Measuring Alpha-Based Performance: Implications for Alpha-Focused Structured Products

Transcription

1 Measuring Alpha-Based Perormance: Implications or Alpha-Focused Structured Products AUTHORS ARTICLE INFO JOURNAL FOUNDER Larry R. Gorman Robert A. Weigand Larry R. Gorman and Robert A. Weigand (2008). Measuring Alpha-Based Perormance: Implications or Alpha-Focused Structured Products. Investment Management and Financial Innovations, 5(2) "Investment Management and Financial Innovations" LLC Consulting Publishing Company Business Perspectives NUMBER OF REFERENCES 0 NUMBER OF FIGURES 0 NUMBER OF TABLES 0 The author(s) This publication is an open access article. businessperspectives.org

2 Larry R. Gorman (USA), Robert A. Weigand (USA) Measuring Alpha-based perormance: implications or Alpha-ocused structured products Abstract We propose that the muted demand or investment innovations such as Portable Alpha arise, at least in part, rom a lack o clarity and transparency regarding the way alpha is deined and measured. We show that the proession has been debating the closely-related issue o alpha/beta separation as ar back as the 1970s, and argue that lack o closure regarding this debate is a natural and expected eature o innovation in money management products. We provide an example o how to measure alpha bias in the context o benchmarking an actively-managed equity portolio, and ind a maximum potential bias rom o 5% per year. Keywords: alpha, beta, Portable Alpha, perormance attribution. JEL Classiication: G11. Introduction 2 Magazines and blogs with thousands o readers a day are devoted to it. The Economist recently called its ever-changing deinition "A looming challenge or the money management industry" ( ). Successul managers "hunt" it, and are not shy about charging or the right to eat at their table when they catch it. It is in short supply, and everyone generally agrees that over time and across markets it is a zerosum game, which means that to earn it you usually have to take it away rom your competitors. We are, o course, talking about the ultimate bottom line o the active money management industry the elusive concept known as alpha. Despite its importance in the investment world, we propose that ad-hoc deinitions and disparate interpretations o alpha have contributed to a state o aairs in which even investment proessionals and their clients rarely bother speciying exactly what alpha is 1. It is not uncommon or a somewhat loose deinition o alpha to be proposed, accepted, and then set aside as operational and legal issues take precedence. However, with many products being marketed around the objective o earning alpha (e.g., Portable Alpha), it is critical to develop a more precise understanding o this key perormance metric. We assert that the status quo has contributed to the muted demand or Portable Alpha and other diicultto-benchmark products, and that moving toward a clearer and more widely-accepted deinition o alpha will have a positive eect on the demand or structured investment vehicles with an alpha ocus. Any discussion o alpha also implies a discussion o beta, as an asset's total return is comprised o two. Larry R. Gorman, Robert A. Weigand, For example, a document available at lists 20 dierent deinitions o what alpha is (gathered rom diverse sources). Signiicant conlicts and variation among the deinitions exist. 16 components, its alpha-return plus its beta-return. Thereore, our inquiry into the question what is the return associated with alpha? also requires that we ask what is the return associated with beta? When viewed rom this perspective, it becomes clear that the issues we address are the same as those driving the current debate regarding the right way to benchmark virtually every actively-managed vehicle, ranging rom 130/30 unds (Lo and Patel, 2007) to hedge und replication products (Lett and Holt, 2007). We urther assert that this is simply the most recent incarnation o a discussion about alpha/beta separation that has been evolving since Fama and French (1993) proposed that value and size were systematic risk actors in the same sense as the global market actor rom the original Capital Asset Pricing Model. In the context o measuring the alpha o an activelymanaged equity portolio, we show that conusion over alpha exists even within a ramework as simple as the Capital Asset Pricing Model (CAPM). The conusion begins with empirical estimation o alpha and beta and extends to interpretation o the estimation parameters. We chronicle how the accepted deinition o beta has changed since the advent o the CAPM, and demonstrate why a nuanced understanding o the way beta continues to evolve is necessary or accurate measurement and interpretation o alpha. To demonstrate the economic signiicance o these issues, we also estimate the maximum possible alpha bias rom using the "wrong" benchmarking model (the CAPM) i another method, in this case the Fama- French 3-Factor Model, provides a better measure o systematic (beta) returns. We ind that, rom , a manager with an extreme tilt toward smallcap value who measured alpha using a single-index CAPM would have had an alpha bias as large as +4.9% per year, while a manager emphasizing an extreme tilt toward large-cap growth would have had a maximum alpha bias o -5.2% per year.

