Next month, I’m hiking the W trek in Torres del Paine at the southern tip of Chile. My travel partner and I ascribe to the do-it-yourself (DIY) mindset, meaning we won’t join a tour group or hire a guide—we will go it on our own. I find that style of travel a lot more fun, but it’s not without its risks—namely, that we end up off-route and lost in the middle of nowhere Patagonia!
Thankfully, travel blogs claim the W trek is well-marked and that it is hard to get lost. Even so, I know I will regularly be scanning the trail for those ever-familiar and comforting hash marks on trees, rocks or posts, noting that we are, in fact, on the right path. The beauty of trail markers is that they are a guiding light of sorts, something to latch onto and follow no matter the circumstances. Hikers find these markers particularly helpful when the terrain is rocky, the landscape unfamiliar or a storm is raging overhead.
As credentialed actuaries, we all have one guiding light in common: our standards of professionalism. Our responsibility to the public, our clients and employers, and the actuarial profession does not exist in a void. Depending on our credentialing body and country of practice, we are subject to a set of professional standards. In the United States, those standards include The Code of Professional Conduct, Actuarial Standards of Practice (ASOPs) and Qualification Standards.
Professionalism standards matter now more than ever as we navigate uncharted territory in the way actuaries work. Our “what” hasn’t changed—we still measure and manage risk to drive better financial outcomes for individuals, organizations and the public. But the “how” certainly has changed, and in no small way. Data creation and availability has exploded. Historically time- and labor-intensive tasks are being automated. New products are entering the scene. Actuaries are partnering with data scientists more often. All of these changes in the “how” raise interesting questions about applying standards of professionalism.
What is my professional responsibility as it relates to data use?
An appropriate tagline for our profession could be: “Actuaries: the OG data scientists.” We have long dealt with large data sets to make predictions. Yet, today’s data reaches far beyond the typical insurance policy application information of age, gender, smoker status and occupation. Readily available data now includes extensive medical history, prescription profiles, motor vehicle reports, voter registration, property records, address history, marriage licenses, credit scores, bankruptcy records, social media profiles—the list goes on. Much of this data could be used for risk classification and improved pricing assumptions for insurance products. But just because the data is available doesn’t mean it’s acceptable for use. Where might we turn for professional guidance on this matter?
As a hopefully obvious ground zero, applicable laws and regulations, which vary based on context and jurisdiction, should be followed. For instance, the General Data Protection Regulation (GDPR) in the EU and the California Consumer Privacy Act in California are two examples of privacy laws put in place to enhance consumer protections. In the insurance context, other important regulations include the Patient Protection and Affordable Care Act of 2010 (ACA), Health Insurance Portability and Accountability Act of 1996 (HIPPA), Genetic Information Nondiscrimination Act of 2008, Fair Housing Act and each U.S. state’s own regulations.
That said, one of the challenges with the speed of technological innovation is that regulations usually do not keep up. As such, ethical frameworks will need to be considered and applied as guidance for the insurance industry. These ethical frameworks are covered extensively in the Society of Actuaries (SOA) certificate program Ethical and Responsible Use of Data and Predictive Models, an excellent ethical primer for those encountering these difficult questions in their actuarial work.
One case study from the course relates to usage-based car insurance, where a company collects granular data on actual driving behavior (e.g., time, location, speed) to offer policyholders potentially lower premiums through improved driving behavior or reduced usage. Despite the benefits for both the insurer and customer, there exists an inherent risk of unintended discrimination against certain policyholders in this model. For instance, if the model suggests driving in an urban area between 1 a.m. and 5 a.m. is risky behavior, this could penalize low-skilled employees who work evening shifts.
Another potential source of discrimination is in the use of proxy variables in modeling. Certain exogenous variables, such as ZIP code, might be correlated with race or other protected classes. Actuaries can avoid discrimination and ensure fairness in their models by assessing outcome equity, where an outcome may be defined as the average level of prediction or errors for each group.1
How do I evaluate output from sophisticated data science models?
The onset of all of this data means that actuaries are increasingly collaborating with data scientists, and in some cases, even taking on the role of data scientists. This partnership allows the introduction of new types of modeling to solve classic actuarial problems, such as experience studies. Data scientists perform the data aggregation and modeling using techniques such as decision trees, random forests or neural networks. Actuaries then analyze the output and own the assumption updates. This puts actuaries in an unfamiliar role where modeling is divorced from analysis and recommendation. How do actuaries appropriately evaluate output from a model they did not build?
ASOP 56 on Modeling provides a strong guiding light here, given that “an actuary using a model developed by others in which the actuary is responsible for the model output is subject to this standard.” In these cases, “the actuary should make a reasonable attempt to have a basic understanding of the model,” which includes purpose, operation, strengths and limitations.
