Вы находитесь на странице: 1из 53

An EDHEC-Risk Institute Publication A Post-crisis Perspective on Diversification for Risk Management May 2011 Institute

The authors are grateful to Professor Lionel Martellini for useful comments and suggestions. 2 Printed in France, May 2011. Copyright EDHEC 2011. The opinions expressed in this study are those of the authors and do not necessarily reflect those of EDHEC Bu siness School. The authors can be contacted at research@edhec-risk.com.

A Post-Crisis Perspective on Diversification for Risk Management May 2011 Table of Contents Abstract ....................................................................... ............................................. 5 Introduction ................... ................................................................................ ......... 7 1. Advantages and Disadvantages of Diversification ................. ................... 11 2. Beyond Diversification: Hedging and Insurance ........ ................................19 Conclusion .................................. ............................................................................29 A ppendices ...................................................................... .......................................31 References ........................... ................................................................................ ...37 About EDHEC-Risk Institute ............................................... ................................41 EDHEC-Risk Institute Publications and Positio n Papers (2008-2011) .........45 An EDHEC-Risk Institute Publication 3

A Post-Crisis Perspective on Diversification for Risk Management May 2011 About the Authors Nol Amenc is professor of finance and director of development at EDHEC Business S chool, where he heads the EDHEC-Risk Institute. He has a masters degree in econo mics and a PhD in finance and has conducted active research in the fields of qua ntitative equity management, portfolio performance analysis, and active asset al location, resulting in numerous academic and practitioner articles and books. He is a member of the editorial board of the Journal of Portfolio Management, asso ciate editor of the Journal of Alternative Investments, member of the advisory b oard of the Journal of Index Investing, and member of the scientific advisory co uncil of the AMF (French financial regulatory authority). Felix Goltz is head of applied research at EDHEC-Risk Institute and director of research and developme nt at EDHEC-Risk Indices & Benchmarks. He does research in empirical finance and asset allocation, with a focus on alternative investments and indexing strategi es. His work has appeared in various international academic and practitioner jou rnals and handbooks. He obtained a PhD in finance from the University of Nice So phia-Antipolis after studying economics and business administration at the Unive rsity of Bayreuth and EDHEC Business School. Stoyan Stoyanov is professor of finance at EDHEC Business School and programme d irector of the executive MSc in risk and investment management for Asia. He has nearly ten years of experience in the field of risk and investment management. H e worked for over six years as head of quantitative research for FinAnalytica. H e also worked as a quantitative research engineer at the Bravo Risk Management G roup. Stoyan has designed and implemented investment and risk management models for financial institutions, co-developed a patented system for portfolio optimis ation in the presence of non-normality, and led a team of engineers designing an d planning the implementation of advanced models for major financial institution s. His research focuses on probability theory, extreme risk modelling, and optim al portfolio theory. He has published nearly thirty articles in academic journal s, contributed to many professional handbooks, and co-authored two books on fina ncial risk assessment and portfolio optimisation. 4 An EDHEC-Risk Institute Publication

Abstract An EDHEC-Risk Institute Publication 5

A Post-Crisis Perspective on Diversification for Risk Management May 2011 Abstract Since the global financial crisis of 2008, improving risk management practices ma nagement of extreme risks, in particular has been a hot topic. The postmodern qua ntitative techniques suggested as extensions of mean-variance analysis, however, exploit diversification as a general method. Although diversification is most e ffective in extracting risk premia over reasonably long investment horizons and is a key component of sound risk management, it is ill-suited for loss control i n severe market downturns. Hedging and insurance are better suited for loss cont rol over short horizons. In particular, dynamic asset allocation techniques deal efficiently with general loss constraints because they preserve access to the u pside. Diversification is still very useful in these strategies, as the performa nce of well-diversified building blocks helps finance the cost of insurance stra tegies. 6 An EDHEC-Risk Institute Publication

Introduction 2. xxxxxxxxxxxxxxxxxx An EDHEC-Risk Institute Publication 7

A Post-Crisis Perspective on Diversification for Risk Management May 2011 Introduction Risk management practices became a central topic after the financial crisis of 2 008. Improvements to the methods of risk measurement, many of them made by indus try vendors, have drawn on the literature on the modelling of extreme events (Du bikovsky et al. 2010; Zumbach 2007). Although there has been extensive research into extreme risk modelling in academe since the 1950s, it is only after difficu lt times that the financial industry becomes more open to alternative methods.1 From an academic perspective, however, risk management decision making goes beyo nd risk measurement and static asset allocation techniques. In fact, it can be a rgued that the non-classical methods are designed to use two basic techniques in financediversification and hedgingin a better way, and with the recent focus on p ost-modern quantitative techniques the role of diversification as a risk managem ent tool has been over-emphasised. Even though it is a powerful technique, diver sification has limitations that must be understood if unrealistic expectations f or the real-world performance of risk management are to be avoided. Although the idea behind it has long existed, a scientifically consistent framework for dive rsification, modern portfolio theory (MPT), was first posited by Markowitz (1952 ). Diversification international diversification, sector and style diversificatio n, and so onhas since become the pillar of many investment philosophies. It has a lso become a very important risk management technique, so much so that it is oft en considered, erroneously, synonymous with risk management. In fact, diversific ation as a 8 An EDHEC-Risk Institute Publication general method is related to risk reduction as much as it is to improving perfor mance and, therefore, it is most effective when it is used to extract risk premi a. In short, it is only one form of risk management. The limitations of diversif ication stem from its relative ineffectiveness in highly correlated environments over relatively shorter horizons. Christoffersen et al. (2010) conclude that th e benefits of international diversification across both developed and emerging m arkets have decreased because of a gradual increase in the average correlation o f these markets. Thus, if international markets are well integrated, there is no benefit in diversifying across them. The variations of correlation are importan t not only across markets but also over time; in the short run, then, relying on diversification alone can be dangerous. Over longer horizons, Jan and Wu (2008) argue that diversified portfolios on the mean-variance efficient frontier outpe rform inefficient portfolios, an argument that adds to the debate that time alon e may not diversify risks. The limitations of diversification mean that, in cert ain market conditions, it can fail dramatically. Using a conditional correlation model, Longin and Solnik (2001) conclude that correlations of international equ ity markets2 increase in bear markets. In severe downturns, then, diversificatio n is unreliable. Furthermore, it is generally incapable of dealing with loss con trol. So enhancing the quantitative techniques behind it by using more sophistic ated risk measures and distributional models can lead to more effective diversif ication but not to 1 - See, for example, the discussion in Sheikh and Qiao (2009) about a framework for static asset allocation based on non-classical models. 2 - Longin and Solni k (2001) base their model on extreme value theory. There are other studies drawi ng similar conclusions through models based on other statistical techniques.

A Post-Crisis Perspective on Diversification for Risk Management May 2011 Introduction substantially smaller losses in crashes. Loss control can be implemented in a so und way only by going beyond diversification to hedging and insurance, two other approaches to risk management. A much more general and consistent framework for risk management is provided by the dynamic portfolio theory posited by Merton ( 1969, 1971). The theory presents the most natural form of asset management, gene ralising substantially the static portfolio selection model developed by Markowi tz (1952).3 Merton (1971) demonstrated that in addition to the standard speculat ive motive, non-myopic long-term investors include intertemporal hedging demands in the presence of a stochastic opportunity set. The model has been extended in several directions: with stochastic interest rates only (Lioui and Poncet 2001; Munk and Srensen 2004), with a stochastic, mean-reverting equity risk premium an d non-stochastic interest rates (Kim and Omberg 1996; Wachter 2002), and with bo th variables stochastic (Brennan et al. 1997; Munk et al. 2004). In addition to these developments, recognising that long-term investors usually have such short -term constraints as maximum-drawdown limits, or a particular wealth requirement , leads to further extensions of the model. Minimum performance constraints were first introduced in the context of constant proportion portfolio insurance (CPP I) (Black and Jones 1987; Black and Perold 1992), and in the context of option-b ased portfolio insurance (OBPI) (Leland 1980). More recent papers (Grossman and Zhou 1996) demonstrate that both of these strategies can be optimal for some inv estors and subsequent papers generalise the model by imposing minimum performance constrain ts relative to a stochastic, as opposed to a deterministic, benchmark. Tepl (2001 ), for example, demonstrates that the optimal strategy in the presence of such c onstraints involves a long position in an exchange option.4 The much more genera l and flexible dynamic portfolio theory leads to new insight into risk managemen t in general and the role of diversification. In this framework, diversification provides access to performance through a building block known as a performanceseeking portfolio (PSP). Downside risk control is achieved by assigning state-de pendentand possibly dynamicweights to the PSP and to a portfolio of safe, or riskfree, assets. In fact, since the latest financial crisis, there has been confusi on among market participants not only about the benefits and limitations of dive rsification as a method for risk management but also about how the methods of he dging and insurance are related to diversification. In this paper, our goal is t o review diversification and clarify its purpose. Going back to the conceptual u nderpinnings of several risk management strategies, we see that, in a dynamic as set management framework, diversification, hedging, and insurance are complement ary rather than competing techniques for sound risk management. The paper is org anised in two parts. The first discusses the benefits and limits of diversificat ion. The second moves on to hedging and insurance and discusses diversification as a method of reducing the cost of insurance. 3 - In fact, extensions of the dynamic portfolio theory concern asset/liability management, but the liability side is beyond the scope of this paper. 4 - See al so Martellini and Milhau (2010) and the references therein for additional detail s. An EDHEC-Risk Institute Publication 9

A Post-Crisis Perspective on Diversification for Risk Management May 2011 Introduction 10 An EDHEC-Risk Institute Publication

1. Advantages and Disadvantages of Diversification An EDHEC-Risk Institute Publication 11

A Post-Crisis Perspective on Diversification for Risk Management May 2011 1. Advantages and Disadvantages of Diversification Diversification and mean-variance analysis Diversification is one of the most widely used general concepts in modern financ e. The principle can be traced back to ancient times, but as far as portfolio co nstruction is concerned, the old saw about not putting all your eggs in one bask et captures the essence of the approach on a more abstract levelreduce portfolio concentration to improve its risk/return profile. Portfolio concentration can be reduced in a number of different ways, from ad hoc methods such as applying equ al weights to methods based on solid scientific arguments. A landmark publicatio n by Markowitz (1952) laid the foundations for a scientific approach to optimal distribution of capital in a set of risky assets. The paper introduced mean-vari ance analysis and demonstrated that diversification can be achieved through a po rtfolio construction technique that can be described in two alternative ways: (i ) maximise portfolio expected return for a given target for variance or (ii) min imise variance for a given target for expected return. The portfolios obtained i n this fashion are called efficient and the collection of those portfolios in th e mean-variance space is called the efficient frontier. Therefore, conceptually, the mean-variance analysis links diversification with the notion of efficiencyop timal diversification is achieved along the efficient frontier. The principles b ehind the Markowitz model can be formalised in the following optimisation proble m Eq. 1 12 An EDHEC-Risk Institute Publication where is the covariance matrix of stock returns, w = (w1,,wn) is the vector of po rtfolio weights, is a vector of expected returns, m is the target portfolio retu rn, and e = (1,,1). The objective function is in fact portfolio variance, the fir st constraint states that portfolio weights should add up to one and the second constraint sets the portfolio return target. The optimisation problem in Eq. 1 i mplies that there are three important inputsthe standalone characteristics repres ented by the vector of expected returns and the variance of stock returns positi oned on the main diagonal of the covariance matrix, as well as the joint behavio ur of stock returns represented by the covariance collected in the off-diagonal elements of . The last input leads to a very important insight indicating that jo int behaviour is crucial to the notion of efficient portfolios; it explains why diversification works.5 In fact, one limitation of the method can be identified by recognising that diversification is less effective when asset returns are mor e highly correlated. This conclusion follows from the decomposition of portfolio variance into two terms Eq. 2 5 - If joint behaviour were unimportant, investing 100% of the capital in the le ast risky stock would always represent the least risky portfolio. where is the corresponding correlation coefficient. The second term is the contr ibution of correlation to total portfolio variance. If ij is close to 1 fo all a ssets, then the e is a single facto d iving the etu ns of all assets. The efo e, dist ibuting capital among many assets is just as effective as investing in o ne asset

