By Andrew Peterson
I recently had the opportunity to attend the SOA’s Investment Symposium event held in New York in mid-March (co-sponsored by PRMIA). For the last two years, this conference has included a specific track of retirement focused sessions, so I was quite interested to see what was being presented and how many people attended. Overall the event was very well done and the retirement track featured a high-caliber of speakers including Bob Merton (a Nobel Prize winning economist at MIT), Olivia Mitchell (executive director of the Pension Research Council) and others.
While there were many interesting ideas, I found the luncheon speech by Emanuel Derman, a professor at Columbia and author of “Models Behaving Badly” (2011) and “My Life as a Quant: Reflections on Physics and Finance” (2004), to be quite interesting. His presentation was based on his most recent book and focused on his views about our modes of understanding in a world where we seem to be enamored by the concept of “big data.”
The concept of analyzing “big data” is being thrown around a lot these days—whether in the context of focused marketing, in setting more precise insurance rates or even in targeting voters for the recent U.S. presidential elections. What I found interesting was his skepticism towards the current view that with computer-aided analysis of patterns in big data, the traditional methods of discovering truth will be replaced in the areas of medicine and social sciences. However, Professor Derman suggested that we need to remember and hold on to the key modes of understanding that have been reflected through-out the centuries. He listed these as:
- Intuition: Here he described the work of scientists like Kepler, Newton and Einstein who used their intuition to develop theories that were based on careful observation and painstaking effort. In this context he described intuition as a merging of the observer with the observed.
- Theories: These are “deep descriptions of the laws of the world” and according to Derman can be right, partially right or totally wrong. Theories are not analogies—they just “are.”
- Models: A model compares something that we don’t understand to something we do. Unlike theories, models are analogies. Black-Scholes option pricing is a model, which is useful to a point, but it is not fact.
- Data & Statistics: This involves the analysis behind big data—using statistics to find past tendencies and correlations in data. But correlation does not imply causation and often people falsely assume that past trends will persist.
The key point made by Professor Derman, as I understood it, is that while using statistical analysis is important and helpful, the three other modes: intuition, theories and models, are necessary to evaluate the results provided through any big data analysis to evaluate cause and interpret results. As I think about this and the application for actuaries working with retirement plans, I believe there are several application points. We receive lots of data points as actuaries, yet we need to use our skills of intuition to evaluate what is important in interpreting results. In addition, it is important not to simply accept results we get from a model (e.g., valuation program) without testing and evaluating whether it fits with our intuition and other relevant data points. Two specific examples for pension actuaries come to mind:
- As we set future economic assumptions, we have significant amounts of historical data that can be analyzed for averages, statistical correlations, variance, etc. Yet simply using the past as a guide for the future may not be the best as we need to add our own intuition about how the future may differ from the past—due to changing economic situations, demographic trends, shifting global economy, etc.
- In developing mortality assumptions, we have typically relied on actual past historical experience to predict future mortality rates, perhaps with an additional improvement factor included. Today we have more sophisticated models that can blend past actual experience with future expectations based on our input and best estimates. We need to combine what those models can do with new sources of information telling us how mortality rates are or aren’t improving depending on the particular demographic or socioeconomic group involved.
This all sounds complex doesn’t it? Yet, this is ultimately a risk management exercise that allows us as actuaries to blend our quantitative analysis with our critical thinking skills…and hopefully that is the key reason we became actuaries.
If you have thoughts or comments on this column, feel free to contact me at the SOA.
Andrew Peterson, FSA, EA, MAAA is staff fellow, retirement systems at the Society of Actuaries headquarters in Schaumburg, Ill. He can be reached at firstname.lastname@example.org.