3 1. Deining and estimating alpha Generally, the alpha (or asset or portolio i) is estimated empirically via a statistical linear regression such as: Rr R r R r R r (1) i 1 F1 2 F2 k Fk where R i is a vector (column o data) representing the last T periods o returns or portolio i. (A common choice is to set T equal to 60 months o data, although daily data are oten used when ocusing on 6-, 12-, or 18-month periods). r is a vector o returns on the risk ree asset (typically the short term T-bill rate or LIBOR) or the last T periods. R F k is a vector o returns on risk actor k or the last T periods. Depending upon the model o risk employed (CAPM or otherwise), there may be only one actor, or there may be several actors. Risk actors are thought to be systematic in nature. That is, it is assumed that virtually all assets are exposed to a relatively small number (= k) o common risks, reerred to as actors. Equation (1) is written in general orm to account or these k actors. The CAPM has only one actor, a maximally-diversiied global portolio o risky assets. Hence, in the CAPM, the actor R F1 is typically written as R MKT. Most models ollowing the CAPM typically employ multiple risk actors (elaborated on in the ollowing section). k is the beta (estimated statistically) associated with the return vector o actor k. Beta measures how sensitive an asset's returns are to movements (returns) in actor k. The CAPM models one beta, whereas there are multiple measures o beta in multi-actor models (one or each risk actor) 1. is the residual vector, indicating deviations between the linear regression line (or response surace or multi-actor models) and the actual returns o asset i. There are T 's estimated in each regression, and they have a mean o exactly zero. The alpha () o portolio i or a certain time span represents the return o the portolio above what would be expected, given the portolio's exposure to the risk actors. Alpha is the money shot number in perormance attribution. A positive value o indicates that the portolio (and most importantly, the manager(s) o the portolio) perormed abnormally well, over and above a certain level o exposure to various systematic risk actors. Alpha is also reerred to as abnormal return, and is interpreted as a direct measure o investment manager skill. 1 Beta is not simply an asset's return volatility relative to market volatility. Beta is a scaled measure o the correlation o returns between the asset and each actor. In the single-index CAPM the scaling actor is the ratio o volatility o asset i to the volatility o the market: i =( i / MKT ) correlation ( i, MKT ). The return resulting rom exposure to the various risk actors is known as the beta component o returns, deined as: Rr R r R r R r (2) i 1 F1 2 F2 k Fk In this depiction + constitute the non-beta return component o portolio i. Over any time period the mean o is always exactly equal to zero, thus the non-beta return equals. Equation (1) is a completely general orm or alpha estimation. It allows or any number o actors and their associated betas (up to k o them). That is, equation (1) provides estimates o and 1, 2, k. For the CAPM (a single-actor model), only and 1 are estimated. Regardless o the number o speciied actors, the estimate o can be interpreted as a measure o abnormal risk-adjusted perormance o an asset, portolio, or manager(s). 2. Models o risk and expected return general results Risk models relate risk to expected returns (ex ante, or beore-the-act estimates), which are almost always dierent than realized returns (ex post, or ater-the-act calculations). One key eature o these models is that all risk is viewed as belonging to one o two categories: either (i) systematic risk or (ii) non-systematic risk 2. Within these models, there is a higher expected reward or exposure to systematic risk actors via higher expected returns. These rewards are known as risk premia, and are thought to expand and contract over time. There is no expected reward or exposure to nonsystematic risk. Expected returns depend upon the magnitude o exposure to systematic risk only. Although non-systematic risk exposure can result in positive or negative returns ex post, the important point is that, ex ante, this type o risk is expected to provide a return o zero. Beyond these commonalities, models dier in what types o risks are considered to be systematic. Some models employ only one systematic risk actor (the CAPM), while other models employ 2, 3, 4 or more. Once the actual returns to the actors are known (rom the historical period o measure, e.g. 60 months), it is possible to compute what the return on a portolio should have been, given the actual actor returns and the portolio's exposure to these actors, as measured by the actor betas. This can also be thought o as a measure o what the portolio would 2 Both systematic and non-systematic risk are commonly reerred to using other terminology as well. Systematic risk is also known as beta risk or actor risk. Non-systematic risk is also known as idiosyncratic risk, diversiiable risk, or asset-speciic risk. 17

