Vol 2, Issue 3. Quarter 4 – 2025

I just completed my read of a recent book from a Senior Economist at a famous financial services firm that develops a somewhat detailed model that was used to guide portfolio construction for 10 year and longer planning horizons. I endured some mental anguish to finish the book out of respect for the author, but I was never able to shake my instinctive response to a sentence that appears rather early in the text which reads:

“we can control market risk by assigning probabilities to different outcomes.”

If I am being completely honest here (and that’s about the only real value that I can claim to add on this site) I have to admit that I have absolutely no idea what the hell that phrase is supposed to mean. Perhaps someone smarter than me can explain how I control something by making up a number. Let me elaborate.

As a practical (not mathematical) matter, Risk here refers to the combination of the probability of a bad outcome and the magnitude of that outcome. Almost no one labels a surprisingly good outcome as a component of risk (although it clearly is the other side of the coin of variability.) So, in this context “managing risk” refers to dealing with the inevitable bad, days, weeks, months, quarters, years, decades, etc. where market returns are less than average.

This particular book is actually built around a paper published on SSRN (which means it was NOT peer reviewed yet. See Davis, et.al. 2024) that explains a model for forecasting asset prices. The approach developed there centers on scenarios which are stories about the future. It then creates a story about what those scenarios imply that market returns will be. They then allocate assets in a way that should do well given those scenarios.  If you have become confused, let me state this in a different way. We create a story about what will happen. We then create a story about how that story will lead to prices. We then create a story about how a portfolio will react to that story based on the first story. Perfectly clear, right?

Of course, a good response may be something like, “its better to have a plan based on a good-faith effort to think things through, as opposed to doing nothing.” Maybe that’s right. Maybe it’s not. I cannot say at this point because planning to do a stupid thing is almost never preferable to the option of doing nothing. But what I can say is simply this. Assigning probabilities is NOT controlling risk. It may very well be acknowledging some dimension of risk, but claiming that you “control” it by sticking a number on your spreadsheet seems like quite a stretch.

This particular model was built using proxies for 15 different factors. How well the proxies match the actual factors remains an open question. The model then uses these numbers to estimate the probabilities of various scenarios. The author then, makes decisions as though these scenarios are the only states of the world that matter.

Let me get to my main point. In the last post we wrote about how hard it was to come up with a good estimate of a single factor – the Equity Risk Premium for the upcoming year. We showed that the estimates are all over the place, and we see no real reason to have a great degree of confidence in any of them. My approach in that discussion is deeply flawed because it is an intentional gross over-simplification of the real world. However, it was made simply to illustrate a point. It was not intended to be the basis for your asset allocation.

By contrast, the approach taken in this book was; instead of estimating a single factor, let’s estimate 15. Let’s then put them into some model and generate some arbitrary number of scenarios that we will plan for. In other words, since we cannot estimate a single factor for a simple model with any degree of credibility, let’s build a model with 15 such factors, that we cannot estimate any better. Surely, that will be much more impressive. Unfortunately, I see no logical reason to believe that this approach will be any better than the approach that I made fun of in the last post, but at least we can show a great deal of effort. The logic seems to follow from an odd type of internal dialog.

• Can you estimate X?
• Not really.
• Can you estimate Y?
• Well, not so much.
• Well what can you estimate?
• I can estimate X^3 – 14Y + log(X/Y) – Y/X factorial?
• Huh?
• Don’t worry – it all works out. Trust me.
• Oh, well in that case, here, take my money.

Alway Beware of Proofiness!!!

“Proofiness” was the title of a famous book by Charles Seife (Proofiness at Amazon). The central theme of the book (as I interpret it) is that we all have a strong temptation to use “mathematical looking things” to make our ideas feel more solid or “Proofy”. As in, “I can prove that I am right – look at this equation!!” This is a delicate matter because mathematical rigor is definitely our friend. But at the same time, many of us are easily impressed by mathematical jargon or statistical values that we don’t really understand, and every charlatan has a great incentive to take advantage of this fact.

Let me be clear. I am not accusing the author of the book that I read of this approach. I am simply saying that taking a bad idea, building a mathematical model around it using data that I find unconvincing, and then telling me that I am managing risk because I use some probability estimate from your “Proofy” model as part of the approach feels, really fishy to me.

So here is where I leave you. If you don’t understand a model – don’t believe it. That doesn’t mean that you argue against it. It simply means that I am not going to give you control over a penny of my money simply because you have a model that you cannot fully explain. Until you show me a model that I can fully buy into, I have a simple approach that I will stick to, stated in 5 parts:

1. The time to invest is NOW. The amount to invest is the money that I do not need to live off of today.
2. The correct share of equities in my portfolio is as much as I can stomach.
3. The correct planning horizon is the retirement of my grandchildren.
4. Diversification is the only free lunch that I know of, and I want it at the lowest possible cost.
5. Until I feel like I have a better model, that I can live with in any possible outcome, I am going to stick with my plan. 

Whatever you are selling that deviates from this plan is of no interest to me. Have a nice day!