Control Strategies For Insurance Fraud With Automobile Insurance Background

Insurance fraud is an important factor that influences the profit of insurance company and increases the risk of insurance industry. Insurance fraud is widespread not only in China but also in abroad. Some literature shows that the level of automobile insurance fraud is about 10-30% in Canada and about 20% in China, respectively. Also, according to statistical data, the total fraud payments in automobile insurance ranges from 96.2 to 208.4 million dollars in Quebec during 1994-1995, and is estimated to be about 3.8 billion RMB in Beijing in 2007-2010. At present, the researches on insurance fraud mainly focus on the estimation on fraud level, the analysis of fraud causes, the identification of fraud and the control strategies. Taking account of the homogeneity of automobile insurance claims and the universality of insurance fraud, this article is from the prospective of insurer and is set in automobile insurance. We build quantitative models of insurance fraud and discuss the optimal control strategies from two aspects:customer classification and optimal investigation.From the perspective of customer classification, we build an optimization model that takes the company profits as objective function and premium income, fraud payments as independent variables. Based on the customer materials of fraud claims, this paper explores optimal strategies of customer classification and of rewards and punishments.The results show that a group of new customers could fall into three classes based on a phase of claim information. The three kinds of customers are named white customer (those who have no or a few fraud claims), grey customer (those who have some fraud claims) and black customer (those who have more fraud claims), respectively. It can be proved that if white and grey customers get an appropriate reward and punishment, and black customers are kicked out of the system directly, then the total profit of company will maximize under certain conditions.In addition, based on the real background that insurance company detects fraud, we take the total investigating expense as objective function and establish a detection model of fraud claim from a coordinating prospective of investigating cost and effect. We find that optimal detection depends on the real fraud level, fraud payment and inspection level. The three detection plans (no detection, partial detection and complete detection) have the alternative advantages and disadvantages and can be applied to different cases. More importantly, under certain scenes, these three detection plans have their own optimal strategy.This paper studies insurance fraud and the control strategies from the prospective of customer classification and cost controlling. The relevant conclusions could provide insurance company theoretical reference and quantitative basis for controlling fraud payments, increasing profit level, decreasing investigating cost and doing customer management.