4 have earned in the period under study i exposure to non-systematic risk had a payo o zero. Thereore, i we have estimates o the k betas and the actual returns on the k actors or the last 60 months, the reward or systematic risk can be computed as: 18 Actual beta return r ( R 2 F 2 ( R r )... ( R k 1 Fk F1 r r ), (3) ) where the 's are estimated via linear regression (equation 1), and are multiplied by the respective means o the actual actor returns above the risk ree rate or the period o measurement 1. Although the expected return or non-systematic risk exposure is zero, the actual return is rarely zero. The actual return to non-systematic risk is, by deinition, alpha. It is the primary measure o investment skill 2. It is computed as: Ex-post non-systematic return (alpha) = (actual total return) (actual beta return), or, equivalently: Rir 1 RF 1r 2 RF2r k RFkr. (4) Note that this computation is the same as equation (1). Also, all parameters (including alpha) at this stage are measured in per period terms (e.g., monthly), and can be annualized later. 3. Speciic models o risk and return The prior section outlined general issues associated with all models o risk and return. In this section we turn to the evolution o these models and discuss speciic models, which we will relate to the general statements made in the prior section. Based on the pioneering work o Sharpe (1964), Lintner (1965), and Mossin (1966), the Capital Asset Pricing Model (CAPM) changed the way we think about risk and return orever. In any model o risk, CAPM or otherwise, total risk (deined as the variance o returns) is comprised o systematic risk plus non-systematic risk. In the CAPM, systematic risk is related to only one actor the global market portolio. Any increase in a portolio's exposure to this systematic risk (measured by beta) is associated with an increase in expected return, while greater 1 It is especially important to subtract the risk ree return in each period rom both R i and all actor returns. Failure to do so will result in a statistical bias in alpha equal to r (1 k ). 2 Note that i an investor has no skill in individual asset selection, but is able to time the market (when to increase or decrease exposure to various systematic actors), this ability will also maniest itsel as positive alpha. exposure to non-systematic risk is not associated with an increase in expected return. The CAPM is stated ormally as: i MKT E R r E R r (5) where E( ) denotes an expected value. When the CAPM is used as the underlying return-generating model, alpha and beta are estimated via a linear regression (equation 1) as shown 3 : i 1 MKT R r R r. (6) Once we obtain an estimate o beta (via equation 6), the historical returns or the market and the risk ree rate, then the actual reward or systematic risk can be computed (using equation 3) as: Actual beta return r R r. (7) 1 MKT Equation (7) is essentially the empirical version o the CAPM (5). This is not surprising, since equation (7) measures the actual return associated with market risk only, and the CAPM implies that only market risk should be rewarded. From the advent o the CAPM in 1964 until the mid-1970s, the model generated relatively little controversy 4. For the most part, statistical tests o the CAPM were supportive o the model's prediction that non-systematic risk should not be rewarded, which is the same as saying that, on average, alpha is not statistically dierent rom zero. The predominant view at the time was that markets were highly eicient, and it was thereore unlikely or anyone to earn alpha consistently over time. Randomness or luck was the common explanation assigned to an organization or individual who demonstrated a consistent ability to earn alpha 5. This view was challenged when Basu (1977) showed that, ater controlling or systematic risk, portolios o low P/E stocks outperormed portolios o high P/E stocks. This inding ran contrary to the predictions o the CAPM. It now seemed possible that alpha could be earned consistently via skill rather than luck. Another deviation rom the CAPM was uncovered in 1981 when two doctoral students, 3 Equation (6) is commonly known as the market model regression. 4 Not all o the initial reaction to the CAPM was avorable, however; e.g., see Bierwag and Grove (1965). 5 Consider the ollowing simple example: I 1024 people lip a coin once a year (and lipping heads is associated with earning positive alpha), then ater ten years we would expect at least one person to have lipped heads ten times in a row. I only one person in a thousand beats the market ten years in a row, the result is just as likely due to chance as skill. This one in a thousand ratio is similar to the historical perormance o mutual und managers (such as Peter Lynch, whose Magellan Fund earned positive alpha in 11 out o 13 years).

