Professional Documents
Culture Documents
Viewpoints - Drawdown Management Strategies - 20190529 - en
Viewpoints - Drawdown Management Strategies - 20190529 - en
Drawdown management
strategies – preparing for
the next market downturn
Multi Asset Boutique –
Alternatives & Multi Manager Solutions
So-called trading strategies, encom- All drawdowns are not alike – an introduction
passing global macro and commodity After equity markets corrected between October and
December 2018, interrupting the soon-to-be longest
trading advisor (CTA) hedge funds, bull market in history, investors are increasingly looking
faced stiff headwinds after an unprece- for portfolio building blocks that provide protection
dented flood of global monetary stim- or a degree of diversification in drawdowns in order to
mitigate the negative impact these drawdowns have
ulus followed the Great Financial Crisis for investors. Drawdowns can occur in all asset classes,
of 2008 and the ensuing Euro Crisis. but since equities tend to be the riskiest asset class
Central bank actions frequently inter- in the portfolios of most institutional investors and thus
have the biggest loss potential, we will focus on equity
rupted trends and created whipsawing drawdowns hereafter. Moreover, while some investors
prices. Global macro and CTAs are might already consider equity corrections in the 5 % range
the only major hedge fund strategies “drawdowns,” we argue that the loss-threshold for a
drawdown definition should be in the –15 % to –20 % range
showing positive bear-market perfor- in order to focus on the occasional but painful events.
mance, positive convexity and the ability Considering the recent Q4 2018 sell-off and the ensuing
to take sizable short positions. These recovery, we also hold that the length and path of a
correction are important for the drawdown definition.
strategies surf capably in rough waters.
The examples in Chart 1 show that the definition of a
drawdown is not straightforward; individual drawdowns
differ in various parameters and are not easily classifiable.
Chart 1 and Table 1, which show all periods with peak-
to-trough losses of more than 20 % of the S&P 500 Price
Index (daily data since 1925), highlight the different shapes
of drawdowns. Since the longest drawdown starting in
September 1929 lasted 25 years, we limit the chart to the
first four drawdown years in order to improve readability.
2/8 Asset Management / Multi Asset Boutique /Alternatives & Multi Manager Solutions / May 2019 Vontobel
For institutional investors only / not for public viewing or distribution
Chart 1: Strong dispersion of historical S&P 500 one should differentiate between longer-stretched draw-
Price Index drawdowns with respect downs on the one hand, and very sharp / abrupt draw-
to duration and magnitude downs on the other hand. Generally speaking, the former
0% can be tackled by deploying medium- to longer-term
– 10 % trend-following CTAs, while the latter form of drawdowns
– 20 % can better be confronted with long volatility strategies.
– 30 %
Trend-following CTAs as drawdown mitigators
Drawdown
– 40 %
– 50 % When it comes to strategies that offer diversification
– 60 % during drawdowns, people often think of trend-following
– 70 % strategies. These strategies are the biggest sub-category
– 80 % of managed futures, also called CTAs (for Commodity
– 90 % Trading Advisors). Pure trend-following strategies rely
– 100 % only on prices and take a long position in a market when
0 100 200 300 400 500 600 700 800 900 000 the price trend goes up and a short position when the
Number of trading days price trend goes down. These strategies were one of the
09.1929 10.2007 03.2000 01.1973 few that performed positively during the 2008 financial
12.1968 08.1987 12.1961 12.1980 crisis as well as during other equity drawdowns. And unlike
02.1966 08.1956 plain-vanilla hedging strategies such as buying index
put options, trend-following does not come with an explicit
Source: Vontobel Asset Management, Bloomberg. ex-ante cost-burden in the form of a payable option
Past performance is no indication of current or future performance.
premium.
Table 1: Notable S&P 500 drawdowns since 1925 For our further analysis, we use the BarclayHedge
BTOP50 CTA Index as proxy for trend-following strategies.
DRAWDOWN NO. MONTHS NO. MONTHS MAX. The BTOP50 Index is an equally weighted index of the
START TO RECOVERY TO TROUGH DRAWDOWN
largest CTA programs which represent at least 50 % of
Sep 1929 301 34 – 86 % the BarclayHedge CTA universe. While there are other
Nov 2007 65 16 – 53 % trend-following indices including the SG Trend Index and
Sep 2000 81 25 – 46 % the HFRI Macro: Systematic Diversified Index, we use
Jan 1973 91 21 – 46 % the BTOP CTA Index because it has the longest history.
