ssrn-id2014430

8/2/2019 SSRN-id2014430

1/16Electronic copy available at: http://ssrn.com/abstract=2014430

1

UBS Alternative Investments 19 December 2011

UBS Financial Services Inc. (UBS FS) is pleased toprovide you with information about alternative

investments. There are a few points we would like toraise with you at the outset.

This article is for educational and informationalpurposes only. It does not constitute investmentadvice. Among other things, it does not take intoaccount your personal financial situation, or yourinvestment goals or strategies. You should notconstrue any information provided here to be aninvestment recommendation for you to follow. Youshould contact your Financial Advisor for informationor recommendations that may be useful specifically toyou. Although all information and opinions expressed

in this document were obtained from sources believedto be reliable and in good faith, no representation orwarranty, express or implied, is made as to itsaccuracy or completeness, and it may not be reliedupon as such.

Any opinions expressed or information provided inthis document is subject to change without notice,and UBS FS has no obligation to update such opinionor information.

Alpha Characteristics of Hedge Funds

Authored by:

Sameer Jain, UBS Alternative Investments

Andrew Yongvanich, UBS Alternative Investments

Xinfeng Zhou, Phd, CFA *

Introduction

Hedge Fund (HF) managers are expected to createexcess investment returns (Alpha) through two primaryskills based sources:

Security selection: buying undervaluedsecurities and selling overvalued securities.

Market timing: entering markets in advanceof, or when they are rising and exiting, orshorting them when they are declining.

In this paper we employ a Kalman Filtering approach tomeasure the skills based component of HF returns. We

separately quantify value generated through markettiming and security selection decisions over variousmarket regimes and detail the characteristics of HFAlpha.

Key Takeaways

HF managers have historically generatedmeaningful, excess, skill based returns (Alpha)through active management.

These excess returns, whilst still verysignificant, have decayed over time as the

industry has grown.

The Alpha in HF returns has consistently comefrom security selection decisions.

The Alpha in HF returns has been reduced bymarket timing decisions.

The benefits of taking risks to generate activeskill based returns outweigh their costs.

In secular equity bear markets, HFs havesignificantly outperformed on both an absolute

as well as on a risk adjusted basis.

In secular equity bull markets, HFs havesacrificed some upside but have been lessvolatile and have outperformed on a riskadjusted basis.

Quantification of time varying Alpha hasimportant implications for manager selection,asset allocation and portfolio construction.

Highlights PageIntroduction 1Separating Sources of Manager Value Creation 2Changing Market Risk Exposure 2Measuring the Value of Timing Decisions 3Challenges in Measuring Changing MarketExposures

3

Measuring Changing Beta Kalman Filtering 5Kalman Filter Results 7Results of Our Analysis 9

Alpha Return by Strategy 9Alpha Volatility by Strategy 10Information Ratio by Strategy 11

Do HFs Create Alpha in Bear Markets 12Do HFs Create Alpha in Bull Markets 13Implications of Absence of Market Timing Alpha 14Conclusion 14

* Non-UBS Contributor

8/2/2019 SSRN-id2014430


2


Separating Sources of Manager Value Creation

HF manager returns are seen to be a combination ofpassive exposure to general market factors (referred toas Beta) and active investing decisions resulting inmarket outperformance (referred to as Alpha). Alpha,in other words, is the skills based return componentaccruing from (i) an ability to time the markets i.e. todirectionally lever and de-lever positions, thusmagnifying or diminishing levels of exposure tomarkets and (ii) from security selection (to go eitherlong or short). While most practitioners agree thatsecurities selection has the potential to generate Alphain investing, not much is known about the valuegenerated from market timing decisions alone.

Surprisingly, it is poorly understood how HFs producemeaningful Alpha over long term market cycles.Generally, when investors and analysts assess manageroutperformance to the market or Alpha they

combine market timing and security selection sources,rather than distinguish the unique contribution of eachto a managers risks and returns. Once one separates amanagers Alpha into market timing and securityselection components, one can determine whethereach component adds value on a stand-alone basis.

In this article, we outline a theoretically soundframework for measuring changing exposures tounderlying markets. By identifying and measuring suchchanging market exposures or dynamic Betas ofHF managers, we are able to break down a HFstrategy's returns and risks into three distinctcomponents:

Beta: the portion created by the passiveaverage long-run market exposure.

Market Timing Alpha: the portion created byproactive variations in market exposure aroundthe passive average exposure over time.

Security Selection Alpha: the portionremaining, due to security selection and stockpicking skills both long and short.

