Category Archives: Skill

Large Hedge Fund Survivor Bias

Why Size Isn’t Everything

Hedge fund survivor bias is especially insidious for the largest firms. Large hedge fund survivor bias overstates expected performance of the biggest firms by nearly half and their risk adjusted return from security selection (stock picking) by 80%. It is impossible to predict the largest funds of the future, but one doesn’t have to – robust skill analytics identify funds that will do even better in the future than tomorrow’s largest.

Past Performance of Today’s Largest Hedge Funds

We follow the approach of our earlier piece on hedge fund survivor (survivorship) bias, which analyzed firms’ long U.S. equity portfolios (HF Aggregate). This dataset spans the long portfolios of all hedge funds active over the past 10 years that are tractable using 13F filings.

We compare group returns to Factor Portfolio – a portfolio with matching factor (systematic) risk. Factor Portfolio captures the return of investing passively in ETFs and index futures with the same risk as the group. This comparison reveals security selection (stock picking) performance, or αReturn – outperformance relative to the Factor Portfolio and the return that would have been generated if markets had been flat.

The following chart compares the performance of the 20 largest U.S. equity hedge fund long portfolios (Large HFs, green) to the Factor Portfolio (black). The security selection performance, or αReturn (blue), is the difference between the two. This is the average past performance of the 20 largest funds of 2015:

Chart of the past total, factor, and residual returns of long U.S. equity portfolios of the 20 largest hedge funds of 2015

Current Largest Hedge Funds: Past Total, Factor, and Residual Long U.S. Equity Returns

Returns (%)

Annualized

10-year Cum.

Total

11.48

215.26

Factor

9.33

154.29

Total – Factor

2.15

60.97

Firms that have grown the largest over the past 10 years have performed exceptionally well: Including the effect of compounding, their long portfolios generated 61% higher return than their passive equivalents. If markets had been flat for the past 10 years, their long equity portfolios would have appreciated by nearly 25%.

The allure of this past performance arouses fund-following, guru-tracking, and billionaire portfolio strategies. But there is one problem: Today’s largest funds represent a top-performing sliver of the thousands of funds active in the past. Of the thousands of funds, some truly are skilled, but many simply got lucky on aggressive bets and became large as a result, irrespective of their skill. This constitutes large hedge fund survivor bias. This performance does not persist and tends to mean-revert.

Future Performance of Yesterday’s Largest Hedge Funds

Most billionaire and guru-following strategies make the assumption that the largest funds are likely to continue generating strong returns. To test this, we tracked the 20 largest long U.S. equity hedge fund portfolios of 2005. Below is the unappealing picture of their average performance:

Chart of the future total, factor, and residual returns of long U.S. equity portfolios of the 20 largest hedge funds of 2005

2005 Largest Hedge Funds: Future Total, Factor, and Residual Long U.S. Equity Returns

Returns (%)

Annualized

10-year Cum.

Total

7.70

116.05

Factor

8.68

138.11

Total – Factor

-0.97

-22.05

The 2005 Large HF Aggregate tracked Factor Portfolio closely until 2010 and has struggled since. Hence, including the effects of compounding, large hedge fund survivor bias overstated security selection returns by 80%.

Size does not always signal quality, nor does it guarantee future performance. Between 2005 and 2015, the forward-looking performance of the largest long hedge fund portfolios of 2005 was just over half the backward-looking performance of 2015’s largest. Why then would the largest hedge funds of 2015 perform differently than the poor showing of the 2005 vintage?

Predicting Top Future Hedge Funds: Stock Picking Skill

Absent a time machine, investors cannot know who will be the future stars. However, they need not despair. Instead of focusing on the largest or top-performing funds of the past, they can turn to those showing the highest evidence of skill. The following chart tracks the long U.S. equity portfolios of 20 hedge funds with the highest 3-year αReturn as of 12/31/2005:

Chart of the future total, factor, and residual returns of long U.S. equity portfolios of the 20 best stock picker hedge funds of 2005

Best 2005 Stock Picker Hedge Funds: Future Total, Factor, and Residual Long U.S. Equity Returns

Returns (%)

Annualized

10-year Cum.

Total

12.60

252.58

Factor

9.11

148.70

Total – Factor

3.49

103.88

The funds above were the best stock pickers of 2005, not the largest. If markets had been flat for the past 10 years, the top stock pickers of 2005 would have returned 40%. For a variety of reasons (scalability constraints, lifestyle preferences), many have not become the largest or best known, but their risk-adjusted returns are strong.

Since active management skills persist, skilled stock pickers of the past continue to generate strong nominal and risk-adjusted returns. The same analysis identifies today’s top stock pickers who will be tomorrow’s outperformers – and without the cost of a time machine!

Conclusions

  • Hedge fund survivor bias is larger for the largest hedge funds.
  • Between 2005 and 2015, large hedge fund survivor bias overstated expected nominal performance by nearly 100% and security selection performance by 80%.
  • Chasing large hedge funds is unnecessary and detrimental. Selecting a fund using robust skill analytics, as illustrated by αReturn, is superior to flawed results hampered by large hedge fund survivor bias.
The information herein is not represented or warranted to be accurate, correct, complete or timely.
Past performance is no guarantee of future results.
Copyright © 2012-2015, 
AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.

Hedge Fund Survivor Bias

And The Flaws of Blind Fund-Following Strategies

Numerous financial data and analytics vendors peddle hedge fund tracking strategies and content. Much of this data is hazardous to investors – Hedge fund survivor bias, a special case of the pervasive survivorship bias, is its key flaw. This artifact overstates nominal fund returns by a fifth and conceals mediocre risk-adjusted performance records.

This post is technical, but it illustrates an important phenomenon and sets up the foundation for upcoming articles. We analyze the long equity portfolios of approximately 1,000 medium and lower turnover non-quantitative hedge funds active over the past 10 years (HF Aggregate). This dataset spans the long portfolios of all non-quantitative hedge funds active over the past 10 years that are tractable using 13F filings.

HF Aggregate consists of two approximately equal sub-sets: HF Surviving Aggregate and HF Defunct Aggregate. HF Surviving Aggregate, similar to the datasets of many vendors, gives a deeply misleading picture of average hedge fund performance. Our HF Aggregate corrects this by including HF Defunct Aggregate – funds that stopped filing 13Fs as their U.S. assets dropped below $100 million.

All Hedge Fund Performance

We compare HF Aggregate to Factor Portfolio – a portfolio with matching factor (systematic) risk. Factor Portfolio captures the return investors would have realized if they had passively invested in ETFs and index futures with the same risk as HF Aggregate. We do this to calculate security selection (stock picking) returns of HF Aggregate.

With the exception of the 2009-2011 period, HF Aggregate generated negative returns from security selection. AlphaBetaWorks’ measure of security selection performance is αReturn – outperformance relative to the Factor Portfolio. αReturn is also the return HF Aggregate would have generated if markets were flat. Since 2011, HF Aggregate’s αReturn was -2%. If markets had been flat, the average medium-turnover long hedge fund portfolio would have lost 2% from its long portfolio. Including the effects of compounding with factor returns, αReturn was -3%.

Putting these elements together, the chart below compares HF Aggregate’s performance (green) to the Factor Portfolio (black). The security selection performance, or αReturn (blue), is the difference between the two. This is the true long performance of the average hedge fund:

Chart of the cumulative total, factor, and residual/security selection performance of all medium turnover hedge fund U.S. equity portfolios, free from hedge fund survivor bias

All Medium Turnover U.S. Hedge Fund Long Portfolios: Factor, Residual, and Total Returns

Performance (%)

Annualized

10-year
Total

8.48

133.57

Factor

8.60

136.39

Total – Factor

-0.12

-2.82

Survivor Hedge Fund Performance – Survivorship Bias in Action

The figures above contrast with those promoted by many data vendors and analytics providers. They typically consider (or provide data on) the survivors only – those funds that are still around, active, and reporting their holdings – HF Surviving Aggregate.