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 1. Advantages and Disadvantages of Dive sification only. Mo e fo mally, if all co elations a e exactly equal to 1, total po tfolio va iance can be ep esented as of the S&P 500 f om the beginning of 2000 to 201 0. The ave age co elation inc eased a ound the dot-com bubble and the 9/11 atta cks and in the financial meltdown of 2008. Figu e 1: The ave age co elation of the secto s in the S&P 500 calculated ove a two-yea olling window meaning that without a etu n ta get the optimal solution to Eq. 1 is a 100% all ocation to the least isky asset. In this situation, dive sification is ineffect ive since the optimal solution is a totally concent ated po tfolio.6 F om an inv esto pe spective, solving the p oblem in Eq. 1 means optimising the isk/ etu n t adeoff because isk is minimised conditional on a etu n ta get. As a esult , dive sification as a gene al method is not only about isk eduction. In fact, assuming the opposite would imply that the most dive sified po tfolio is the gl obal minimum va iance (GMV) po tfolio, which is obtained by d opping the second const aint in Eq. 1. This statement is a guable, howeve , as GMV po tfolios can be concent ated on the elatively lowe -volatility stocks, which also implies co ncent ation in such secto s as utilities.7 In fact, building well-dive sified po tfolios is mo e about efficient ext action of isk p emia than about me e isk minimisation. This conclusion, howeve , assumes that dive sification is designed to wo k ove the long un ac oss diffe ent ma ket conditions. Along with the in fluence of co elation on dive sification oppo tunities, this assumption is anot he d awback of the app oach. In a ma ket c ash, fo example, asset etu ns beco me highly co elated and the sho tcomings of dive sification a e highlighted. Th is empi ical esult is illust ated in figu e 1, in which we show the ave age co elation of the secto indices

In these conditions, as illust ated in figu e 2, in which we compa e the in-samp le pe fo mance of two optimised st ategies the maximum Sha pe atio (MSR) and the GMV po tfoliosto that of the equally weighted (EW) po tfolio and the cap-weighte d S&P 500, dive sification is unhelpful. In all cases, the unive se consists of the secto indices of the S&P 500. The plot shows that all st ategies, even the optimised ones, post la ge losses du ing the c ash of 2008. These losses a e ef lected in table 1, which shows the maximum-d awdown statistics fo the st ategie s in the pe iod between Janua y 2007 and Septembe 2010. Table 1: The maximum d awdown expe ienced by the st ategies in figu e 2 between Janua y 2007 and Septembe 2010 St ategy MSR GMV EW S&P 500 Max d awdown 24.33% 24.45% 49.43% 52.56% An EDHEC-Risk Institute Publication 13

6 - We assume that the po tfolio is long-only. If unconst ained sho ting is allo wed, then it is possible to const uct a ze o-volatility po tfolio f om any pai of pe fectly positively co elated assets having diffe ent volatilities. Since isk can be hedged completely using only two assets, it follows that the e is no point in building a dive sified po tfolio unde these assumptions as well. 7 - S ee appendix 1 fo a theo etical ema k on the st uctu e of GMV po tfolios.

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 1. Advantages and Disadvantages of Dive sification Figu e 2: Even though optimised po tfolios such as MSR and GMV a e well dive sif ied, they suffe ed la ge losses du ing the 2008 c isis. Fo compa ison, the EW p o tfolio and the cap-weighted S&P 500 a e also shown. dist ibuted o if investo s have quad atic utility functions; both of which assu mptions a e ove ly simplistic. Empi ical esea ch has fi mly established thatespe cially at high f equenciesasset etu ns can be skewed, leptoku tic, and fat-taile d and quad atic utility functions a ise in the model as a second-o de Taylo se ies app oximation of a gene al utility function. Using va iance as a p oxy fo isk is also cont ove sial. A disadvantage often pointed out is that it penalise s losses and p ofits symmet ically while isk is an asymmet ic phenomenon associ ated mo e with the left tail of the etu n dist ibution. The efo e, a ealistic isk measu e would be mo e sensitive to the downside than to the upside of the etu n dist ibution. At a given confidence level , V lue- t-Risk (V R), downside risk me sure widely used in the industry, is implicitly defined s threshold loss such th t the portfolio loses more th n V R with prob bility equ l to 1 m inus the confidence level,

where X is r ndom v ri ble describing the portfolio return distribution. Since diversific tion s concept goes beyond me n-v ri nce n lysis, it h s been r gued th t f ilure in m rket cr shes is c used m inly by the in ppropri te ssump tions m de by the M rkowitz model. If downside risk me sure is used inste d of v ri nce, the portfolio m y perform better during severe cr shes. Which downsid e risk me sure is ppropri te, however, is not cle r nd V R is h rdly the only ltern tive. Although different w ys of me suring risk h ve been discussed since the 1960s,

Diversific tion nd gener l ltern tive risk models Even though diversific tion is generic concept, we use me n-v ri nce n lysis to exemplify its dv nt ges nd dis dv nt ges. Me n-v ri nce n lysis is b sed o n the ssumption th t risk- verse investors m ximise their expected utility t t he investment horizon nd t ke into ccount only two distribution l ch r cterist ics me n nd v ri nce. This ssumption is re listic either if sset returns re n orm lly 14 An EDHEC-Risk Institute Public tion

There re, however, good re sons for the f ilure of diversific tion to reduce lo sses in sh rp m rket downturns. Incre sed correl tion, common in downturns, limi ts diversific tion opportunities. Perh ps more import ntly, diversific tion is d esigned to extr ct risk premi in n efficient w y over long horizons, not to co ntrol losses over short horizons. Misunderst nding the limit tions of the ppro ch c n misle d investors into concluding th t, since diversific tion did not pro tect them from big losses in 2008, it is useless concept.

9 - See Stoy nov et

l. (2011) nd the references therein.

A Post-Crisis Perspective on Diversific tion for Risk M n gement M y 2011 1. Adv nt ges nd Dis dv nt ges of Diversific tion n xiom tic ppro ch w s t ken in the 1990s10 with the development of firm-wide risk me surement systems. The first xiom tic construction w s th t of coherent risk me sures by Artzner et l. (1998). The xiom th t gu r ntees th t diversif ic tion opportunities would be recognised by ny coherent risk me sure is th t o f sub- dditivity, Eq. 3 where denotes the measu e of isk and X and Y a e andom va iables desc ibing the etu ns of two assets, i.e., the isk of a po tfolio o f assets is less than o equal to the sum of the isks of the assets. It is poss ible to efo mulate the po tfolio selection p oblem in Eq. 1 with any isk measu e satisfying Eq. 3 in the objective function; that is, instead of minimising va iance, we can minimise a sub-additive isk measu e subject to the same const ai nts. An axiomatic app oach, howeve , implies that the e could be many isk measu es satisfying the axioms, and sub-additivity axiom in pa ticula . As a conseque nce, the choice of a pa ticula isk measu e fo the po tfolio const uction p ob lem becomes difficult and must be made on the basis of additional a guments. Sta nda d deviation, fo example, satisfies the sub-additivity axiom. This conclusio n is appa ent f om equation Eq. 2the second te m, which involves the co elations , is the eason sub-additivity holds. VaR is gene ally not sub-additive, but it is obust, easy to inte p et, and equi ed by legislation and, as a consequence, it is widely used. Fu the mo e, ecent esea ch11 indicates that sub-additivity holds when the confidence level is high enough and the etu ns a e fat-tailed. A isk measu e suggested as a mo e info mative, cohe ent (and the efo e sub-addi tive) alte native to VaR is Conditional Value-at-Risk (CVaR). It measu es the av e age loss as long as the loss is la ge than the co esponding VaR. We a e inte ested in whethe o not adopting a downside isk measu e esults in d amaticall y diffe ent pe fo mance in ma ket c ashes. Figu e 3: The in-sample pe fo mance of GM CVAR and GM VaR po tfolios, both isk measu es at the 95% confidence level, du ing the c ash of 2008, togethe fo compa isonwith the cap-weighted S&P 500 10 - Ma kowitz (1959) suggested semi-va iance as a bette alte native to va ianc e as a p oxy fo isk, as it conce ns only adve se deviations f om the mean. 11 - See Danielsson et al. (2010). Although using a downside isk measu e may help fine-tune the benefits of dive s ification, it clea ly does not help much in seve e ma ket downtu ns. Figu e 3 an d table 2 p ovide an illust ation fo the pe iod f om Janua y 2007 to Septembe 2010, the same pe iod as that in figu e 2. Since the point of this illust ation is to compa e esults in times of la ge ma ket downtu ns, we limit the compa iso n to this time pe iod only. Table 2: The maximum d awdown expe ienced by the st ategies in figu e 3 between Janua y 2007 and Septembe 2010. St ategy GM CVAR GM VaR Max d awdown 22.92% 29. 15% An EDHEC-Risk Institute Publication 15

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 1. Advantages and Disadvantages of Dive sification Holding eve ything else equal, we conside CVaR and VaR alte native isk measu e s at a standa d confidence level of 95% fo both. Figu e 3 shows the values of t he global minimum CVaR (GM CVaR) and the global minimum VaR (GM VaR) po tfolios th ough time and table 2 shows the co esponding maximum-d awdown statistics. Th e losses in table 2 a e significant, though the GM CVaR po tfolio leads to d awd own ma ginally lowe that that of the GMV po tfolio (see table 1). That table 2 shows no significant eduction in d awdown is unsu p ising. By building the GM V aR po tfolio, we a e actually minimising the loss occu ing with a given p obabi lity (5% in the example in the example in figu e 3). The e is no gua antee that la ge losses will not be obse ved. Likewise, by building the GM CVaR po tfolio, we a e minimising an ave age of the ext eme losses. Again, having a small ave ag e ext eme loss does not necessa ily imply an absence of la ge losses in ma ket c ashes. In fact, it is possible to make a mo e gene al statement that is indepen dent of the choice of isk measu e. In the p evious section, we a gue that dive sification oppo tunities disappea when the co elation of asset etu ns is clos e to 1. Leaving the multiva iate no mal wo ld complicates the analysis, but it i s possible to demonst ate12 that dive sification oppo tunities disappea if asse t etu ns become comonotonic (inc easing functions of each othe ), which co esp onds to pe fect linea dependence in the Ma kowitz f amewo k.13 In Eq. 3 the joi nt dist ibution of X and Y can be any; the p ope ty is assumed to hold fo all p ossible multiva iate dist ibutions and, 16 An EDHEC-Risk Institute Publication