Dec 1968 42 19 – 33 % The chart below highlights the performance of the
Sep 1987 23 3 – 30 % BTOP50 CTA Index during all S&P 500 corrections with
Dec 1980 25 20 – 24 % a peak to trough performance of at least –5 % (since
Jan 1962 20 6 – 23 % inception of the BTOP50 in 1987). The blue bar shows the
Aug 1956 26 17 – 19 % performance of the BTOP50 CTA Index during the peak
Feb 1966 15 8 – 18 % to trough period of the S&P 500, the yellow bar shows the
Jun 1990 9 5 – 16 % performance of the BTOP50 CTA Index from peak until
Jul 1998 5 2 – 16 % recovery of the S&P 500.
Source: V
ontobel Asset Management, Bloomberg. Chart 2: BTOP50 CTA Index performance was positive
Past performance is no indication of current or future performance.
during the major drawdowns of S&P 500,
but not necessarily during smaller corrections
One striking distinction is the length of the “under-water” 100 %
period. For example, while it took 25 years to recover 80 %
the drawdown starting in 1929 purely in nominal terms 60 %
(please note that we do not use total returns here, i.e. 40 %
dividends are not reinvested), recent drawdowns had 20 %
shorter durations. Another difference is the various paths 0%
of the individual drawdowns. In 1987, it took only slightly – 20 %
more than three months to get from peak to a 32 % trough. – 40 %
On the other hand, the drawdown starting in 1968 only – 60 %
reached its trough after roughly one and a half years. Both
11.2007
09.2000
09.1987
06.1990
07.1998
10.2018
06.2015
02.1994
01.2000
01.1990
07.1999
02.2018
08.1997
04.2000
09.1991
Performance of the BTOP50 CTA Index was positive in Based on the often noticed positive performance of trend-
the five major S&P 500 corrections since 1987, while following strategies during longer equity drawdowns,
performance during smaller corrections (including the but also during extended upward trends, the term “CTA
one starting in October 2018) was mixed. smile” has become established. We get this pattern
by plotting the monthly returns of the S&P Index on the
We pick the two worst drawdowns of the S&P 500 Index horizontal and the corresponding monthly returns of
to show the evolution of the performance of the BTOP50 the BTOP 50 CTA Index on the vertical axis of a scatter
CTA Index during the S&P drawdown. In the November chart and then fitting a second order polynomial. The
2007 drawdown, the BTOP50 CTA Index generated charts below show the “CTA smile” with data since 1987,
positive performance during the initial phase of the draw- first with monthly, then with quarterly returns.
down and managed to keep these gains during the
recovery phase of the S&P 500 Index. Chart 5: The “CTA smile” with monthly data shows
the tendency of CTAs to perform well
Chart 3: In the November 2007 S&P 500 drawdown, in the most positive or negative months
the BTOP50 CTA Index was positive during in equity markets
the initial phase and protected its gains during
in same period
80 % 15 %
60 % 10 %
40 % 5%
20 % 0%
0% –5 %
– 20 % –10 %
– 40 % –20 % –10 % 0% 10 % 20 % 30 %
60 %
Source: Vontobel Asset Management, Bloomberg.
40 %
Past performance is no indication of current or future performance.
20 %
0%
– 20 % The smile is more pronounced, i.e. shows a higher curva-
– 40 % ture or convexity, for quarterly returns. It is worth noting
– 60 % that most other hedge fund strategies show a negative
2000 2001 2002 2003 2004 2005 2006 smile or smirk with respect to equity returns.
S&P 500 BTOP50 CTA
The “CTA smile” highlights two features of trend-followers:
Source: Vontobel Asset Management, Bloomberg, BarclayHedge. First, trend-followers increase their exposure when trends
Past performance is no indication of current or future performance.
extend, i.e. let the winners run, and cut losing positions
quickly, be it on the long or the short side. In mathematical
The reasons why trend-following performance was not finance jargon we would say that they are long gamma.
equally strong in the above examples are manifold and may Second, trend-followers need the trends to last a certain
be due to factors such as different trend opportunities in time in order to build their profits. We consider the follow-
asset classes beyond equities (fixed income, currencies ing factors important when assessing a trend-follower’s
and commodities), the initial positioning of the index funds, reaction to a drawdown: holding time, positioning at the
the composition of the index and strategy changes in onset of a drawdown, and asset allocation.
the index funds.
4/8 Asset Management / Multi Asset Boutique /Alternatives & Multi Manager Solutions / May 2019 Vontobel
For institutional investors only / not for public viewing or distribution
The holding time measures the reaction time or speed of This contrasts with many other active strategies as well
a trend-follower. This varies between individuals from as equities which exhibit fat left tails, i.e. negative skewness.