We outline here a Kalman Filtering approach (described

in detail later) for measuring value generated throughmarket timing and security selection decisions.

Our analysis suggests that HF managers havehistorically generated Alpha across most strategiesespecially through security selection, but they havedetracted value through their market timing decisions.This however does not necessarily indicate that all HFmanagers are bad market timers, for our study wasdone at an aggregate industry level by strategy type;

when we tested selectively on proprietary data for abiased sample of manager returns, we did detectmarket timing Alpha, albeit on an inconsistent basis.

Our analysis also suggests that during bear markets, inparticular September 2000 September 2002 andNovember 2007 February 2009, HF managers havesignificantly outperformed the broad equity index byprotecting downside risk.

Changing Market Risk Exposure

HFs pursue absolute return strategies in an attempt togenerate attractive returns in a variety of marketconditions. Accomplishing this goal requires significantflexibility including the ability to increase or decreaseexposure very quickly, short sell markets and securities,invest in illiquid securities, shift strategies as well asrotate across sectors or asset types. Skillful managers

are expected to anticipate favorable marketmovements by increasing exposures in advance of, orwith, rising markets or decreasing exposures inadvance of declining markets i.e. adjust their Betas inanticipation of broad market moves.

These short to medium term shifts to preempt marketsare essentially market timing decisions. The questionfor investors, therefore, is to understand to whatextent such market timing decisions add value. If a HFmanager is good at market timing, investors may feelcomfortable allowing them to change exposures todifferent markets as they see fit. On the other hand, ifa manager is a poor market timer, such tactical betscould add significant risk without a commensurateincrease in returns.

Active Management

Why do market exposures of HF strategies change overtime?

In many cases, managers actively manage marketexposures as an explicit way to create value. If amanager believes that a certain equity market isundervalued, then for a period of time, he willpotentially increase exposure to those equities in that

market as a means of creating value, or in other wordsgenerate market timing Alpha. In other cases,however, a change in market exposure is a by-productof a strategy to create value, say through securityselection. In this case, changing exposures happenswithout regard to explicit market timing value-creation;the value-creating strategy is based on some otherpremise, often valuations based security selection. Asan example, it may be the case that an equity long-short manager is changing her Beta over time simply as

8/2/2019 SSRN-id2014430


3


a result of changes in the attractiveness of valuations inhigh Beta (small cap) versus low Beta (large cap) stocks.

Measuring the Value of Timing Decisions

In this paper we have dispensed with most of themathematics that we used in building our models,except for a few equations that are integral to theexplanation. Most readers may however, skim theseequations without loss of comprehension, or directtheir attention to the results of our analysis on page 9.

In order to decompose the various aspects of HF Alpha,we first need a way to quantify the market timingcomponent of returns. One conceptually simpleapproach is to examine a managers exposure to themarkets that he trades at every point in time. If onaverage the manager increases his exposure to marketswhen they go up, and decreases exposure to these

markets when they go down, the manager willgenerate positive returns through market timing.

Comparing the returns from a managers averagemarket exposures to the returns from themanagers time varying market exposures cantherefore help determine value added frommarket timing.

More formally, we can express a managers returns asfollows:

, , , , ,t t m t m t t t m m t m t m m t t R R R R (1)

where R t is the managers excess return over cash at

time t, t is the expected Alpha at time t, m,tis a vectorof the managers exposure to the market at time t, Rm,tis the markets excess return over cash at time t, and tis the residual variation in the managers return.

The first equality term in the above equation is a fairlystandard representation but things get moreinteresting when we examine the second equality term,which is in fact, a measure of Active Beta or markettiming.

Estimating Active Beta: If we define m as the

managers average Beta exposure,mt m

is themanagers Active Beta exposure based on themanagers timing decisions. Equation (1) thereforestates that a managers return at each point in timeconsists of four components:

1. Expected Alpha based on security selectiondecisions, t

2. A return based on the managers average

exposure to the market, ,m m tR

3. A return based on the managers market

timing decisions, ,mt m m t R

4. A residual component with zero mean,corresponding to the risk generated fromsecurity selection decisions, t

Computing ,mt m m t R , we generate a time seriesof the managers market timing returns and estimatethe expected return and standard deviation of markettiming returns; i.e., the Alpha and Alpha volatility from

market timing decisions.

Challenges in Measuring Changing MarketExposures

The challenge that investors face, and much of theliterature on the subject ignores, is in measuringmarket exposures on a time varying basis. Standardfactor models assume that market factor exposures areconstant through time, as shown in the fairly popularequation (2).