12 - See, fo example, Ekeland et al. (2009) and Rschendo f (2010). 13 - Comonoto nicity is in fact a cha acte istic of the uppe F chet-Hoeffding bound of any mul tiva iate dist ibution. Since in this analysis we hold the ma ginal dist ibution s fixed, it follows that the comonotonic behaviou is a p ope ty of the dependen ce st uctu e of the andom vecto , o the so-called copula function. As a conseq uence, the p esence of dive sification oppo tunities is a copula p ope ty. This statement is in line with the conclusion that dive sification oppo tunities a e a function of co elations in the Ma kowitz f amewo k since the copula function in the multiva iate Gaussian wo ld is uniquely dete mined by the co elation mat ix. 14 - We need the technical condition sup(X,Y) (X + Y) = (X) + (Y) whe e the su p emum is calculated ove all biva iate dist ibutions (X,Y) with fixed ma ginals . This condition is int oduced as a sepa ate axiom in Ekeland et al. (2009). See appendix 2 fo additional details. by design, fo all cohe ent isk measu es. As a esult, the dependence st uctu e of the asset etu ns dete mines the p esence of dive sification oppo tunities, whe eas the function of the isk measu e is to identify them and t ansfo m them into actual allocations.14 Fo the wo st possible dependence st uctu e, which is that of functional dependence, the inequality in Eq. 3 tu ns into an equality, which means that it is not possible to find a po tfolio whose isk is smalle th an the weighted ave age of the standalone isks. Intuitively, unde these ci cum stances, a 10% d op in one of the assets dete mines exactly the changes in the o the assets, since they a e inc easing functions of each othe . In a situation s uch as this one, holding a b oadly dive sified po tfolio is just as good as hold ing only a few assets. As a consequence, we can a gue that gene alising the mean -va iance f amewo k leads to the conclusion that, if secu ities a e nea ly funct ionally dependent in ma ket c ashes, then the e a e no dive sification oppo tuni ties. Unde these conditions, choosing a isk measu e is edundant because the a gument is gene ic (see appendix 2 fo additional details). Statistical a gument s p ovide evidence fo this conclusion as well. Figu es 2 and 3 show the in-samp le pe fo mance of the optimised st ategies. In this calculation, we assume pe fe ct knowledge of the mean and va iance in the Ma kowitz analysis and pe fect know ledge of the multiva iate dist ibution fo the GM CVaR and GM VaR examples. Yet

in these pe fect conditions, none of the optimised st ategies is able to p ovide easonable loss p otection in 2008. In eality, the optimal solutions would be

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 1. Advantages and Disadvantages of Dive sification influenced by the noise coming f om ou impe fect knowledge of these pa amete s, suggesting that the esults may be even wo se. Howeve , ou esults with pe fec t pa amete knowledge show that attempts to imp ove the pa amete estimato s, o the model fo the multiva iate dist ibution, will be of little help in educing the d awdown of optimally dive sified po tfolios in seve e ma ket c ashes. The additional info mation, howeve , comes at a cost. The coskewness and coku tosis pa amete s inc ease significantly the total numbe of pa amete s that need to be estimated f om histo ical data. Thus, a po tfolio of 100 assets would equi e e stimation of mo e than 4.5 million pa amete s. Compa ed to accounting fo highe -o de moments when coskewness and coku tosis pa amete s a e estimated without p ope ly handling estimation isk, a simple mean-va iance app oach thus tends to lead to bette out-of-sample esults since it avoids the e o -p one estimation of highe -o de dependencies. Neve theless, Ma tellini and Ziemann (2010) demons t ate that, fo lowe -dimensional p oblems, if the pa amete estimation p oblem is p ope ly handled, including highe -o de comoments adds value to the po tfoli o selection p oblem and can lead to highe isk-adjusted etu ns, indicating tha t it p ovides access to additional dive sification oppo tunities. As fo p otect ion f om losses in ext eme ma ket conditions, howeve , this app oach is no mo e helpful than any of the othe s discussed in the p evious sections.

Dive sification and highe -o de comoments Anothe way to extend the f amewo k beyond the mean-va iance analysis is to cons ide highe -o de Taylo se ies app oximations of investo s utility function. The highe -o de app oximation esults in highe -o de moments in the objective fun ction of the po tfolio optimisation p oblem given in Eq. 1 (Ma tellini and Ziema nn 2010). Using the fou th-o de app oximation, fo example, means inco po ating po tfolio skewness and ku tosis in addition to po tfolio va iance. In this way, the objective function becomes mo e ealistic in the sense that it takes into a ccount the empi ical facts that asset etu ns a e asymmet ic and exhibit excess ku tosis. This p oblem setup makes it possible to identify dive sification oppo tunities othe than those available in the co elation mat ix because po tfolio skewness and ku tosis depend on the coskewness and coku tosis of asset etu ns t hat ep esent statistical measu es of dependence of the asymmet ies and the peak edness of the stock etu n dist ibutions. The coskewness and coku tosis appea i n addition to cova iance and desc ibe othe aspects of the joint behaviou . An EDHEC-Risk Institute Publication 17

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 1. Advantages and Disadvantages of Dive sification 18 An EDHEC-Risk Institute Publication

2. Beyond Dive sification: Hedging and Insu ance An EDHEC-Risk Institute Publication 19

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 2. Beyond Dive sification: Hedging and Insu ance The discussion in the p evious section illust ates the benefit of dive sificatio n, which is to ext act isk p emia, and two key sho tcomings: (i) it is un eliab le in highly co elated ma kets and (ii) it is not an efficient technique of los s cont ol in the sho t te m. Complaints that dive sification has failed a e some what misleading, as it was neve meant to p ovide downside p otection in ma ket c ashes. F om a p actical viewpoint, it is impo tant to t anscend dive sificatio n and to identify techniques that can complement it and offset its sho tcomings. One potential technique is hedging, gene ally used to offset pa tially o compl etely a specific isk. Hedging can be done in a va iety of ways; the best exampl e, pe haps, is th ough a position in futu es. Suppose that a given po tfolio has a long exposu e to the p ice of oil, a isk the po tfolio manage is unwilling to take ove a given ho izon. One possibility is to ente into a sho t position in an oil futu es cont act. If the po tfolio has an undesi able long exposu e to a given secto (financials, say), anothe hedging st ategy is to sho t sell the co esponding secto index. Depending on the ci cumstances, the hedge can be pe fect, if the co esponding isk is completely emoved, o impe fect (pa tial), leading to some esidual exposu e. In the following section, we discuss the adva ntages and disadvantages of combining hedging and dive sification. The limitatio ns of this combination stem la gely f om the static natu e of hedging. Insu ance , which is dynamic in natu eand the second topic of this sectioncan be used to ove come these limitations. Hedging: fund sepa ation and isk eduction The mean-va iance f amewo k int oduced by Ma kowitz (1952) does not conside a isk-f ee asset; the investable unive se consists of isky assets only. Tobin (19 58), howeve , a gued that, in the p esence of a isk-f ee asset, investo s shoul d hold po tfolios of only two fundsthe isk-f ee asset and a fund of isky assets . The fund of isky assets is the maximum Sha pe atio (MSR) po tfolio const uct ed f om the isky assets. Fu the mo e, the isk ave sion of investo s does not c hange the st uctu e of the efficient MSR fund; it affects only the elative weig hts of the two funds in the po tfolio. This a angement is the esult of a so-ca lled two-fund sepa ation theo em, which posits that any isk-ave se investo can const uct po tfolios in two steps: (i) build the MSR po tfolio f om the isky a ssets and (ii) depending on the deg ee of isk-ave sion, hedge pa tially the is k p esent in the MSR po tfolio by allocating a f action of the capital to the i sk-f ee asset. Figu e 4: The in-sample efficient f ontie of the isky assets (in blue) and the CML (capital ma ket line) togethe with the tangency po tfolio, the GMV po tfol io, and the po tfolio with the same isk as the GMV on the CML. The annualised isk-f ee ate is set to 2%.15 15 - This ate is used in all calculations unless stated othe wise. F om a geomet ic pe spective, adding a isk-f ee asset to the investable unive s e esults in a linea efficient f ontie called the 20 An EDHEC-Risk Institute Publication

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 2. Beyond Dive sification: Hedging and Insu ance capital ma ket line (CML), a line tangential to the efficient f ontie gene ated by the isky assets. Since the point of tangency is the MSR po tfolio, it is al so known as the tangency po tfolio. Figu e 4 illust ates the geomet ic p ope ty. 16 Int oducing a isk-f ee asset and pa tial hedging as a technique fo isk ed uction aises the following question. Fo a given isk const aint, which po tfol io const uction technique is bette ? Taking advantage of dive sification, maximi sing expected etu n subject to the isk const aint and choosing the po tfolio o n the efficient f ontie of the isky assets, o taking advantage of the fund-se pa ation theo em and, instead of building a customised po tfolio of isky assets , pa tially hedging the isk of the MSR po tfolio with the isk-f ee asset to me et the isk const aint? F om a theo etical pe spective, the second app oach is s upe io because the isk-adjusted etu n of all po tfolios on the CML is not sma lle than those on the efficient f ontie of the isky assets. Figu e 5: The pe fo mance and the dynamics of the maximum d awdown of the GMV po tfolio and the GMV match on the capital ma ket line with the same isk on the CML. The in-sample pe fo mance of the two po tfolios i s shown in figu e 5. Both po tfolios a e equally isky in te ms of volatility bu t the one on the CML pe fo ms bette . The components of the po tfolio account fo its bette isk/ etu n t adeoff. The efficient MSR po tfolio is const ucted to p ovide the highest possible isk-adjusted etu n. The efo e, it is in the cons t uction of this po tfolio that we take advantage of dive sification to ext act p emia f om the isky assets. The MSR po tfolio is in fact esponsible fo the p e fo mance of the ove all st ategy. The isk-f ee asset, by cont ast, is the e t o hedge isk. In fact, the fund-sepa ation theo em implies that the e is also a functional sepa ationthe two funds in the po tfolio a e esponsible fo diffe ent functions. Although volatility is kept unde cont ol, both the GMV po tfolio an d the GMV match on the CML (see figu e 4) post heavy losses in the c ash of 2008 . Unlike dive sification, howeve , hedging can be used to cont ol ext eme losses . In theo y, the isk-f ee asset has unive sal hedging p ope ties. If the po tfo lio is allocated enti ely to the isk-f ee asset, then, in theo y, it g ows at t he isk-f ee ate. App op iate allocation to the isk-f ee asset can thus hedge pa tially all aspects of isk a ising f om the unce tainty in the isky assets. We can easily, fo example, const uct a po tfolio on the CML with an in-sample m aximum d awdown of no mo e than 10%. Fo ou dataset, it tu ns out that a po tfo lio with this p ope ty is obtained with a 40% allocation to the MSR po tfolio. E xplicit loss cont ol of this type is not possible if the investo elies only on dive sification. 16 - The isky assets gene ating the efficient f ontie on the plot a e the sect o indices of the S&P 500. We conside the ten-yea pe iod f om 2000 to 2010. Th e weights in the optimisation p oblem a e between -40% and 40%. The isk-f ee as set is assumed to yield an annual etu n of 2%, a etu n ep esentative of the a ve age th ee-month T easu y bill ate f om 2000 to 2010. To check this conclusion in p actice, we choose the GMV po tfolio on the efficie nt f ontie of the isky assets and the po tfolio An EDHEC-Risk Institute Publication 21

