Investment and trading rules abound. In a realm as complex as investing and trading, not having a set of rules by which to operate can be paralyzing. So a necessary part of thriving is to abide by a set of rules that relieve you of having to decide the same thing several times. “Disciplined” and “systematic” are shorthand to mean an investor abides by certain rules automatically.
Models are different from rules. CAPM is a model. Setting aside whether you believe in it, it doesn’t tell you how much to hold in the market portfolio, never mind when to buy, when to sell, and how much to diversify.
Other models imply whether assets are relatively expensive or cheap, even if they don’t give explicit rules on buying and selling. The “Fed Model” relates the earnings yield on stocks to yields on 10-year Treasuries. It’s a popular one, despite Cliff Asness’ efforts to dispel it. Asness points out that stocks’ earnings can adjust with inflation, while the Treasuries’ yield is nominal, so investors relying on the Fed Model are relying on a misunderstanding. The forecasting power of the Fed Model, Asness demonstrates, is poor.
If a model’s forecasting power is poor, then what to make of trading rules based on it that would have been profitable in the past?
Recently I investigated some new software a client had purchased. It related asset prices to certain exogenous factors, and included tools to allow you to make certain investing decisions based on those relationships. The asset/factor relationship was the model, and the tools allowed you to come up with your own rules. Certain canned rules were pre-programmed to guide you.
This investigation was frustrating. I sought evidence of the model’s forecasting ability; instead I received back-tested results from using the tools. Basing investing rules on a model with unknown forecasting ability is not a discipline; it’s a superstition. I continue to hold out hope that the model contains something useful, but to convince me will require the vendor to stop dwelling on the results of the rules and start dwelling on the forecasting ability of the model itself.