|
Abstract:
|
It is widely held that healthcare spending in the US closely follows the "80-20 rule", whereby 20% of individuals are responsible for 80% of total expenditures. However, detailed research investigations of the upper tails of healthcare spending distributions are limited due to data privacy, and most have been confined to survey or subsample data. We apply recently-developed statistical tools from extreme value theory to data from a sample of aggregated commercial insurance claims of over one million Americans in an effort to quantify the varying nature of tail risk in annual health expenditures, as a function of several covariates. We study both marginal behavior and year-over-year tail dependence, and examine implications for tail risk at the level of whole insurance pools.
|