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 2. Beyond Dive sification: Hedging and Insu ance Figu e 6: The in-sample pe fo mance of the GMV (in g een), the GMV match on the CML (in blue), and a po tfolio on the CML const ucted such that it has a maximum d awdown of 10% (in ed) symmet ically the ight tail of the etu n dist ibution. As a consequence, this app oach can lead to limited d awdown but at the cost of lowe upside potential. Insu ance: dynamic isk management In the p evious example, the eason fo the lowe upside potential is the fact t hat hedging is a static technique. The enti e analysis takes place in a single i nstance and the optimal po tfolio is, essentially, a buy-and-hold st ategy. As a consequence, the weight of the MSR does not depend on time o on the state of t he ma ket. Ideally, investo s would demand an imp oved downside and an imp oved upside at the same time. This, howeve , is not feasible with a static technique. Simple fo ms of dynamic isk management, also called po tfolio insu ance, we e suggested in the late 1980s. Black and Jones (1987) and Black and Pe old (1992) we e the fi st to suggest constant p opo tion po tfolio insu ance (CPPI). This s t ategy is a dynamic t ading ule that allocates capital to a isky asset and ca sh in p opo tion to a cushion that is the diffe ence between the cu ent po tfol io value and a selected p otective floo . The esulting payoff at the ho izon is optionlike because the exposu e to the isky asset app oaches ze o if the value of the po tfolio app oaches the floo . The ove all effect is simila to that of owning a put optionCPPI gua antees that the floo will not be b eached. Anothe popula insu ance st ategy is optionbased po tfolio insu ance (OBPI) (G ossman a nd Vila 1989). This st ategy basically consists of buying a de ivative inst umen t so that the left tail of the payoff dist ibution at the ho izon is t uncated a t a desi ed th eshold.

A compa ison of the pe fo mance of th ee po tfoliosthe GMV po tfolio, the GMV mat ch on the CML, and a po tfolio on the CML with an in-sample maximum d awdown of 10% is shown in figu e 6. Hedging makes it possible to match in-sample any maximu m d awdown, ega dless of its size. Since the po tfolio etu n dist ibution is a weighted ave age of the etu n dist ibution of the MSR po tfolio and a constant , Eq. 4 whe e 0 v 1 is the weight of the MSR po tfolio and the isk-f ee ate, it follows that by changing v the po tfolio etu n dist ibution is scaled up o down. Using Chebychevs inequality, it is possible to demonst ate that the p obabi lity of la ge losses can be made infinitely small by educing v, in which is the va iance of the MSR po tfolio. Even though this app oach is capa ble of cont olling the downside of the etu n dist ibution,17 the e is a caveat. Along with the left tail, scaling influences 22 An EDHEC-Risk Institute Publication

17 - In the pa ticula case of the dataset used fo figu e 6, v = 0.4 the po tfolio with a 10% in-sample maximum d awdown.

esults in

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 2. Beyond Dive sification: Hedging and Insu ance The de ivative inst ument can be a simple Eu opean call option o an exotic p od uct depending on additional path-wise featu es we would like to enginee . Even t hough CPPI and OBPI a e conceptually simple, they seem to be based on sepa ate t echniques athe than on a mo e basic f amewo k. Neve theless, since the option can, in theo y, be eplicated dynamically, both CPPI and OBPI can be viewed as m embe s of a single family of models. In fact, a much mo e gene al extension is v alid. The dynamic po tfolio theo y developed by Me ton (1969, 1971) can be exten ded with absolute o elative const aints on asset value and it is possible to s how that both CPPI and OBPI a ise as optimal st ategies fo investo s subject to pa ticula const aints (Basak 1995, 2002). The t eatment of the const aints in continuous-time dynamic po tfolio theo y is gene ic; they a e int oduced in te m s of a gene al floo . The floo s can be absolute o elative to a benchma k po t folio. An absolute floo , fo instance, can be any of the following: A capital-g ua antee floo . The floo is calculated by the fo mula A olling-pe fo mance flo o . This floo is defined by whe e t* is a p edefined pe iod of time, twelve months, fo example. The olling pe fo mance floo gua antees that the pe fo mance will stay positive ove pe iod t*. A maximum-d awdown floo . A d awdown const aint is implemented by 18 - See Amenc et al. (2010b) fo additional info mation in the context of the d ynamic co e-satellite app oach. whe e is positive p r meter less th n 1 nd At portfolio we lth t time t. A m ximum-dr wdown floor implies th t the v lue of the portfolio never f lls below cert in percent ge, 100(1 )%, of the m ximum v lue tt ined in the p st. This c onstr int w s initi lly suggested s n bsolute constr int but c n be reformul ted s rel tive one.18 A rel tive-benchm rk floor. This rel tive floor is defi ned by where r is the risk-free r te, T-t c lcul tes the time to horizon, A0 is the init i l portfolio we lth, nd k < 1 is positive multiplier. This floor is usu lly used in CPPI nd non-viol tion of this floor gu r ntees th t the str tegy will p rovide the initi l c pit l t the horizon. where k < 1 is positive multiplier nd Bt is the v lue of benchm rk t time t. This floor gu r ntees th t the v lue of the portfolio will st y bove 100k% o f the v lue of the benchm rk. Sever l floors c n be combined together in singl e floor by c lcul ting their m ximum, . The new floor c n then be dopted s s ingle floor in the dyn mic portfolio optimis tion problem. It follows from the d efinition th t if is not viol ted, then none of the other floors will be, either . An EDHEC-Risk Institute Public tion 23

A Post-Crisis Perspective on Diversific tion for Risk M n gement M y 2011 2. Beyond Diversific tion: Hedging nd Insur nce Solving dyn mic sset lloc tion problem with n implicit floor constr int res ults in n optim l lloc tion of the following form, Eq. 5 weights of the buildi ng blocks. In Eq. 4, the weights re st tic, where s in Eq. 5 they re st te- n d, potenti lly, time-dependent. This is the improvement th t m kes insur nce n dequ te gener l ppro ch to downside risk m n gement. Figure 7: The in-s mple perform nce of the 10% m ximumdr wdown str tegy on the C ML nd dyn mic str tegy with 10% m ximum-dr wdown constr int where PSP is the generic not tion for the weights of perform nce-seeking portf olio, SAFE the weights in the s fe ssets, the de ree of risk aversion, Ft the v alue of the selected floor at time t, and the value of the optimal constrained p ortfolio (see appendix 3 for additional details). The solution in Eq. 5 is a fun d-separation theorem in a dynamic asset allocation settin . The optimal wei ht e quals a wei hted avera e of two buildin blocks constructed for different purpos es. The PSP is constructed for access to performance throu h efficient extractio n of risk premia; in fact, under fairly eneral assumptions it is the MSR portfo lio. The eneral oal of the SAFE buildin block is to hed e liabilities. In the very simple example of the previous section, SAFE consists of a overnment bond maturin at the investment horizon. In a dynamic settin , dependin on the inst itution constructin the strate y, SAFE has a different structure. For example, critical factors for pension funds are interest rates and inflation. As a result , the SAFE portfolio for a pension fund would contain assets hed in interest ra te risk and inflation risk (see appendix 3 for additional details). Even thou h Eq. 5 is much more eneral than Eq. 4, considerin only the buildin blocks, the reatest difference is in the 24 An EDHEC-Risk Institute Publication

An illustration of the improvement of insurance strate ies on hed in is provide d in fi ure 7. In the upper part of the fi ure, we compare the in-sample perform ance of the 10% maximum-drawdown strate y obtained throu h the static methods of hed in 19 and a dynamic strate y20 with a maximum-drawdown floor of 10%. The lo wer part of the fi ure shows a plot of the dynamics of the allocation to the MSR portfolio and illustrates how insurance strate ies control downside losses. Whe n there is a market downturn and the value of the portfolio approaches the floor , the allocation to the PSP buildin block, or the MSR portfolio in this case, d ecreases. When the value of the portfolio hits the floor, as it nearly does in t he crash of 2008 (see fi ure 7), allocation to the risky MSR portfolio stops alt o ether and the portfolio is totally invested in the SAFE buildin block. Since the SAFE asset is supposed to carry no risk, it is not possible, in theory, to b reach

19 amic lier c et

The same portfolio is represented by the red line in fi ure 6. 20 - The dyn portfolio is implemented as a dynamic core-satellite strate y with a multip of six and a risk-free instrument yieldin an annual return of 2%. See Amen al. (2010b) for further details on core-satellite investin .

