Eureka! Fund Managers Do Have Stock-Picking Skills! But Does That Translate into High Returns for Clients?" Ron Bird Paul Woolley Centre for the

Size: px

Start display at page:

Download "Eureka! Fund Managers Do Have Stock-Picking Skills! But Does That Translate into High Returns for Clients?" Ron Bird Paul Woolley Centre for the"

Dayna McKenzie
6 years ago
Views:

1 Eureka! Fund Managers Do Have Stock-Picking Skills! But Does That Translate into High Returns for Clients?" Ron Bird Paul Woolley Centre for the Study of Capital, UTS Tuesday 20 May,

2 Malkiel on the Funds Management Industry From 1990 to 2006 the US financial services sector grew from 4.9% to 8.3% of GDP...a substantial share of this increase comprised increases in asset management fees In 1990 US active equity mutual funds managed $132B at an expense ratio of 85 basis points and by 2010 FUM had risen to $2,474B with an expense ratio of 91 basis points One could argue that the increase in total fees (141 times) could prove to be socially useful if it reflected increased returns to investors and/or improved efficiency in the market Neither of these arguments are supported by the data Index funds still outperform passive funds No change in the efficiency of the market from 1980 to 2011 Paul Woolley Centre for the Study of Capital 2

3 Some Reflections Evidence like this led me to ponder numerous interesting questions: Why do investors favour active managers? Why are active managers paid so much? Do the active managers actually have some skills? Are the clients themselves to blame for the poor return from active management? Can we improve the way that we utilise the managers? And many more Paul Woolley Centre for the Study of Capital 3

4 A Series of Papers A selection of the papers being produced by the Centre throw some light on these issues: The contribution of stock picking and portfolio construction to fund performance Optimal management of a funds management firm through its life cycle (and the implications for clients How various active decisions contribute to fund performance The traits of a superior fund manager Paul Woolley Centre for the Study of Capital 4

5 Some Background to Active Management Each fund has a stated investment style and a benchmark index consistent with that style E.g small cap growth manager (Russell 2000 Growth) Active management involves taking bets relative to the benchmark We use the size of these bets to measure: The manager s preference for each stock A fund s active position which reflects the level of aggression with which the manager implements the fund s investment process (Cremers and Petajisto, 2009) The style tilts that are built into the Fund s portfolio (Wermers, 2012) Paul Woolley Centre for the Study of Capital 5

6 Ex-ante Measures of Active Management (1) Active positions (Cremers and Petajisto) Active position = 0.5 N j= 1 abs( w f w where the active position is one half of the aggregate of the absolute differences between the actual fund weighting in a particular stock, wf,j, and the benchmark weighting for that stock, wi,j. the benchmark is determined as that index that minimises the funds active position on most occasions, j i, j ) Paul Woolley Centre for the Study of Capital 6

7 Ex-ante Measures of Active Management (2) Style Divergences (Wermers extended) Style divergence in dimension l = a where w j is the weight of stock j in the funds portfolio, is the weight of stock j in the fund s benchmark l portfolio and C j is stock j s score in dimension l. Each month each stock is ranked on each dimension and given a quintile ranking which becomes it score for that dimension that period The dimensions included are: Size, growth/value and momentum N a l i ( w C w j j jc j= 1 l j ) i w j Paul Woolley Centre for the Study of Capital 7

8 Data Data Sample period: 1995 to 2012 Institutional US equity mutual funds Mutual fund data: CRSP US Mutual Fund Database Thomson Reuters S12 Mutual Fund Holdings Index data Russell (9) and Standard & Poor (9) Value, growth, style neutral Small cap, medium cap, large cap Market and accounting data: CRSP/Compustat Merged Database Paul Woolley Centre for the Study of Capital 8

9 Paper 1: Diversification versus Concentration... and the Winner is? 9

10 Diversified v Concentrated Portfolios The two tasks of fund managers are: Stock selection Portfolio construction Post-Markowitz investors and managers alike have placed high emphasis on portfolio construction resulting in a compromise when building a portfolio between the contribution that a stock makes to return and to risk As a consequence a manager who may be good at identifying mispriced stocks may not deliver high returns to his clients because he is including in his portfolio stocks that he has no reason to like. Paul Woolley Centre for the Study of Capital 10

