When everything goes back to where it started, we normally forget what happened

Not too long ago, I found myself looking at a graph for a stock on Fineco, which had fallen substantially, but then climbed all the way back up to its original level. Net profit, obviously, nil. Yet the sense of accomplishment which that graph communicated was definitely not “no action took place.” And thus, a simple question arose in my mind, which I discussed with friends who happen to be economists, and which I am now sharing here, in the spirit of sparking a good conversation.

Observation

When our investments, after a period of high volatility, reach their starting level, we tend to consider such event as non-existent. Our account is back to square one; no need to adjust our sense of the real risks involved. However, the fact is that between the starting point and the end point, there has been some movement – a drawdown, possibly painful, possibly just observed externally – which contains some information about the volatility of the asset, about our ability to stand through it, and which is at risk of being forgotten when we focus only on the two endpoints.

In order to give this phenomenon a temporary name, I would suggest calling it path amnesia (or perhaps more directly, round-trip discount): the bias towards systematically discounting risk information associated with the path when the destination is the same as the origin.

A bit of SOTA

I’ve never enjoyed writing the SOTA section, but it’s necessary to acknowledge the reality here, since it is easy to misinterpret what cognitive psychology has already extensively described in a more rigorous way than I could ever come up with.

Two relevant references, in particular:

• The peak-end rule by Kahneman et al. means that an experience will be evaluated retrospectively in terms of the peaks (either positive or negative) and the ending point of the experience, not in terms of the integral of the path. The experience of a deep drawdown and recovery back to zero can be seen as the very experience predicted by the heuristic.
• The break-even effect, introduced by Thaler and Johnson in 1990 (well known as “gambling with the house money”) demonstrates that people tend to increase their risk preference after a losing streak and recovery just to get back to even.

Thus, in combination, these two lines of research have already explained much of what I observed. It would thus be wrong to suggest that the suggested bias is totally new. I make this observation explicit because, in my view, the main value of such an exercise is not the creation of yet another term, but rather the identification of the borderline between what is already there and what is not.

There is no margin here, if at all, in the understatement of risk perception but rather a more operational one, namely: at equal historical volatility on an integrated basis, will the path shape impact future behavior, in particular, the size of re-entry into a position?

In other words: Two paths, both having equal overall volatility but different shapes, a V-shape, consisting of one deep drawdown compared to W-shape, comprising of two smaller drawdowns, do they lead to equal re-entry sizing or is the shape more important than the pure volatility? And will this effect remain relevant even if the person looking at the graph has not experienced the drawdown personally, eliminating thus the motivational element of the break-even effect?

But if that were true, we would identify a mechanism beyond peak-end and break-even: not just “we base our evaluation of the experience on its extremes” and not just “whichever one lost is more eager to take the risk to make up for it,” but “the actual path is such that it changes how future risks are estimated even by an impartial observer.” Incremental or revolutionary, it has potential for application to practical risk communication to retail investors.

How to test it

Before even considering an experimental approach using subjects, which takes time and effort, the simpler method is to examine existing market data: retail inflows/ outflows, ETF positioning, sentiment measures, contrasting behavior after V-shaped paths compared with flat ones with the same total return. Any signal would appear here before it would even be possible to design a questionnaire.

However, I don’t believe that I have found an entirely new bias; my earlier research before writing this essay convinced me quite the contrary; however, the process of discerning what is known and what is yet to be found is, by itself, useful. If there is anyone who has seen such study, one which separates the effect of path shape at a constant variance, I would like to hear about it; otherwise, it’s just another theory that could be proven by available data.