Indeed, the performance of surviving hedge funds is superior: their nominal return is 26% higher than HF Aggregate’s and their security selection performance is positive. Not surprisingly, surviving funds have consistently generated positive risk-adjusted returns from security selection, outperforming the replicating Factor Portfolio. This is the performance investors typically see:

Chart of the cumulative total, factor, and residual/security selection performance of surviving medium turnover hedge fund U.S. equity portfolios, affected by the hedge fund survivor bias

Surviving Medium Turnover U.S. Hedge Fund Long Portfolios: Factor, Residual, and Total Returns

Performance (%)

Annualized

10-year

Total

9.54

159.57

Factor

9.06

147.37

Total – Factor

0.48

12.20

Defunct Hedge Fund Performance

The disconnect between these two pictures of average hedge fund performance is due to survivor bias. Of the approximately 1,000 medium turnover hedge funds tractable using 13Fs that have been active filers over the past 10 years, only half remain. The defunct half dropped out of many databases and out of HF Surviving Aggregate. HF Defunct Aggregate struggled under low factor returns and poor security selection. This is the under-performance swept under the rug:

Chart of the cumulative total, factor, and residual/security selection performance of defunct medium turnover hedge fund U.S. equity portfolios, excluded to cause hedge fund survivor bias

Defunct Medium Turnover U.S. Hedge Fund Long Portfolios: Factor, Residual, and Total Returns

Performance (%)

Annualized

10-year

Total

6.07

83.52

Factor

7.14

104.12

Total – Factor

-1.06

-20.60

The difference in performance between surviving and defunct funds is especially dramatic post-2008:

  • Surviving and defunct hedge funds’ long portfolios show similar nominal returns through 2008. Surviving hedge funds are slightly ahead with a 5% higher αReturns.
  • The 2008 draw-down for surviving and defunct hedge funds is similar. Both groups generate negative αReturns: widespread portfolio liquidation devastates crowded hedge fund bets across both groups.
  • From 2009 the survivors decouple from the defunct: Defunct funds trim exposures. Surviving funds boost exposures.
  • Since 2009 HF Surviving Aggregate outperforms HF Defunct Aggregate by over 70%. Approximately half is due to higher systematic risk and half is due to security selection.
  • Survival is mostly a matter of exposure and stock picking.

Absent a time machine, investors and fund followers cannot know who will be the future survivors. HF Defunct Aggregate consists of survivors that did well enough to last until 2005, but subsequently perished. Unfortunately, many strategies are built on a swampy foundation – the assumption that the average hedge fund is the same as the average surviving hedge fund. True fund performance is a fifth lower.

Consequently, robust skill analytics developed with the understanding of hedge fund survivor bias are critical to keep investors out of yesterday’s winners that tend to become tomorrow’s losers.

Conclusions

  • Historical performance of surviving hedge funds overstates actual average returns by a fifth.
  • Hedge fund survivor bias boosts 10-year nominal returns by 26%, primarily post-2008.
  • Hedge fund survivor bias boosts 10-year security selection returns by approximately 15%.
  • Fund performance and holdings studies that ignore survivor bias will deliver misleading conclusions and disappointing performance.
The information herein is not represented or warranted to be accurate, correct, complete or timely.
Past performance is no guarantee of future results.
Copyright © 2012-2015, 
AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.

Berkshire’s Energy Investment Skills

Should Investors Follow Buffet out of XOM?

Berkshire Hathaway’s year-end 2014 Form 13F showed the liquidation of the approximately $4 billion Exxon Mobil (XOM) position. This sale has generated considerable discussion. Absent data on Berkshire’s Energy Sector record, the sale is uninformative; we provide this data here.

Investors typically treat all ideas of excellent managers with equal deference. This is usually a mistake – even the most skilled managers are seldom equally skilled in all areas. However, Berkshire Hathaway has excellent track record of security selection in the energy sector. Investors should take note of this particular sale.

Berkshire Hathaway’s Security Selection

The risk-adjusted return of Berkshire’s long equity portfolio, estimated from the firm’s 13F filings, is spectacular. We estimate an approximately 60% cumulative return from security selection (stock picking) over the past 10 years. This is αReturn, a metric of security selection performance – the estimated annual percentage return the portfolio would have generated if markets were flat. Berkshire’s cumulative αReturn is shown in blue in the chart below. For comparison, the group of U.S. hedge funds generated slightly negative long security selection returns over the period (in gray):

Chart of the historical return from security selection (stock picking) of Berkshire Hathaway

Berkshire Hathaway’s Security Selection Return

Berkshire Hathaway’s Energy Security Selection

Berkshire’s risk-adjusted return in the Energy Sector is also excellent, though less consistent. If markets were flat over the past 10 years, the long energy portfolio would have returned over 125% compared to a greater than 10% loss for the average hedge fund:

Chart of the historical risk adjusted return from security selection (stock picking) of Berkshire Hathaway  in the Energy Sector

Berkshire Hathaway’s Energy Security Selection Return

All five energy investments over the past 10 years generated positive residual returns un-attributable to the market. These are the sources of the risk-adjusted returns from security selection:

Return (%)

Symbol Name

Total

Factor

Residual

COP ConocoPhillips

94.54

88.12

6.42

PSX Phillips 66

40.33

13.03

27.29

PTR PetroChina Co. Ltd. Sponsored ADR

187.96

104.42

83.54

SU Suncor Energy Inc.

106.19

94.98

11.21

XOM Exxon Mobil Corporation

60.52

50.65

9.86

Stock picking performance persists. Therefore, the sale of XOM by Berkshire is indeed a negative indicator.

Berkshire Hathaway’s Energy Market Timing

While Berkshire shows significant skill in selecting energy stocks, it does not appear skilled in timing the overall energy market. There is no statistically significant relationship between Berkshire’s exposure to the Energy Factor and the factor’s subsequent return:

Chart of Berkshire Hathaway 's Energy Factor timing: the relationship between energy factor exposure and return

Berkshire Hathaway’s Energy Market Timing

Therefore, Berkshire’s sale of XOM is not a bearish indicator for the overall energy sector.

Summary

  • Managers’ trades are predictive only in areas where the managers display statistically significant investment skills (or lack thereof).
  • Berkshire Hathaway has a strong record of energy security selection (stock picking). Consequently, the sale of XOM is a bearish indicator for this particular stock.
  • Berkshire Hathaway does not have a consistent record of energy market timing. Consequently, the sale of XOM is not an indicator for the sector in general.
The information herein is not represented or warranted to be accurate, correct, complete or timely.
Past performance is no guarantee of future results.
Copyright © 2012-2015, AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.

Hedge Funds’ Best and Worst Sectors

Due to the congestion of their investor base, crowded hedge fund stocks are volatile and vulnerable to mass selling. The risk-adjusted performance of consensus bets tends to disappoint. In two past pieces we illustrated the toll of crowding on exploration and production as well as internet companies. We also reviewed two specific crowded bets: SanDisk and eHealth.

While crowded hedge fund ideas do poorly most of the time, they don’t always. Market efficiency varies across sectors, and some industries are more analytically tractable than others. In this article we survey the sectors with the best and worst hedge fund performance records. We will illustrate when investors should stay clear of crowded ideas and when they can embrace them.

Analyzing Hedge Fund Performance and Crowding

To explore performance and crowding we analyze hedge fund sector holdings (HF Sector Aggregate) relative to the Sector Market Portfolio (Sector Aggregate). HF Sector Aggregate is position-weighted, and Sector Aggregate is capitalization-weighted. This follows the approach of our earlier articles on aggregate and sector-specific hedge fund crowding.

Hedge Funds’ Worst Sector: Miscellaneous Metals and Mining

Historical Hedge Fund Performance: Miscellaneous Metals and Mining

Hedge funds’ worst security selection performance for the past ten years has been in the Miscellaneous Metals and Mining sector. The figure below plots historical HF Miscellaneous Metals and Mining Aggregate’s return. Factor return is due to systematic (market) risk. It is the return of a portfolio that replicates HF Sector Aggregate’s market risk. The blue area represents positive and the gray area represents negative risk-adjusted returns from security selection (αReturn).

Chart of the historical total, factor, and security selection performance of the Hedge Fund Miscellaneous Metals and Mining Sector Aggregate

Hedge Fund Miscellaneous Metals and Mining Sector Aggregate Historical Performance

Even without adjusting for risk, crowded bets have done poorly. They consistently underperformed the factor portfolio, missing out on over 300% in gains.

