Is it really different this time?

Yes… and no.

This was partly a test to see if the article would preview… they have it set up to preview on LinkedIn, but it didn’t work in Twitter and it doesn’t look like it worked here.

Have a table:

Nice article.

Back in my ALM-days we had TAS models. The mortality followed whatever table was loaded and then we’d adjust it (the whole thing) by a factor to get it in line (approximately & crudely) with recent experience.

Further, we’d do sensitivity analysis, increasing or shocking the mortality or lapses by various percents.

Investment scenarios would just come from our investment team with no input by the actuaries except for interest rate changes/shocks.

It seemed that the only tool in my tool box was a hammer, so the assumptions were hammered down as necessary but certainly not finessed.

When reading your article, it seems that some modelers are have a more-thorough, multi-variate method of assessing and modeling those aspects. Is this a valid observation? Are there some easy-to-describe ways of how this is done?

In what sort of ways would one deal with/account for “most models assume that mortality and economic factors are independent.”? Is that part of the secret sauce of your model & modeling software?
Indeed, “the complicated dynamics of correlations can be difficult to capture.”

“…there may be surprising results” reminded me of the spurious correlations website. How sure are we that there’s a causation here?

“[ALM actuaries] should be communicating with a broad range of experts in their organizations to capture multiple views on risk…”
Besides interest rate/investment risk, our concerns were mortality and lapses. The actuaries were certainly the experts on mortality (we built the tables!) and lapses were easy enough to observe. In what ways would this broad range of experts advise us? What sorts of things could they contribute?

Thanks for your insight.

For operational risks, emerging risks, strategic risks, etc. – softer, hard-to-quantify risks – sometimes going to the frontline experts is all you can do.

How much do you know about the legal risks facing your company? The industry? What about reputational risks to your company? Strategic concerns? How reasonable is next year’s sales plan, and where is it likely to fail? Do you know all the dirty secrets in the asset & liability portfolios that might be getting downplayed in the hopes they blow over?

Plus people just focus on different things. If you asked about pandemic mortality risk before 2020 you’d probably get several different perspectives, depending on who you asked:

  • What it means for the company’s product portfolio
  • What it means in terms of business continuity
  • What it means for employee benefit costs

Now, of course, you’d get even more. How many groups considered the impact a pandemic might have on the sales process? If they didn’t, maybe more sales experts should have been involved in the risk discussions.

Good answer. Thank you for all of that.

Now, how do you go about quantifying some of those things? How would you incorporate them into your projection model? …especially something like legal & reputational risks. Sure, we can acknowledge them, but how do you place a number on something like that? And if you don’t place a number on it, what do you do with that information?

fwiw, the modeling I was involved in was only on inforce business, so sales projections didn’t come into play.

You’re never going to get great numbers for that sort of stuff, but you usually still need to put something together so that if nothing else you know what to prioritize for monitoring. Asking for severity estimates at a few “practical” frequencies (i.e. what’s their view of a 1-in-5 year loss, a 1-in-10 year loss, etc.) can let you put together enough of a distribution to decide whether certain types of events represent tail risk or just a normal cost of doing business. Of course, since people focus on past events – which hopefully already resulted in mitigations being put in place – they may end up understating the risks. So it might take a little prodding to get them to thinking about what the next “new” event could look like.

Some risks (like reputational risk) don’t immediately translate into dollar amounts, so you may need to first get estimates of the knock-on effects (impacts on broker relationships, policyholder retention, employee morale, etc.) and then talk to others about the quantitative effects those might have.

It’s all a lot squishier than things like (normal non-pandemic) mortality and interest rates, where there’s a lot of hard data on hand. So they tend to get less of a focus. That “squishiness” extends to pandemic risk – the data for pandemic mortality is extremely thin, and generally based on very different circumstances. If the Spanish flu happened today, with modern medicine and no world war, how bad would it be? Heck, where is the Spanish flu on the theoretical “pandemic distribution”? And what if it’s not the Spanish flu? (What if it’s AIDS 2, or a less aggressive Ebola?) What’s the mode & rate of transmission? What’s the age, gender or economic breakdown of mortality look like, and how does that change things? There’s no right answer to any of that – all you can really do is think broadly, put together some believable narratives using the best information available, turn that into some quantifiable consequences and use those to get folks thinking about contingency plans, exposure monitoring, mitigation, etc.

(Sometimes in the end the numbers are just there to get folks’ attention and demonstrate you’ve done the due diligence.)