A Post-Crisis Perspective on Diversification for Risk Mana ement May 2011 2. Beyond Diversification: Hed in and Insurance the floor.21 In a recovery, the return from the safe asset can be used to build up a new cushion and invest a ain in the MSR portfolio. In this way, exposure to extreme risks is limited and access to the upside is preserved throu h the MSR portfolio because it is desi ned to extract premia from risky assets by takin f ull advanta e of the method of diversification. Table 3: The maximum and the avera e drawdown of the two strate ies in fi ure 7 between January 2007 and September 2010 Strate y Dynamic strate y Static strate y Avera e drawdown 2.9% 2% Max drawdown 9.2% 10% 21 - In a practical implementation, a breach of the floor may occur because, as a result of turnover constraints, tradin may need to be less frequent, which ca n result in a breach occurrin between rebalancin dates, or because a perfect h ed e with the SAFE portfolio may not be possible as a result of market incomplet eness, which implies that there may be residual risks in the portfolio. Neverthe less, dynamic asset allocation is the ri ht eneral approach to controllin down side risks. The drawdown characteristics of the two strate ies are shown in table 3. To all appearances, they both exhibit similar in-sample avera e and maximum drawdown. T he dynamic strate y, however, has reater upside potential, a result of the desi n of the MSR portfolio. The difference in the properties of the static and the dynamic approaches are best illustrated in a Monte-Carlo study. Fi ure 7 compare s the performance of only two paths, but in practice we need more than two to a in insi ht into the difference in the extreme risk exposure of the two strate ie s. We fitted a eometric Brownian motion (GBM) to the MSR sample path and enera ted 5,000 paths with a ten-year horizon. For each path, which represents one sta te of the world in this settin , we calculated the dynamic strate y with a 10% m aximum drawdown. The static strate y is a fixed-mix portfolio with a 40% allocat ion to the MSR portfolio. Then, in each state of the world, we calculated the an nualised returns and the maximum drawdown of the two strate ies. The results are summarised in fi ure 8. The plot on the left shows the histo ram s of the annualised return distribution for the two strate ies superimposed. The blue histo ram indicates better access to the upside performance of the dynamic strate y. The plot on the ri ht shows the correspondin histo rams for the maxi mum-drawdown distribution. The reat difference stems from the inability of the static approach to keep losses under control. In some states of the world, the m aximum drawdown reaches more than 20%, even thou h the same static strate y was desi ned to have a 10% in-sample maximum drawdown. In contrast, there is no sin le state of the world in which the dynamic strate y has a maximum drawdown reat er than 10%. Fi ure 8: The annualised return distribution and the maximumdrawdown distributio n of the dynamic and the static strate ies calculated from 5,000 sample paths An EDHEC-Risk Institute Publication 25

A Post-Crisis Perspective on Diversification for Risk Mana ement May 2011 2. Beyond Diversification: Hed in and Insurance Table 4: The risk-return characteristics of the dynamic and the static strate ie s calculated from the distributions in fi ure 8 Strate y Annualised avera e retu rn 9.56% 8.26% Avera e max drawdown 8% 10.32% Lar est max drawdown 9.64% 28.3% We did a Monte-Carlo study to illustrate this effect on an insurance strate y wi th a maximum-drawdown constraint. The PSP buildin block is modelled as a GBM, E q. 6 where is the Sharpe ratio of the strategy. We adopt the parameter va ues ca ibrated in Munk et a . (2004)23 and have the Sharpe ratio be = 0.24, which corr esponds to the ong-term ratio for the S&P 500.24 We generated 5,000 samp e path s from the mode in Eq. 6 with an investment horizon of ten years. For each samp e path, we ca cu ated the dynamic insurance strategy and computed its average a nnua return, as we as the average annua return of the PSP component. Figure 9: The return distribution of a dynamic strategy compared to that of the PSP component. The top pair of p ots is produced with the defau t va ue of = 0.2 4 and the bottom pair of p ots is produced with = 0.36, which is a 50% improveme nt on the defau t va ue. SP is shortfa probabi itythe probabi ity that the annu a ised average return wi be negative. Dynamic strategy Static strategy The risk-return characteristics ca cu ated from the distributions shown in figur e 8 are shown in tab e 4. The annua ised average return of the dynamic strategy is higher than that of the static strategy, as expected, and the big difference in the maximumdrawdown distributions is apparent. The average maximum drawdown o f the static strategy is near the in-samp e figure of 10%. 22 - Diversification can invo ve the transaction costs arising from additiona t rading. 23 - The mode in Munk et a . (2004) is more genera as it a ows for a stochastic interest rate. The parameter va ues used in the simu ation are S = 14. 68% and r = 3.69%, the value for r being the long-term mean in the mean-rever ion model fitted by Munk et al. (2004). 24 - See Amenc et al. (2010a). Diver ification and the co t of in urance Diver ification can be implemented, at lea t in theory,22 at no co t, but in ura nce alway ha a co t. The co t of in urance i ea ie t to pot in the OBPI tra tegie in which a certain amount of capital i inve ted in a derivative in trume nt. In thi ca e, the co t i the price of the derivative. Since the derivative can u ually be replicated by a dynamic portfolio, it i clear that uch co t ca n be pre ent in other type of dynamic in urance trategie . In uch ca e , howe ver, they materiali e a implicit opportunity co t . One way to illu trate the c o t of in urance i to look at the return di tribution of the dynamic trategy a t the inve tment horizon and the corre ponding hi togram of the PSP building blo ck. The opportunity co t of in urance appear a a lower expected return for the dynamic trategy. 26 An EDHEC-Ri k In titute Publication

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 2. Beyond Diver ification: Hedging and In urance Although cap-weighted indice are popular in the indu try, there i ample empiri cal evidence that they are poorly diver ified and highly inefficient (Haugen and Baker 1991; Grinold 1992; Amenc et al. 2006). The rea on i that capitali ation weighting lead to high concentration in a handful of tock . In fact, equally weighted portfolio , although navely diver ified, have been found to provide high er ri k-adju ted return .25 Although it ha been hown that even navely diver ifi ed portfolio dominate the corre ponding cap-weighted portfolio , equal weightin g provide optimal diver ification from the tandpoint of mean-variance analy i if and only if all ecuritie have identical expected return , volatility, and if all pair of correlation are the ame. Since thi hypothe i i highly unreal i tic, there i a clear indication that, by carefully e timating the ri k and re turn parameter , it would be po ible to con truct ri k-efficient MSR portfolio providing uperior ri k-adju ted return . Succe ful implementation of an MSR p ortfolio i critically dependent on the quality of the parameter e timator . Ame nc et al. (2010a) do an empirical tudy for the S&P 500 univer e from January 19 59 to December 2008. They how that u ing parameter e timation technique re ult ing in robu t e timate of the ri k and the return parameter lead to an optimi ed trategy with a Sharpe ratio more than 50% higher than the Sharpe ratio of t he S&P 500 index. If improving diver ification make po ible a 50% improvement in the Sharpe ratio of the PSP, it i intere ting to ee to what An EDHEC-Ri k In titute Publication 25 - See, for example, De Miguel et al. (2009). The top pair of plot in figure 9 compare the two di tribution . The annuali ed expected return of the dynamic trategy i 6.72%, wherea that of the PSP build ing block i 8.46%. Although the difference in the annuali ed return di tributio n eem large on the plot, it mu t be kept in mind that drawdown protection re u lt in good path-wi e propertie that are hard to pot in the hi togram of the d ynamic trategy in figure 9. The good path-wi e propertie materiali e a a ign ificantly maller hortfall probability (SP). One way to off et the co t of in u rance i to improve the building block of the dynamic trategy. Since the PSP i devoted to performance, it mu t be con tructed a a well-diver ified portfolio . In practice, the common approach i to adopt a tandard tock market index, a cap-weighted portfolio. 27

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 2. Beyond Diver ification: Hedging and In urance degree it can off et the implicit co t of in urance. So we regenerated the cena rio from the model in Eq. 6, keeping the ame parameter value and increa ing t he Sharpe ratio to 0.36. The hi togram of the return di tribution of the dynam ic trategy before and after the Sharpe ratio improvement are compared in the bo ttom pair of plot in figure 9. The annuali ed expected return of the dynamic t rategy improve from 6.72% to 8.19%, a jump that, in thi context, implie that improving the Sharpe ratio of the PSP by 50% very nearly compen ate for the co t of in urance. 28 An EDHEC-Ri k In titute Publication

Conclu ion An EDHEC-Ri k In titute Publication 29

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 Conclu ion The global financial cri i of 2008 ha hifted the attention of all inve tor t o ri k management. In a broad context, ri k management i about maximi ing the p robability of achieving certain objective at the inve tment horizon while tayi ng within a ri k budget. Diver ification, hedging, and in urance can be relied o n to make optimal u e of ri k budget . The e three technique involve different a pect of ri k management, but they are complementary technique rather than co mpeting one . Diver ification provide inve tor with the be t reward per unit o f ri k through a mart combination of individual a et . It i de igned to work in the long run acro different market condition and i , therefore, helple i n uch pecific condition a evere market downturn . Since the purpo e of dive r ification i efficient extraction of ri k premia, it i mo t effective in the con truction of performance- eeking portfolio . Hedging can be combined with div er ification to reduce ri k that cannot be diver ified away. Hedging i achieve d through a portfolio of afe a et , or imply through ca h, which i another d edicated building block. A non-diver ifiable ri k that can be handled in thi wa y i the ri k of a large drawdown. In urance, unlike diver ification and hedging , combine the afe building block and the PSP optimally to comply with the corr e ponding ri k budget . So down ide ri k control i be t achieved through dynami c a et allocation. Thi technique make it po ible to control the down ide of the return di tribution while 30 An EDHEC-Ri k In titute Publication pre erving acce to the up ide through the PSP. In thi context, a well-diver i fied portfolio i a building block of crucial importance. A carefully de igned P SP with an improved Sharpe ratio re ulting from good diver ification can reduce the implicit co t of in urance.

Appendix An EDHEC-Ri k In titute Publication 31

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 Appendix Appendix 1: Are GMV Portfolio Generally Concentrated in Low-Volatility Stock ? If a non-degenerate covariance matrix i a umed, the optimi ation problem in Eq . 1 without the expected return con traint make po ible the following analytic olution, Eq. 7 where i the inver e of the covariance matrix. The factor in Eq . 7 en ure that the weight add up to one and, therefore, the optimal olution i proportional to the vector . Thi olution i , for all intent and purpo e , the GMV portfolio. Suppo e that return of the a et in the portfolio are not c orrelated. The inver e of the covariance matrix then ha the following very imp le tructure, to tilt the portfolio further toward the lower-volatility a et . When all a e t are perfectly po itively correlated, the GMV long-only portfolio i concentra ted entirely in the lowe t-volatility tock. A way to illu trate thi idea i to con ider, for the ake of implicity, a con tant correlation model and differen t range for the tock volatilitie . The ca e in which all tock have one and t he ame volatility i not of intere t becau e the GMV portfolio are trivial equa lly weighted for any correlation. The expectation i that the more we allow the tock volatilitie to vary, the tronger the departure from equal weighting will be. In thi illu tration, we choo e the inver e of the Herfindahl index a a me a ure of concentration which, in a vector notation, take the form, where jj i the variance of the return of the j-th a et. In thi ca e, the optim al olution a given in Eq. 7 i proportional to the inver e of the quared vola tilitie . A a re ult, the more diver e the volatilitie in the univer e are, th e more highly the GMV portfolio i concentrated in the lowervolatility a et . I t i expected that increa ing correlation from zero to a po itive number will l ead to even greater concentration in the lowervolatility a et ince, a a re u lt of ome common factor , it would be optimal 32 An EDHEC-Ri k In titute Publication where w i a vector of weight (e.g., the optimal olution in Eq. 7). For an equ ally weighted portfolio, H-1 = n, which tand for the number of tock in the u niver e, and for a portfolio totally concentrated in one a et, H-1 = 1. The con centration metric H-1 i between 1 and n for any other long-only portfolio. To p re erve thi interpretation, we calculate the GMV portfolio u ing the optimi at ion problem in Eq. 1 with the additional con traint that the weight hould be n on-negative. We con ider a hypothetical univer e of 100 tock , the annuali ed v olatilitie of which are equally paced in the following range : [16%, 18%], [15 %, 19%], [13%, 21%], and [10%, 24%]. In the three ca e , the average volatility i one and the ame; the only difference i the di per ion of the volatilitie a round the

