We received a comment on my VaR vs. Tracking Error: The flawed debate post that is worth its own post as an answer.
Reader "P" asked:
"What do you consider to be the most significant dangers of using a short-term, downside measure (e.g., VaR) for a fund with a long term investment horizon? Is there much literature on this? Presumably much of it is behavioural?"
P, thank you for your question. It sparked a spirited debate internally amongst our blog contributors. Collectively, this is our answer to your question.
First, we believe there are two questions implicit in the one you asked:
1. Why would a long horizon manager need to look at short-term risk?
Long horizon models have long half-life by definition. As a result, when a market reversal occurs after a period of low or moderate volatility, long-term models become wildly out of sync with a market for at least 3-4 months. In this case, the only model that can give reasonable risk readings is a short-term model. This has nothing to do with a provider, but is inherent in a long horizon model's construction. If one could afford to be without any risk readings in times of turbulent market reversals, then one could stick with only a long-term model.
2. Is there a danger of whip-sawing?
We tend to agree with you that this problem is behavioral. If the portfolio has a long-term horizon and is down significantly in the short-term, it is reasonable to wonder whether the portfolio manager, senior management, and/or the client will remain committed to its long-term horizon. If not, then the short-term downside risk can’t be ignored.
So, if the danger of whip-sawing is real, this turns into a question of how a long horizon manager uses the short-term model. Obviously, short-horizon managers will use a model in a different way, because their bets are on the same horizon with risk, so they can weigh risk-return tradeoffs on the same frequency. For a long horizon manager, a short-term model is not a day-by-day investment tool, but rather yet another reading on a dashboard of relevant market indicators. It is an FYI when markets are tranquil. It becomes crucial and a source of actionable intelligence in turbulent times.
We welcome your questions! Please leave comments on any post and we will respond here or to you directly.
Posts
Thank you for your answer.
ReplyDeleteI agree that a short term model can be informative as an indicator, but that it should not be used as the key risk model for a long term process.
One question remains: what will be the negative impact of using a ST model as the fundamental risk model for a LT process? Behaviourally, will a manager tend to take less risk if they are confronted by a model that is ST in construct, even if their investment horizon is LT?
Hi P,
ReplyDeleteWorking with our clients, it is our experience that they need to just decide whether they are going to consider short-term (ST) risk or not. In a portfolio managed with a long-term process, if all involved (managers, clients, senior management, etc.) are committed to ignoring ST risk then a ST model shouldn't be considered. Why waste time on the analysis and absorb an additional, unnecessary data cost? Most of our clients with a long-term investment process can't realistically ignore ST risk. Many wish they could. Most feel it is unwise to ignore ST risk if they know that they will be held accountable. There are potential portfolio construction implications to considering ST risk. There are potentially far uglier practical implications of simply ignoring ST risk.
FactSet integrates third party risk models of varying horizons because there is not a clear single answer for all portfolios.