,t t m m t t R R

(2)

As a result, returns from market timing are captured inthe coefficient. If a manager has generated positiveAlpha (), investors have no way of knowing how

much of this comes from market timing versussecurity selection. However, this distinction is crucial tounderstanding howmanagers have generated returns,and to what types of risks (whether market timing orsecurity selection) they have been exposed to.

As an example, a long short equity manager who ispoor at market timing may have substantial exposure

to the equity market during a downturn. Investors withsignificant equity positions in their portfolios may wantto avoid this manager, or at the very least, activelymanage their equity exposure in a way that offsets themanagers market timing decisions.

We tested multiple approaches to isolate the markettiming component of manager returns and present ourfindings in Figure A and Figure B (page 7 and 8).

8/2/2019 SSRN-id2014430

4/16

8/2/2019 SSRN-id2014430

5/16

5


approaches did not provide any way to model thesensitivities of factor exposures in the returns timeseries. They measured past (weighted) averagesensitivities. In addition, all sensitivity coefficients in theregression equation were identically affected by theweighting, regardless of their rate of change as theweights only depend on the position of the observationin the time series.

A Better Approach to Measuring Changing Beta Kalman Filtering

We propose that measuring a managers market timingability requires a new estimation framework, one thatadapts to and dynamically reflects the managers actualexposures. After extensively testing, we discardedregression based approaches and chose to estimateHFs dynamic Betas using a filtering procedure. Thisapproach, borrowed from engineering, treats Betas as

state variables that follow an autoregressive process.

As is fairly well known in signal processing, when itcomes to separating impurities from valuableinformation, a key feature of any high quality filter is tokeep as much useful data as possible while eliminatingless meaningful data. However, due to the difficulty indifferentiating between the two, it is almost alwaysimpossible to obtain 100% pure information. The mostacceptable compromise is to adopt an approach thatestablishes the minimum estimation errorwith respectto the true value. The Kalman Filter is an approach thathas been shown not only to work well in practice, but

proves to have smaller estimation error than mostother techniques in a broad array of situations.

We used the Kalman Filter to generate estimates of atime series of hidden true information using observabledata that was accompanied by a lot of noise. Bycombining this filtering technique with statisticalestimation methodology, we were able to obtain thehidden true information HF market risk exposure andtime varying Beta - with a fair degree of precision.

The basic structure of the Kalman Filter has two parts:(i) a measurement module, and (ii) a process module.While the measurement module connects the

unobservable information (HF market risk exposure) tothe observable data (monthly returns data of the HFindex), the process module describes a model thatallows the unobservable data to evolve through time.Once we started with a good estimate of market riskexposure with some degree of confidence in thatestimate, we could predict a better estimate offuture market risk exposure - through the process andmeasurement module by using only expectations andnot the variance associated with the predicted values.

After this, the next step was to compare the predictedvalue with the realized value and minimize the error toimprove the predicted value iteratively for the nextperiod.

Technical Description

Our filter can be expressed using an observationequation and measurement equation, as shown below:

Observation Equation

,t t m t t t R X (3)

where RNt ,0~

Process Equation

1

, , 1

t t

t

m t m t

M

(4)

where Mis a 2 2 state transition matrix that modelsthe dynamics of expected Alpha and market Beta over

time,t

is a vector of serially uncorrelated disturbance

with mean zero and covariance matrix Q,

~ 0,t N Q .

Estimates for t are derived using the followingequation:

1 1( )t tt t t t t t K R R

(5)

where tt| is the estimate of the managers time t

market exposures at time t, 1|

tt is the estimate of the

managers time t market exposures at time t 1, Rt is

the managers actual return at time t, 1|

ttR is the

managers time t predicted return at time t 1, and Ktis the Kalman Gain.

Conceptually, the time tBeta estimate depends on the

previous Beta estimate, the difference between thepredicted and actual return, and the Kalman Gain. TheKalman Gain, in turn, is a function of the observationnoise (t) and the process variation (t). If theobservation noise is high relative to the processvariation, the Beta estimates will be relatively stable.On the other hand, if the process variation dominatesthe observation noise, the Beta estimates will fluctuatesubstantially.

8/2/2019 SSRN-id2014430

6/16

6


Kalman Filter Provided Superior Results

We established that the Kalman Filter approachoutperformed both time weighted and rollingregressions, for it optimally used all information abouta managers returns. Exponentially weightedregressions or rolling regressions, by contrast, madearbitrary judgments about how much importance toattach to each observation, and lead to less reliableestimates.