The HF Sector Aggregate’s risk-adjusted return from security selection (αReturn) is the return it would have generated if markets were flat – all market effects on performance have been eliminated. This idiosyncratic performance of the crowded portfolio is a decline of 87%. Crowded bets in this sector are especially dangerous, given their persistently poor performance:

Chart of the historical security selection performance of the Hedge Fund Miscellaneous Metals and Mining Sector Aggregate

Hedge Fund Miscellaneous Metals and Mining Sector Aggregate Historical Security Selection Performance

In this sector, hedge funds lost $900 million to other market participants. In commodity industries, the recipients of this value transfer are usually private investors and insiders.

Current Hedge Fund Bets: Miscellaneous Metals and Mining

The following stocks contributed most to the relative residual (security-specific) risk of the HF Miscellaneous Metals and Mining Sector Aggregate as of Q3 2014. Blue bars represent long (overweight) exposures relative to the Sector Aggregate. White bars represent short (underweight) exposures. Bar height represents contribution to relative stock-specific risk:

Chart of the top contributors' contribution to the Hedge Fund Miscellaneous Metals and Mining Sector Aggregate's risk

Crowded Hedge Fund Miscellaneous Metals and Mining Sector Bets

The following table contains detailed data on these crowded bets. Large and illiquid long (overweight) bets are most at risk of volatility, mass liquidation, and underperformance:

Exposure (%) Net Exposure Share of Risk (%)
HF Sector Aggr. Sector Aggr. % $mil Days of Trading
ZINC Horsehead Holding Corp. 72.74 2.41 70.33 148.5 15.6 80.55
SLCA U.S. Silica Holdings, Inc. 0.30 9.68 -9.39 -19.8 -0.2 6.45
LEU Centrus Energy Corp. Class A 4.54 0.22 4.32 9.1 17.2 4.85
SCCO Southern Copper Corporation 7.69 70.19 -62.51 -132.0 -2.3 4.18
CSTE CaesarStone Sdot-Yam Ltd. 0.00 5.18 -5.18 -10.9 -0.8 1.14
MCP Molycorp, Inc. 3.84 0.84 3.01 6.3 1.7 0.92
MTRN Materion Corporation 7.15 1.82 5.33 11.3 2.1 0.69
HCLP Hi-Crush Partners LP 0.49 2.90 -2.41 -5.1 -0.2 0.35
CA:URZ Uranerz Energy Corporation 2.00 0.27 1.72 3.6 11.7 0.29
IPI Intrepid Potash, Inc. 0.36 3.38 -3.02 -6.4 -0.5 0.22
OROE Oro East Mining, Inc. 0.00 0.52 -0.52 -1.1 -39.9 0.05
CANK Cannabis Kinetics Corp. 0.00 0.10 -0.10 -0.2 -2.7 0.05
UEC Uranium Energy Corp. 0.00 0.33 -0.33 -0.7 -0.4 0.02
FCGD First Colombia Gold Corp. 0.00 0.09 -0.09 -0.2 -19.0 0.02
MDMN Medinah Minerals, Inc. 0.00 0.16 -0.16 -0.3 -4.8 0.01
QTMM Quantum Materials Corp. 0.00 0.13 -0.13 -0.3 -6.3 0.00
ENZR Energizer Resources Inc. 0.00 0.12 -0.12 -0.3 -11.7 0.00
AMNL Applied Minerals, Inc. 0.00 0.20 -0.20 -0.4 -18.5 0.00
LBSR Liberty Star Uranium and Metals Corp. 0.00 0.03 -0.03 -0.1 -4.9 0.00
Other Positions 0.61 0.21
Total 100.00

Hedge Funds’ Best Sector: Real Estate Development

Historical Hedge Fund Performance: Real Estate Development

Hedge funds’ best security selection performance has been in the Real Estate Development Sector. The figure below plots the historical return of HF Real Estate Development Aggregate. Factor return and αReturn are defined as above:

Chart of the historical total, factor, and security selection returns of the Hedge Fund Real Estate Development Sector Aggregate

Hedge Fund Real Estate Development Sector Aggregate Historical Performance

Since 2004, the HF Sector Aggregate outperformed the portfolio with equivalent market risk by approximately 200%. In a flat market, HF Sector Aggregate would have gained approximately 180%:

Chart of the historical security selection (residual) return of the Hedge Fund Real Estate Development Sector Aggregate

Hedge Fund Real Estate Development Sector Aggregate Historical Security Selection Performance

In this sector, hedge funds gained $1 billion at the expense of other market participants. The Real Estate Development Sector appears less efficient but tractable, providing hedge funds with consistent stock picking gains.

Current Hedge Fund Real Estate Development Bets

The following stocks contributed most to the relative residual (security-specific) risk of the HF Real Estate Development Sector Aggregate as of Q3 2014:

Chart of the contribution to the residual (stock-specific) risk of the various hedge fund Crowded Hedge Fund Real Estate Development Sector bets

Crowded Hedge Fund Real Estate Development Sector Bets

The following table contains detailed data on these crowded bets. Since in this sector hedge funds are “smart money,” large long (overweight) bets are most likely to outperform and large short (underweight) bets at most likely to do poorly:

Exposure (%) Net Exposure Share of Risk (%)
HF Sector Aggr. Sector Aggr. % $mil Days of Trading
HHC Howard Hughes Corporation 28.47 15.98 12.49 326.5 17.5 36.73
CBG CBRE Group, Inc. Class A 52.28 26.54 25.74 672.7 10.8 27.58
JLL Jones Lang LaSalle Incorporated 0.14 15.21 -15.07 -393.9 -8.5 12.86
JOE St. Joe Company 0.04 4.94 -4.91 -128.2 -13.5 8.82
ALEX Alexander & Baldwin, Inc. 0.00 4.71 -4.71 -123.2 -13.5 5.38
HTH Hilltop Holdings Inc. 1.35 4.86 -3.51 -91.8 -18.3 4.29
KW Kennedy-Wilson Holdings, Inc. 3.60 6.11 -2.51 -65.6 -7.6 1.19
TRC Tejon Ranch Co. 3.36 1.55 1.81 47.2 37.9 0.77
EACO EACO Corporation 0.00 0.22 -0.22 -5.7 -436.1 0.65
FOR Forestar Group Inc. 0.62 1.66 -1.05 -27.3 -5.3 0.42
FCE.A Forest City Enterprises, Inc. Class A 8.78 10.56 -1.78 -46.5 -1.9 0.35
SBY Silver Bay Realty Trust Corp. 0.07 1.68 -1.61 -42.0 -8.4 0.23
AVHI A V Homes Inc 0.26 0.87 -0.61 -15.8 -28.7 0.20
MLP Maui Land & Pineapple Company, Inc. 0.00 0.29 -0.29 -7.5 -132.0 0.10
CTO Consolidated-Tomoka Land Co. 0.16 0.77 -0.61 -15.9 -24.5 0.09
RDI Reading International, Inc. Class A 0.02 0.54 -0.52 -13.7 -14.2 0.08
ABCP AmBase Corporation 0.00 0.15 -0.15 -3.8 -130.1 0.06
AHH Armada Hoffler Properties, Inc. 0.00 0.59 -0.59 -15.5 -9.4 0.06
OMAG Omagine, Inc. 0.00 0.07 -0.07 -1.9 -24.7 0.05
FVE Five Star Quality Care, Inc. 0.26 0.49 -0.23 -6.1 -5.1 0.04
Other Positions 0.01 0.07
Total 100.00

Real Estate Development is not the only sector where hedge funds excel. Crowded Coal, Hotels, and Forest Product sector ideas have also done well. Skills vary within each sector: The most skilled funds persistently generate returns in excess of the crowd, while the least skilled funds persistently fall short. Performance analytics built on robust risk models help investors and allocators reliably identify each.

Conclusions

  • With proper data, attention to hedge fund crowding prevents “unexpected” volatility and losses.
  • Market efficiency and tractability vary across sectors – crowded hedge fund bets do poorly in most sectors, but do well in some.
  • Investors should avoid crowded ideas in sectors of persistent hedge fund underperformance, such as Miscellaneous Metals and Mining.
  • Investors can embrace crowded ideas in sectors of persistent hedge fund outperformance, such as Real Estate Development.
  • Funds with significant and persistent stock picking skills exist in most sectors, even those with generally poor hedge fund performance. AlphaBetaWorks’ Skill Analytics identify best overall and sector-specific stock pickers.
The information herein is not represented or warranted to be accurate, correct, complete or timely.
Past performance is no guarantee of future results.
Copyright © 2012-2015, AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.