40 to 120 days for the majority. This indicates that trend- The chart below shows the historical evolution of the roll-
followers generally do not profit from drawdowns that ing 36-month skewness of the BTOP50 CTA Index. After
are brief. In recent years, initial signs of equity draw- being positive for a long time, it turned negative in February
downs have often resulted in central bank actions and 2018, and overall we notice a downtrend in skewness.
rhetoric that stopped the drawdowns and generated
“whipsaw” price behavior. This environment was detri- Chart 7: The rolling 36-month skewness of the
mental to trend-following performance; the shorter the BTOP50 CTA Index was mostly positive
holding period and more reactive the strategy, the higher in the past, but a downtrend is noticeable
the risk of being “whipsawed” by repeatedly sharp price 1.5
Skew (absolute)
reversals. Not surprisingly, some trend-followers have 1.0
therefore increased the holding-period of their programs, 0.5
thus slowing the reaction time and reducing the risk of 0.0
being whipsawed. These trend-followers’ reduced trading – 0.5
speeds, however, come at the expense of higher market
12.2009
12.2003
12.2005
12.2007
12.1999
12.1993
12.2001
12.1989
12.1995
12.1997
12.1991
12.2011
12.2013
12.2015
12.2017
beta and reduced convexity since they take longer to
adapt to emerging trends.
The positioning at the onset of drawdowns is another Source: Vontobel Asset Management, Bloomberg, BarclayHedge.
Past performance is no indication of current or future performance.
factor to consider when looking at the behavior of trend-
followers in a drawdown. If a drawdown follows just on
the back of a long-lasting, smooth uptrend, trend-followers While the trend-follower features were widely acknowl-
will have built up substantial long equity positions. In the edged in the past, the recent period of muted trend-
first phase of the equity correction, trend-followers will follower performance has caused much discussion about
therefore suffer losses on their equity allocation which the return potential for trend-following going forward.
can, absent of profit-generating trends in other classes, The main points of criticism are:
lead to initial losses in parallel with equities. The period –– The strategy’s presumed simplicity makes it easy to
January / February 2018 exemplified such an environment replicate. This has led to big inflows and an over-
and highlighted the path-dependency of trend-followers’ crowding in the space.
positioning and correlation to equity markets. –– Markets have become more efficient, hence the
behavioral inefficiencies that trend-followers rely on
Finally, an important part of trend-followers’ edge comes to make profits have diminished.
from allocating across asset classes, i.e. including fixed –– The unprecedented liquidity flood and increased
income, commodities and currencies. The number of market control and coordination of central banks
futures contracts invested in can range from 50 to 150 has interrupted emerging trends, especially on the
across the globe, and the individual allocations are broadly downside, and generated whipsaws which have
diversified across contracts, so profitable trends may led to losses for trend-followers.
arise from various sources. Consequently, historical trend-
following contributions during equity drawdowns came With regards to the first point, a September 2018 analysis
not only from short positions in equities, but also from by AQR1 concluded that while trend-following assets
other asset classes. For example, a substantial part of peaked in 2008 at USD 211 billion, they have since
the 2008 trend-following performance came from fixed decreased and were running at USD 124 billion by mid-
income and commodities. The allocation to the asset 2018.
classes and individual contracts is highly dependent
on the portfolio construction process of the individual Chart 8: Estimated Managed Futures AUM
managers. As a result, the performance dispersion January 2004 – June 2018
of trend-followers is significant, although the strategy 230
$211 B
seems relatively homogeneous at first glance. This favors 210
investing in a diversified portfolio of high-quality trend- 190
followers rather than a single CTA, in order to optimize
AUM ($ Billions)
170
overall risk-reward. 150
130
Trend-following: Recent pro and con debates 110 $124 B
Returning to the CTA smile, it also shows evidence of 90
positive skewness, a feature of trend-followers which is 70
highly sought-after for protecting against equity draw- 50
downs and complementing negatively skewed strategies. 2004 2006 2008 2010 2012 2014 2016 2018
This means that compared to a normal distribution,
Source: AQR
the return distribution of trend-followers has more pro-
nounced positive returns than negative ones, mainly
an effect of the “let winners run, cut losers quickly” rule. 1
AQR (2018), “Trend Following in Focus”
5/8 Asset Management / Multi Asset Boutique /Alternatives & Multi Manager Solutions / May 2019 Vontobel
For institutional investors only / not for public viewing or distribution
At the same time, AQR points out that futures markets Trend-followers tend to generate substantial profits unex-
have increased in size over the same period. pectedly in a short time after long drought periods, for
which the first quarter of 2019 may serve as an example.