However, one of the drawbacks in using KalmanFiltering, was that it required a significant amount ofdata. While an OLS regression may be reliablyestimated with as few as 30 data points, KalmanFiltering requires an order of magnitude more data,depending on the number of factors. Fortunately, wewere able to limit this drawback to some extent byadding more structure to our model. For example, weassumed that Alpha and Beta is not correlated i.e

while both Alpha and Beta have their own variance thecovariance between Alpha and Beta- i.e. the off-diagonal elements in the Q matrix are zero - and wereable to reduce the amount of data required to properlyestimate the Kalman model.

Kalman Filter Results

One of the practical advantages of the Kalman Filterwas that due to the filtering property, informationfrom the previous steps was accumulated in thevariance of best estimate, and the variance of bestestimate was reduced relative to the variance ofbetter estimate. In essence this filter recursivelyminimizes estimation error without observing thehidden (market factor exposure) information.

In Figures A and B, we provide empirical evidence ofthe minimum error property by filtering the HFRI EquityHedge index's excess returns into US market excessreturns for the S&P 500.

8/2/2019 SSRN-id2014430

7/16

7


Kalman Filter Results (continued)

Figure A: HFRI Equity Hedge Beta Exposure

Source: HFR Industry Reports, HFR, Inc., www.hedgefundresearch.com

While normal regressions show a relatively flatter and smoother pattern of Beta changes- as seen in the green and redlines- , the filtering technique produces much more dynamic picture as seen by the more volatile blue line - of changein market exposures. When we compared the mean square error, the statistical measure of how much the forecasted

value differed from the realized value, the filtering technique turned out:

(i) to have 27% lesser error than from normal regression and

(ii) to have 18% lesser error than from the weighted regression technique.

By making the error component as small as possible the Kalman filter was able to pick up changing Beta exposuresfaster which, as described earlier helped us determine Active Beta, and market timing Alpha.

Jan1993 Jan1995 Jan1997 Jan1999 Jan2001 Jan2003 Jan2005 Jan2007 Jan2009 Jan20110

0.2

0.4

0.6

0.8

1HFRI Equity Hedge Market Beta Exposure

Kalman Filter

Ordinary Regression

Weighted Regression

8/2/2019 SSRN-id2014430

8/16

8


Kalman Filter Results (continued)

Figure B: HFRI Equity Hedge Annualized Alpha

Source: HFR Industry Reports, HFR, Inc., www.hedgefundresearch.com

Figure B plots expected Alpha over time. It suggests that, except for a brief time period in 2008, over nearly all othertime periods HF managers have generated Alpha. The skills based component of returns generation has been ofparticular value to investors, as we will later demonstrate.

Figure B also suggests a discernable trend in diminished Alpha over time, though. Our calculations suggest that after2003 the average Alpha in the Equity Hedge strategy has been around 3% to 4 % while for the six years following1993 the average Alpha was around 9%. Despite the decay, our analysis suggests that HFs have significant andpersistent excess returns that are from security selection based, rather than market timing based. We do not know thereasons for Alpha decay. It is entirely possible, that as HFs have proliferated over the years, the underlying markets fortheir securities may have become more efficient. Also, growth in assets under management may have curtailed theirability to uncover arbitrage opportunities or have reduced their capacity for implementation. Another reason could bethat the greater institutionalization of the industry, as well as risk from outflows/ investor redemptions, may havereduced their propensity for risk-taking.

Another point, which bears with intuition is that, one would not expect long term expected Alpha to fluctuate as muchas Beta (as compared with Figure A), and indeed our results bear that out.

We further conducted this exercise for HFRI Event Driven, HFRI Macro, HFRI Market Neutral, HFRI Quantitative

Directional and HFRI Short Bias indices and found results that were not too different.

Jan1993 Jan1995 Jan1997 Jan1999 Jan2001 Jan2003 Jan2005 Jan2007 Jan2009 Jan2011

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

HFRI Equity Hedge Annualized Alpha

Kalman Filter

Ordinary Regression

Weighted Regression

8/2/2019 SSRN-id2014430

9/16

9


Results of Our Analysis

Figure 1 and 2: Alpha Return by Strategy

Figure 1: Alpha Return by Strategy (1993-2011)

6.7%

7.0%

5.3%

6.7%

5.9%

0.4%

-0.8%

-0.5%

0.8%

-0.8%

1.3%

-0.3%

-2% -1% 0% 1% 2% 3% 4% 5% 6% 7% 8%

HFRI Equity Hedge

HFRI Event Driven

HFRI Macro

HFRI Market Neutral

HFRI Quantitat ive

Directional

HFRI Short Bias

Security Selection Market Timing

Source: UBS Al, HFR Industry Reports, HFR, Inc.,www.hedgefundresearch.com

Figure 2: Alpha Return by Strategy (2003-2011)

