In this section, we first use the Pessimistic Assumption that sets the degree of suspected fraud at 100%, which means that the total claim payments by the industry for these suspected cases represent detected fraud. We also assume that the multiplicative factor (3.4), obtained from our Best Guess Assumption for the 2,772 coverages, applies to the 2,454 claims for which information on claim payments is available.
An interesting corollary of the present study is the finding of “p” is equal to roughly 113. Again “p” is the conditional probability for claim adjustment staff to detect fraud, given the claim is fraudulent. This can be seen as a significantly low index of efficiency for the entire verification process. An important question therefore arises: Why is this index of efficiency so low?
There are countless answers to that question: 1. It can reflect the incompetence of claim adjustment staff to efficiently identifying fraud cases. They may not have the adequate experience or training to detect fraud, which in fact is not necessarily their main preoccupation. 2. It can yield serious doubts about the relevance of fraud indicators used to flag possible fraud cases. 3. It may be related to the low quality or quantity of investigations.
The results can also reflect an induced laxity by insurers because of the low anticipated benefits of fighting fraud. Choosing the right answer cannot be made without a proper study of the real incentives of each participant in the market to fight against fraud. In Dionne and Belhadji’s study, they found that a large proportion of the fraud cases (93%) were not prosecuted. The main reason for nonprosecution was “insufficient proof’ (59%). This high percentage of unprosecuted claims for that particular reason naturally triggers a question.
Why was the investigation not pushed further A possible answer may reside in the fact that many of these claims represent low monetary values. If the claim amount is too small to justify the costs of further investigations, then maybe higher deductibles are in order. Higher deductibles would raise claim levels to the point where investigations could be worth pursuing for the insurers. However, higher deductibles may also increase the benefits of build-up by insureds. Investigations and prosecutions have also been seen as bad publicity for the investigating and prosecuting firms.
The fraud problem is not only a problem of robbery but endangers the very principle of insurance. The question remains with the industry. As researchers, we will focus our attention on finding some statistical and management tools in order to isolate the main causes and improve the claims-premiums ratio in that market.
Our Best Guess Estimator roughly yields a 10% fraud rate, and this result is found to be quite stable. However, the fraud rate is found to have a 16.5% upper bound. The findings in Dionne-Belhadji (1996) are multiplied by 3.4, which were given as a floor estimate, or observed fraud rates. In monetary values, this means that total fraud payments by the industry in 1994-1995, ranged from 96.2 to 208.4 million dollars instead of 28.4 to 61.3 million dollars. In other words, our results indicate that 10 to 21.8% of all claim payments are fraudulent instead of 3 to 6.4%.