|
Abstract:
|
The taint of an insurance claim is the dollar amount the insurance company overpaid on the claim (owing to poor claim management, fraud, corruption, or other errors). During a claims audit, companies are often interested in estimating the total taint across all claims they've managed. Since it's not feasible to audit all claims, sampling methods and estimators need to be devised. We developed a series of R programs to simulate the bias and variability of ratio estimators under three different sampling schemes: Simple Random Sampling, Stratified Random Sampling, and Probability Proportional to Size Sampling. We also varied the simulations across a variety of sample sizes, regions, and assumed taint distributions. We selected an estimator that minimized the variability and bias of the estimate of the total population taint.
|
