Predictive analytics (PA) has become an integral part of Society of Actuaries (SOA) education, both for those seeking one of our designations and those pursuing continuing education. Let’s review where we are and what is on the horizon for SOA education as it relates to PA.
Beginning in 2018, two requirements for the ASA designation were added. The first was the Statistics for Risk Modeling Exam, which is a multiple-choice exam offered three times per year where candidates learn key concepts that underly all modeling activities, as well as the details of time series, generalized linear models, decision trees, clustering and principal components. As a multiple-choice exam, its purpose is to ensure candidates understand the technical details of these methods before applying them in the next component.
The second requirement was the Predictive Analytics Exam, which is the first SOA exam to use advanced tools in a fully proctored setting. During this exam, candidates have access to Microsoft Word and Excel, and statistical software R (via RStudio), which are used to analyze a business problem using a realistic data set. Unlike previous exams, there is no unique best solution (as is true of modeling in general). Candidate success is based on making appropriate choices and providing a concise justification for these choices.
This enhanced education not only ensures that our new ASAs have skills that are directly relevant to actuarial work, but it also provides an opportunity to work on less structured problems that more closely reflect actuarial work in the business environment.
As more candidates acquire PA skills, there is an opportunity for the SOA to add specific applications within a candidate’s track. A good example is the Financial Modeling Module for candidates on the Quantitative Finance and Investment track. In this module, among other things, candidates use R to fit time series models to financial data.
An early foray into bringing PA education to existing SOA members was the creation of the PA certificate program. Each cohort of candidates spends about five months working through an e-Learning program and then appears (virtually, these days) at a closing seminar followed by an assessment where their newly acquired skills are applied to a business problem. Successful candidates earn the certificate in PA.
The success of the certificate program led to a partnership between the SOA and Lincoln Financial Group, where the SOA is facilitating a corporatewide upskilling training program to bring big data and PA competencies to all Lincoln Financial actuaries, as well as select nonactuarial staff.
At the time of writing, the SOA also is working on developing a certificate in the ethical and responsible use of big data and predictive models in insurance applications. Advances in techniques and the availability of new and larger data sources bring the challenge of ensuring that our work conforms to the ethical standards expected of actuaries and the demands of regulators.
In September, we offered Predictive Analytics 4.0, the fourth offering of a symposium that is solely devoted to PA across all practice areas and all levels of engagement, from relative beginners to those managing a team of analysts.
Figure 1 shows the frequency and number of participants for each of the educational components discussed in this article.
Figure 1: PA Education by the Numbers
|Statistics for Risk Modeling Exam||6||2,284|
|Predictive Analytics Exam||4||5,756|
|Predictive Analytics Certificate||9||384|
|Lincoln Financial Upskilling||1||346|
|Predictive Analytics Symposium||4||806|
Actuaries are well-known for their technical skills and abilities to solve complex problems. In the evolving world of big data, it is essential for actuaries to have the expertise to efficiently leverage data in creating solutions that drive positive business results. The SOA is pleased to have been a part of these developments that have significantly enhanced the capabilities of current and future members.
Copyright © 2020 by the Society of Actuaries, Schaumburg, Illinois.