Beruflich Dokumente
Kultur Dokumente
Liquidity:
Alpha
Generator or
Risk Factor?
How funds with impaired liquidity profiles
fare in times of market stress.
Faryan Amir-Ghassemi,
Colleen White, &
Michael Perlow
2 Hedge Fund Liquidity: Alpha Generator or Risk Factor? / August 2015 NOVUS
Portfolio
50 30, 65%
haves much the same as a yield curve: 40
30
20
10
Portfolio A
0
Sample Portfolio A's Liquidity Profile 1 5 10 30 60 90 120
Portfolio A
Liquidity Days
100%
90
80
70
Portfolio
60
50 30, 65%
40 profile, as a greater percentage of the portfolio
30 is illiquid, irrespective of the comparable 30-day
20
10 liquidity metric.
0
In order to normalize each of the 1000+ funds
1 5 10 30 60 90 120
liquidity curves, we approached this indexing
Liquidity Days
issue in two ways: using an integral function and
performing a principal component analysis.
The key difference between a yield curve The integral function simply looks at the weight-
and a liquidity curve is that the liquidity curve is ed total area under the curve at each time step,
additive, whereas a yield curve can invert or kink providing an aggregate numeric which describes
further down the time-horizon. the total liquidity curve. Integral functions only
The importance in analyzing the full liquidity work with additive distributions. The principal
curve versus an arbitrary point is how indicative component analysis is the methodology most risk
the arbitrary point may or may not be to the overall providers use to create factor inputs, and it is what
4 Hedge Fund Liquidity: Alpha Generator or Risk Factor? / August 2015 NOVUS
we use internally to condense yield curve sensi- nearly identical index results:
tivities into singular values (i.e., DV01). Principal With this data now indexed across all of our
components are necessary for non-additive distri- hedge funds at every time step, we have the
butions, as twist is a characteristic of a curve that necessary inputs required to regress each funds
an integral cannot capture. Having tested both liquidity profile against their simulated returns in
methodologies on the liquidity profiles of all of our periods of market stress.
portfolios in the HFU, we found both tools provide
3
Y = 3E-05X-0.6141 -05x
-0.6141
2
Y = 3E-05X - 1.5849 y = 3E-05x
-1
1
Normalized Data
Normalized Data
-1
-2
-3
-4
PC1 Integral Linear (PC1) Linear (Integral)
5 Hedge Fund Liquidity: Alpha Generator or Risk Factor? / August 2015 NOVUS
Historical VIX
385.65% VIX
360.00% LEHMAN
320.00%
280.00%
160.00%
EUROZONE
120.00% CRISIS 2.0
OIL CRASH
80.00% MOMENTUM
REVERSAL
40.00%
0.00%
-40.60%
8/19/04 5/26/05 3/2/06 12/7/06 9/13/07 6/19/08 3/26/09 12/31/09 10/7/10 7/14/11 4/19/12 1/25/13 11/1/13 8/13/14 5/21/15
6 Hedge Fund Liquidity: Alpha Generator or Risk Factor? / August 2015 NOVUS
Quant Meltdown
0.15
0.1
0.05
Performance for August 2007
-0.05
-0.1
-0.15
-0.2
-0.25
-4 -3 -2 -1 0 1
Liquidity in July 2007
return half a percent higher than a fund with an average liquidity profile.
-0.1
10/08)
-0.2
2 m C u m u la t iv e P e r fo rm a n c e (9
-0.3
-0.4
-0.5
-0.6
-0.7
-3 -2 -1 0 1
Liquidity
Grexit 1.0 May 2010 a US based event. Our HFU was comprised of 82%
Here we have a case where our regression pro- US securities as of the time of this event. None-
vides contradictory results. Perhaps it is due to theless, the slope of the regression was clearly
the event being a global (European) stress, versus influenced by certain outliers which may have
0.15
0.1
0.05
0.0
Cumulative Per 5-6/2010
-0.05
-0.1
-0.15
-0.2
-0.25
-0.3
-0.35
-0.4
-3 -2 -1 0 1
Liquidity 4/2010
Liqudity Liquidity
Coef P Value Intercept P Value2 R2 n Date Performance Date
US Debt Downgrade September this period of market stress. Managers that held
2011 steady during this V-Shaped market swing gen-
This was a period where many funds were caught erally did fine, but many cut risk near the bottom
flat-footed, and underperformed the market heav- when market fear hit an apex. Unfortunately, our
ily. However, the regression above proves incon- simulation assumes a buy-hold from 06/30/2011
clusive, compared to Lehman. There is actually 09/30/2011 and so the negative trading compo-
a slightly negative slope to the regression which nent cannot be captured from public regulatory
belies conventional wisdom. We believe the filings without stretching assumptions. This data
anomalous results are due to an inability to cap- set in particular provides a worthwhile opportuni-
DebtthatDowngrade
ture the trading component occurred during (Cumulative Per) returns with the simulated
ty to compare realized
0.2
2m Cumulative Performance (8-9/11)
0.1
0.0
2m Cumulative Performance (8-9/11)
-0.1
-0.2
-0.3
-0.4
-0.5
-3 -2 -1 0 1
Liquidity (7/11)
buy-hold to better understand the negative impact that trading under duress can have.
