SOA Launches Predictive Analytics Initiative

By Ian Duncan, December 2015

Although Predictive Analytics has not been a mainstream focus of actuaries, there are many of us who have been practicing in this area for a number of years, and a number of universities teach relevant courses in regression, GLM, time-series and other relevant subjects. Recent actions by the SOA leadership and Board promise to move Predictive Analytics (PA) front and center for actuaries of the future.

• Identify education/training needs for PA actuaries.An SOA survey has raised concern that actuaries are inadequately trained in the type of statistics and models that are required for PA. There are many roles in a PA project, however, from data management, through modeling to implementation (where business knowledge and skills are important), so the interested actuary has opportunities that do not necessarily involve advanced statistics and modeling. A workgroup is, however, currently considering (see below) what knowledge and techniques will be required for actuaries in the future.

  The SOA has offered PA continuing education for some time, including the Advanced Business Analytics Seminar. A second seminar specifically aimed at Health Actuaries will be offered beginning in 2016, using the ACTEX textbook, "Healthcare Risk Adjustment and Predictive Modeling." A second edition of this text is planned for 2016.

• A marketing communications campaign to promote actuaries in these roles, target potential employers and inform members of these opportunities is planned.
• Providing educational opportunities for members in predictive modeling:
  • University courses/preliminary exams/Fellowship track.Many universities teach students the fundamentals of modeling (Time Series, Regression and Generalized Linear Modeling, for example). It is often difficult to fit practical applications of this material into the undergraduate syllabus, given the SOA's and universities' course requirements. Once they graduate and enter actuarial student programs, students frequently do not have opportunities to apply their knowledge of PA, either. Thus the strategy has to be two-pronged: encourage more hands-on modeling at the university level, and encourage more rotations and jobs at actuarial firms that apply these models.

    One of the key recommendations of the committee was that Predictive Analytics should be added to the ASA syllabus. A syllabus re-design committee is currently considering this issue, which is not without challenges: PA is not a subject that can be adequately tested in a multiple-choice environment. Simultaneously with the SOA's decision to enhance the syllabus, other actuarial bodies have similar initiatives (e.g. the CAS with its new exam S and the International Actuarial Association's recommendation to include data and predictive modeling on the syllabus for the "qualified actuary" (essentially the SOA's ASA).

    Fellowship exam committees will be encouraged to add practical applications of PA to exam tracks. Actuaries at the fellowship level will not be required to perform the type of modeling that will be expected from Associates, but they will be expected to know how the models are applied in practice.

  • The SOA has significantly increased its research budget in recent years and this will be directed at projects in PA, particularly in Life.
  • Publicizing actuarial capabilities: the SOA has considerable collateral and research in this area. It isn't well-known, especially by non-actuaries. One of the campaign objectives is to raise awareness of work done by actuaries in predictive analytics.

As you can see, there are lots of developments coming. I am sure ACTEX will continue to be a leader in providing materials for actuaries in this area.

Ian Duncan, FSA, MAAA
Professor of Actuarial Science, University of California, Santa Barbara

Adapted from an article in the SOA Predictive Analytics and Forecasting Section newsletter of the SOA and published here with permission from the SOA.