As data analytics is playing a more important role in insurance industry, we will interview actuaries and professionals in data science to share their perspectives on Data Scientists and Actuaries. A series of Q&A's will be posted, and we welcome any additional comments.
What questions do you have for actuaries in predictive analytics? To start the conversation, we talked to Louise Francis, consultant and founder of Francis Analytics and Actuarial Data Mining.
At the 2015 CAS Annual Meeting in Philadelphia, Francis participated in a panel discussion on “how data scientists and actuaries can work in harmony”.
Q: Tell us about your consultancy. What kinds of predictive modeling projects do you work on (such as pricing, reserving, claims, risk selection, etc.)?
A: I work by myself, but I reach out to other small consultancies if I need more resources. My company does a mixture of reserve opinions for captives, self-insured retention, and large deductible portfolios. As we evolve and embrace more technology, we find a need for “predictive modeling - lite” approaches and simulations for reserve variability and reinsurance estimates.
For larger accounts, I advise insurance companies on predictive modeling. The applications of my modeling solutions span multiple areas, including underwriting, fraud, claims, and marketing.
Q: When you are consulting with insurance companies, do your clients include actuaries or predictive analytics professionals?
A: My clients tend to be in predictive analytics, and I typically act as the actuary on the project.
Q: How long have you been working on predictive analytics?
A: Since the early 2000s.
Q: That’s a very long time, especially for a topic that seems to have become more popular over the past few years. Is “Big Data” just a new term for something that has really been around for a while?
A: There certainly is a recent trend in favor of embracing new techniques. However, “big data”, as defined by Google and the tech industry, does not apply to insurance.
In insurance, “big data” can refer to large volumes of transactional data, but this is not the level of data that we need. Instead, we often must aggregate the transaction-level details to the element or policy level. Therefore, despite growing interest in using massive datasets, we are often left with a much smaller amount of data when doing the actual analysis.
I still see a huge issue with the quality of the data, rather than the size. We need usable records that we can aggregate to the level we need.
Q: In an article recently published on the Actuarial Review (“Come Together: Experts Explore How Data Scientists and Actuaries Can Work in Harmony”), you predicted that some actuaries will break into data science, while some data scientists will switch to the actuarial profession. Would this crossover continue when the CAS Institute (iCAS) launches its new credential in predictive analytics and data science?
A: I think actuaries will continue to acquire skills, education, and experience in data science. For example, CAS members may learn through limited attendance seminars, along with the rest of the continuing education program. This is where CAS Institutes may come in, to help people pick up the knowledge they need. Outside the CAS, online courses like Coursera may provide additional information.
Data scientists will continue to work in insurance companies, and they are interested in working with actuaries. Hopefully this will continue.
Q: If data scientists wanted to gain more insurance knowledge, should they take the actuarial exams or pursue the iCAS credential?
A: I have no firsthand experience with data scientists wanting to start taking exams. It might happen, but the long travel time on exams can be a barrier.
The new iCAS credential is intended for both actuaries and non-actuaries. Data scientists may get a waiver for the analytics portion, but will need to do insurance component. Actuaries may get a waiver for the business knowledge, but will need to do the analytics component.
Q: What do actuaries need to do to adapt to rising interest in data scientists?
A: We still have useful business skills that others may not have. However, if the actuarial profession were to remain stagnant and not respond to the industry’s demand for data scientists, then actuaries will decrease in size relative to other insurance professions. Certain tasks, like ratemaking, may have already gone to data scientists. That could have been a result of our own stubbornness.
We are still evolving. We are revising our basic education -- especially with Exam S -- to include rigorous analytics. The iCAS credential hopefully will attract actuaries, and we will continue to learn through continuing education programs, limited attendance seminars, and webinars in machine learning.
Q: In what ways have data scientists enriched or improved the actuarial profession?
A: Most of the actuaries who do data science today have picked up books written by data scientists. We are doing this because we have paid attention to what is happening outside our professions.
Q: Do you think graduate education is necessary? Would you encourage an actuary to pursue a graduate degree?
A: I wouldn’t discourage graduate education. If you have a master’s degree, you wouldn’t need to take actuarial exams. Most people doing data science now picked it up on their own, and the graduate school programs in analytics are relatively new.
Actuaries are good at self-studying. If you can pass exams, you can study on your own. Can you read the literature and learn the software and tools to get the job done? I think that’s more important.