5 Rol Banz and Mark Reinganum, working independently at the University o Chicago, discovered that portolios o small capitalization stocks outperormed portolios o large capitalization stocks, ater controlling or their exposure to the market risk actor 1. Like Basu's P/E discovery, this inding was an anomaly, at least rom the viewpoint o the CAPM. In the ollowing decade, much research was aimed at better understanding the P/E and size eects. It was a time o transition, but throughout this period, although the CAPM was repeatedly challenged, it was largely let unchanged 2. That is, rom 1964 to the early 1990s, systematic risk was considered to be based on only one actor, the market, while the P/E and market capitalization anomalies were largely viewed as puzzles yet to be solved. In 1993, more than ten years ater the discoveries o Basu, Banz and Reinganum, Eugene Fama and Ken French published a paper that reuted the one actor structure o the CAPM in avor o a three actor model o systematic risk. The new actors (in addition to the market) were a value actor (similar in spirit to Basu's P/E ratio, but based instead on the ratio o book value o equity to stock price, a.k.a. the book-to-market ratio) and a size actor (capturing the Banz and Reinganum market capitalization eect). The new model came to be known as the Fama-French Three-Factor Model. Inclusion o these two additional actors was initially controversial, as this was the irst time that actors previously deined as alpha-components were recast as beta-components. Eventually, however, volumes o research papers, much lively and intelligent debate, and innovations in index products such as exchange-traded unds (which signiicantly lowered the cost o obtaining exposure to the value and size actors), resulted in academics and (most) practitioners accepting these actors as legitimate beta-risks. Expressed in expectation orm, the Fama-French Three-Factor model is E( R ) r 2 i E( R 1 ) r E( R ) E( R ) SMB 3 MKT HML, (8) where R SMB is the return to a portolio o small cap stocks minus the return to a portolio o large cap stocks (Small Minus Big), and R HML is the return to a portolio o high book to market stocks minus the return to a portolio o low book to market stocks 1 The inding has since come to be known as the size eect. 2 During this period, Eugene Fama requently commented "It takes a model to beat a model". (High Minus Low) 3. The Fama-French alpha and its three betas are estimated empirically via a linear regression such as 4. Rr R r R R. (9) i 1 MKT 2 SMB 3 HML As the one-actor CAPM declined rom avor and the Fama-French Three Factor Model gained acceptance, the deinition o systematic risk also changed 5. This spawned an evolution in the way we conceptualize equity expected returns, and the way we measure alpha. The next section elaborates urther on this idea. We show that ailure to account or these changes results in biased estimates o alpha, which in turn aects perormance attribution, ees, managers' compensation, and investor perceptions o the money management industry. 4. Bias in alpha estimations Next consider the measurement o alpha under both the one-actor CAPM and the Fama-French model. Following equation (4), under the CAPM, alpha is estimated as CAPM CAPM R i r 1 R r (10) MKT whereas under the Fama-French Three-Factor model, alpha is estimated via equation (9) as FF 3 r ( ) 3 1 RMKT r FF Ri. (11) FF 3 FF 3 ( ) ( ) 2 RSMB 3 RHML I there are actually three systematic risk actors that drive expected returns, but the measurement o alpha or an actively-managed equity portolio is conducted according to the one-actor CAPM, a bias will be induced in the estimation o alpha. The bias is equal to: Alpha Bias = False Alpha True Alpha = CAPM FF3 FF3 CAPM AlphaBias ( 1 1 ) ( R FF3 FF3 ( R ) ( R ). 2 SMB 3 HML MKT r ) (12) 3 There is no risk ree rate subtracted rom the SMB or HML portolios because the actor returns are constructed rom the dierence between two portolios (each with a risk ree rate subtracted) and in the dierencing process, the risk ree rates cancel out. 4 Historic returns or all three actors are available at Ken French's web site: 5 Carhart (1997) suggests that a ourth actor (price momentum) should also be considered as a systematic inluence. This model, known as the Fama-French/Carhart Four Factor model, remains somewhat controversial. The ocus o this paper is limited to the Fama French Three Factor model. Risk consulting irms such as Barra and Axioma employ numerous risk actors (including, but not limited to, SMB, HML and momentum). The additional risk actors are employed not necessarily or their systematic characteristics, but rather because some industries are exposed to speciic risks, and asset managers are interested in measuring these risks. 19

