Exams in Real Life: MAS-II
Have you ever been studying for an exam and thought, "Why am I learning this stuff? When am I ever going to use any of it?" If so, you are in luck! This is the second in a series of articles examining how content from CAS exams are used in real life.
This issue we're focusing on Modern Actuarial Statistics-II (MAS-II). This is the exam candidates are probably least familiar with, as it has only been offered twice so far and has had about 100 takers total. It replaced the CAS Exam 4 requirement, which most candidates fulfilled via the old SOA Exam C. MAS-II does not have exactly the same material as Exam 4/C, however. The CAS took the opportunity with the exam change to update the syllabus to reflect more current (and more advanced) statistical methods, thereby helping the exam program keep pace with the rapidly evolving insurance environment.
For an actuary's everyday life, the usefulness of the subjects covered on this exam will be most evident to actuaries who fit and design predictive models. In the last issue of Future Fellows, Nate Willilams explained very well* how the MAS-I syllabus prepares an actuary for predictive modeling. MAS-II builds on this idea, but instead of focusing on time series and extended linear models, it covers Markov Chain Monte Carlo and linear mixed models, to name a couple of the central items on the exam. Giving the actuary exposure to more topics in predictive modeling particularly helps in diagnosing issues in their models. Understanding how concepts like regularization, bagging, boosting and basic feature engineering/selection affect the performance of a learning model can help the actuary figure out problems they encounter when training a model. MAS-II introduces several tools that can be used to help implement these concepts. The exam also familiarizes the actuary with R output, which continues to grow in popularity in day-to-day actuarial work.
But what about other actuaries — those who might not be fitting models themselves? As I was reviewing the syllabus to write this article, a quote from Harry Potter and the Chamber of Secrets occurred to me. Near the end of the book, Mr. Weasley says to Ginny, "Never trust anything that can think for itself if you can't see where it keeps its brain." A similar idea is true for insurance models — actuaries cannot be content to let a model be a black box. While we can rely on the work of other actuaries (with documentation of course — remember your ASOPs!), merely running an existing model program and getting a result is not good enough. We must be able "to understand a model and to evaluate the resulting goodness of fit," which is a direct quote from the MAS-II syllabus.
As more and more actuarial work involves predictive modeling, actuaries must be ready to participate in the conversation of modeling, even if it isn't in their exact job descriptions. I often sit in staff meetings where we discuss the variables used, the way a certain model will work and how well the model performs. These types of discussions will likely become even more important in the future and fit perfectly with the learning objectives of MAS-II, which also contain the directive that "candidates should focus on understanding the design choices made in modeling, the output from those [statistical software] packages, and how that output was interpreted."
Some of the other concepts on the MAS-II Exam are used more broadly in the actuarial field, such as credibility, which is covered in depth in the exam. Considering the credibility of data is second nature as we analyze trends and other metrics. Extending credibility weighting into modeling is an important advancement in actuarial statistics.
I hope this article has given a little insight into the purpose and utility of MAS-II. In the next issue, we will discuss Exam 5-Basic Techniques for Ratemaking and Estimating Claim Liabilities.