I just finished reading a fascinating article from Nautilus, Issue 003, titled “Unhappy Truckers and Other Algorithmic Problems“. The article struck home with a bit of nostalgia for my undergraduate days in Computer Science, studying toy problems like “the traveling salesman.” Where the article leaped to the real world was in exploring the necessity of incorporating the human element into the math of transportation optimization. A particular quote struck me:
Allowing for human behavior in a quantitative way was key
In the context of the article, I interpret the quote in several ways.
The first way I interpret the quote is to understand that humans will behave in mathematically sub-optimal ways. Assuming rationality can lead to solutions, that while white-board correct, are real-world incorrect. As an example, the article discusses a mathematically optimal route that was avoided because it required driving over a bridge that had collapsed and been rebuilt several years before.
This reminds me of A. Gary Shilling’s famous quote, “the markets can remain irrational far longer than you and I can remain solvent.” As quantitative modelers, it is key that we incorporate irrationality and habit into our models. We should expect it. Markets, at the end of the day, are auction systems driven by humans. While there are certain rules of arbitrage that tend to keep prices within rational expectations, we should never underestimate the ability for markets to get out of line. To me, this is where a balance of model intuition and product design is important: understanding the markets your models will fail in and how you can account for it with product design is paramount to long term consistency in performance and risk management.
Secondly, there is a modeling distinction between feasible and implementable. To quote the article, “[t]ractors can be stored anywhere, humans like to go home at night.” Incorporating the happiness of the drivers into the modeling may lead to mathematically sub-optimal results, but implementable superior ones. As quantitative product developers, we can create investment solutions with superior total return expectations — but may lead to investor dissatisfaction due to other idiosyncratic behaviors along the way. This is why we believe in outcome-oriented investment products. By designing a product to meet a specific need, and clearly and transparently communicating the purpose of the product, we believe we can align investor expectation with product performance and behavior, leading to greater satisfaction.