6 The bias arises rom three sources, and the direction o the bias can be positive or negative. The irst source o bias is due to the dierence between the Fama-French and CAPM market-actor betas multiplied by the market risk premium. This bias is expected to be relatively small, as a market-actor beta estimated with and without the other actor betas would be expected to return slightly dierent values. The remaining two sources o alpha bias are the most interesting in the context o our discussion. These arise rom the betas on the size and value actors. For portolios with a small-cap emphasis, a CAPM-based alpha will have a positive bias. In other words, the CAPM-measured alpha is too large because some o the additional return earned rom overweighting small-cap stocks a systematic betarisk in the Fama-French model is bundled into the measurement o alpha. Portolios emphasizing large-cap stocks suer rom the opposite problem their CAPM-measured alphas are negatively biased, because less exposure to small-cap risk reduces the expected risk premium. I a portolio with a largecap emphasis outperorms relative to these lower return expectations, managers should be credited with earning a higher alpha. For portolios with a value emphasis (high book to market ratio), a CAPM alpha will also be positively biased because it ails to account or the value risk premium (the extra return earned rom overweighting risky value stocks). Alternatively, or portolios emphasizing growth (low book to market ratios) the bias will be negative, as less exposure to the value risk actor decreases the portolio's expected return. Collectively, these eects imply that small-cap value portolios will be especially susceptible to positive alpha bias, while large-cap growth portolios will be susceptible to negative alpha bias. Intuitively, the alpha bias arises rom Fama-French based beta returns being imprecisely included in (or subtracted rom) a CAPM-based alpha. Incorrectly measuring the alpha and beta components o returns in this manner is sometimes reerred to as dirty alpha or alpha contamination. The alpha is contaminated because it is an inaccurate measure o perormance and manager skill. 20 Table 1. Estimates o Alpha bias Portolio CAPM alpha Fama-French alpha Alpha bias Large-cap growth 2.34% 2.85% 5.20% Small-cap value 18.99% 14.08% +4.91% A brie numerical example will illustrate the potential magnitude o alpha bias. Using monthly data rom , we estimated alphas using both the single-index CAPM and the Three-Factor Fama-French models or two portolios: one with an extreme tilt toward small-cap value, the other with an extreme tilt towards large-cap growth (all return and actor data were obtained rom Ken French's online data library). As expected, the large-cap growth portolio had an annual alpha that was 5.2% higher when estimated using the Fama-French model (which also accounts or the size and value actor risk premia). The extreme large-cap portolio has low exposure to these risk actors, which justiies a lower expected return. A greater amount o the returns earned by this portolio thereore register as alpha in the Fama-French depiction o risk and expected return. Also as expected, the small-cap value portolio had an annual alpha that was 4.9% lower when estimated with the Fama- French model. With signiicant exposure to the size and value risk actors, the small-cap value portolio is held up to a higher expected return standard thereore less o the returns earned by this portolio registers as alpha using the Fama-French model. 