It’s that time again! This is the time when we talk about what’s happening in the wild, woolly world of CAS research. As usual, there are a lot of different things going on, but if there’s one theme that emerges, it’s building for the future. A lot of our time and thinking on research has been on new ways of sourcing, conducting and disseminating it. To be clear, that’s not all we’ve been doing. All of that forward thinking has taken place against the backdrop of our standing research committees and publishing areas continuing the work that they’ve been doing for years.
New Research Council
Probably the most tangible manifestation of this development effort comes in the form of a new operational body. By now, you have heard about the structural changes that the CAS has embarked on in our effort to remain efficient and resilient. Committees are now working groups; working groups may create task forces for specific projects, and so on. For research, this means a significant change. Over the course of many years, the CAS Executive Council has served as the authoritative body which monitors and approves research activity. This mechanism has worked quite well, but we felt that there were ways it could be improved. Beginning soon, the role of the Executive Council in research will be phased out and replaced with a Research Council.
So, what does this mean? One specific effect is a shift in the direction of research ideas. Historically, the CAS Executive Council had a passive role that involved approving proposals that originated from any of our standing research committees, CAS leadership or from the membership at large. This ensured that there was a control mechanism in place to ensure that our research budget was being used prudently. The Research Council will still have authority over how funds are spent, but it will go beyond that. In the new configuration, we expect that they will establish research priorities for each fiscal year. To be clear, the Research Council will set a research agenda that is informed by input from our standing research working groups, but they will broaden our reach and incorporate the needs and wants of stakeholders outside the profession. This is something that we have always tried to address, but we are now acting in a very deliberate way.
There are three principal benefits to this revision. First, there will be greater clarity around research priorities for the CAS as an organization. The Research Council will serve as the single conduit through which ideas are vetted and overall strategy for research takes shape. Second, by bringing additional perspectives onto the Research Council, we will ensure that our research work will continue to reinforce the relevance and market value of the work done by CAS members. Finally, by placing the Research Council in a proactive, decisive role, we hope that we can shorten the turnaround time between idea and publication.
A few of our current projects
Looking at some of the work underway in 2021, it feels as though a shift in direction has already begun. One example is the new way in which we’re producing content. Research is typically conducted in one of two ways: The CAS will contract with a third party to carry out research that we deem relevant, or alternately, volunteers will submit completed work to appear in either the E-Forum or Variance.
One of our current projects uses a different approach. The CAS research actuary and several volunteers have contracted with a data provider to examine social inflation in liability claims. This is unique for several reasons. First, the idea and plan grew from a series of conversations among CAS leadership and staff. Second, rather than outsourcing the work to an external researcher, we looked for volunteers to carry it out. This meant adding volunteers to the service agreement between the CAS and the data vendor, which is a first for the vendor. Until now, the vendor had only engaged in two-party contracts. Third, and finally, this represents something of an experiment in how to utilize volunteer capacity. Instead of passively accepting work from volunteers (which we love to support and will continue to do!) the CAS, as an organization, developed a specific research project and then went looking for volunteers to carry it out.
We have also been proactive in identifying research opportunities related to race and insurance pricing. The events of the recent past have amplified the importance of this issue in the United States and the rest of the world, and the CAS has responded. We are aware that this is a sensitive and complicated topic with a long history. Moreover, it demands a fusion of domain knowledge, data-centric analysis and an awareness of the interplay between public institutions and private enterprise — sounds like a perfect research topic for a CAS actuary.
We have also been proactive in identifying research opportunities related to race and insurance pricing. The events of the recent past have amplified the importance of this issue in the United States and the rest of the world, and the CAS has responded.
At the moment, we have no fewer than four workstreams generating content that can help our members and the broader public understand the issues surrounding race and insurance pricing. This includes an explanatory paper that walks through various definitions of disparate treatment and disparate impact, work that examines the history of regulation of rating factors and an overview of how companies may develop methods to gauge the level of disparate impact in their models.
Every CAS member and candidate should have seen a request to participate in a survey on how they use technology in their work. The questions we addressed were meant to be clear and pragmatic. For example, where do actuaries stand on adoption of scripting tools like R, Python or MATLAB? Are newer members more likely to use emerging platforms like Google Sheets, Tableau and the like? What skills do our members want to develop over the next 12 months? We had over one thousand responses and are beginning to comb through the results. You should see a high-level summary of the data shortly.
This has been christened the first annual technology survey, meaning that we are expecting to see this survey carried out every year. Though we try, actuaries can’t predict the future. However, there’s one thing we’re certain of: The tools we use will continue to change. The feedback we’ve gotten for this inaugural survey will inform and improve the survey when it returns next summer.
Curious to know more about what projects are hosted by the CAS on GitHub? You can always visit the page itself at https://github.com/casact. We’ve also begun publishing a basic set of metrics about activity. As I write this, we have over 1,400 commits across 18 repositories.
By the time you read this, the number of repositories should number at least 19 as we add another package, which may be used to analyze loss reserves. Caesar Balona and Ron Richman have taken the techniques described in their paper “The Actuary and IBNR Techniques: A Machine Learning Approach” and implemented them in a Python package called “tryangle.” If you are already familiar with the Python “chainladder” package, this one is for you. If you’re not familiar with the Python “chainladder” package, what are you waiting for?
In May of this year, the paper “AGLM: A Hybrid Modeling Method of GLM and Data Science Techniques” was awarded the 2021 Hachemeister Prize. This paper demonstrates a novel technique that adapts the GLM framework to handle instances when the data set has a large number of features. The modeling method is highly flexible, yet retains explanatory power that can sometimes be lost when one uses non-linear techniques like random forests or gradient-boosted machines. The net effect is to unite the attractive elements of GLMs and more modern statistical learning methods.
The Hachemeister Prize Committee also noted the following as particularly strong papers:
- “Discrimination-Free Insurance Pricing” by Mathias Lindholm, Ronald Richman, Andreas Tsanakas and Mario V. Wüthrich.
- “Silent Cyber Assessment Framework” by the Institute and Faculty of Actuaries’ Cyber Risk Investigation Working Party (Simon Cartagena, Visesh Gosrani, Jasvir Grewal and Justyna Pikinska).
- “Cyber Claim Analysis Through Generalized Pareto Regression Trees with Applications to Insurance” by Sébastien Farkas, Olivier Lopez and Maud Thomas.
These projects are just a few of the noteworthy developments happening with CAS research. Space doesn’t permit us to mention everything that’s going on, so be sure to scroll through the latest E-Forum or have a look at our calendar of upcoming webinars. As always, if you have questions, comments or research ideas that you’d like us to explore, you can reach me at firstname.lastname@example.org.