Chart 9: Futures Markets Open Interest Timing is therefore nearly impossible and there is a risk
January 2004 – June 2018 of leaving the strategy just before another performance
24 boost, which happened to quite a few investors losing
faith in trend-following just prior to the 2014 performance
spike. Managers like AQR 2 or CFM3 found that trend-
Rolling 3-Month Median Open Interest
20
(Number of contracts, in millions)
2
urst B., Ooi Y. H., Pedersen L. H. (2017),
H
“A Century of Evidence on Trend-Following Investing”
3
apital Fund Management (2015),
C
“Trend Following: A Persistent Market Anomaly“
6/8 Asset Management / Multi Asset Boutique /Alternatives & Multi Manager Solutions / May 2019 Vontobel
For institutional investors only / not for public viewing or distribution
Global macro strategies: A good complement to CTAs for There is no “hard and fast” definition of global macro
building a holistic drawdown management portfolio strategies. The main common denominator of global
Global macro strategies tend to share a few characteristics macro managers is their forward-looking focus on broader
with CTAs which are crucial for drawdown management economic, regional and asset class shifts. Other than
purposes: Positive convexity/skew as well as the ability to that, the global macro scene is extremely diverse, as illus-
go net short entire asset classes in size. trated in the table below. This also explains why global
macro indices are of rather limited use when assessing
On the other hand, global macro strategies differ from global macro’s suitability as drawdown management
(trend-following) CTAs in important ways, making them a strategy. Rather than looking at global macro from an
valuable complement for drawdown management pur- extremely heterogeneous (and thus diluted) index level,
poses. The main differences to trend-following CTAs are it seems more fruitful to look at individual styles (and
global macro’s discretionary (non-systematic) use of fun- managers): Some of them prove rather valuable as draw-
damental information to trade a generally broader suite of down mitigators, while others might even turn out counter-
instruments (often options) in a forward-looking manner. productive.
Table 2: Global macro strategies vary widely in approach and are not always preferable for drawdown management
Linear vs. Ranging from completely linear expressions (e.g. futures, Mix of linear and options-based implementations (crucially
Non-linear forwards) to completely options-based implementations only long options), with a bias to options- and options-like
(i.e. long gamma) trading styles
Amount of Ranging from strongly positive net vega to moderately Prefer positive vega portfolios
net vega negative vega portfolios
Geography Ranging from truly global approaches to regional focuses Build portfolio of regionally diversified approaches,
(e.g. emerging or developed markets) with sufficient emphasis on major markets
Asset Class(es) Ranging from truly multi-asset class (incl. equities, Mix across asset classes, conditional on good liquidity
commodities, credit) to more rates- or FX-focused also during stress scenarios
approaches. Ranging from purely intra-asset class to
overwhelmingly cross-asset class expressions
Trading speed Ranging from very strategic 1-2 year trading horizons Mix of time horizons, with a tilt to relatively faster trading
to very tactical, even intra-day approaches speeds – even more so, when medium to long-term
trend-following CTAs are already deployed in the portfolio
Sources of Ranging from purely fundamental data-driven to over- Mix between price and non-price input data
information whelmingly technical approaches.
Originality of Ranging from mainstream/equilibrium views to more Tilt to managers with relatively less mainstream, more
views “long shot” (i.e. less probable, less priced-in) variant views (given convincing arguments and sufficient
investment ideas risk controls)
Risk bias Ranging from significantly long-biased (balanced portfolio) Avoid long-biased managers, in all their explicit and implicit
approaches to moderately short-biased styles forms. We prefer a mix between unbiased and (tactically)
short-biased managers/strategies
Degree of Ranging from one theme / trade to dozens of themes / trades Mix of diversification degrees. This is conditional on the
diversification more concentrated managers / strategies being sufficiently
robust /convex to withstand adverse market moves
Momentum vs. Ranging from mainly momentum / break-out driven to Prefer momentum and momentum-like strategies (i.e.
mean-reversion largely mean-reversion / value-oriented approaches positive gamma), for which disciplined risk management
is a key feature
Degree of Ranging from overwhelmingly intuitive (synergistic) Mix of information processing approaches. Avoiding overly
intuition approaches to analytically refined, more quantitative optimized/fragile approaches, which might be overly
approaches reliant on the more recent past. Preference for “skeptics”
as well as risk-conscious and nimble traders
Number of Ranging from single-risk taker model to dozens Tend to avoid overly diversified multi-PM strategies, as their
risk-takers of risk-takers portfolios often tend to be somewhat diluted, for the sake
of higher performance-stability
7/8 Asset Management / Multi Asset Boutique /Alternatives & Multi Manager Solutions / May 2019 Vontobel
For institutional investors only / not for public viewing or distribution
Global macro style selection: Separating drawdown market, so-called straddle and strangle option strategies.
management “wheat from chaff” Although this simple strategy literally guarantees positive
First of all, one has to avoid falling into an overly narrow, convexity, it unfortunately comes with a hefty price tag:
myopic and backward-looking bias often associated with Substantial premium time decay on the options acquired.
crisis-protection strategies. This tendency of “fighting Hence, this approach to creating convexity only tends to
the last war” led many investors to flock to conventional be sensible if the market prices for volatility (implied vola-
tail-protection strategies (e.g. buying short-dated equity tility) are very cheap and / or the timing of an anticipated
volatility) right on the back of the Great Financial Crisis larger price move is spot-on, thus losing less time value.
