Tuesday, May 20th, 2014

Scenario Analysis

One of our prospective clients asked an interesting question this week. How does Harlyn do its scenario analysis? How do we reduce all the possible permutations of the future to a manageable number of options? The short answer is that we don’t. The reason is that scenario analysis is a forecast in disguise and we have a well-known aversion to forecasting. We would not change our investment process if the price action disagreed with what the scenario specified, so there is not much point in constructing the scenario.

Nonetheless it is clearly important to prepare for the future as much as possible. Just because we have a systematic approach to investing, this does not mean we have to look backward all the time. This is where our concept of the probability curve comes in. Originally we developed the curve to avoid the base effects inherent in using a single period to construct our probability estimates. We did not want to commit to a new position because something which happened 53 weeks ago was no longer included in our sample. Hence we base our estimates on the average of different sample periods ranging from 52 to 14 weeks. We do not expect these probability estimates to be consistent, but frequently we observe that they are, and that’s when they become interesting.

The current 50-week sample is a subset of the current 52-week sample, and so on. If our estimate of equities beating bonds becomes steadily more favourable as we shorten the sample period, we regard this as bullish for equities, reflecting better momentum of returns, or falling volatility, or a combination of the two. It’s a simple matter to draw a trend line through these estimates and work out whether the fit with the underlying data is good enough to be significant. Normally it is. We allow for the possibility of a reversal in the trend, by using a three stage polynomial (‘S’ curve).

By definition every data point in this week’s 14-week sample will be in next week’s 15-week sample, and so on. This means that the shape of the curve will only change gradually. If we see that a particular asset class has an upward-sloping trend, preferably with some convexity, we can say with confidence that it is likely to outperform its comparator on a risk-adjusted basis. We publish over 100 of these probability curves every week, side by side with the time-series of the simple average. In other words we show the current situation and our best estimate of how it is going to evolve. It may not be conventional scenario analysis, but it is forward-looking and grounded in the data.

So, what do we expect the near-term future to look like? Three of our most interesting curves are (1) EM Equities vs Global Equities and (2) Small Caps vs Large Caps in the UK and the US. EM Equities have a powerful, upward-sloping and convex curve. This suggests that they may not fall as much as feared, even if there is a general correction in equity markets (led by the US which has a downward-sloping curve). Small caps in the UK and the US have powerful, downward-sloping and concave curves which suggest that investors will stay defensive in equities, even if there is no correction.

Leave a Reply

Full Blog Archive:


Other Recent Posts