Momentum Reversal January 2014 Ocwen Financial, CVR Energy, Baidu, General Mo-
January 2014 acted as a reset for many popular tors and Apple were particularly hurt this month.
hedge fund names after a torrential 2013 saw mo- Given that we see little correlation between illi-
mentum stocks vastly outperform value. This was a quidity and underperformance, this period appears
period of market undercurrent, as many hedge fund more symptomatic of crowding as a precipitator for
specific names heavily underperformed the broad- underperformance than fund illiquidity. Perhaps
er market. Hedge fund centric names such as Best this is why we see a wider dispersion of results than
Buy, JC Penney, YPF, Nationstar Mortgage, Sprint, other regressions, and this will be a period of focus
January 2014
January 2014
0.2
0.15
0.1
Performance for Jan 2014
0.05
0.0
-0.05
-0.1
-0.15
-0.2
-3 -2 -1 0 1
Liquidity in Dec 2013
Oil Dive October 2014 recovery for the broader market (outside of En-
Finally, the Oil Dive of 2014 was in many ways a ergy) to close out the month. As a result, trading
miniature version of the US Debt Ceiling crisis of is likely the key factor that we could not capture,
2011. While there was significant market stress by which would likely have amplified our results.
the middle of October 2014, we saw a V-shaped
October 2014
October 2014
0.25
0.2
0.15
Performance for Oct 2014
0.1
Performance for Oct 2014
0.05
0.0
-0.05
-0.1
-0.15
-0.2
-3 -2 -1 0 1
Liquidity Before Oct 2014
Key takeaways was probably the best testing ground for the per-
ils of illiquidity, and it coincidentally occurred on a
Public regulatory data likely is not the best model period where 13-F analysis provides a reasonable
fit for definitive results, as trading is a huge psy- sense of manager trading patterns during stress.
chological component that weighs on how a man- Lehman showed that impaired liquidity is a factor
ager handles such periods of market stress. The that can put a portfolio at risk during a severe
US Debt downgrade and October 2014 perhaps market stress.
best exemplify this, as many of our private hedge We conclude that illiquidity has been a
fund clients use our proprietary skill-set simula- powerful alpha-generator for strategies such as ac-
tions on their daily position-level data to under- tivism in this market cycle. However, as one of our
stand how impactful these buy/sell decisions had clients has put it, liquidity is one of three bullet to
been during stress events. the head risks, which cannot be accurately teased
As mentioned before, the post-Lehman market out in a rising tide market. Being cognizant of
rally has provided little comfort for correction- liquidity constraints should we transition to a new
ists (or bears), as every stress event has simply market regime, or undergo a severe correction is
provided an opportunity to buy the dip. Lehman undoubtedly a critical risk parameter to monitor.
About Novus Novus was founded in early 2007 by a group of
investors, data scientists and engineers to build a foundation
that helps the worlds top investors generate higher returns.
Simply stated, we believe investors need to innovate, adapt and
increase the value they deliver. Today, almost 100 of the worlds
top managers and allocators managing approximately $1 trillion
are using a single foundation that leverages all of their ideas and
produces innovative technology on a regular, iterative cycle.
DISCLAIMER
This paper was prepared for informational and discussion purposes only. This paper does not purport to be all-inclusive or to necessarily contain all
the information that a recipient thereof may desire or require nor does it purport to give legal, tax or financial advice. Novus does not undertake to
correct, update or revise this paper and the conclusions and thesis herein may change without notice. Novus shall not in any way be liable for any
claims relating to this paper, any information herein or any errors or omissions with respect to any of the foregoing and makes no express or implied
representations or warranties as to the accuracy or completeness of this paper or any information herein. Thus, investors and others should conduct
their own independent investigation and analysis in connection with any of the matters set forth herein. By accepting this paper, the recipient there-
of agrees to keep confidential this paper and the information contained herein and not to further disclose or distribute this paper or any information
contained herein. This paper may not be photocopied, reproduced or distributed to others at any time, in whole or in part, without the prior written
consent of Novus. The delivery of this paper at any time subsequent to its date does not imply that information herein is correct as of anytime subse-
quent to such date. As used in this paragraph, except to the extent the context otherwise requires, the term "Novus" includes its affiliates and its and
their respective partners, directors, officers and employees.