3.1%

5.8%

3.2%

6.5%

4.2%

-2.5%

-0.9%

-0.9%

1.4%

-0.5%

-0.2%

-0.6%

-4% -2% 0% 2% 4% 6% 8%

HFRI Equity Hedge

HFRI Event Driven

HFRI Macro

HFRI Market Neutral

HFRI Quantitative

Directional

HFRI Short Bias



We analyzed Alpha characteristics for three major HFstrategy groupings equity hedge, event driven,macro as well as three sub-strategies in the EquityHedge index. Those being, market neutral, quantitativedirectional, and short bias strategies.

Figures 1 through 6 compare the level of Alpha return,Alpha volatility or risk, as well as the Information Ratiofrom security selection versus market timing for these

six strategies.

With the exception of global macro and shortbiased, all other strategies had negative markettiming returns. We conclude that in aggregate, HFmanagers have not been able to produce positiveAlpha return through market timing over the longrun. This result is not particularly surprising as veryfew managers, both in the traditional active long-only space, or in mutual funds too, have beensuccessful at persistently timing markets correctlyover different market cycles.

Global macro managers have produced Alpha

return through market timing over both the longand short run. This is perhaps because macromanagers are given a wider mandate enablingthem to participate across all asset classesincluding equities, fixed income, interest rates,currencies and commodities. Macro strategies areprimarily centered on directional Beta timinginvestments which may explain their ability togenerate market timing Alpha returns.

However, all HF manager strategies havegenerated significant excess return throughsecurity selection.

While there has been some Alpha return decaythrough security selection in the past decade, mostHF managers still produce meaningful Alphareturn.

Market Neutral is the largest Alpha return producerthrough security selection at approximately 6.5%,suggesting that HF managers have unique skills ingoing both long and selling short.

Equity Hedge strategies have seen the most declinein Alpha return across time. This could beattributed to the fact that it is the most popularstrategy by number of firms, funds and largestAUM which may have resulted in a drag fromunderperforming managers.

The analysis suggests that over the past 20 years,HF managers destroy value through market timing,very similar to findings from active long-onlymanagers. However, due to HFs ability toparticipate in a wider range of investmentinstruments and markets, ability to short, andhedge, HF managers have delivered Alpha returnover the long run.

8/2/2019 SSRN-id2014430

10/16

10


Results of Our Analysis (continued)

Figure 3 and 4: Alpha Volatility by Strategy

Figure 3: Alpha Volatility by Strategy (1993-2011)

5.5%

3.9%

5.4%

4.1%

6.6%

11.1%

1.9%

2.1%

3.1%

2.4%

3.1%

5.4%

0% 2% 4% 6% 8% 10% 12%

HFRI Equity Hedge

HFRI Event Driven

HFRI Macro

HFRI Market Neutral

HFRI Quantitative

Directional

HFRI Short Bias



Figure 4: Alpha Volatility by Strategy (2003-2011)

4.1%

3.4%

4.2%

4.1%

4.5%

4.5%

1.7%

1.8%

2.3%

2.3%

2.3%

1.7%

0% 1% 2% 3% 4% 5%

HFRI Equity Hedge

HFRI Event Driven

HFRI Macro

HFRI Market Neutral

HFRI Quantitative

Directional

HFRI Short Bias



Security selection, across all strategies, contributedmore volatility to the funds when compared tomarket timing.

Alpha volatility from market timing has rangedapproximately between 2-4% while securityselection has had a higher Alpha volatility between4-7%. Even though both Alpha volatilitycomponents added additional risk to overall

returns, returns derived from taking securityselection risks outweigh the negatives as explainedby an increase in the Information Ratios.

It appears, as seen in Figure 3 and Figure 4, thatalmost all HF strategies have reduced their riskprofiles since 2003. This may explain why most HFstrategies have generated lower returns. Due tothe recent market environment uncertainty andposture favored towards capital preservation, HFsappear to be taking less risk resulting in lowerreturns.

Historically, quantitative directional strategies haveproduced the most Alpha volatility while eventdriven strategies have added the least. Morerecently, risk appears to be more evenlydistributed. Because of this, we next examine theInformation Ratio to determine Alpha per unit ofrisk taken. Ostensibly, if an investor is strategyagnostic, the strategy with the highest InformationRatio, is the one that ought to be chosen.

