, and coverage limit, and mandatory purchase requirements. The results for flood insurance and wind insurance are similar. There is evidence that the following are all significant and associated with higher probability of purchasing insurance—lower premium, lower deductible, more recent previous hurricane experience, location in a flood- plain or closer to the coast, higher income, and younger homeowners. However, demand is relatively inelastic with respect to premium and deductible, and the willingness to pay for a $1 reduction in deductible varies throughout the population with some willing to pay more than $1, a behavioral anomaly. The recency of the last hurricane experience is more influential for homeowners who experienced damage than for homeowners who did not. Results suggest that insurance purchase and home retrofits are complements, not substitutes. Finally, statistical models that can be used to predict insurance penetration rates for a region under different premium levels are presented to be used in the framework.