Man Institute road ahead reflections

Drawdown profiles vary dramatically across major asset classes Diversification may help, but won’t solve all your problems all the time. Over the past century, the major asset classes haven’t all crashed at once. But seldom have all been simultaneously in the clear Pairing assets thoughtfully could make the difference between a portfolio that diversifies and one that compounds its own pain I’ve written a fair amount on the nature of equity drawdowns: the pain they cause, their anatomy, what things look like at the peak.1 Here’s an attempt to expand the exercise to other portfolio assets and strategies. After all, you likely have other stuff in your portfolio, and you are therefore likely interested in what their drawdown profiles might look like, and how they interact.

You’ve got to have a code, so they say, here’s mine: A drawdown is 50% of annualised volatility, over an undetermined period of time Drawdowns are felt daily, monthly data should not be used to examine them Two drawdowns separated by a big jump higher are two drawdowns not one ‘A big jump’ is 100% of annualised volatility. Imagine equities fell from 100 to 90 (that’s a 10% decline, in case your mental maths is rusty), then rose to 108 (+20%), then fell again to 85 (-21%). That’s two episodes ‘A big jump’ is 100% of annualised volatility. Imagine equities fell from 100 to 90 (that’s a 10% decline, in case your mental maths is rusty), then rose to 108 (+20%), then fell again to 85 (-21%). That’s two episodes Defining drawdowns is as much art as science You try and implement this creed in code – or indeed any other systematic formulation, I’ve tried a lot – you will always end up with some episodes that simply don’t pass the sniff test. The algorithm is the starting point, but I’ve then gone through and winnowed them, old school You try and implement this creed in code – or indeed any other systematic formulation, I’ve tried a lot – you will always end up with some episodes that simply don’t pass the sniff test. The algorithm is the starting point, but I’ve then gone through and winnowed them, old school I’ve done this for seven portfolio blocks: equities (proxied by the S&P 500, as per the prior notes mentioned), fixed income (proxied by rolling 10-year Treasury futures), gold (physical), Trend (15 vol, all asset), Value (Fama-French HML), Momentum (Fama-French Momentum) and Quality (AQR QMJ). You can see the results in graphical form in the Appendix, but more intuitively in the table in Figure 1.

Past performance is not indicative of future results. This analysis is based on our research and is intended for illustrative use based on considerations listed in this paper and should not be construed as a recommendation and should not be relied on. Source: Bloomberg and Man Group. Date range: July 1926 - May 2026.

Figure 1 makes for interesting perusal. But for those of you who prefer a picture, Figure 2 shows the profile of an illustrative typical drawdown, using the average realised characteristics (depth, length, volatility on the way down and up, and so on). Allow me a brief word on each.

Equity drawdowns are relatively severe, but brief. The average nadir is the second deepest (after Gold) but the average descent length is also the second fastest (after Momentum). On the other side, they are unrivalled in terms of the speed of the recovery, at usually less than five months on average. This means that, taking the fall and the bounce back as one combined event, their average episode is the shortest of all the assets in our survey. They are also distinctive in terms of the contrast between volatility in the drawdown, and outside of it. Stocks run at 15% volatility in normal times, but 24% in sell-offs, a nine-point spread well in excess of all the comparators. Practical implication: investors should be particularly wary of false dawns during red zones in their portfolio’s equity beta.

Bond drawdowns are interesting in terms of the proportion of history they cover. On our definition, bonds are in an episode almost half the time, even if the absolute numerical frequency, at one every 23 months, is in line with that of Equities and Trend. These declines tend to be moderate, and not particularly volatile, realised risk in the down market is almost identical to normal times risk. This tallies with the market folklore that bonds, in contrast to stocks, grind lower and crash up, at least with regard to the move down. Monetary tightening tends to be well telegraphed (while loosenings often come by surprise), and these frequent but gentle performance declines are within that grain.

Gold drawdowns are arguably the most unpleasant of all, tending to be both longer (1.5 years on average) and deeper (down almost a third) than other assets. Moreover, recovery time is significantly slower; a mean of 1.3 years is double any of the others, with the exception of Momentum (0.9). The one redeeming feature, if you can call it that, is that the downdraft tends to be relatively orderly. In fact, the volatility of the yellow metal in the drawdown (18%) is actually lower than in normal times (20%), a counterintuitive result, but speaking to the fact that this is the kind of pain that is long and grinding, rather than a short sharp one. Is a marathon more painful than the 1500 metres? You pays your money, you takes your choice. In any event, these characteristics make sense for an asset largely lacking in economic fundamentals, and therefore highly susceptible to real or perceived regimes, and narrative trading. Some stories can last a long time.

Trend drawdowns are distinctive in that they are heavily tilted toward smaller sell-offs. Despite having overall volatility not hugely dissimilar to equities, defining small / medium / large thresholds at the same level (10/20/30%) puts almost 90% in the ‘small’ category. This likely speaks to the dynamic nature of the strategy: like Kipling’s leopard, it can change its spots as the environment changes, although it should be noted that its few large bear markets do come around major regime inflection points, and none more profound than the start of World War II. There is also a similar pattern to Gold, in that volatility within down-moves tends to be less than that outside them, only here the pattern is more extreme, at 160 basis points (bps) below (10.9% plays 12.5%). This is largely by design, given that vol. scaling is meant to reduce exposures as the market’s overall risk patina increases.

Like Fixed Income, Value also tends to spend a lot of its time in drawdown posture, at 45%, only two points short of the former. Remember that value-investing is a contrarian bet on intrinsic book value being recognised in market value, and with greater emphasis than long term expected cashflows. Anyone who’s spent any time in the school playground will have learned the hard way that, most of the time, being contrarian doesn’t work, even if the few times it does are spectacular.

Momentum, on the other hand, usually spends the least proportion of its time in drawdown (20%), for exactly the same reason. Unlike Trend (to which it is positively correlated over long periods, as well as in drawdowns, as we shall see later), its drawdown profile is skewed towards the larger nadirs (which represent 60% of the total). These tend to be extremely short and sharp, with the average length of six months – the quickest in our list. The average magnitude is in line with stocks, and the biggest fall of 63% (the market pivot at the tail-end of the Global Financial Crisis [GFC]) was deeper than any, bar equity. Watch someone go through a mid-life crisis; when a contrarian pivot comes, it is often fast and profound.

Quality drawdowns have a bit of a regime-feel to them. While in general moderate and relatively rare, the strategy has both the longest drawdown (the seven years in the 1960’s tech hype cycle prior to the Nifty Fifty boom) and the longest gap between drawdowns (over a decade, from 1980 to 1992) of any that we have surveyed. It is also an outlier in terms of the containment of the extremes: its largest decline (the 31% fall through the post-GFC recovery) is smaller than any of the other maximums. Given a disciplined commitment to elevate metrics of income statement or balance sheet quality, there’s only so wrong you can go it seems. An average covers a multitude of sin, so, of course, there are exceptions to the characterisations of these drawdowns. Try telling an investor going through the almost three year equity drawdown between 1929-1932 that stock market crashes are sharp but short. Still, hopefully the above can provide a helpful base rate to cue off when you’re in the teeth of one.

Past performance is not indicative of future results. This analysis is based on our research and is intended for illustrative use based on considerations listed in this paper and should not be construed as a recommendation and should not be relied on. Source: Bloomberg and Man Group. Date range: July 1926 - May 2026.