A close partnership with data science colleagues is paramount to ensure alignment on the modeling effort. Yet it is not enough to simply rely on the data scientists’ technical expertise. Fortunately, there is extensive literature and training available on data science, artificial intelligence (AI) and predictive analytics. For starters, the SOA has added Predictive Analytics to the exam pathway for its candidates. For credentialed actuaries, Willis Towers Watson recently partnered with the SOA to build out the SOA’s Predictive Analytics Certificate Program. There are also many excellent courses available online—in some cases at no cost. edX offers Python Basics for Data Science, and Khan Academy has an approachable Welcome to SQL offering. The world is your oyster when it comes to learning this new content!
How do I train my actuarial students when routine work has been automated?
A 2019 McKinsey study performed at the request of the SOA found that 35 percent of actuarial work could be disrupted by 2030 due to automation and a shift of work to data scientists. In the absence of data-intensive calculation work, actuaries will spend more of their time on problem-solving and insight generation. This is an overall positive shift for actuaries—it allows us to differentiate ourselves from other professions with our business acumen and ability to drive solutions to life’s financial risks.
However, this change will require adaptation. Actuaries have long followed somewhat of an apprenticeship model. An actuarial student’s work might be tedious tasks, such as uploading new assumption tables, conducting policy-level model validations in Excel or executing a 22-step financial-close reserving process. These routine and even rote tasks have been considered a rite of passage of sorts, a way to learn the complex work by seeing it deep in the trenches. As those tasks are streamlined, eliminated or automated, how do we develop our actuarial students into experts on our products and calculations?
The first (and perhaps obvious) answer is we rely on the rigorous and comprehensive SOA exam system. Just as the nature of our work has evolved, so has the nature of the exam system. The introduction of computer-based testing for fellowship-level exams means that questions are no longer constrained by calculations that can be performed on the TI-30XS Multiview. Instead, questions can be answered using Excel, allowing the exam writers to draft questions that better simulate real-life work scenarios. The modules and corresponding assessments interspersed throughout the exam pathway are another opportunity for students to acquire practical experience with the nitty-gritty actuarial calculations. We can trust our exam process to educate our students.
Even so, there is a natural premium on real-world experience, so it’s worth considering which types of experiences will be equally, or dare I say more, meaningful for actuarial students in the absence of data manipulation and number crunching. As our collective role as actuaries shifts toward insight generation, we would do well to bring our actuarial students along for the ride. It’s one thing to load variable universal life in-force files into an Access database and observe the fields of account value, cost of insurance deductions and interest credited. It’s another matter altogether to identify the unusual increase in interest and account value, then tie it back to what happened last quarter in the equity market. If we teach our actuarial students to build this muscle early in their careers, they will be well-positioned to be the next generation of effective decision-makers.
How do I become qualified to perform actuarial services in new or emerging practice areas?
This is an age-old question, one that actuaries have been asking repeatedly as new practice areas emerge, due to the strict requirements in Precept 2 of The Code of Professional Conduct: “An Actuary shall perform Actuarial Services only when the Actuary is qualified to do so.” It’s such a popular question, in fact, that the U.S. Qualification Standards include a section on “Changes in Practice and Application.” Section 4.4 addresses actuaries practicing in emerging or nontraditional areas, noting that they can satisfy the continuing education (CE) requirements by maintaining knowledge of applicable standards of practice, actuarial concepts and techniques relevant to the topic of the Statement of Actuarial Opinion (SAO).”2
This may be more difficult in practice than in theory. If your company is one of the pioneers of, say, parametric insurance, it’s likely that there aren’t yet webcasts, articles and events to attend on the topic. But this is a great example of how professionalism serves as a guiding light—but perhaps not a perfect roadmap—for how to get there. In cases like these, we may need to flex our professionalism standards to new applications.
One source of CE may be reading the ASOPs and determining relevance to the new practice area. Another idea is to be the first to author an article on the topic—time researching for the article counts toward CE (although not as “structured” credit). Finally, consider what constitutes “job-relevant” education—this will look different for actuaries outside of traditional actuarial functions or outside of the insurance context. Just as actuaries use actuarial judgment in their day-to-day work, the application of professional standards requires judgment. When in doubt, consult other actuaries, or even the Actuarial Board for Counseling and Discipline (ABCD), for guidance.
WALK IN THE LIGHT
Whether you are swimming in your newly created data lake, partnering with data scientists, training actuarial students sans rote processes or practicing in a shiny new field, you are not alone. As actuaries, we share a professional responsibility to the public, our clients and employers.
But we do not do our work in the dark. The professionalism resources available to actuaries, and the principles underlying them, provide a guiding light for actuaries dealing with a constantly changing environment. So, join me, as we walk in the light.
Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries or the respective authors’ employers.
- 1. Ruiz, Amber, and Kim Steiner. ACLI Webinar Series: Future of Predictive Analytics and Innovation. American Council of Life Insurers, October 20, 2021. ↩
- 2. Beuerlein, Bob. Professionalism in Action—Shiny, New … and Professional: Professionalism and New and Emerging Practice Areas. Contingencies, July/August 2017. ↩
Copyright © 2021 by the Society of Actuaries, Schaumburg, Illinois.