By Matthew Leatherman and Allen He

Boards and executives of long-term funds, such as pension plans, sovereign wealth funds, and endowments, have a challenging problem. They need to manage those portfolios to meet their long-term purpose, which may be decades or more into the future. Yet no fund has the luxury of looking only to that long-term time horizon. Each must also meet expectations in the near term in order to continue in its role and with its investment strategy.

Currently, investors tend to treat risk management as tradecraft – expert, but repetitive – and neglect the critical thinking required to tailor their risk management to their organization’s purpose. As a result, nearly all of these critical decisions are made by defaults encoded into statistical formulas. The problem is many of these formulas default to a short-term frame.

There are common short-term assumptions around annualized monthly volatility (GIPS), within-horizon risk, the normal distribution, and Value-at-Risk (VaR), all of which make sense in general, but may be less accurate for long-term investors. These defaults can each be summarized through the following flight-path metaphors:

1. Takeoff (annualized monthly volatility): The incorrect initial assumption (as mandated by GIPS) is that annualizing monthly volatility by scaling by the square root of time is a standard ‘check-off-the-box’ procedure, and that time period doesn’t matter. In actuality, this construct assumes correlations don’t change over time and ignores momentum.
2. Flight lanes (within-horizon risk): The incorrect assumption is that only the end-of-horizon risk matters and that the path taken along the way doesn’t. However, some paths that ultimately meet their target may face huge interim drawdowns that not everyone can stomach.
3. Forecasting turbulence (the normal distribution): The incorrect assumption is that each data point is counted equally for a distribution and volatility is averaged. However, distribution should be split into two: a ‘normal’ regime and a ‘turbulent’ regime.
4. Final destination (Value-at-Risk): The incorrect assumption is that it is the total probability the portfolio will lose at least $X, where in reality, it is the end-of-horizon probability; the actual shortfall probability is higher.

In addition, there is a growing body of work that already challenges the status quo, urging risk professionals not to take the default assumptions for granted.

1. The Divergence of High and Low Frequency Estimationhighlights the dangers of scaling up monthly volatility by multiplying by the square root of time, a technique required by GIPS as part of the standard global risk-reporting framework. While perfectly fine for short-term investors, embedded assumptions regarding momentum and correlation across asset classes mean that the longer your time horizon, the more this technique misstates risk for an investor.
2. The Mismeasurement of Risk highlights the dangers of long-term investors anchoring risk to metrics like Value-at-Risk (VaR) and probability of loss. The default metrics only give us the risk at the end of the horizon but tell us nothing about the interim pathway taken to get there. The longer our time horizon, the lower our probability of shortfall at the end, but also the greater chance that we dip below a certain threshold at some point along the path.

   Trade-off of shortfall and potential drawdown

   

3.  Principal Components as a Measure of Systemic Risk highlights the dangers of assuming a single, normal return distribution for all periods of history. In practice, the longer your time horizon, the greater chance you experience volatile events in the markets full of “black-swan” events clustered together. In these times, the volatility can perhaps be better explained by a separate, “turbulent” distribution instead of the default normal distribution used during periods of calm.
4. Back in late 2018, FCLTGlobal published Balancing Act: Managing Risk Across Multiple Time Horizonsto call attention to some of these issues and influence change. While a lot has been achieved, we hope to address some additional issues in this iteration, and there is opportunity to do more by way of implementation.

Since our initial work on risk, we have found that there are some potential hurdles to implementation. Simply spreading the word about default inputs is not enough: the framing must shift so that long-term investors are inspired to make change. Such hurdles include:

1. Lack of clarity: Investment risk management defaults may be unstated or nested within complex formulas. People cannot be expected to think critically about material they don’t know exists.
2. Impracticality: Alternate defaults may not appear feasible. People who know to think critically about defaults may be under the impression that alternative defaults do not exist.
3. Unavailability: Data needed to use alternate defaults may be less available. Those who know about alternate defaults skill may think the data needed to make them work does not exist.
4. Inconsequentiality: The impact of resetting defaults and collecting data differently may feel like more work than it’s worth.
5. Difficulty: Resetting defaults and collecting data differently is hard and involves up-front fixed costs. This is a behavioral tendency – status-quo bias – that is amenable to being nudged so that resetting defaults can become a higher priority.

Through the second phase of our risk research, FCLTGlobal hopes to learn from members why implementation is not happening, what is needed to increase it, as well as to draft a list of hypothetical tools to change behavior. Preliminary tools being considered include, but are not limited to, real-life case studies of portfolio changes, introductory educational videos, fill-in-the-blank templates, and a primer document comparing and contrasting default vs. alternative risk measurements.

If you have any thoughts relevant to the adoption or implementation of long-term risk measurements, or questions about our work, please contact [email protected].