Pricing of insurance products

Abstract

The US insurance industry totaled $1.22 trillion in 2018, with about 51% of the share being that of Property and Casualty, which has automobile insurance and commercial Insurance as its main two components. Considering the colossal insurance industry, for the purpose of this report, we have chosen the pricing policy of automobile insurance. According to the American Automobile Association's (AAA) 2019 statistics, the average cost to own a 2019 model vehicle is approximately $9,282, and the average Insurance on a medium sedan is $1,251. The prices for average Insurance have also risen from $944.36 in 2016 to $1004.58 in 2017, with the highest being in the state Louisiana at an average of 1443.72. Thus, a high variation is observed in the pricing of the automobile insurance premium in the USA. The premium calculation for insurance policy of automobiles generally relies on a linear regression model with a limited number of parameters usually taken into consideration. These parameters like the speed of driving, state of intoxication, the person behind the wheel, etc. are generally reported by the policyholder, and trust forms a significant factor in these. The other parameter, like the number of insurances claimed last year, the model of the vehicle into consideration, is previously updated. However, with vehicle prices and model specifications changing with each new model introduced, the real-time effects of these introductions onto the Insurance and pricing are not considered in the policy. In such a situation, with the fierce competition arising between the many insurers in the market and the low switching costs, an adequate prediction of the future insurance costs on a case-by-case basis for each individual becomes a business imperative. In this study, data collection will be facilitated by taking the data of various policyholders and insurance claimants into view. Monitoring of specific data such as the speed of the vehicle and proper maintenance schedules can also be done by collaborating the data from the devices installed in the policyholder's car. Using all the above data, along with the comparative vehicle specification and prices, machine learning algorithms will be used to risk premium estimates and policy amounts on a real-time basis of the policyholder data. In such a case, the policy insurers can also opt-in to provide specific discounts to safe drivers or charge people based on the various parameters mentioned above. Data has been considered from the Google Cloud and the Kaggle datasets. This data is for the frequency of accidents as well as the automobile policy claims in the USA. Initially, we assess the correlations of the various parameters that lead to automobile accidents. Apart from the ones traditionally considered, few of the new parameters considered are colour of the vehicle, the number of children as well as consumption of alcohol, proximity to liquor stores, etc. These parameters, which lead to higher accidents, are then introduced in the formulation for the policy premium calculations, with a higher premium being charged for the people with a higher risk of accidents.