In a paper published in the journal Scientific Reports, researchers investigated earthquake insurance uptake in Oklahoma after the 2011 seismic events. Using supervised machine learning on 812 respondents, they discovered age, gender, race, ethnicity, residency duration, and past earthquake experiences significantly influenced insurance decisions.
Study: Understanding Oklahoma's Earthquake Insurance Landscape: Influential Factors and Predictive Modeling. Image credit: metamorworks/Shutterstock
Older individuals, males, longer-term residents, and those with quake exposure were more inclined to buy coverage than renters. The study highlighted the significance of insurance in managing environmental risks, advocating for awareness and advanced machine learning tools like random forests for predictive modeling in similar scenarios.
Background
The surge in global natural disasters emphasizes the need for robust insurance mechanisms to manage risks associated with events like hurricanes, earthquakes, and floods. However, understanding people's perception of disaster consequences and the factors influencing their risk perception, such as demographics and local contexts, needs to be addressed in earthquake insurance design.
Previous studies across various regions highlight how risk perception, experience, and societal factors drive individuals' decisions to purchase disaster insurance, underlining its significance in mitigating economic and societal losses. The staggering costs and devastation wrought by events like Hurricane Katrina and the Sichuan Province earthquake stress the criticality of effective disaster management strategies and policies.
Determinants of Oklahoma Earthquake Insurance
The research gathered data through a 2017 online earthquake survey conducted in Oklahoma, using Qualtrics, forming part of a study examining the liability of the oil and gas industry for earthquake damage. This survey, approved by the Oklahoma State Institutional Review Board, involved 1,153 participants, with 813 completing the questionnaire.
The data encompassed respondents' attitudes towards earthquakes and socio-demographic information, constituting the variables used for analysis. These variables were classified into socio-demographics and attitudes toward earthquakes to understand their influence on residents' decisions to acquire earthquake insurance.
Independent variables analyzed included socio-demographic factors like age, gender, race, and attitudes toward earthquakes in Oklahoma. The descriptive statistics illustrated differences between respondents with and without earthquake insurance. For instance, individuals with insurance tended to be older, reflecting potential correlations between various demographic factors and insurance uptake. Anticipations from existing literature suggested diverse relationships between demographic factors and insurance ownership, such as minority groups' wariness about environmental threats and older individuals' support for specific industry activities.
In addition to socio-demographics, respondents' duration of residence in Oklahoma, property ownership status (renters vs. property owners), political affiliations, and past earthquake damage experiences were hypothesized to impact insurance decisions. Expectations anticipated that individuals residing in Oklahoma for an extended period would demonstrate a higher likelihood of insuring their property, attributing this inclination to their increased exposure to earthquakes since 2009. Anticipations suggested that renters would show lower insurance uptake. Political affiliations, earthquake damage experience, and beliefs regarding the role of oil and gas companies in Oklahoma were also considered influential factors in predicting insurance acquisition behavior.
The study employed supervised machine learning techniques, including logit models, ridge regression, Least Absolute Shrinkage and Selection Operator (LASSO), decision trees, and random forests to analyze and predict earthquake insurance uptake. These techniques aimed to understand the influential variables affecting insurance decisions and predict individuals likely to purchase earthquake insurance, reflecting the complexity and nuances within the dataset. Researchers utilized each method, considering its suitability in classification problems and its capacity to decipher influential factors and make predictions, thereby offering a comprehensive understanding of the determinants of insurance acquisition.
The evaluation of these models involved a range of performance measures, including accuracy, sensitivity, specificity, and precision. This multi-faceted assessment allowed for a comprehensive review of the decision trees and random forests' effectiveness in identifying individuals likely to purchase earthquake insurance. The estimation strategies utilized various R software packages and data division into training and testing sets to ensure reliable model estimation and prediction. Researchers employed techniques like cross-validation to determine optimal parameter values, prune decision trees to prevent overfitting, and optimize model performance. These methods provided robust analyses, unraveling the determinants of insurance uptake in Oklahoma's context of earthquake risk.
Oklahoma Earthquake Insurance: Influential Factors
This study delves into the factors impacting Oklahoma earthquake insurance uptake, examining influential determinants and predictive modeling outcomes. The investigation involved diverse methodologies, including logit models, ridge regression, LASSO, decision trees, and random forests. These approaches collectively identified influential factors such as socio-demographics, attitudes toward earthquakes, and connections to the oil and gas industry, shedding light on who tends to acquire earthquake insurance.
The findings underscore several trends: age, gender, ethnicity, and political affiliation significantly influence insurance uptake. Surprisingly, Democrats showed lower ownership compared to Independents, contrary to initial expectations. Additionally, tenure in Oklahoma, housing status, education level, income, earthquake experience, and attitudes toward environmental regulation surfaced as crucial factors shaping insurance decisions.
Predictive modeling, employing decision trees and random forests, showcased robust capabilities in foreseeing individuals likely to obtain earthquake insurance. The decision tree offered approximately 90% accuracy, whereas the random forest excelled with 100% accuracy. These models highlighted intricate patterns, revealing nuances in property rental status, oil and gas company presence, tenure, income, and beliefs regarding wastewater injection as pivotal factors impacting insurance uptake.
Conclusion
To sum up, this study illuminates the multifaceted landscape of factors influencing earthquake insurance uptake in Oklahoma. It unravels significant determinants such as socio-demographics, attitudes toward earthquakes, and ties to the oil and gas industry through various methodologies. The findings underscore the nuanced interplay of age, gender, ethnicity, political affiliation, tenure, housing status, education, income, past earthquake experiences, and environmental attitudes in shaping insurance decisions.
Moreover, predictive models like decision trees and random forests demonstrated robust capabilities, elucidating intricate patterns and highlighting the pivotal impact of factors like property rental status, oil and gas company presence, tenure, income, and beliefs regarding wastewater injection on insurance uptake.
Machine Learning Powering Breakthroughs in Climate Forecasting and Modeling