'The end of Behavioral Economics,' declares a tweet from Nassim Nicholas in his typically dramatic fashion. He's retweeting a forbiddingly (for me!) technical article in the journal 'Nature Physics' with the title 'The ergodicity problem in economics.' The article is written by a statistician named Ole Peters who is a fellow at the London Mathematical Laboratory.
Even though, at a cursory glance, the paper may appear to be about some arcane matters of mathematics and/or economics, it actually has a direct connection to how you and I absorb investment news and analysis and how we react to it. For example, you will find a direct connection between the recent turmoil in fixed-income mutual funds and this discussion.
It so happens that almost all the investment-related information that we are subjected to is in the form of averages and other forms of aggregations which combine information over periods of time into a single measurement. This is a problem, a huge one. As it happens, just some months ago, I wrote a column that discussed what appeared to be a mathematical trick described in a tweet. The tweet asked if you were offered an investment that had a 50 percent chance of returning 0.6x (40% loss) and a 50 percent chance of returning 1.5x (50%) gain, should you take it? The answer would appear to be an obvious yes because (0.5 * 0.6) + (0.5 * 1.5) = 1.05. You can, on average, expect a 5 percent gain so why not.
However,it so happens that if you actually did it repeatedly, you would lose a lot of money. These two options average to 1.05 but if taken repeatedly would cause guaranteed losses. The trick is that the above equation executes both sides of the experiment simultaneously. However, if you first lose 40 percent, and then gain 50 percent, you are left with a 10 percent loss. Obviously, the same thing happens when the gain comes first and the loss come later. So when averaged, this gets you 1.05x, when done sequentially, its 0.9x.
In fact, as your loss increases, the gain needed to compensate for it goes up exponentially. If, instead of 0.6x, you got 0.5x, then the 1.67x shoots up to 2. At 0.4x, it has shot up to 2.5x. At 0.1x (something that happens to lots of stocks), you'll need 10x to compensate! The moral of the story is that investing losses are hard to compensate with gains and so it's disproportionately important to avoid losses.
However, there's a deeper issue here, which is the 'death of behavioral economics' bit that Taleb is talking about. The basic idea behind behavioral economics is that human beings are not rational and are subject to numerous irrational 'biases.' In investing, for example, 'loss aversion' is supposed to be one huge irrationality that human beings have. Daniel Kahneman, the psychologist who got an Economics Nobel for his work, calculated that people feel losses at a multiple of 2.25. That would mean that a loss of Rs 1 lakh would feel as bad as a gain Rs 2.25 lakh would feel. This is supposed to be an irrationality. But is it?
In an individual's life, a single large loss can wipe out a large chunk of the gains made before that, as well as make future gains smaller or impossible. The average of those individual gains and losses is irrelevant, no matter what Nobel Prize winners tell you. Here's a key idea from Peter's paper: "Observed behaviour deviates starkly from model predictions. ..., this has led to a narrative of human irrationality in large parts of economics. ... the models were exonerated by declaring the object of study irrational."
Even the trivial example above shows that a larger aversion to losses is rational and justifiable. As Warren Buffett's famous two rules of investing state: No. 1: Never lose money. No. 2: Never forget rule No. 1.