In order to estimate likely returns and risks in the financial markets, our general approach is to identify a set of historical instances that match current conditions on a broad range of important dimensions (in practice, using an "ensemble approach" that randomizes over scores of subsets of historical data). We then look at various features of that cluster, including the average return that followed over various horizons, the deepest loss over various horizons, and the overall spread of those outcomes. In general, the clusters include a mix of both positive and negative outcomes, resulting in moderate estimates of expected return and moderate estimates of risk. In some cases, the average return across the cluster of instances is very positive, and the individual instances show few negative outcomes at all. That sort of condition justifies a very aggressive investment position. In contrast, since the late-1990's, the average returns of the clusters have been quite poor, with a preponderance of negative outcomes in historical instances having similar characteristics. There have been a few exceptions, including the bulk of 2003 (and on the basis of the ensemble methods we presently use, the period between early-2009 and early 2010 - see Notes on Risk Management for details on our unpleasant "miss" during that period). But generally speaking, market conditions since the late-1990's have supported a defensive investment stance much more often than is typical on a historical basis. Of course, the near-zero total return of the S&P 500 since the late-1990's, coupled with two separate market losses of more than 50%, is a reflection that on average, concerns about poor return and high risk during this period have been well-placed.
In reviewing market conditions this week, what strikes me most is the pattern that emerges when we look across various horizons, from 2 weeks out to 18 months. When we examine the average 2-week outcome that has historically followed periods that cluster with present con! ditions, the average outcomes are negative, but not strikingly so. Specifically, the expected return is in the lowest 26% of all historical observations, but that average return is only about -1%, a figure that is overwhelmed by typical short-term noise. That's another way of saying that guessing the market's outcome over the next couple of weeks is like guessing the throw of a very slightly biased pair of dice.
But the profile starts to change significantly as we move out the investment horizon. Looking out 5 weeks, for example, the prospective return falls into the lowest 8% of historical observations. Now, this could certainly change on the basis of shifts in various market conditions, but here and now, the 5-week horizon is more defensive than we've seen in the other 92% of historical data.
Most striking, though, is what we observe on the basis of prospective drawdown (the deepest loss the market experiences within a given horizon) looking out over the coming 18 months. On that front, the present drawdown estimate is in the worst 1.5% of all historical observations.
Keep in mind the distinction between the drawdown and the return over a given period. The drawdown over an 18-month period is the deepest loss experienced by the market from the current point to the lowest point within that horizon, even if the deepest loss occurs fairly early in that window. In contrast, the return over a given period is measured from the starting point to the ending point. Importantly, once we observe conditions that associate with a significant risk of drawdown, we can almost always find some point later on that provides a better entry opportunity to accept market risk.
Continue reading.
In reviewing market conditions this week, what strikes me most is the pattern that emerges when we look across various horizons, from 2 weeks out to 18 months. When we examine the average 2-week outcome that has historically followed periods that cluster with present con! ditions, the average outcomes are negative, but not strikingly so. Specifically, the expected return is in the lowest 26% of all historical observations, but that average return is only about -1%, a figure that is overwhelmed by typical short-term noise. That's another way of saying that guessing the market's outcome over the next couple of weeks is like guessing the throw of a very slightly biased pair of dice.
But the profile starts to change significantly as we move out the investment horizon. Looking out 5 weeks, for example, the prospective return falls into the lowest 8% of historical observations. Now, this could certainly change on the basis of shifts in various market conditions, but here and now, the 5-week horizon is more defensive than we've seen in the other 92% of historical data.
Most striking, though, is what we observe on the basis of prospective drawdown (the deepest loss the market experiences within a given horizon) looking out over the coming 18 months. On that front, the present drawdown estimate is in the worst 1.5% of all historical observations.
Keep in mind the distinction between the drawdown and the return over a given period. The drawdown over an 18-month period is the deepest loss experienced by the market from the current point to the lowest point within that horizon, even if the deepest loss occurs fairly early in that window. In contrast, the return over a given period is measured from the starting point to the ending point. Importantly, once we observe conditions that associate with a significant risk of drawdown, we can almost always find some point later on that provides a better entry opportunity to accept market risk.
Continue reading.
No comments:
Post a Comment