Returns-Based Style Analysis – Overfitting and Collinearity

Plagued by overfitting and collinearity, returns-based style analysis frequently fails, confusing noise with portfolio risk.

Returns-based style analysis (RBSA) is a common approach to investment risk analysis, performance attribution, and skill evaluation. Returns-based techniques perform regressions of returns over one or more historical periods to compute portfolio betas (exposures to systematic risk factors) and alphas (residual returns unexplained by systematic risk factors). The simplicity of the returns-based approach has made it popular, but it comes at a cost – RBSA fails for active portfolios. In addition, this approach is plagued by the statistical problems of overfitting and collinearity, frequently confusing noise with systematic portfolio risk. 

Returns-Based Style Analysis – Failures for Active Portfolios

In an earlier article we illustrated the flaws of returns-based style analysis when factor exposures vary, as is common for active funds:

  • Returns-based analysis typically yields flawed estimates of portfolio risk.
  • Returns-based analysis may not even accurately estimate average portfolio risk.
  • Errors will be most pronounced for the most active funds:
    • Skilled funds may be deemed unskilled.
    • Unskilled funds may be deemed skilled.

These are not the only flaws. We now turn to the subtler and equally critical issues – failures in the underlying regression analysis itself. We use a recent Morningstar article as an example.

iShares Core High Dividend ETF (HDV) – Returns-Based Style Analysis

A recent Seeking Alpha article provides an excellent illustration of problems created by overfitting and collinearity. In this article, Morningstar performed a returns-based style analysis of iShares Core High Dividend ETF (HDV).

Morningstar estimated the following factor exposures for HDV using the Carhart model:

Morningstar: Returns-Based Analysis of the iShares Core High Dividend ETF (HDV) Using the Carhart Model

iShares Core High Dividend ETF (HDV) – Estimated Factor Exposures Using the Carhart Model – Source: Morningstar

The Mkt-RF coefficient, or loading, is HDV’s estimated market beta. A beta value of 0.67 means that given a +1% change in the market HDV is expected to move by +0.67%, everything else held constant.

The article then performs RBSA using an enhanced Carhart + Quality Minus Junk (QMJ) model:

Morningstar: Returns-Based Analysis of iShares Core High Dividend ETF (HDV) Using the Carhart + Quality Minus Junk (QMJ) Model

iShares Core High Dividend ETF (HDV) – Estimated Factor Exposures Using the Carhart + Quality Minus Junk (QMJ) Model – Source: Morningstar

With the addition of the QMJ factor, the market beta estimate increased by a third from 0.67 to 0.90. Both estimates cannot be right. Perhaps the simplicity of the Carhart model is to blame and the more complex 5-factor RBSA is more accurate?

iShares Core High Dividend ETF (HDV) – Historical Factor Exposures

Instead of Morningstar’s RBSA approach, we analyzed HDV’s historical holdings using the AlphaBetaWorks’ U.S. Equity Risk Model. For each month, we estimated the U.S. Market exposures (betas) of individual positions and aggregated these into monthly estimates of portfolio beta:

Chart of the historical market exposure (beta) of iShares Core High Dividend ETF (HDV)

iShares Core High Dividend ETF (HDV) – Historical Market Exposure (Beta)

Over the past 4 years, HDV’s market beta varied in a narrow range between 0.50 and 0.62.

Both of the above returns-based analyses were off, but the simpler Carhart model did best. It turns out the simpler and a less sophisticated returns-based model is less vulnerable to the statistical problems of multicollinearity and overfitting. Notably, the only way to find out that returns-based style analysis failed was to perform the more advanced holdings-based analysis using a multi-factor risk model.

Statistical Problems with Returns-Based Analysis

Multicollinearity

Collinearity (Multicollinearity) occurs when risk factors used in returns-based analysis are highly correlated with each other. For instance, small-cap stocks tend to have higher beta than large-cap stocks, so the performance of small-cap stocks relative to large-cap stocks is correlated to the market.

Erratic changes in the factor exposures for various time periods, or when new risk factors are added, are signs of collinearity. These erratic changes make it difficult to pin down factor exposures and are signs of deeper problems:

A principal danger of such data redundancy is that of overfitting in regression analysis models.
-Wikipedia

Overfitting

Overfitting is a consequence of redundant data or model over-complexity. These are common for returns-based analyses which usually attempt to explain a limited number of return observations with a larger number of correlated variable observations.

An overfitted returns-based model may appear to describe data very well. But the fit is misleading – the exposures may be describing noise and will change dramatically under minor changes to data or factors. A high R squared from returns-based models may be a sign of trouble, rather than a reassurance.

As we have seen with the HDV example above, exposures estimated by RBSA may bear little relationship to portfolio risk. Therefore, all dependent risk and skill data will be flawed.

Conclusions

  • When a manager does not vary exposures to the market, sector, and macroeconomic factors, returns-based style analysis (RBSA) using a parsimonious model can be effective.
  • When a manager varies bets, RBSA typically yields flawed estimates of portfolio risk.
  • Even when exposures do not vary, returns-based style analysis is vulnerable to multicollinearity and overfitting:
    • The model may capture noise, rather than the underlying factor exposures.
    • Factor exposures may vary erratically among estimates.
    • Estimates of portfolio risk will be flawed.
    • Skilled funds may be deemed unskilled.
    • Unskilled funds may be deemed skilled.
  • Holdings-based analysis using a robust multi-factor risk model is superior for quantifying fund risk and performance.
The information herein is not represented or warranted to be accurate, correct, complete or timely.
Past performance is no guarantee of future results.
Copyright © 2012-2015, AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.

Upgrading Fund Active Returns

And Not Missing Out

Maybe your fund took extra risk to keep up with its benchmark. Maybe your fund should have made more – much more – given the risks it took. By the time market volatility reveals underlying exposures, it may be too late to avoid severe losses. There is a better way: Investors can continuously monitor a fund’s risk, the returns it should be generating, and the value it creates. This value should matter most to investors and allocators. Regrettably, most fund analysis tools and services pay no attention to it.

To illustrate, we analyze two funds: one that did much worse than it should have, and one that did better.

PRSCX – Negative Active Returns

The T. Rowe Price Science & Technology Fund (PRSCX) manages approximately $3 billion. This fund generally tracks its benchmark and it gets 3 star rating from a popular service. Notwithstanding this, PRSCX has produced persistently negative active returns. Given its historical risk, PRSCX should have made investors far more money: Over the past ten years, an investor would have made 50-80% more owning a passive portfolio with PRSCX’s risk profile.

Chart of the historical cumulative passive and active returns of T. Rowe Price Science & Technology Fund (PRSCX)

T. Rowe Price Science & Technology Fund (PRSCX) – Passive and Active Return History

While we seem to bolster arguments for passive investing, reality is more complex: Active returns (both positive and negative) persist over time. Thus, upgrading from PRSCX to a fund with persistently positive active returns is a superior move. We will provide one candidate.

PRSCX – Historical Risk

The chart below shows PRSCX’s historical risk (exposures to significant risk factors). The red dots indicate monthly exposures (as a percentage of assets) over the past 10 years; the black diamonds indicate latest exposures:

Chart of the historical exposures of T. Rowe Price Science & Technology Fund (PRSCX) to significant risk factors

T. Rowe Price Science & Technology Fund (PRSCX) – Exposure to Significant Risk Factors

PRSCX varied its exposures over time. U.S. Market is the most important exposure, reaching 200% (market beta of 2) at times. As expected for a technology fund, its U.S. Technology exposure has been near 100%. Also note PRSCX’s occasional short bond exposure. Many equity funds carry large hidden bond bets due to the risk profile of their equity holdings. Most investors and portfolio managers are not aware of these bets. Yet for these funds, bond risk is a key driver of portfolio returns and volatility.