8/2/2019 SSRN-id2014430

11/16

11


Results of Our Analysis (continued)

Figure 5 and 6: Information Ratio by Strategy

Figure 5: Information Ratio (1993-2011)

1.2

1.8

1.0

1.6

0.9

0.0

0.9

1.3

0.9

1.2

0.6

0.1

0.0 0.5 1.0 1.5 2.0

HFRI Equity Hedge

HFRI Event Driven

HFRI Macro

HFRI Market Neutral

HFRI Quantitative

Directional

HFRI Short Bias

Only Security Selection Security Selection & Market Timing


Figure 6: Information Ratio (2003-2011)

0.8

1.7

0.8

1.6

0.9

-0.6

0.5

1.2

0.9

1.1

0.7

-0.6

-1.0 -0.5 0.0 0.5 1.0 1.5 2.0

HFRI Equity Hedge

HFRI Event Driven

HFRI Macro

HFRI Market Neutral

HFRI Quantitative

Directional

HFRI Short Bias

Only Security Selection Security Selection & Market Timing


The Information Ratio here is a measure of Alphareturn per unit of Alpha risk undertaken to generatethat excess skill based return. It removes Beta ormarket risk and isolates just the active risk the riskthat HF managers are paid to take. Similar to theSharpe ratio, the higher the value, the more attractivethe strategy is on a risk adjusted basis.

Our analysis theorizes that the Information Ratio of

security selection alone is higher than that ofcombined returns from security selection andmarket timing. In other words, market timingactually destroys efficiency, even after taking intoaccount that it provided diversification againstsecurity selection.

We submit that if HF managers remove the markettiming component and increase the risk exposureof security selection decision, one can achievehigher expected Alpha. In this context, therefore,our analysis provides a clear guide on whether thetime variations in market exposures should behedgedin case where value is being destroyed,our answer is unreservedly yes.

Historically, as seen in Figure 5 and Figure 6,almost all strategies have produced a positiveInformation Ratio; this advances our thesis thatHFs, in the long term through multiple marketregimes, outperform traditional investingstrategies.

8/2/2019 SSRN-id2014430

12/16

12


Do Hedge Funds create Alpha in Bear Markets?

Figure 7: Bear Market Returns (09/00-09-02)

-12.4%

4.0%

-8.4%

-30.0%

0.1%

-35% -30% -25% -20% -15% -10% -5% 0% 5% 10%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Selection


Figure 8: Bear Market Returns (11/07-02/09)

-28.1%

-21.5%

-2.8%

-3.8%

-52.0%

-60% -50% -40% -30% -20% -10% 0%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Selection


Figure 9: Bear Market Risks (09/00-09-02)

7.6%

7.3%

3.6%

17.7%

1.9%

0% 5% 10% 15% 20%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Selection


Figure 10: Bear Market Risks (11/07-02/09)

11.5%

8.1%

6.2%

19.5%

3.3%

0% 5% 10% 15% 20% 25%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Selection


Now that we have established that HFs, in aggregate,have demonstrated an ability to create Alpha, weexplore their performance in falling markets. We selectEquity Hedge strategies as an example due to theirlarge sample size and commercial popularity.

As shown in Figure 7 and Figure 8, we tested for twoperiods of falling returns, September 2000 September 2002 and November 2007 - February 2009.

During periods of falling markets, Equity Hedgeprotected up to 50% in 2008/2009 and almost72% in 2000/2002 of downside risk; we posit thatthese are not insignificant numbers. Whencomparing the different return drivers, Beta

exposure in swift falling markets is the largestdetractor of returns. HFs, for they are by definitionhedged to Beta (a long term average of 0.41 forEquity Hedge) protect investors.

One would expect that HF managers who cut netexposure to the equity markets when they arefalling would outperform the markets still further.We however do not see this happening, as seen inFigure 7 and Figure 8; market timing has notcreated Alpha. Rather the Alpha of around 4%comes from security selection decisions in2000/2002.

8/2/2019 SSRN-id2014430

13/16

13


In 2008/2009 security selection decisions was not apositive contributor as correlations peaked, andindividual securities moved inline with the marketsirrespective of sector, industry or geography. Pricesdeeply diverged from fundamentals as marketswere largely driven by fear and panic. Stock pickers

were rewarded for their fundamental analysis inthe earlier bull market but failed to provide anymeaningful return in the most recent crisis thiswas not unexpected, for when markets stopfunctioning as they did in 2008/2009 pricediscovery broke down.

By hedging on the downside, HF managers areable to lower their overall risk profile leading tolower volatility and greater diversification forinvestor portfolios. When comparing Equity Hedgerisk to equity market risk, they appeared to bealmost 45% less.

