The HPCDM procedure is a new, experimental high-performance procedure in SAS/ETS. The HPCDM procedure estimates a compound distribution model (CDM), which is the distribution of an aggregate loss that you expect to see in a given period of time. The aggregate loss depends on the number of loss events that occur in a given period of time and the severity (magnitude) of each loss event.
If you have estimated the probability distribution model of the frequency of loss events by using the COUNTREG procedure in SAS/ETS, and if you have estimated the probability distribution model of the severity of each loss by using the SEVERITY procedure in SAS/ETS, then PROC HPCDM combines the frequency and severity models to estimate the distribution of an aggregate loss. The probability distribution model of the aggregate loss is referred to as the compound distribution model (CDM).
At its core, PROC HPCDM uses the Monte Carlo simulation method to generate a random sample of the aggregate loss. The random sample is used to compute empirical estimates of various summary statistics and percentiles of the compound distribution. The HPCDM procedure also has the following features:
PROC HPCDM performs a scenario or what-if analysis when the distributions of severity and count depend on a set of exogenous variables. You can simulate a scenario for one entity that is subject to loss events. For example, you can estimate the distribution of the aggregate loss that one insurance policyholder might incur by using the policyholder’s characteristics. You can also simulate a scenario that consists of multiple entities that are subject to the loss events. For example, you can estimate the distribution of the aggregate loss that is incurred together by a portfolio of financial instruments by using the characteristics of each instrument as well as the macroeconomic factors that affect each of them.
PROC HPCDM can perform a parameter perturbation analysis to assess the effect of the uncertainty in the estimates of severity and frequency model parameters on the uncertainty of various statistics of the compound distribution. For example, if you are interested in computing the value-at-risk (VaR), which is usually the 97.5th or the 99.5th percentile of the aggregate loss distribution, then by using perturbation analysis of PROC HPCDM, you can get an estimate of not only the mean expected VaR but also the standard error that is associated with VaR. Both estimates can help you maintain an adequate amount of capital for regulatory compliance as well as a healthy, solvent business.
PROC HPCDM estimates the compound distribution of an aggregate adjusted loss when you specify SAS programming statements to adjust the magnitude of each simulated loss. For example, this is helpful in estimating the distribution of the total amount paid by an insurance company to its policyholders in a given period of time. The amount paid for each loss depends on the magnitude of the loss as well as the various provisions in the insurance policy, such as the deductible and policy limit.