PRSCX – Historical Active Returns

The above exposures define a passive replicating portfolio matching PRSCX’s risk. The fund manager’s job is to outperform this passive alternative by generating active returns.

To isolate active returns, we quantify passive factor exposures, estimate the passive return, and then calculate the remaining active return – αβReturn. We further break down αβReturn into risk-adjusted return from security selection, or stock picking (αReturn), and market timing (βReturn):

Component 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Total 1.6 2.46 7.1 11.88 -43.8 67.83 21.25 -4.51 6.25 43.7 9.39
Passive -0.69 1.4 5.45 19.26 -46.13 77.52 23.21 -3.42 20.36 51.54 13.61
αβReturn 0.98 -0.69 -2.2 -7.45 2.71 -11.61 -2.22 -8.24 -13.76 -14.14 -5.84
αReturn -1.95 -3.29 -6.45 -2.79 7.92 0.78 4.17 -12.31 -11.47 -4.54 1.11
βReturn 2.94 2.6 4.25 -4.66 -5.2 -12.39 -6.39 4.07 -2.29 -9.6 -6.95
Undefined 1.3 1.75 3.85 0.07 -0.38 1.92 0.26 7.15 -0.35 6.3 1.61

Note that we are unable to account for trades behind some of the returns – the “Undefined” component. It may be due to private securities or intra-period trading; it may be passive or active. Yet, even if we assume that all undefined returns above are active, PRSCX still delivered persistently negative αβReturn over the past ten years. Furthermore, the compounding of negative αβReturn leaves investors missing out on 50-80% in gains.

FSCSX – An Upgrade Option with Similar Historical Risk

While a passive portfolio would have been superior to PRSCX, it is not the best upgrade. Allocators and investors can do better owning a fund with consistently positive αβReturns, since αβReturns persist. One candidate is Fidelity Select Software & Computer Services Portfolio (FSCSX):

Chart of the historical exposures of Fidelity Select Software & Computer Services Portfolio (FSCSX) to significant risk factors

Fidelity Select Software & Computer Services Portfolio (FSCSX) – Exposure to Significant Risk Factors

Currently, FSCSX and PRCSX have similar exposures. AlphaBetaWorks’ risk analytics estimate the current annualized tracking error between the two funds at a 5.29% (about the same volatility as bonds, and less than one half of market volatility).

FSCSX – Historical Active Returns

FSCSX’s 3-year trailing average annual return of 23% is slightly ahead of PRSCX’s 20%. But most importantly, given its lower historical risk, FSCSX has delivered positive αβReturns versus PRSCX’s significantly negative ones. The chart below shows FSCSX’s ten-year performance. The purple area is the positive αβReturn. The gray area is FSCSX’s passive return:

Chart of the historical cumulative passive and active returns of Fidelity Select Software & Computer Services Portfolio (FSCSX)

Fidelity Select Software & Computer Services Portfolio (FSCSX) – Passive and Active Return History

FSCSX is superior to a passive portfolio with similar risk and to PRSCX. Mind you, this is not a sales pitch for FSCSX but merely a consequence of its positive αβReturn and αβReturn persistence.

Few fund investors and allocators possess the tools to quantify active returns. Yet, this knowledge is an essential competitive advantage, leading to improved client returns, client retention, and asset growth. Unfortunately, many are content to pick funds based on past nominal returns and to suffer the consequences: picking yesterday’s winners tends to pick tomorrow’s losers. AlphaBetaWorks spares clients from the data processing headaches, financial modeling, and statistical analysis of thousands of portfolios, delivering predictive risk and skill analytics on thousands of funds.

Conclusions

  • Analyzing a fund’s performance relative to a benchmark ignores the most important question: What should you have made given its risk?
  • Some mutual funds produce persistently negative active returns; others produce persistently positive active returns.
  • Upgrading from a fund with persistently negative active return (αβReturn) to a replicating passive portfolio tends to improve performance.
  • Upgrading from a passive portfolio to a fund with persistently positive αβReturn also tends to improve performance.
  • Tools that accurately estimate fund risk and active returns provide enduring competitive advantages for investors and professional allocators, leading to improved client returns, client retention, and asset growth.
The information herein is not represented or warranted to be accurate, correct, complete or timely.
Past performance is no guarantee of future results.
Copyright © 2012-2014, 
AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.

Hedge Fund E&P Crowding – Q2 2014

U.S. hedge funds share a few systematic and idiosyncratic long bets – a phenomenon called “crowding.” Hedge fund crowding within specific sectors can be heavy; bets on exploration and production (E&P) companies are particularly crowded. Hedge fund E&P bets are the subject of this article. Eight stocks are responsible for three quarters of the herding.

Crowding is costly to investors, fund managers, and allocators: over the past 10 years the aggregate hedge fund E&P portfolio underperformed the market E&P portfolio by 23% while taking more risk. The risk-adjusted return was even worse – a loss of 52%.

Identifying Hedge Fund E&P Crowding

To evaluate hedge fund (HF) exploration and production (E&P) herding, we followed the approach of our earlier work on aggregate and sector-specific hedge fund crowding: We created an aggregate position-weighted portfolio (HF E&P Aggregate) consisting of all exploration and production long equity positions reported by over 400 U.S. hedge funds with medium to low turnover. We then evaluated HF E&P Aggregate’s risk relative to the capitalization-weighted portfolio of E&P equities (Market E&P Aggregate) using AlphaBetaWorks’ Statistical Equity Risk Model.

Crowded E&P Stocks Underperform

Crowding hurts performance. HF E&P Aggregate had poor returns following the peak of the last energy cycle in 2008. Consequently, understanding Hedge Fund E&P crowding is vital to investors, fund managers, and allocators.

When the broad market and the E&P sector are doing well, crowded hedge fund E&P stocks generally outperform. However, these stocks generally underperform in the down cycle. This is a tell-tale sign of flocking to higher-risk stocks. This crowding towards higher market and sector betas is consistent with the aggregate systematic crowding of hedge funds:

Chart of the historical returns of Hedge Fund Exporation and Production (E&P) Aggregate and Market E&P Aggregate

Historical Return for Hedge Fund E&P Aggregate vs. Market E&P Aggregate

Indeed, we will see later that hedge fund E&P aggregate has both higher market exposure (market beta) and higher E&P sector exposure (E&P sector beta) than the market E&P Aggregate.

When we adjust for the systematic (factor) risk, we discover that the residual return of HF E&P aggregate due to security selection is even worse. Investors would have made 52% more over the past 10 years if they had invested in an ETF portfolio with similar factor risk:

Chart of the historical nominal, factor, and residual (risk-adjusted returns due to security selection ) of Hedge Fund Exploration and Production (E&P) Aggregate

Historical Hedge Fund E&P Aggregate, Factor, and Residual Returns

Hedge Fund E&P Risk

HF E&P Aggregate has a 5.0% estimated future tracking error relative to Market E&P Aggregate, primarily due to stock-specific bets:

Chart of the factor and residual contribution to the hedge fund E&P crowding, measured as the relative variance of Hedge Fund Exploration and Production (E&P) Aggregate

Sources of Relative Risk for Hedge Fund E&P Aggregate

Source Volatility (%) Share of Variance (%)
Factor 2.77 24.67
Residual 4.31 75.33
Total 4.96 100.00

The 5.0% tracking error means that HF E&P Aggregate’s annual return is forecasted to differ from Market E&P Aggregate’s by less than 5.0% two thirds of the time.

Hedge Fund Factor (Systematic) E&P Crowding

On the chart below, HF E&P Aggregate’s factor exposures (red) are similar to Market E&P Aggregate’s (gray), but tend to be higher. Hedge Funds tend to invest in names with higher market and sector betas, perhaps as a logical consequence of their compensation structure:

Exposure of Hedge Fund Exploraton and Production (E&P) Aggregate to factors contributing most to hedge fund E&P crowding, or risk relative to Market E&P Aggregate

Hedge Fund E&P Aggregate’s Exposure to Significant Risk Factors

Hedge Fund Residual (Idiosyncratic) E&P Crowding

Over 75% of the relative risk of HF E&P Aggregate is due to stock-specific (residual) bets. Below are the sources of HF E&P Aggregate’s relative residual variance. Three quarters of the estimated relative residual risk is due to only eight stocks:

Stocks contributing most to hedge fund E&P crowding: their contribution to the relative residual variance of Hedge Fund E&P Aggregate

Stocks Contributing Most to Relative Residual Variance of Hedge Fund E&P Aggregate

As with HF Energy Aggregate, some of the largest bets are not the stocks hedge funds own, but the stocks they don’t own. (For example, hedge funds are underweight COP, OXY, and EOG.) The crowded bets are likely to deliver negative risk-adjusted returns in flat or declining oil and gas producer cycle.