2008. Buying the kind of “protection” which worked best
during the last crisis often turned out to be the most Trading strategies, on the other hand, tend to spend a
expensive protection. Just as important, both market relatively small amount of capital on directly buying
participants’ as well as regulators’ various reactions to a options. Depending on their specific trading styles, they
past crisis often render a crisis in the same “spot” (e.g. deploy various means to generate positive convexity
market or country) less likely. There are some similarities in less expensive ways:
to the situation after a plane crash occurred: Risk aware-
ness and mitigation measures often lead to the post- Disciplined risk management
crash risks actually decreasing rather than increasing – Most global macro managers use pre-defined stop-
while consumers’ (behavioral) reaction goes the other way. losses on both the portfolio as well as on a position /
theme level, and most CTAs have systematically built-in
A key aspect when investing in global macro strategies sliding stop-losses as well. This key feature of most trading
for drawdown management purposes is to look at each strategies of adding to winners and cutting losers is a
manager and style individually, given their high heteroge- crucial source of convexity, coming at the cost of being
neity. For example, the January 15, 2015 Swiss Franc “whipsawed” every once in a while, when market prices
de-peg from the Euro saw some macro managers score oscillating in a range lead to the trading strategy getting
big, while others suffered significant losses. If the losers stopped out repeatedly at the (in hindsight) wrong spots.
and winners among macro managers during this particu-
lar episode had been randomly distributed, there would Selectively investing in fundamentally
be no point in trying to select specific managers. How- asymmetric market constellations
ever, the outcome for specific macro managers (and their An extreme example of this was going long credit-default
related style of macro trading) was anything but random: swaps protection around 2006 / 07, due to their being
There was a clear tendency of managers with long con- very cheap, i.e. only having priced a minimal likelihood of
vexity profiles to score big, in some instances in the order a credit meltdown. Similarly, shorting the Euro against
of +30 %. These managers’ positive convexity was cre- the Swiss franc ahead of the January 2015 peg break
ated mostly by being long options and positioned for a also offered a strongly asymmetric opportunity; as the
divergence of the Euro-Swiss Franc trading range, as well exchange rate came ever closer to the 1.20 floor, the
as by being skeptical about the Swiss National Bank’s Swiss National Bank got nervous, and the positioning of
stamina (and willingness) to stay their course and “walk investors was crowded in the belief of the status quo
their talk.” holding.
Disclaimer
This document is for information purposes only and nothing contained in this document should constitute a solicitation, or offer, or recommen-
dation, to buy or sell any investment instruments, to effect any transactions, or to conclude any legal act of any kind whatsoever. Past perfor-
mance is not a reliable indicator of current or future performance. The return may go down as well as up due to changes in rates of exchange
between currencies. The value of invested monies can increase or decrease and there is no guarantee that all or part of your invested capital
can be redeemed. Although Vontobel Asset Management AG (“Vontobel”) believes that the information provided in this document is based
on reliable sources, it cannot assume responsibility for the quality, correctness, timeliness or completeness of the information contained in this
document. Except as permitted under applicable copyright laws, none of this information may be reproduced, adapted, uploaded to a third party,
linked to, framed, performed in public, distributed or transmitted in any form by any process without the specific written consent of Vontobel.
To the maximum extent permitted by law, Vontobel will not be liable in any way for any loss or damage suffered by you through use or access to
this information, or Vontobel’s failure to provide this information. Our liability for negligence, breach of contract or contravention of any law as
a result of our failure to provide this information or any part of it, or for any problems with this information, which cannot be lawfully excluded, is
limited, at our option and to the maximum extent permitted by law, to resupplying this information or any part of it to you, or to paying for the
resupply of this information or any part of it to you. Neither this document nor any copy of it may be distributed in any jurisdiction where its dis-
tribution may be restricted by law. Persons who receive this document should make themselves aware of and adhere to any such restrictions.
In particular, this document must not be distributed or handed over to US persons and must not be distributed in the USA.