In response to “Beyond Triangles Capturing Insights From New Analytic Technology” (AR September-October 2021), I agree that improving technology and a growing set of analytical tools at hand means that we have more options today to do reserve analysis than we had a few years ago. The article seemed to imply though that P&C reserving actuaries have a choice between staying with link ratios and aggregate triangles or moving on to use more advanced analytics at an individual claim level for reserving both to identify segments of the business that behave materially different than other segments for a given coverage and to do the forecasting work to create reserve estimates. I believe there is a third option that has advantages over either of those approaches. One could do the segmentation work for a given coverage at an individual claim level using claim characteristics that link to policy-level characteristics and then do the forecasting work to create the reserve estimates using Bayesian Markov chain Monte Carlo (MCMC) methods on incremental aggregated data using those selecting segments in a single model. Using a Bayesian MCMC approach lends itself to credibility weighting the segments to help with reliability in forecasting the total payments as well as providing a readily available means to deal with correlation issues in forecasting and tools to test the total forecast reliability of different models as opposed to devising ad hoc routines.
—Michael R. Larsen, FCAS, MAAA