SCORviews hero

Changing Actuarial Role: Applying the Theoretical
October  2019

Has your career followed the path you had in mind as an actuarial student?
Yes and no. You don’t have clear expectations when you start out – I certainly didn’t. I was just happy to get my first job and use the quantitative skill set that I had built in college in an applied environment. I didn’t like applied math in college, but in the real world it’s a lot of fun. You can use it to solve actual real world problems instead of doing proofs all day — which I thoroughly enjoyed but you can only do one thing if you stay in that world and that’s more proofs! 

When I got my first job at Milliman after passing my first exam, I could not have imagined how perfect a fit actuarial science was with my personality. I wouldn’t have predicted my move into modeling, but that’s the path that my career took.

For me — and for a lot of actuaries — being at the intersection of predictive modeling and actuarial science is the most exciting place to be. I think the Society of Actuaries knows that. Today it’s considered sort of a hybrid, nontraditional role for an actuary, but I don’t think it will be for long. Predictive modeling and the data science skills that go with it are going to be a core part of what actuaries do. It already is in a lot of ways.

What’s the appeal of life insurance to data scientists?
There’s been a natural flow from Silicon Valley to the banking industry and, now, to insurance. Data scientists have already revolutionized many ways of doing business in the financial arena, and life insurance is a fresh piece of that market to apply their skill set to solve business problems. There was initially a perception that the industry had a lot of low hanging fruit; that methods, skillsets and data sources used were outdated. While that’s true, progress has not been as fast as it could be because of the wave of outside industry experts attempting to reinvent the wheel. Ultimately actuaries need to remain the decision makers and continue to be held accountable for business outcomes.

How did you get started down this path?
In my earlier actuarial jobs, I tried to do as much programming as I could. It’s important if you want to make that transition to start programming at work — just start solving real problems. The first thing I built that was used by someone other than myself was a fairly simple collection of Python scripts that grabbed data from all over the company and cleaned it up in a structured way for use by underwriters throughout the renewal process. There was nothing predictive about it, but it was widely used and very helpful. It allowed underwriters to immediately see the performance of their block all the way down to the case level, which helped inform decisions.

Once I started programming, I knew just as clearly as when I found actuarial science that this was what I wanted to do. So I needed to find a way to bring the two pieces together. And my timing has been really lucky because that’s where a significant piece of the industry is headed.

What’s driving the increased role of predictive modeling within the actuarial profession?
Well, there’s a need for it across the life insurance value chain. More and more data sources are available that require a totally different skill set to incorporate into what we do. The availability of affordable and scalable computing has eliminated barriers which existed in the past. Algorithms that used to be interesting on paper in a computer science research setting are now widely used and easily scalable.

With natural language processing we can extract meaning from text that already exists in our databases. At SCOR we have already shown this information to an important predictor for claims. With object character recognition (OCR) technologies we can begin incorporating the vast pool of handwritten and other sources of unstructured text into our models — think APS documents. Actuaries are well positioned to bring data science and predictive modeling into insurance — if we can keep up with these technologies.

Competitive pressure is another driver. There’s an influx of core data scientists from outside the industry who have these skills — machine learning, AI, knowledge of software and programming, cloud compute infrastructure — and they’ve had successes. They’ve worked to streamline underwriting and claims processes and learned how to incorporate natural language processing into many different functional areas, for example.

Actuaries have the business knowledge, and the data scientists coming into our industry will absorb the business knowledge over time. They’re not going to simply be given a problem or solution and just code it up. They very much want to think things through from the ground up. They will acquire the business knowledge, no doubt. What we need to do is learn how to work together and cross train each other’s skill sets, because each group of professionals bring something to the table.

Do you see any resistance among actuaries to data science and predictive modeling?
Most actuaries are very receptive. The most common reaction is “We already do that.” And that’s correct. I think the point is that other professions have different methods that are worth taking a look at. Data scientists bring a lot of outside skills, a fresh way of looking at things that can be a big value-add. But we can’t ignore the years of research from the SOA, the years of actuarial wisdom, the underwriting wisdom. Time and time again there’s the data scientist who comes in just to give you a result that is obvious to people who’ve been in the business. The key is to channel that fresh energy and those fresh eyes onto the right problems — and often an actuary is the best person to do that.

The attraction — or resistance — often depends on the individual. When I was taking exams, we heard a lot about predictive modeling, but it wasn’t necessarily built into the curriculum. Today there is a predictive modeling exam that includes real programming and a certificate program for credentialed actuaries. Students are being taught open source languages like R during the credentialization process. In the near future, I believe the typical actuary will have the skill sets to drive these data science initiatives.

How difficult is it to go from a more traditional actuarial track to a predictive modeling/data science role?
James: If you’re a pricing or valuation actuary now and you’ve been in your career for a while, you might be interested but you might not have opportunities at work to develop the necessary skills. Whereas previously you could further your education through the business side — for example, you could develop knowledge of regulations on the job. Now, in order to stay technically relevant, you need to start picking up core programming languages or stay abreast of the new machine learning algorithms so that, at the very least, you have the vocabulary and can insert yourself when you need to. And you do need to be inserted.

But in other ways it’s easy because actuaries are self-teaching experts. We’re the type who can study for an exam, and after 300 hours go by, you sit and hopefully you pass. That frame of mind, that skill is what’s most transferrable, in my opinion. Because the data science world is less structured than the actuarial world, you’ve got to find your pathway yourself, so you need to be a bit more resilient. This is something we need to rectify — and the Society of Actuaries should play a role here.

Can a more traditional actuary play in areas of innovation and data science without developing strong computer science skills?
James: I believe so. A fairly senior level actuary may not take an entry level data science position, but you can ease your way into these emerging roles. One way is to be involved in leading projects where you provide the business insights. These are critical roles because you help structure the project so that the data science team is solving the right problem. Whether it involves underwriting, new distribution channels, claims triage or something else, I think actuaries are best positioned to lead data science initiatives, because they can think holistically about the insurance value chain and bring all the necessary functions together. The data science space is ripe with opportunities for actuaries at any level to get involved.