5. The cost o obtaining beta returns inluences alpha Properly-measured alpha is in low supply and in high demand. It thereore ollows that the price associated with obtaining alpha is relatively high. Alternatively, systematic returns associated with each o the three Fama-French actors are plentiul and relatively inexpensive to obtain. Stock index utures, total return swaps (TRSs) and exchangetraded unds (ETFs) are traded on numerous market indices in highly-liquid markets, making access to the market actor easy to obtain at low cost. Exposure to the small minus big (SMB) market cap actor can be synthetically created by taking a long position in the Russell 2000 and a short position in the Russell 1000, as utures, TRSs and ETFs are available on these indices. Exposure to the high minus low (HML) book to market actor (value versus growth) can be obtained by taking a long position in the Russell 1000 Value index and a short position in the Russell 1000 Growth index. These indices also trade in the orm o utures, TRSs and ETFs. With access to systematic (beta) returns being relatively easy to obtain at low cost, there is no reason to pay anything but the most nominal o ees or exposure to these beta components o equity returns. O course, paying relatively large ees or access to alpha-based returns makes sense, as long as the ee is less than the alpha. With a two-tier pricing structure between alpha and beta returns, a contaminated alpha (with a potentially unknown direction o bias) is likely to be mispriced compared to the beta returns investors can obtain on their own.

7 This potential mispricing is not in the best interest o the money management industry (at least overall), as it is likely to reduce total investor demand or any product purporting to supply alpha. Demand or alpha-ocused structured products, such as Portable Alpha, should rise as the money management industry provides greater transparency regarding the measurement and interpretation o alpha. 6. Relevance or benchmarked portolios There are many unds and investment products that or various reasons are benchmarked against a speciic index. For example, a long-only equity manager may be judged on whether they outperorm the Russell 3000, an active extension (i.e. 130/30) manager may be compared against the perormance o the Russell 1000, and a hedge und manager, choosing to invest via more exotic strategies, may select (possibly with an intent to game his/her perormance) a speciic benchmark or comparison. I a und has a mandate (imposed by upper management, investors, or sel-imposed by the actual asset managers) to benchmark itsel against, say, the Standard and Poor's 500, then what measure o the market should be employed in the estimation o alpha in equation (9)? The S&P 500, or the broad market comprised o all available assets? From the investor's point o view (contemplating investment in a benchmarking und) the proper market measure or perormance assessment is the universe o all assets available to the investor. However, rom the viewpoint o assessing the skill o the portolio manager (constrained to invest only in assets contained in the benchmark) the measure o the market in equation (9) should be assets available to the manager that is, the benchmark 1. Hence, the choice o the deinition o the market return actor used in equation (9) depends upon whose perspective the assessment is being made rom: the manager's or the investor's. For assessment rom the investor's view (assuming they have broad latitude in the types o assets they can invest in), then equation (9) should employ the broadly-deined market return, no matter what the benchmark constraints are on a und under consideration. For assessment o manager skill, equation (9) should employ the benchmark. Some may argue that or purposes o measuring manager skill, only one actor should be employed the index (i.e. use equation (6) with the market 1 For a manager who is benchmarked to an index, but has latitude to invest outside the index, then or the purpose o measuring the skill o the manager, the measure o the market return in equation (9) should be expanded beyond that o the index to relect the set o all assets in which they could invest. deined as the index). However, consider a manager benchmarked to the S&P 500 who tilts the portolio toward small-cap value stocks. The manager will likely earn positive alpha via (CAPM style) equation (6), and a lower (possibly negative) alpha via Fama-French style equation (9). The alpha o equation (6) is spurious because the manager tilted his/her holdings toward systematic risk, and equation (6) is mis-speciied rom a systematic risk perspective. This contaminates the alpha measure o equation (6) by including beta returns into the alpha metric. I the Fama-French measure o alpha equals zero, the manager demonstrated no skill in employing this tilt. The implication is that he/she earned the air market return or additional