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 Appendix average. On the e a umption , we calculate numerically the GMV portfolio for d egree of correlation ranging from 0 to 0.99 and we e timate the concentration m etric H-1 for each of them. Figure 10: The inver e of the Herfindahl index of long-only GMV portfolio a a function of correlation in a con tant correlation model. The volatilitie are eq ually paced between the lower and the upper bound , which are provided in annua li ed term . volatility tock increa e monotonically, an increa e that i off et by a monot onic decrea e of the total weight allocated to the group of higher-volatility t ock . For correlation more than 0.6, the portfolio i concentrated entirely in t he group of lower-volatility tock . Figure 11: The um of the GMV weight corre ponding to three group of tock of equal ize ranked by their volatility. The volatilitie of the tock are equal ly paced in the interval [16%, 18%]. The plot of H-1 a a function of correlation i provided in figure 10. A expect ed, the relation hip i monotonic in the four ca e and, all other thing equal, a wider range of volatility corre pond to greater concentration. The plot in f igure 10 indicate that concentration increa e , but it i not clear if it doe o a a re ult of the higher relative weight of the low-volatility tock . To ex plore thi que tion, we rank the GMV tock into three group of equal ize and then calculate the um of the weight of the tock in each group. Figure 11 ho w the um of the e weight a a function of correlation, a uming that the toc k volatilitie are in the interval [16%, 18%]. If the tock return are non-corr elated, the total weight allocated to the three group of tock i almo t the ame, which i con i tent with the value of H-1 being near 100 in figure 10. A c orrelation increa e , the total weight allocated to the lowerThi numerical illu tration corre pond to the fairly homogeneou ca e in which tock volatilitie do not vary much. The conclu ion would be tronger if we all owed a greater degree of non-homogeneity. Appendix 2: Diver ification Opportunitie Are Determined by the Multivariate Dep endence Structure of A et Return In thi appendix, we examine more clo ely the claim that the functional dependen ce of a et return leave no room for diver ification. The analy i i general in the en e that we do not limit the di cu ion to the mean-variance etting al one. Rather, we work with a general ri k mea ure, denoted by he e, making as few assumptions as possible while extending the Ma kowitz setting. An EDHEC-Risk Institute Publication 33

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 Appendix Suppose, to sta t, that the investable unive se consists of n assets, the joint behaviou of the etu ns of which is desc ibed by the andom vecto (X1,, Xn). We distinguish between ma ginal dist ibutions, desc ibing the elements of the and om vecto on a standalone basis, and the joint dist ibution, desc ibing the join t behaviou of all stocks. Let be a gene al isk measu e that is sub-additive (i .e., it satisfies Eq. 3) and positively homogeneous26 (aX) = a(X) fo positive a; assume that it satisfies the st uctu e neut ality condition (Ekeland et al. 2009 ), Eq. 8 26 - Technically, these two p ope ties gua antee convexity of . 27 - See, fo exa mple, the o iginal publications of Hoeffding (1940) and F chet (1951). 28 - See Rs chendo f (2010) and the efe ences the ein. Eq. 9 whe e U is unifo mly dist ibuted in [0, 1], o 2. The multiva iate dist ib ution function of Y can be ep esented as Eq. 10 The definition in Eq. 9 indicat es that the components of a comonotonic andom vecto can be ep esented as inc easing functions of each othe . Eq. 10 shows that the uppe F chet-Hoeffding boun d, Eq. 11 which is an uppe bound of any multiva iate dist ibution function, is the dist ibution function of a comonotonic andom vecto . Since the ma ginal dis t ibutions of the uppe bound a e the same, it follows that comonotonicity is, e ssentially, a p ope ty of the copula, o the dependence st uctu e, of the andom vecto . In fact, the wo st possible dependence st uctu e tu ns out to be that o f comonotonicity. It is possible to show that fo any convex isk measu e that s atisfies Eq. 8, the following inequality holds,28 whe e the sup emum is taken with espect to all possible joint dist ibutions; ho weve , the ma ginal dist ibutions a e fixedthat is, we va y only the dependence s t uctu e. This condition implies that the dependence st uctu e fo which the sup emum is attainedcall it the wo st possible dependence st uctu e leads to a linea decomposition of the isk measu e. If this condition does not hold, then by the sub-additivity p ope ty fo the wo st possible dependence st uctu e. Howeve , t he st ict inequality means that we must exclude volatility and the multiva iate no mal dist ibution as a possible setup. Additional motivation fo this conditio n can be found in Ekeland et al. (2009). To cha acte ise the wo st possible depe ndence st uctu e, we need the notion of comonotonicity. A andom vecto Y is sai d to be comonotonic if,27 1. The following ep esentation holds

That is, holding the ma ginal dist ibutions fixed, any isk measu e satisfying t he assumptions above has an absolute 34 An EDHEC-Risk Institute Publication

which can be estated in te ms of po tfolio

isk and standalone

isks as Eq. 12

A Post-C isis Pe spective on Dive sification fo Risk Management May 2011 Appendix maximum when stock etu ns a e inc easing functions of each othe . Fu the mo e, fo this dependence st uctu e the isk of any po tfolio equals the weighted ave age of the isks of the po tfolio constituents, which means that the isk measu e becomes a linea function of po tfolio weights. As a esult, unde the assumpt ion of comonotonicity, the extension of the mean-va iance analysis in Eq. 1 with a gene al isk measu e in the objective function tu ns into a linea p og ammin g p oblem. And assuming that no sho tselling is allowed, any point on the effici ent f ontie gene ated by the extended ve sion of Eq. 1 is a po tfolio of no mo e than two assets,29 indicating that dive sification is pointless. T ivially, th e global minimum isk po tfolio is totally concent ated in the stock with the sm allest standalone isk. Likewise, the global maximum etu n po tfolio is totally concent ated in the stock with the maximum expected etu n. The esult fo the inte mediate points follows f om the linea ity of the isk measu e. The easonin g is independent of the choice of . If the vecto of etu ns is not comonotonic, then dive sification oppo tunities a e dete mined by the deg ee to which the mul tiva iate dist ibution function can deviate f om its uppe F chet-Hoeffding bound (see Eq. 11). This is one facto that dete mines the potential fo po tfolio i sk to deviate f om its uppe bound, which is the weighted ave age of the standal one isks (see Eq. 12). The othe facto is, of cou se, the isk measu e and its p ope ties. T ivially, if is constant fo any po tfolio, then it is unable to i dentify any dive sification oppo tunities

Appendix 3: Dynamic Po tfolio Choice in Continuous Time with an Implicit Lowe B ound In this appendix, we p ovide in detail the assumptions and the p oblem setup lea ding to the pa ticula solution given in Eq. 5. In te ms of assumptions, we do n ot conside the most gene al setting. Conside an economy whose unce tainty is ep esented th ough a standa d p obability space (, A, P) and a finite time span d enoted by T.30 Investo s t ade n assets, the p ices of which, ep esented by the elements of a vecto St, evolve as 29 - The statement efe s to the gene al case of distinct standalone isks. 30 See Ma tellini and Milhau (2010) fo additional details. in which z is an n-dimensional B ownian motion and diag(St) is a diagonal mat ix with the elements of St on the main diagonal, t is a non-stochastic isk-f ee ate, t is the market price of risk vector, and t i a time-dependent n x n matrix . The market i a umed to be complete. A portfolio trategy i de cribed by a v ector proce of weight wt adapted to the filtration of the probability pace a ugmented with the natural filtration of the n-dimen ional Brownian motion. Letti ng the value proce of the trategy be A, we An EDHEC-Ri k In titute Publication 35

We can conclude that dive sification oppo tunities a e dete mined essentially by the joint behaviou of stock etu ns. The function of isk measu es in this mo e gene al setting is to p ovide an objective to t anslate those oppo tunities in to actual allocations. If the e a e no dive sification oppo tunities, howeve , t he choice of objective is i elevant.

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 Appendix obtain the following expre ion de cribing it dynamic , . In addition, we a um e that there i a benchmark evolving according to the equation in which i the a et value of the optimal olution and: Eq. 14 Eq. 15 Eq. 16 and that the floor i a multiple of the benchmark, Ft = kBt. Becau e we a ume that the market i c omplete, it follow that the benchmark can be replicated by the traded ecuritie . The inve tor utility function take the following form, The expre ion in Eq. 14 tand for the weight of the PSP, Eq. 15 for the weigh t of the benchmark replicating portfolio, and Eq. 16 define a the tocha tic proce of the benchmark-replicating portfolio con tructed uch that . where i the CRRA utility function and is the risk-aversion parameter. This defi nition results in infinite disutility if x < k. Investors maximise expected util ity at the terminal time instant T by solvin the followin expected utility max imisation problem subject to a bud et constraint: Eq. 13 where A0 is the initial capital and MT the pricin kernel. In addition, there is also a liquidity constraint AT 0 in an almost-sure sense for all t T. On these assumptions, the solution to the expected utility maximisation problem in Eq. 13 is iven by 36 An EDHEC-Risk Institute Publication

References An EDHEC-Risk Institute Publication 37

A Post-Crisis Perspective on Diversification for Risk Mana ement May 2011 References Amenc, N., F. Goltz, and A. Gri oriu. 2010a. Risk control throu h dynamic core-s atellite portfolios of ETFs: Applications to absolute return funds and tactical asset allocation. Journal of Alternative Investments 13 (2): 47-57. Amenc, N., F . Goltz, and V. Le Sourd. 2006. Assessin the quality of stock market indices. E DHEC-Risk Institute Publication (September). Amenc, N., F. Goltz, L. Martellini, and P. Retkowsky. 2010b. Efficient indexation: An alternative to cap-wei hted i ndices. EDHEC-Risk Institute Publication (January). Forthcomin in Journal of In vestment Mana ement. Artzner, P., F. Delbaen, J.-M. Eber, and D. Heath. 1998. Co herent measures of risk. Mathematical Finance 6: 203-228. Basak, S. 1995. A ene ral equilibrium model of portfolio insurance. Review of Financial Studies 8: 105 9-90. Basak, S. 2002. A comparative study of portfolio insurance. Journal of Eco nomic Dynamics and Control 26 (7-8): 1217-41. Black, F., and R. Jones. 1987. Sim plifyin portfolio insurance. Journal of Portfolio Mana ement 14 (1): 48-51. Bla ck, F., and A. Perold. 1992. Theory of constant proportion portfolio insurance. Journal of Economic Dynamics and Control 16:403-426. Brennan, M., E. Schwartz, a nd R. La nado. 1997. Strate ic asset allocation. Journal of Economic Dynamics an d Control 21 (8-9):1377-1403. Christoffersen, P., V. Errunza, K. Jacobs, and X. Jin. 2010. Is the potential for international diversification disappearin ? (Mar ch 16). Available online at SSRN: http://ssrn.com/ abstract=1573345 Danielsson, J., C. de Vries, B. Jor ensen, S. Mandira, and G. Samorodnitsky. 2010. Fat tails , VaR and subadditivity. Available online at: http://www.riskresearch.or /?papid =35&catid=0 De Mi uel, V., L. Garlappi, and R. Uppal. 2009. Optimal versus naive diversification: How inefficient is the 1/N portfolio strate y? Review of Finan cial Studies 22 (5): 1915-53. Dubikovsky, V., M. Hayes, L. Goldber , and M. Liu. 2010. How well can the risk of financial extremes be forecast? Whitepaper, MSCI Barra, Research Insi hts. Ekeland, I., A. Galichon, and M. Henry. 2009. Comonot onic measures of multivariate risks. Available online at SSRN: http://ssrn.com/a bstract=1115729 Frchet, M. 1951. Sur les tableaux de corrlation dont les mar es so nt donnes. Ann. Univ. Lyon Sect. A Srie 3 (14): 5377. Grinold, R. C. 1992. Are benc hmark portfolios efficient? Journal of Portfolio Mana ement 19 (1): 34-40. 38 An EDHEC-Risk Institute Publication