Position (%)
Symbol Name HF E&P
Aggregate
Market E&P
Aggregate
Relative Share of
R
isk (%)
ATHL Athlon Energy, Inc. 7.95 0.54 7.41 16.64
CHK Chesapeake Energy Corporation 9.63 2.41 7.23 14.96
COP ConocoPhillips 0.51 12.30 -11.79 12.21
OXY Occidental Petroleum Corporation 0.36 9.33 -8.97 9.18
EOG EOG Resources, Inc. 0.87 7.46 -6.60 6.65
CIE Cobalt International Energy, Inc. 3.31 0.87 2.44 5.11
CNQ Canadian Natural Resources Limited 6.06 0.00 6.06 4.58
EPE EP Energy Corp. Class A 11.31 0.66 10.65 4.55
TLM Talisman Energy Inc. 3.96 0.00 3.96 2.69
SD SandRidge Energy, Inc. 2.67 0.41 2.26 2.68
CLR Continental Resources, Inc. 0.09 3.43 -3.34 2.09
CA:CTA Crocotta Energy Inc. 0.18 0.00 0.18 1.93
WLL Whiting Petroleum Corporation 4.08 1.12 2.95 1.92
NBL Noble Energy, Inc. 0.33 3.29 -2.96 1.09
PXD Pioneer Natural Resources Company 6.12 3.83 2.28 1.02
MRO Marathon Oil Corporation 0.08 3.13 -3.05 0.95
OAS Oasis Petroleum Inc. 2.48 0.66 1.83 0.84
PRMRF Paramount Resources Ltd. Class A 1.26 0.00 1.26 0.81
APA Apache Corporation 1.63 4.49 -2.86 0.78
ATHOF Athabasca Oil Corporation 1.13 0.00 1.13 0.77

Of course, the poor long-term performance of the HF E&P aggregate does not mean that some individual equities will not do well. In fact, in the third quarter of 2014, ATHL was acquired. With half of the shares owned by hedge funds, ATHL was one of the largest positions of HF E&P Aggregate and the top idiosyncratic bet.

Summary

  • Hedge Fund Exploration and Production (E&P) Aggregate tends to have higher risk than Market E&P Aggregate.
  • Crowded Hedge Fund E&P stocks tend to underperform market aggregate over the long-term, in spite of higher risk.
  • Crowded Hedge Fund E&P stocks tend to generate negative risk-adjusted returns.
  • There is both factor (systematic/market) and residual (idiosyncratic/security-specific) crowding of long hedge fund E&P portfolios.
  • Over 80% of recent crowding is attributable to eight stocks: ATHL, CHK, COP, OXY, EOG, CIE, CNQ, and EPE.
The information herein is not represented or warranted to be accurate, correct, complete or timely.
Past performance is no guarantee of future results.
Copyright © 2012-2014, 
AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.

Top-Performing Hedge Fund Profile – Pershing Square

A Survey of Pershing Square’s Security Selection and Market Timing

Not all outperformance is true outperformance. There are many funds whose performance looks spectacular on the surface, but whose risk-adjusted performance is poor. This article takes a closer look at Pershing Square Capital Management, nominally one of the top performing funds throughout 2014 and over the long-term. We show that the risk-adjusted performance of Pershing’s reported long equity positions is, in fact, impressive. We also delve into sectors and factors behind these outstanding risk-adjusted results.

Risk-Adjusted Performance Defined

The AlphaBetaWorks (ABW) Performance Analytics Platform identifies risk-adjusted performance as αβReturn – performance relative to a passive replicating portfolio. αβReturn consists of performance due to security selection (αReturn) and market timing (βReturn). αReturn is the return a fund would have generated if markets were flat. βReturn is the return a fund generated by varying its factor exposures.

Pershing Square – Risk-Adjusted Returns

The fund’s 10-year long αβReturn is around 300%, while passive long return is over 500%. Below we illustrate Pershing’s overall active return (αβReturn) and passive return over the past ten years. The purple area highlights cumulative positive active return; cumulative passive return is the gray area underneath:

Chart of the historical active returns (market timing and security selection) of Pershing Square's reported long equity portfolio

Pershing Square Capital Management Historical Passive and Active Returns – Long Equity Portfolio

Below we provide the components of Pershing’s estimated annual returns. Active returns (αβReturns) are broken down into security selection, or stock picking, (αReturn) and market timing (βReturn). Most of the 2014 return is active: roughly two thirds from stock picking and one third from market timing.

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Total 130.78 31.65 29.23 -0.61 -33.27 52.18 32.09 2.74 21.18 35.03 31.21
Passive 25.06 5.20 15.92 4.62 -35.87 41.23 24.38 5.68 19.00 34.16 7.95
αβReturn 105.72 26.45 13.31 -5.23 2.60 10.94 7.71 -2.94 2.17 0.87 23.26
αReturn 124.96 39.02 10.13 -0.81 2.69 11.09 12.22 -5.25 -1.85 -0.36 14.62
βReturn -19.24 -12.57 3.19 -4.42 -0.09 -0.15 -4.51 2.31 4.02 1.24 8.63

We will now delve deeper into the sources of Pershing’s high risk-adjusted returns – the sectors and the factors that contributed most to αβReturn.

Pershing Square – Long Equity Security Selection

αReturn is the risk-adjusted return from security selection – the return a fund would have generated if markets were flat. Pershing’s recent and 10-year results are stellar:

Chart of the historical security selection return of Pershing Square's reported long equity portfolio

Pershing Square Capital Management Historical Security Selection Return – Long Equity Portfolio

Furthermore, Pershing’s trailing three-year annualized αReturn of 5% exceeds αReturns of 81% of its peers:

Chart of the distribution of security selection returns of U.S. medium turnover hedge funds' long equity portfolios relative to Pershing Square

Pershing Square Capital Management 3-year Sector Security Selection Return vs. Peers

Sector Security Selection

Digging deeper, most of the 2014 αReturn came from the Industrial and Consumer sectors. Over history, there have been a few years of especially large positive (and negative) sector-specific αReturns:

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Communications 1.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Consumer 60.30 0.21 7.63 -0.18 0.75 2.75 8.74 1.01 -14.08 -3.97 3.11
Finance 63.47 38.81 2.50 -0.46 0.64 3.88 3.51 -10.26 2.67 -3.43 1.35
Health 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.80
Industrial 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3.96 9.55 7.03 8.37
Miscellaneous 0.00 0.00 0.00 -0.17 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Technology 0.00 0.00 0.00 0.00 1.30 4.47 -0.03 0.03 0.00 0.00 0.00

The Industrial sector is an area of consistent strength:

Chart of the security selection return of Pershing Square's reported long equity positions in the industrial sector

Pershing Square Capital Management Industrial Sector Security Selection Return – Long Equity Portfolio

Pershing’s recent long positions in the industrial sector are:

APD Air Products and Chemicals, Inc.;
CA:CP Canadian Pacific Railway;
GB:PAH Platform Specialty Products Corp.

Pershing’s Consumer sector αReturn has been mixed over the past 3 years; it has been excellent over the long-term and in 2014:

Chart of the security selection return of Pershing Square's reported long equity positions in the industrial sector

Pershing Square Capital Management Consumer Sector Security Selection Return – Long Equity Portfolio

Pershing’s recent long position in the consumer sector is:

BKW Burger King Worldwide, Inc.

Additional security-selection insights, such as position sizing skill and the analysis of portfolio scalability/overcapitalization, are available but beyond the scope of this article.

Pershing Square – Long Equity Market Timing

βReturn is the risk-adjusted return from market timing – the return due to variation in factor exposures over time. Positive βReturns occur when a manager takes on larger exposures to factors that subsequently generate larger-than-typical returns.