exposure to systematic risk. Summary and conclusions Recently Portable Alpha, an alpha-ocused absolute return product with tremendous potential, has met with somewhat muted demand. There are several reasons or the lack o robust demand including complexity o implementation, transparency, the ability to identiy consistent alpha generating sources, and competing products such as active extension or 130/30 unds. Beyond these commonly-mentioned reasons, skepticism, a general lack o understanding and conusion have also contributed to the ambiguous demand or Portable Alpha products. Much o the conusion arises rom a lack o clear consensus regarding a strict deinition o alpha. We argue that this diminishes demand or all alpha-ocused products, with Portable Alpha products possibly aected more than others. One partial remedy is to move toward greater clarity o exactly what alpha is, why it is best measured against the most relevant benchmarking model (which evolves over time), and why, when properly measured, it is worth paying or. Ongoing debate and transparency regarding these issues are in the best interest o the proession. Relative return products such as active portolio extensions (130/30 unds) pose an additional challenge or Portable Alpha, as they serve as substitute goods. Direct comparisons between Portable Alpha and active extension products are diicult to obtain, in large part due to asymmetric perormance methodologies alphas and/or inormation ratios can only be computed or relative return products such as active portolio extensions. The inability to directly compare the perormance o absolute and relative return products is problematic or the investor. Another diiculty arises because perormance metrics are usually computed rom the viewpoint o assessing und manager skill, and not rom the viewpoint o assessing portolio enhancement to the investor. Reconciling these 21

8 perormance measurement issues and moving to a common assessment methodology in which both absolute and relative return products are assessed via investor-ocused metrics would do much to improve investor demand or both types o products. We urther assert that money managers should expect to revisit the debate over "What's alpha and what's beta?" regularly, and that it is natural to view this question as one that will never be ully resolved. The issue o alpha/beta separation began as early as the 1970s with the discovery o the va- Reerences lue and size anomalies, and will require additional consideration with each wave o innovation in investment management. The latest incarnation o this debate, o course, concerns the best way to benchmark 130/30 unds and hedge und replication products (Lo and Patel (2007) and Lett and Holt (2007)). Managers should embrace these discussions they are necessary or the proession to continue moving toward the transparency investors want regarding perormance attribution and ees. 1. Banz, R., "The Relationship Between Return and Market Value o Common Stock", Journal o Financial Economics 9: Basu, S., "Investment perormance o common stocks in relation top their price earnings ratios: A test o market eiciency", Journal o Finance 32, June, Bierwag, G. and M. Grove, 1965". On Capital Asset Prices: Comment," Journal o Finance 20, Carhart, M. 1997", On Persistence in Mutual Fund Perormance," Journal o Finance 52, The Economist, "And God Created Alpha," January Fama, E.F. and K.R. French, "Common Risk Factors in the Returns on Stocks and Bonds", Journal o Financial Economics 33 (1), French, K., Grinold, R., "The Fundamental Law o Active Management", Journal o Portolio Management (Spring) 30: Hubrich, S., "An Alpha Unleashed: Optimal Derivative Portolios or Portable Alpha Strategies," Working Paper, T. Rowe Price Associates, Baltimore. 10. Lett, C. and C. Holt, "Two Solitudes", Canadian Investment Review (Fall), Lintner, J., "The Valuation o Risk Assets and the Selection o Risky Investments in Stock Portolios and Capital Budgets", Review o Economics and Statistics 47: 1, Lo, A. and P. Patel "130/30: The New Long-Only", Working Paper, AlphaSimplex Group, LLC and Sloan School o Management, MIT. 13. Mossin, J., "Equilibrium in a Capital Asset Market". Econometrica 34, Reinganum, M. R., "Abnormal Returns in Small Firm Portolios". Financial Analysts Journal 37, Sharpe, W.F., "Capital Asset Prices: A Theory o Market Equilibrium under Conditions o Risk", Journal o Finance 19,