A Post-Crisis Perspective on Diversification for Risk Mana ement May 2011 References Grossman, S. J., and J.-L. Vila. 1989. Portfolio insurance in complete markets: A note. Journal of Business 62:473-76. Grossman, S., and Z. Zhou. 1996. Equilibr ium analysis of portfolio insurance. Journal of Finance 51 (4): 1379-1403. Hau e n, R. A., and N. L. Baker. 1991. The efficient market inefficiency of capitaliza tion wei hted stock portfolios. Journal of Portfolio Mana ement 17 (3): 35-40. H oeffdin , W. 1940. Masstabinvariante Korrelationstheorie. Schriften des mathemat ischen Instituts und des Instituts fr an ewandte Mathematik der Universitt Berlin 5:179-233. Jan, Y.-C., and Y.-L. Wu 2008. Revisit the debate of time diversifica tion. Journal of Money, Investment and Bankin 6:27-33. Kim, T., and E. Omber . 1996. Dynamic nonmyopic portfolio behavior. Review of Financial Studies 9 (1): 1 41-61. Leland, H. 1980. Who should buy portfolio insurance? Journal of Finance 3 5 (2): 581-94. Lioui, A., and P. Poncet. 2001. On optimal portfolio choice under stochastic interest rates. Journal of Economic Dynamics and Control 25 (11): 18 41-65. Lon in, F., and B. Solnik. 2001. Extreme correlation of international equ ity markets. Journal of Finance 56 (2): 649-676. Markowitz, H. M. 1952. Portfoli o selection. Journal of Finance 7 (1): 77-91. Markowitz, H. M. 1959. Portfolio s election: Efficient diversification of investments. New York: Wiley. Martellini, L., and V. Milhau. 2010. Hed in versus insurance: Lon -term investin with sho rt-term constraints. Workin paper, EDHEC-Risk Institute. Martellini, L., and V. Ziemann. 2010. Improved estimates of hi her-order comoments and implications fo r portfolio selection. Review of Financial Studies 23 (4): 1467-1502. Merton, R. C. 1969. Lifetime portfolio selection under uncertainty: The continuous time ca se. Review of Economics and Statistics 51: 247-57. Merton, R. C. 1971. Optimum c onsumption and portfolio rules in a continuous-time model. Journal of Economic T heory 3:373-413. Munk, C., and C. Srensen. 2004. Optimal consumption and investme nt strate ies with stochastic interest rates. Journal of Bankin and Finance 28 (8): 1987-2013. Munk C., C. Srensen, and T. Vinther. 2004. Dynamic asset allocati on under mean-revertin returns, stochastic interest rates, and inflation uncert ainty: Are popular recommendations consistent with rational behaviour? Internati onal Review of Economics and Finance 13:141-66. Rschendorf, L. 2010. Worst case p ortfolio vectors and diversification effects. Forthcomin in Finance and Stochas tics. An EDHEC-Risk Institute Publication 39

A Post-Crisis Perspective on Diversification for Risk Mana ement May 2011 References Sheikh, A. Z., and H. Qiao. 2009. Non-normality of market returns: A framework f or asset allocation decision makin . Whitepaper, J.P. Mor an Asset Mana ement, J PMor an Chase & Co. Stoyanov, S. V., S. Rachev, B. Racheva-Iotova, and F. Fabozz i. 2011. Fat-tailed models for risk estimation. Journal of Portfolio Mana ement 37 (2): 107-17. Tepl, L. 2001. Optimal investment with minimum performance constr aints. Journal of Economic Dynamics and Control 25 (10): 1629-45. Tobin, J. 1958 . Liquidity preference as behavior towards risk. Review of Economic Studies 25:6 5-86. Wachter, J. 2002. Optimal consumption and portfolio allocation under meanrevertin returns: An exact solution for complete markets. Journal of Financial and Quantitative Analysis 37:63-91. Zumbach, G. O. 2007. The RiskMetrics 2006 me thodolo y. Available online at SSRN: http://ssrn.com/abstract=1420185 40 An EDHEC-Risk Institute Publication

About EDHEC-Risk Institute An EDHEC-Risk Institute Publication 41

A Post-Crisis Perspective on Diversification for Risk Mana ement May 2011 About EDHEC-Risk Institute Founded in 1906, EDHEC is one of the foremost French business schools. Accredite d by the three main international academic or anisations, EQUIS, AACSB, and Asso ciation of MBAs, EDHEC has for a number of years been pursuin a strate y for in ternational excellence that led it to set up EDHEC-Risk in 2001. With sixty-six professors, research en ineers, and research associates, EDHEC-Risk has the lar est asset mana ement research team in Europe.

The Choice of Asset Allocation and Risk Mana ement EDHEC-Risk structures all of its research work around asset allocation and risk mana ement. This issue corresponds to a enuine expectation from the market. On the one hand, the prevailin stock market situation in recent years has shown th e limitations of diversification alone as a risk mana ement technique and the us efulness of approaches based on dynamic portfolio allocation. On the other, the appearance of new asset classes (hed e funds, private equity, real assets), with risk profiles that are very different from those of the traditional investment universe, constitutes a new opportunity and challen e for the implementation of allocation in an asset mana ement or assetliability mana ement context. This str ate ic choice is applied to all of the Institute's research pro rammes, whether they involve proposin new methods of strate ic allocation, which inte rate the alternative class; takin extreme risks into account in portfolio construction; studyin the usefulness of derivatives in implementin asset-liability mana emen t approaches; or orientin the concept of dynamic core-satellite investment mana e ment in the framework of absolute return or tar et-date funds. An Applied Research Approach In an attempt to ensure that the research it carries out is truly applicable, ED HEC has implemented a dual validation system for the work of EDHEC-Risk. All res earch work must be part of a research pro ramme, the relevance and oals of whic h have been validated from both an academic and a business viewpoint by the Inst itute's advisory board. This board is made up of internationally reco nised rese archers, the Institute's business partners, and representatives of major interna tional institutional investors. Mana ement of the research pro rammes respects a ri orous validation process, which uarantees the scientific quality and the op erational usefulness of the pro rammes. Six research pro rammes have been conduc ted by the centre to date: Asset allocation and alternative diversification Styl e and performance analysis Indices and benchmarkin Operational risks and perfor mance Asset allocation and derivative instruments ALM and asset mana ement These pro rammes receive the support of a lar e number of financial companies. The re sults of the research pro rammes are disseminated throu h the EDHEC-Risk locatio ns in London, Nice, and Sin apore. 42 An EDHEC-Risk Institute Publication

A Post-Crisis Perspective on Diversification for Risk Mana ement May 2011 About EDHEC-Risk Institute In addition, EDHEC-Risk has developed a close partnership with a small number of sponsors within the framework of research chairs or major research projects: Re ulation and Institutional Investment, in partnership with AXA Investment Mana e rs Asset-Liability Mana ement and Institutional Investment Mana ement, in partne rship with BNP Paribas Investment Partners Risk and Re ulation in the European F und Mana ement Industry, in partnership with CACEIS Structured Products and Deri vative Instruments, sponsored by the French Bankin Federation (FBF) Dynamic All ocation Models and New Forms of Tar et-Date Funds, in partnership with UFG-LFP A dvanced Modellin for Alternative Investments, in partnership with Newed e Prime Brokera e Asset-Liability Mana ement Techniques for Soverei n Wealth Fund Mana ement, in partnership with Deutsche Bank Core-Satellite and ETF Investment, in p artnership with Amundi ETF The Case for Inflation-Linked Corporate Bonds: Issuer s and Inve tor Per pective , in partner hip with Roth child & Cie Advanced Inve t ment Solution for Liability Hedging for Inflation Ri k, in partner hip with Ont ario Teacher Pen ion Plan Exploring the Commodity Future Ri k Premium: Implicat ion for A et Allocation and Regulation, in partner hip with CME Group Structur ed Equity Inve tment Strategie for Long-Term A ian Inve tor , in partner hip wi th Socit Gnrale Corporate & Inve tment Banking The Benefit of Volatility Derivative in Equity Portfolio Management, in partne r hip with Eurex Solvency II Benchmark , in partner hip with Ru ell Inve tment The philo ophy of the In titute i to validate it work by publication in inter national journal , a well a to make it available through it po ition paper , publi hed tudie , and conference . Each year, EDHEC-Ri k organi e a major inte rnational conference for in titutional inve tor and inve tment management profe ional with a view to pre enting the re ult of it re earch: EDHEC-Ri k In ti tutional Day . EDHEC al o provide profe ional with acce to it web ite, www .edhecri k.com, which i entirely devoted to international a et management re e arch. The web ite, which ha more than 42,000 regular vi itor , i aimed at prof e ional who wi h to benefit from EDHEC analy i and experti e in the area of a pplied portfolio management re earch. It monthly new letter i di tributed to m ore than 700,000 reader . EDHEC-Ri k In titute: Key Figure , 2009-2010 Nbr of permanent taff Nbr of re earch a ociate Nbr of affiliate profe or Ov erall budget External financing Nbr of conference delegate Nbr of participant at EDHEC-Ri k Indice & Benchmark eminar Nbr of participant at EDHEC-Ri k In titute Ri k Management eminar Nbr of participant at EDHEC-Ri k In titute Exe cutive Education eminar 66 18 6 9,600,000 6,345,000 2,300 582 512 247 An EDHEC-Ri k In titute Publication 43

