“You get what you pay for” is both an idiom and the most common argument I hear against using open-source tools. I’d argue that’s an unfortunate oversimplification. In this article, and with the help of two actuaries with experience in the open-source realm, I’d like to highlight why I think open-source tools will continue to be an important piece of the actuary of the future’s toolkit. Along the way we’ll take a  look at three key questions regarding the open-source environment:  

1. What is open source?
2. Why should both prospective and credentialed actuaries care?
3. How can skills with open-source tools be developed?

What is open source?

Oxford defines “open source” as an adjective that “denotes software for which the original source code is made freely available and may be redistributed and modified.” To help localize that definition to actuarial work, I reached out to Brian Fannin, ACAS, CAS Staff Research Actuary, who helped lead the internal charge for the CAS to create and maintain an organizational GitHub account (which is an exceptional resource for those looking for sample projects and data). In expanding on the definition, Fannin opined that “beyond the major programming languages [R and Python] the tools that actuaries will most often touch are likely Git and open-source databases,” while the full open-source world also includes “PostgreSQL, the Linux operating system, and open versions of the Microsoft office suite.” The open-source world is a wide one, reaching out to even more exotic options like Hugging Chat, the ChatGPT competitor maintained by open-source AI platform Hugging Face.

Why should both prospective and credentialed actuaries care?

The function of any tool is to improve the efficiency of a given task or process. Open-source tools are no different, and when they prove they can improve the efficiency of actuarial tasks or processes they should be readily embraced. Open-source tools are particularly suited to tasks that are often repeated, require transparency, would benefit from research or support from an outside community, and are sensitive to cost.

To explore those ideas further I met with Stephen Mildenhall, FCAS, a frequent CAS presenter and former member of the CAS Board of Directors who now spends a large part of his time developing his own open-source package Aggregate. After talking with Mildenhall about the biggest selling point of open-source tools, (“The biggest benefit, of course, is that they’re free.”) Mildenhall highlighted that open-source models foster a collaborative community and that collaborative community extends far outside the actuarial space. He was quick to point out that while the CAS recently (and deservingly!) celebrated the major milestone of 10,000 active members, that’s a small amount when compared to number of contributors to Python’s package Index (PyPI). Since 2022, the number of Python projects available on PyPI has grown by over 125,000, all created under open-source licenses. Who knows what great advancements are hidden in that vast bank of knowledge?

Transitioning from the hypothetical to the actual, a clear example of the benefits of open-source tools can be found within the CAS itself. The CAS GitHub page is connected to Chainladder-Python, a Python R package designed for estimating and modeling outstanding reserves. There’s an R version as well that predates Chainladder-Python. Reserving is a task that’s often repeated, best performed with transparency, honed by research and a vibrant outside community, and can be sensitive to cost. Sound familiar? I’d heartily recommend exploring chainladder distribution as a brief introduction to how an open-source model can immediately impact your actuarial processes. While reserving appears as an easy application, it’s worth mentioning again that the open-source world is an exotic one. There’s a bevy of other articles that can tout the possible benefits of large language models (LLM) like ChatGPT, a tool that is not open source and is already facing scrutiny for a lack of transparency. Open-source LLMs could not only better maintain transparency but also be trained on bespoke data to create models.

To answer why a prospective actuary should devote some time to open-source tools, Fannin summed it up best: “It doesn’t particularly matter which one you learn [Python or R]; any actuary [or prospective hire] with skills in either is more valuable than one with neither.”

How can skills with open-source tools be developed?

Both Mildenhall and Fannin are uniquely positioned to opine on this question. I first met both while attending a series of virtual Python workshops they were co-teaching for the CAS. Fannin still teaches the R workshop regularly and has been impressed by the continued demand for the course over the years. “We’re easily getting 30 attendees for each workshop, and relative to the number of candidates and actuaries, that’s not an insignificant amount,” he said.

While the CAS does provide limited training, there’s also an abundance of free or self-guided training materials out for virtually all open-source tools, particularly R and Python. A quick Google search can reveal a wide world of free training and walkthroughs, but ultimately familiarity can only come with practice. Mildenhall’s advice to all learners is a great distillation of the process: “Learning to program is not a spectator sport. It’s a hands-on activity. Have a problem you’d like to solve and get to it!”

I heartily agree that there’s no substitute for getting your own keyboard clacking. As a student, that familiarity can often be built into introductory programming classes. Take them if you can! Of all the classes I took during my college years, introductory and applied R programming courses were arguably the credit hours most applicable to my day-to-day work. As a professional, it can be a little tougher, but a formal development plan with clear goals and a supportive manager can go a long way, as will mentors and coding study buddies. I’ve found that those experiences not only expand an actuary’s skills but also often open doors for employment and collaborative research opportunities.

Open-source tools turn “you get what you pay for” on its head, promoting a freer exchange of ideas and driving forward progress as community. Next time you begin a project, look for a training goal or start some research, I’d highly recommend a thorough investigation of any open-source options. At the end of your project, your expense ratio might even thank you.