Algorithm Aversion: On Models and The Donald

Whatever your politics, one can’t help but be fascinated by the Trump phenomenon. One interesting aspect is how pollsters got Trump wrong. We can get insight into that mistake from Nate Silver at fivethirtyeight. To his immense credit, Silver admits he got Trump all wrong, and went back to have a look at the why and the how…. it boils down to a case of algorithmic aversion (see here and here).

Silver’s approach to forecasting, which has been extremely successful across a wide range of subjects, is based on aggregating data from multiple polls to build as big a sample as possible. It is a mechanical, quantitative, model based, statistical method. The model spits out an answer, and that is forecast. But in the Trump case, he over-rode the model. He had a bias, and in this instance he applied it to his forecast, and voila, a bad outcome. Even Silver, a devoted algorithmist if there ever was one, fell into the trap. He did not trust the model, even though it had been right so many times before.

So what? This is an investing blog. We have been doing model based investing since 1987 (or longer, pre-Mount Lucas). What Silver was confronting was the difference between model trading and discretionary trading. In quantitative decision making, you are creating a vast distribution of outcomes, the results of many shots on goal. One really does not know the outcome of a particular forecast, only that it is the practitioner’s attempt to construct the least biased way to make that forecast. Discretionary decision making is all about bias, about taste, about conditions, about attitude. It’s a one off, artistic, poetic insight into a particular situation. It makes us feel human. It’s also really hard and hardly repeatable. There’s a parallel with quantitative trading of bonds here to illustrate the Silver situation in finance. A few years ago the idea that interest rates could go to zero, or long duration bonds could trade at negative yields seemed ludicrous, not a single textbook or expert countenanced the possibility. As yields went lower, we wondered about altering our own models to incorporate a yield floor, our own bias creeping in – in our hearts we thought it couldn’t happen. And yet here we are. Silver knew in his heart that Trump could not win, and he married his model to that position.

Suppose you want to do both, you want to run models and you want to use poetic license. How does one resolve the conflicts? Ever wonder why there is never a track record for TV talking heads? Because it is likely to be horrible, and if they kept a record, no one would listen – and therein lies the resolution. If you are going to approach decision making both ways, keep the results separate. At least you can see, in red and green, if your judgment has any value. Many years ago, we were discussing a capital allocation with a fund of funds. After many meetings, and lots of questions, they finally said that they could not invest with us as we did model based trading. (We always break out our results, model and discretionary). We were puzzled – they had investments in other funds that used models. When we pressed they said – “We know, but they don’t tell us about them.”  Bias confirmed. Information is king.