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 About EDHEC-Ri k In titute Re earch for Bu ine The In titute activitie have al o given ri e to executive education and re earc h ervice off hoot . EDHEC-Ri k executive education programme help inve tment profe ional to upgrade their kill with advanced ri k and a et management t raining acro traditional and alternative cla e . The EDHEC-Ri k In titute PhD in Finance www.edhec-ri k.com/AIeducation/PhD_Finance The EDHEC-Ri k In titute PhD in Finance i de igned for profe ional who a pire to higher intellectual level and aim to redefine the inve tment banking and a et management indu trie . It i offered in two track : a re idential track for high-potential graduate tudent , who hold part-time po ition at EDHEC, and an executive track for practitioner who keep their full-time job . Drawing it fac ulty from the world be t univer itie and enjoying the upport of the re earch c entre with the greate t impact on the financial indu try, the EDHEC-Ri k In titu te PhD in Finance create an extraordinary platform for profe ional development and indu try innovation.

improved ri k/reward efficiency compared to cap-weighted indice . The weighting of the portfolio of con tituent achieve the highe t po ible return-to-ri k ef ficiency by maximi ing the Sharpe ratio (the reward of an inve tment per unit of ri k). The e indice provide inve tor with an enhanced ri k-adju ted trategy in compari on to cap-weighted indice , which have been the ubject of numerou c ritique , both theoretical and practical, over the la t few year . The index er ie i ba ed on all con tituent ecuritie in the FTSE All-World Index Serie . C on tituent are weighted in accordance with EDHECRi k portfolio optimi ation, re flecting their ability to maximi e the rewardto-ri k ratio for a broad market in dex. The index erie i rebalanced quarterly at the ame time a the review of the underlying FTSE All-World Index Serie . The performance of the EDHEC-Ri k E fficient Indice are publi hed monthly on www.edhec-ri k.com. EDHEC-Ri k Alternative Indexe www.edhec-ri k.com/indexe /pure_ tyle FTSE EDHEC-Ri k Efficient Indice www.edhec-ri k.com/indexe /efficient FTSE Group, the award winning global index provider, and EDHEC-Ri k In titute la unched the fir t et of FTSE EDHEC-Ri k Efficient Indice at the beginning of 20 10. Offered for a full global range, including All World, All World ex US, All W orld ex UK, Developed, Emerging, USA, UK, Eurobloc, Developed Europe, Developed Europe ex UK, Japan, Developed A ia Pacific ex Japan, A ia Pacific, A ia Pacific ex Japan, and Japan, the index erie aim to capture equity market return wit h an 44 An EDHEC-Ri k In titute Publication The different hedge fund indexe available on the market are computed from diffe rent data, according to diver e fund election criteria and index con truction m ethod ; they un urpri ingly tell very different torie . Challenged by thi hete rogeneity, inve tor cannot rely on competing hedge fund indexe to obtain a true and fair view of performance and are at a lo when electing benchmark . To add re thi i ue, EDHEC Ri k wa the fir t to launch compo ite hedge fund trateg y indexe a early a 2003. The thirteen EDHEC-Ri k Alternative Indexe are publ i hed monthly on www.edhec-ri k.com and are freely available to manager and inv

e tor .

An EDHEC-Ri k In titute Publication 45

EDHEC-Ri k In titute Publication and Po ition Paper

(2008-2011)

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 EDHEC-Ri k In titute Publication (2008-2011) 2011 Amenc, N., L. Martellini, F. Goltz, and D. Sahoo. I there a ri k/return tradeoff acro tock ? An an wer from a long-horizon per pective (April). Amenc, N., L. Ma rtellini, F. Goltz, and L. Tang. Improved beta? A compari on of indexweighting cheme (April). Sender, S. The elephant in the room: Accounting and pon or ri k in corporate pen ion plan (March). Martellini, L., and V. Milhau. Optimal de ig n of corporate market debt programme in the pre ence of intere t-rate and infla tion ri k (February). 2010 Amenc, N., and S. Sender. The European fund management indu try need a better gra p of non-financial ri k (December). Amenc, N., S, Focardi, F. Goltz, D. Schrder, and L. Tang. EDHEC-Ri k European private wealth management urvey (November). Am enc, N., F. Goltz, and L. Tang. Adoption of green inve ting by in titutional inv e tor : A European urvey (November). Martellini, L., and V. Milhau. An integrat ed approach to a et-liability management: Capital tructure choice , pen ion fu nd allocation deci ion and the rational pricing of liability tream (November) . Amenc, N., F. Goltz, Martellini, L., and V. Milhau. New frontier in benchmark ing and liability-driven inve ting (September). Martellini, L., and V. Milhau. F rom determini tic to tocha tic life-cycle inve ting: Implication for the de ig n of improved form of target date fund (September). Martellini, L., and V. Mil hau. Capital tructure choice , pen ion fund allocation deci ion and the ration al pricing of liability tream (July). Sender, S. EDHEC urvey of the a et and l iability management practice of European pen ion fund (June). Goltz, F., A. Grig oriu, and L. Tang. The EDHEC European ETF urvey 2010 (May). Martellini, L., et V. Milhau. A et-liability management deci ion for overeign wealth fund (May). Am enc, N., and S. Sender. Are hedge-fund UCITS the cure-all? (March). Amenc, N., F. Goltz, and A. Grigoriu. Ri k control through dynamic core- atellite portfolio o f ETF : Application to ab olute return fund and tactical a et allocation (Jan uary). Amenc, N., F. Goltz, and P. Retkow ky. Efficient indexation: An alternative to cap-weighted indice (January). Goltz, F., and V. Le Sourd. Doe finance theor y make the ca e for capitali ation-weighted indexing? (January). 46 An EDHEC-Ri k In titute Publication

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 EDHEC-Ri k In titute Publication (2008-2011) 2009 Sender, S. Reaction to an EDHEC tudy on the impact of regulatory con traint on the ALM of pen ion fund (October). Amenc, N., L. Martellini, V. Milhau, and V. Zi emann. A et-liability management in private wealth management (September). Amenc, N., F. Goltz, A. Grigoriu, and D. Schroeder. The EDHEC European ETF urvey (May ). Sender, S. The European pen ion fund indu try again be et by deficit (May). Ma rtellini, L., and V. Milhau. Mea uring the benefit of dynamic a et allocation trategie in the pre ence of liability con traint (March). Le Sourd, V. Hedge fu nd performance in 2008 (February). La ge tion indicielle dan l immobilier et l in dice EDHEC IEIF Immobilier d Entrepri e France (February). Real e tate indexing an d the EDHEC IEIF Commercial Property (France) Index (February). Amenc, N., L. Mart ellini, and S. Sender. Impact of regulation on the ALM of European pen ion fund (January). Goltz, F. A long road ahead for portfolio con truction: Practitioner view of an EDHEC urvey. (January).

2008 Amenc, N., L. Martellini, and V. Ziemann. Alternative inve tment for in titutiona l inve tor : Ri k budgeting technique in a et management and a et-liability m anagement (December). Goltz, F., and D. Schroeder. Hedge fund reporting urvey (No vember). DHondt, C., and J.-R. Giraud. Tran action co t analy i A-Z: A tep toward be t execution in the po t-MiFID land cape (November). Amenc, N., and D. Schroed er. The pro and con of pa ive hedge fund replication (October). Amenc, N., F. G oltz, and D. Schroeder. Reaction to an EDHEC tudy on a et-liability managemen t deci ion in wealth management (September). Amenc, N., F. Goltz, A. Grigoriu, V. Le Sourd, and L. Martellini. The EDHEC European ETF urvey 2008 (June). Amenc, N. , F. Goltz, and V. Le Sourd. Fundamental difference ? Comparing alternative inde x weighting mechani m (April). Le Sourd, V. Hedge fund performance in 2007 (Febru ary). Amenc, N., F. Goltz, V. Le Sourd, and L. Martellini. The EDHEC European inve tment practice urvey 2008 (January). An EDHEC-Ri k In titute Publication 47

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 EDHEC-Ri k In titute Po ition Paper (2008-2011) 2010 Amenc, N., and V. Le Sourd. The performance of ocially re pon ible inve tment and u tainable development in France: An update after the financial cri i (Septem ber). Amenc, N., A. Chron, S. Gregoir, and L. Martellini. Il faut pr erver le Fond de R erve pour le Retraite (July). With the EDHEC Economic Re earch Centre. Amen c, N., P. Schoefler, and P. La erre. Organi ation optimale de la liquidit de fo nd dinve ti ement (March). Lioui, A. Spillover effect of counter-cyclical market regulation: Evidence from the 2008 ban on hort ale (March).

2009 Till, H. Ha there been exce ive peculation in the US oil future market ? (Nove mber). Amenc, N., and S. Sender. A welcome European Commi ion con ultation on the UCITS depo itary function, a ha tily con idered propo al (September). Sender, S. IAS 19: Penali ing change ahead (September). Amenc, N. Quelque rflexion ur la rg ulation de la ge tion d actif (June). Giraud, J.-R. MiFID: One year on (May). Lioui , A. The unde irable effect of banning hort ale (April). Gregoriou, G., and F. -S. Lhabitant. Madoff: A riot of red flag (January). 2008 Amenc, N., and S. Sender. A e ing the European banking ector bailout plan (Dec ember). Amenc, N., and S. Sender. Le me ure de recapitali ation et de outien la liquidit du ecteur bancaire europen (December). Amenc, N., F. Ducoulombier, and P. Foulquier. Reaction to an EDHEC tudy on the fair value controver y (December) . With the EDHEC Financial Analy i and Accounting Re earch Centre. Amenc, N., F. Ducoulombier, and P. Foulquier. Raction apr ltude. Ju te valeur ou non : un dbat ma l po (December). With the EDHEC Financial Analy i and Accounting Re earch Centr e. Amenc, N., and V. Le Sourd. Le performance de linve ti ement ocialement re p on able en France (December). Amenc, N., and V. Le Sourd. Socially re pon ible inv e tment performance in France (December). Amenc, N., B. Maffei, and H. Till. Le cau e tructurelle du troi ime choc ptrolier (November). Amenc, N., B. Maffei, an d H. Till. Oil price : The true role of peculation (November). 48 An EDHEC-Ri k In titute Publication

A Po t-Cri i Per pective on Diver ification for Ri k Management May 2011 EDHEC-Ri k In titute Po ition Paper (2008-2011) Sender, S. Banking: Why doe regulation alone not uffice? Why mu t government in tervene? (November). Till, H. The oil market : Let the data peak for it elf (Octo ber). Amenc, N., F. Goltz, and V. Le Sourd. A compari on of fundamentally weighted indice : Overview and performance analy i (March). Sender, S. QIS4: Significan t improvement , but the main ri k for life in urance i not taken into account i n the tandard formula (February). With the EDHEC Financial Analy i and Account ing Re earch Centre. An EDHEC-Ri k In titute Publication 49

For more information, plea e contact: Carolyn E id on +33 493 187 824 or by e-m ail to: carolyn.e id@edhec-ri k.com EDHEC-Ri k In titute 393 promenade de Angl ai BP 3116 06202 Nice Cedex 3 France EDHEC Ri k In tituteEurope 10 Fleet Place Ludgate London EC4M 7RB - United Kingdom EDHEC Ri k In tituteA ia 1 George Stree t #07-02 Singapore 049145 www.edhec-ri k.com

Вам также может понравиться