These results reiterate the important role of havinga low Beta and the potential for Alpha generation,when compared to passive long only equity indexreplication.

In summary, we conclude that HFs have performedbetter than the equity market in drawdown periods.This outperformance however has little to do withsuperior market timing abilities; we were not able toconclusively establish that they increased their Betaexposures ahead of market rises or decrease their Betabefore market falls. Rather, their outperformance is afunction of their hedged characteristic being hedged;

they do not rely on market exposure to create returns,which protects them in falling markets.

Do Hedge Funds create Alpha in Bull Markets?

We isolated three periods of rising returns, these beingfrom January 1993 August 2000, October 2002 October 2007, and March 2009 October 2011. Thefirst and the most recent period data can be viewed inthe appendix. We believe it is too premature to analyzethe returns from the most recent 2 years as anindication of long term performance.

From the data in Figure 11, we can see significantupside appreciation during rising equity markets.As noted in the commentary earlier, securityselection and not market timing is the main Alphagenerator during these successful times. Markettiming actually reduced performance during theserallies while security selection provided an equalamount of return as Beta exposure. Thisdemonstrates that there is skill generating theexcess returns opposed to solely Beta exposure.

HFRI Equity Hedge index captured 78% of theupside with 56% of the equity market risk. Byanalyzing the Information Ratio, or the Alpha riskadjusted return, the HFRI EH Index outperformedthe broader markets. The results establish that

even if the upside is only capturing a portion of therally, EH is a superior investment option due to itsdecreased risk profile.

Figure 11: Bull Market Returns (10/02-10-07)

9.6%

5.0%

4.9%

12.2%

-0.4%

-2% 0% 2% 4% 6% 8% 10% 12% 14%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Selection


Figure 12: Bull Market Risks (10/02-10-07)

5.5%

4.0%

3.6%

9.7%

1.5%

0% 2% 4% 6% 8% 10% 12%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Selection


8/2/2019 SSRN-id2014430

14/16

14


Figure 13: Bull Market Information Ratio (10/02-10-07)

1.7

-0.3

1.4

1.3

-0.5 0 0.5 1 1.5 2

S&P

HFRI EH (Total)

Market Timing

Security Selection


Conclusion

With reduced long-term return expectations fortraditional assets, HFs, often perceived as a pure formof active management, have drawn increasing investorattention. HF managers have a great deal of freedomin exploiting trading approaches and utilize thatfreedom to change their strategy in order to capitalizeon opportunities as they see them. It is precisely thisfreedom to actively manage their portfolio that hasbeen viewed by many as the primary advantage HFshave over more traditional styles of activemanagement; indeed, investors often want theirmanagers to access opportunities where they seethem. That said, this freer form of active managementis not without its costs. To the extent that HF managerscan employ dynamic trading strategies, and have moreflexibility to execute market timing, it is important forboth managers and investors to understand the risksthey are exposed to; and whether those risks are onesthat generate returns.

In this paper we provided a Kalman Filter basedframework to better understand and measure ActiveBeta - the changing exposures that HFs have tounderlying market factors in a more rigorous manner.We compared the returns from a managers average

market exposures to the returns from the managerstime varying market exposures to separately determineAlpha or value added from market timing and securityselection. We explored the properties of HF Alpha andhighlight implications.

We conclude that HF managers, as an industry, havedemonstrated near persistent ability to create skillsbased returns in almost all market environmentsthrough security selection.

Implications of Absence of Market Timing Alpha

Conventional wisdom views market timing ability witha healthy dose of skepticism, and our analysis supportsthis view. However in some ways, this result may seemsurprising. After all, why should timing markets beharder than selecting securities? There might be atleast two reasons for this result.

First, HFs in general may be focused much more onsecurity selection than on market timing. To the extentthat market timing decisions are implicit, or unrelatedto a managers views about a particular market, wewould expect market timing to increase risk without acommensurate increase in return.

Second, even managers who have some market timingability will have difficulty demonstrating this abilityacross just one or two markets, when the market

timing component of their strategies may lacksufficient breadth. The fundamental law of activemanagement states that a managers Information Ratiodepends on the quality of his forecasts, as well as onthe breadth of the investment strategy (i.e., thenumber of independent investment decisions hemakes). Even a highly skilled manager will have troublegenerating a high Information Ratio if he only makestiming decisions across one or two markets. Thisbecomes more obvious when analyzing global macrostrategies, which seemingly generates excess returnsthrough market timing, as they focus on multiplemarkets and underlying instruments.