Pershing Square has varied factor and sector exposures dramatically over time. In the chart below, the red dots depict historical long portfolio factor exposures, while the diamonds show current exposures. Critically for skill evaluation, ABW’s non-market risk factors exclude all market effects, which enables sound performance attribution:

Chart of the historical exposures to systematic risk factor of Pershing Square's reported long equity portfolio

Pershing Square Capital Management Historical Factor Exposures – Long Equity Portfolio

Pershing Square’s factor timing performance (βReturn) was negative from 2004 to 2006, largely flat through 2012, and positive since 2012:

Chart of the historical return due to market timing (factor exposure variation) of Pershing Square's reported long equity portfolio

Pershing Square Capital Management Historical Market Timing Return – Long Equity Portfolio

Similar to analyzing αReturns by sector, AlphaBetaWorks analyzes specific sources of βReturn to identify which factors the fund has timed successfully, or unsuccessfully. For example, In 2014 Pershing’s positive market-timing returns were due to:

  • Increased exposure to the U.S. Health, Canada Market, and Canada Industrial factors;
  • Decreased exposure to the U.S. Consumer factor.

Pershing Square’s significant βReturn components, by year:

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
U.S. Consumer Sector -2.33 3.61 -0.57 -3.48 1.25 0.13 1.56 2.88 -0.47 -1.69 1.70
U.S. Health Sector 0.03 -0.07 0.05 -0.06 -0.02 -0.04 0.08 -0.14 -0.09 -0.23 2.11
USD FX 8.30 -8.77 -1.21 -1.41 -0.01 0.40 -1.23 -0.40 0.11 0.41 -1.05
Canada Market (Beta) -4.19 -3.70 -2.07 0.53 0.55 0.46 -2.37 0.98 0.61 4.18 2.40
Canada Industrial Sector -1.18 0.59 0.14 -1.28 -0.10 -0.42 -0.93 -0.80 1.27 4.67 3.68

Conclusion

  • Risk-adjusted performance is return above a passive portfolio replicating a fund’s typical risk profile.
  • Pershing Square exhibited strong risk-adjusted return from overall security selection (αReturn) and, recently, strong market/factor timing (βReturn).
  • Pershing Square recently exhibited especially strong risk-adjusted security selection returns (αReturns) in Consumer and Industrial sectors.
  • Pershing Square recently exhibited strong risk adjusted market timing returns (βReturns) due to increased exposures to the U.S. Health, Canada Market, and Canada Industrial factors, and reduced exposure to the U.S. Consumer factor.
  • Among its many advantages, our analysis provides allocators insights into manager core competencies. Portfolio managers, too, benefit from deeper understanding of their own – and their teams’ – strengths and weakness.
The information herein is not represented or warranted to be accurate, correct, complete or timely.
Past performance is no guarantee of future results.
Copyright © 2012-2014, 
AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.

Hedge Fund Energy Crowding – Q2 2014

U.S. hedge funds share a few systematic and idiosyncratic long bets – a phenomenon called “crowding.” Crowding exists within aggregate portfolios and within specific sectors. Energy bets are particularly crowded and are the subject of this article. Crowded bets are the main sources of hedge funds’ collective and many individual funds’ energy sector returns. Four risk factors (systematic bets) and six stocks (idiosyncratic bets) are behind three quarters of the herding.

Systematic (factor) hedge fund energy crowding primarily consists of:

  1. A bet against integrated energy companies;
  2. A bet on oil refiners;
  3. A bet on cheap international and against expensive U.S. producers;
  4. A bet on independent oil and gas producers.

Idiosyncratic (residual) hedge fund energy sector crowding primarily consist of exposures to CVI, XOM, CHK, ATHL, CVX, and YPF.

Combined, these bets account for three quarters of aggregate hedge fund long energy sector risk (tracking error) relative to the overall energy equity market.

Crowding has two vital implications: First, hedge fund impatience contributes to greater volatility in crowded bets. Second, hedge fund long energy equity portfolios tend to generate negative risk-adjusted returns.  Thus, consensus stock and factor picks are likely to disappoint.

Identifying Hedge Fund Energy Crowding

We created an aggregate position-weighted portfolio (HF Energy Aggregate) consisting of all energy sector equities held by over 400 U.S. hedge funds with medium to low turnover. The size of each position is the dollar value of its ownership by hedge funds. This process is similar to our earlier analysis of aggregate long hedge fund crowding.

We then evaluated HF Energy Aggregate’s risk relative to the capitalization-weighted portfolio of U.S. energy equities (Market Energy Aggregate) using AlphaBetaWorks’ Statistical Equity Risk Model. Finally, we analyzed HF Energy Aggregate’s systematic and idiosyncratic bets and looked for evidence of crowding.

Hedge Fund Energy Risk

HF Energy Aggregate has a 7.8% estimated future tracking error relative to Market Energy Aggregate, mostly due to factor bets:

Chart of the sources of relative variance of aggregate portfolio of hedge funds' long energy holdings

Sources of Relative Risk for Hedge Fund Energy Aggregate

Source Volatility (%) Share of Variance (%)
Factor 6.55 70.10
Residual 4.28 29.90
Total 7.83 100.00

HF Energy Aggregate’s annual return will differ from Market Energy Aggregate’s by more than 7.8% about one third of the time. Combined, hedge funds’ energy portfolios are active, and closet indexing is not a concern.

Hedge Fund Factor (Systematic) Energy Crowding

Below are HF Energy Aggregate’s most significant relative factor exposures (red). The benchmark (gray) is Market Energy Aggregate:

Chart of the factor exposures of the aggregate portfolio of hedge funds' long energy holdings to most influential risk factors

Factors Contributing Most to the Relative Risk for Hedge Fund Energy Aggregate

Below is the contribution of various factors to HF Energy Aggregate’s relative systematic risk. These are the components of the red “Factor Variance” in the first chart:

Chart of the contribution to systematic (factor) hedge funds energy crowding from the most significant risk factors

Factors Contributing Most to Relative Factor Variance of Hedge Fund Energy Aggregate

Four factors are responsible for 75% of the relative factor risk of Hedge Fund Energy Aggregate:

Exposure (%)
Factor HF Energy Aggregate Market Energy Aggregate Relative Share of Risk (%)
Integrated Oil 2.62 38.75 -36.13 23.01
Oil Refining and Marketing 30.87 13.18 17.68 17.79
Stat Factor 1 11.86 0.26 11.59 17.08
Oil and Gas Production 65.53 47.23 18.30 16.48

These bets have the following meanings:

  1. Integrated Oil – short bet on integrated energy companies;
  2. Oil Refining and Marketing – long bet on oil refiners;
  3. Stat Factor 1 – systematic risk not captured by the standard market risk factors;
  4. Oil and Gas Production – long bet on non-integrated oil and gas producers.

AlphaBetaWorks’ Statistical Factors (Stat Factors) capture systematic risks overlooked by traditional risk models. Statistical factors and exposures to them are estimated using factor analysis. Stat Factor 1 turns out to be a combination of bets on cheap international, and against expensive U.S. producers. Securities with the largest positive exposure to Stat Factor 1 are LukOil ADR (LUKOY), Rosneft GDR (GB:ROSN), and BP ADR (BP). Securities with the largest negative exposure to it are Range Resources (RRC) and Cabot Oil & Gas (COG):

Chart of exposure to statistical factors of various positions of hedge funds' aggregate long energy  portfolio

Hedge Fund Energy Positions’ Exposures to Stat Factors

Hedge Fund Residual (Idiosyncratic) Energy Crowding

Below are the sources of HF Energy Aggregate’s relative residual variance. These are the components of the blue “Residual Variance” in the first chart. Six stocks are responsible for over three quarters of the relative residual (idiosyncratic) risk of HF Energy Aggregate:

Chart of the contribution to total residual risk of the most significant positions of the aggregate hedge fund energy portfolio

Stocks Contributing Most to Relative Residual Variance of Hedge Fund Energy Aggregate

These stocks will have the most sway on HF Energy Aggregate, and many individual funds. Conversely, the funds’ impatience will also affect these stocks the most. Investors should be ready for seemingly inexplicable volatility, especially among smaller companies:

Position (%)
Symbol Name HF Energy Aggregate Market Energy Aggregate Relative Share of Risk (%)
CVI CVR Energy, Inc. 8.93 0.23 8.70 46.67
XOM Exxon Mobil Corporation 0.50 23.90 -23.39 16.98
CHK Chesapeake Energy Corporation 5.78 0.91 4.87 5.49
CVX Chevron Corporation 0.22 13.50 -13.28 4.64
ATHL Athlon Energy, Inc. 4.77 0.34 4.43 4.37
YPF YPF SA Sponsored ADR Class D 1.89 0.00 1.89 3.05
EPE EP Energy Corp. Class A 6.78 0.26 6.53 2.29
OXY Occidental Petroleum Corporation 0.22 4.47 -4.25 1.70
CIE Cobalt International Energy, Inc. 1.99 0.33 1.66 1.64
COP ConocoPhillips 0.31 5.61 -5.30 1.64
CNQ Canadian Natural Resources Limited 3.64 0.00 3.64 1.18
MWE MarkWest Energy Partners, L.P. 3.65 0.85 2.80 1.03
EOG EOG Resources, Inc. 0.52 3.23 -2.71 1.01
SD SandRidge Energy, Inc. 1.60 0.13 1.48 0.86
TLM Talisman Energy Inc. 2.38 0.00 2.38 0.78
RGP Regency Energy Partners LP 3.25 0.74 2.52 0.71
PXD Pioneer Natural Resources Company 3.67 1.68 1.99 0.62
WLL Whiting Petroleum Corporation 2.44 0.55 1.89 0.56
WPZ Williams Partners L.P. 3.21 1.39 1.83 0.35
ATLS Atlas Energy, L.P. 0.62 0.14 0.48 0.31

As with factor crowding, there is significant stock-specific crowding into a handful of popular names.

Summary

  • There is both factor (systematic/market) and residual (idiosyncratic/security-specific) crowding of long hedge fund energy portfolios.
  • The main sources of systematic crowding are three bets on energy sub-sectors and a relative value bet on international vs. domestic producers.
  • The main sources of stock-specific crowding are CVI, XOM, CHK, ATHL, CVX, and YPF.
  • Investors in crowded stocks may experience elevated volatility.
  • Collectively, hedge funds’ long U.S. energy equity portfolios tend to generate negative risk-adjusted returns. Consequently, their crowded bets in this sector tend to disappoint.
The information herein is not represented or warranted to be accurate, correct, complete or timely.
Past performance is no guarantee of future results.
Copyright © 2012-2014, 
AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.

Hedge Fund Crowding Trends

The March to Uniformity – Illustrated and Quantified

We examined the evolution of systematic, idiosyncratic, and total risk of long equity hedge fund portfolios relative to each other.  We found decreasing differentiation and increasing herding over time. In summary, over the past 10 years total differentiation declined by 30% while systematic (factor) differentiation declined by 39%. As capital increasingly flows to undifferentiated managers, uninformed allocators will find themselves in crowded bets that are destined, at best, for index returns. With proper tools, investors can identify plenty of skilled and differentiated managers.

Hedge Funds Going Mainstream

The rise of hedge funds as an institutional asset class has been hazardous to diversified hedge fund portfolios. Some researchers find the asset class has under-performed a passive equivalent over the past ten years, while cumulative real investor profits over the period may have been negative. Even the industry-backed rebuttal shows similar post-2004 performance (Table 4: Investing in hedge funds vs T-bills).

A larger pool of capital chasing a limited set of opportunities is commonly blamed. The systematic (factor) and idiosyncratic (residual) return dispersion of currently active funds’ long equity portfolios has indeed decreased dramatically over the past 15 years. Even during the 2008-2009 volatility spike, the active returns of today’s funds were less differentiated than in the lower-volatility regime of 2003-2004:

Chart of the dispersion of returns of currently active  hedge funds' long equity portfolios

Return Dispersion History – Long Hedge Fund Portfolios

In spite of this herding, many differentiated and skilled funds remain. The key is finding them. Skilled and differentiated managers are likely to generate positive idiosyncratic returns in the future. Allocators who lack the proper tools may find themselves invested across many funds that are not nearly as distinguished as they seem.

Analyzing Hedge Fund Crowding Trends

We used the AlphaBetaWorks (ABW) North America Statistical Equity Risk Model to analyze the long equity positions of today’s medium to low turnover hedge funds at two points in time: year-end of 2004 and 2013. We combined position factor exposures to estimate fund factor exposures. We then compared the factor exposures and residual risk of every fund relative to every other fund, estimating the future relative volatility (tracking error) of every fund pair. At each date, we estimated the relative tracking errors for approximately 100,000 portfolio pairs. We then aggregated estimates of funds’ differences into a broad picture of differentiation within the asset class. We further separated differences among funds into market (factor) and security-specific (residual) bets. The higher the expected relative tracking error between two funds, the more different they are from one another.

This approach is more robust than returns-based style analysis, which fails for funds that vary factor exposures over time. The ABW Equity Risk Models also address flaws common to simpler holdings-based and fundamental analyses.

Total Hedge Fund Differentiation

Between 2004 and 2013, the average expected tracking error between two hedge fund long portfolios declined from 21% to 16%:

Chart of hedge fund crowding as estimated by the change in the distribution of equal-weighted relative tracking error of long hedge fund portfolios for the funds currently active

Equal-Weighted Relative Tracking Error Distribution – Long Hedge Fund Portfolios

However, this relative tracking error is not representative of the differentiation investors will realize – it ignores fund size.

A better measure of investor outcomes uses asset-weighted tracking error – the expected relative volatility of two dollars invested in different portfolios. This measure declined from 21% to 14%:

Chart of hedge fund crowding as estimated by the change in distribution of asset-weighted relative tracking error of long equity hedge fund portfolios for currently active funds

Asset-Weighted Relative Tracking Error Distribution – Long Hedge Fund Portfolios

The largest funds – and dollars invested in them – have become even more crowded than the average fund.

Systematic (Factor) Differentiation

The primary driver of hedge fund crowding is the herding of market (factor) bets. The expected volatility of relative factor returns for capital invested in different funds dropped from 15% in 2004 to 9% in 2013.

Chart of hedge fund crowding as estimated by the change in asset-weighted relative factor (systematic) tracking error distribution for long equity hedge fund portfolios of currently active funds

Asset-Weighted Relative Factor Tracking Error Distribution – Long Hedge Fund Portfolios

We estimate that in 2014, differences in factor returns of hedge funds’ long equity portfolios will be below 9% about 2/3 of the time.

While much capital is pursuing uncorrelated systematic bets, the bulk is relatively less differentiated and crowded into similar factors. Our earlier Insight discussed these shared factor bets.

Hedge Fund Skill, Size, and Differentiation

The long equity portfolios of today’s hedge funds have become markedly less differentiated between 2004 and 2013, primarily due to the crowding of factor bets. Crowding is greatest among the largest funds, as evidenced by the larger circles in the lower half of the figure below – a map depicting hedge fund skill and differentiation:

Chart of the distribution of skill, size, and differentiation for long equity hedge funds' portfolios

Hedge Fund Skill, Size, and Differentiation

Skilled, differentiated managers are in the upper-right. Unskilled, un-differentiated managers are in the bottom-left.

In spite of increased hedge fund herding, plenty of skilled and differentiated managers remain. Especially attractive are those who stand apart from the crowd and make large bets in areas where they show significant evidence of skill. We discuss general and specific fund skills in earlier articles. AlphaBetaWorks Analytics enable allocators to identify both and quantify crowding within existing fund portfolios.

Summary

Hedge fund crowding has increased over the past 10 years due to declines in market (factor) and stock-specific (residual) differentiation:

  • Between 2004 and 2013, total differentiation declined by 30%.
  • Between 2004 and 2013, systematic (factor) differentiation declined by 39%.
  • New capital has been predominantly committed to less differentiated managers.
  • Many skilled and differentiated managers exist and can be identified with the proper tools.
  • ABW identifies skilled and differentiated managers, including those most likely to outperform in the future.
Copyright © 2012-2014, AlphaBetaWorks, a division of Alpha Beta Analytics, LLC. All rights reserved.
Content may not be republished without express written consent.