Although the reason why HFs have historicallydetracted value through market timing is unclear, theimplications for investors are relatively straightforward.Either HF managers are poor market timers, or they donot actively manage their exposures to differentmarkets. In either case, investors cannot rely on HFmanagers to avoid large corrections in the markets thatthey trade. Investors may also want to consider moreactively managing their market exposures to offset therisks that HF managers generate by changing theirBetas over time.

8/2/2019 SSRN-id2014430

15/16

15


Appendix: Hedge Fund Alpha in Bull Markets

Figure A1: Bull Market Returns (01/93-08/00)

17.0%

6.1%

11.6%

14.7%

-0.7%

-5% 0% 5% 10% 15% 20%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Selection


Figure A3: Bull Market Risks (01/93-08/00)

9.5%

5.5%

6.9%

1.9%

13.2%

0% 2% 4% 6% 8% 10% 12% 14%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Selection


Figure A5: Bull Market Information Ratio (01/93-08/00)

1.8

-0.3

1.1

1.7

-0.5 0 0.5 1 1.5 2

S&P

HFRI EH (Total)

Market Timing

Security Selection


Figure A2: Bull Market Returns (03/09-10/11)

11.6%

9.8%

2.3%

-0.6%

23.7%

-5% 0% 5% 10% 15% 20% 25%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Selection


Figure A4: Bull Market Risks (03/09-10/11)

9.9%

7.3%

3.4%

17.5%

1.8%

0% 5% 10% 15% 20%

S&P

HFRI EH (Total)

Beta Return

Market Timing

Security Se lection


Figure A6: Bull Market Information Ratio (03/09-10/11)

1.2

-0.3

0.7

1.4

-0.5 0 0.5 1 1.5

S&P

HFRI EH (Total)

M arket Timing

Security Selection


8/2/2019 SSRN-id2014430

16/16

16


Alternative Investment Funds Risk DisclosureInterests of Alternative Investment Funds (the Funds) are sold only to qualified investors, and only by means ofoffering documents that include information about the risks, performance and expenses of the Funds, and whichClients are urged to read carefully before subscribing and retain. This communication is confidential, is intended solelyfor the information of the person to whom it has been delivered, and should not be reproduced or otherwise

distributed, in whole or in part, to third parties. This is not an offer to sell any interests of any Fund, and is not asolicitation of an offer to purchase them. An investment in a Fund is speculative and involves significant risks. TheFunds are not mutual funds and are not subject to the same regulatory requirements as mutual funds. The Funds'performance may be volatile, and investors may lose all or a substantial amount of their investment in a Fund. TheFunds may engage in leveraging and other speculative investment practices that may increase the risk of investmentloss. Interests of the Funds typically will be illiquid and subject to restrictions on transfer. The Funds may not berequired to provide periodic pricing or valuation information to investors. Fund investment programs generally involvecomplex tax strategies and there may be delays in distributing tax information to investors. The Funds are subject tohigh fees, including management fees and other fees and expenses, all of which will reduce profits. The Funds mayfluctuate in value. An investment in the Funds is long-term, there is generally no secondary market for the interests ofthe Fund, and none is expected to develop. Interests in the Funds are not deposits or obligations of, or guaranteed orendorsed by, any bank or other insured depository institution, and are not federally insured by the Federal DepositInsurance Corporation, the Federal Reserve Board, or any other governmental agency. Prospective investors shouldunderstand these risks and have the financial ability and willingness to accept them for an extended period of time

before making an investment in a Fund. Investors should consider a Fund as a supplement to an overall investmentprogram. In addition to the risks that apply to alternative investments generally, there are risks specifically associatedwith investing in hedge funds, which may include those associated with investing in short sales, options, small-capstocks, junk bonds, derivatives, distressed securities, non-U.S. securities and illiquid investments.

This document is for educational and informational purposes only. It does not constitute investment advice. Amongother things, it does not take into account your personal financial situation, or your investment goals or strategies. Youshould not construe any information provided here to be an investment recommendation for you to follow. You shouldcontact your Financial Advisor for information or recommendations that may be useful specifically to you. Although allinformation and opinions expressed in this document were obtained from sources believed to be reliable and in goodfaith, no representation or warranty, express or implied, is made as to its accuracy or completeness, and it may not berelied upon as such. Any opinions expressed or information provided in this document is subject to change withoutnotice, and UBS Financial Services Inc. has no obligation to update such opinion or information.

UBS 2011. UBS Financial Services Inc., a subsidiary of UBS AG.

ssrn-id2014430

Documents