11 Three Great Men As times goes on, I get more and more convinced that the right method of investment is to put large sums into enterprises which one thinks one knows something about..... It is a mistake to think one limit s one s risk by spreading too much between enterprises about which one knows little and has no reason for special confidence. (Keynes, 1934) Wide diversification is only required when investors do not understand what they are doing. Diversification is a protection against ignorance. It makes very little sense for those who know what they re doing. (Buffett) The academics have done a terrible disservice to intelligent investors by glorifying the idea of diversification. Because I just think the whole concept is literally almost insane. (Munger) To which I would add: Fund managers spend 90% of their time on stock-picking and there lies their expertise They generally have little or no special expertise in portfolio construction Time and time again I have seen managers give back their advantage when translating their stock preferences into diversified portfolios Their behaviour would seem to be at variance with Keynes, Buffett and Munger!!! Paul Woolley Centre for the Study of Capital 11

12 Objective of this Study Determine how best to use manager skills The method that we employ is to: identify a manager s preferred stocks This is done on the basis of the active bets that they take on individual stocks (as a measure of their level of conviction) build concentrated portfolios based on these preferred stocks using conviction weights Compare the performance of these concentrated portfolios with the manager s actual performance We then compare two strategies: Build a diversified portfolio of funds that hold concentrated portfolios Build a concentrated portfolio of managers who hold diversified portfolios Paul Woolley Centre for the Study of Capital 12

13 Incremental Returns a/transaction costs Total and Net Returns - Full Sample Conviction Weights Against Own Index Portfolios Annualised Returns Transaction Costs Annualised Net Returns Incremental Portfolios Annualised Returns Transaction Costs Annualised Net Returns Top %** 1.60% 12.75% Top 1 to %** 1.60% 12.75% Top %** 1.45% 11.50% Top 6 to %** 1.25% 9.85% Top %** 1.35% 10.93% Top 11 to %** 1.04% 9.26% Top %** 1.28% 10.50% Top 16 to %** 0.98% 8.62% Top %** 1.24% 10.17% Top 21 to %** 0.94% 8.25% Top %** 1.18% 9.91% Top 26 to %** 0.80% 7.66% All Funds 10.04%** 0.00% 10.04% All Funds 10.04%** 0.00% 10.04% Own Index 9.14%** 0.20% 8.94% Own Index 9.14%** 0.20% 8.94% We estimated transaction costs using a method developed by Lesmond (1999) As you see almost all of the added value comes from the five favoured stocks with no added value beyond about 15 stocks This has potential implications for how to benefit from the skills of active managers 13

14 1 Fund 5 Fund An Alternative Strategy Rebalancing every three years Concentrated Portfolios Returns Std Dev Sharpe Ratio Top %*** Top %*** Top %*** Top %*** Top %*** Top %*** Any 1 Fund 10.06%*** Any 5 Fund 10.07%*** On an after-cost and risk-adjusted basis a strategy of investing in a portfolio of concentrated funds rather than in a diversified fund leads to only slightly better outcomes where you are randomly choosing from the whole population of managers 14

15 Implications and Extensions of Findings Managers stock selection skills are eroded by requiring them to hold diversified risk-controlled portfolios However, an obvious strategy of building a portfolio based on the preferred stocks sourced from a number of managers does only slightly better than current practice of investing via diversified managers This highlights that there are two elements to the staements of Keynes and Buffett: Hold concentrated portfolios but only if you are good A few other insights: Growth managers are by far the best stock pickers Worst beets perform even better than the best bets A long/short portfolio of best/worst bets has returned after/costs about 10% pa over the last 15 years. Paul Woolley Centre for the Study of 15 Capital

16 Paper 2: Fund Self-Inflicted SAD and its Impact on Clients Paul Woolley Centre for the Study of Capital 16

As funds grow in terms of FUM, they become less aggressive in the

17 Some Important Stylised Facts Small and /or good funds take the largest active positions. As funds grow in terms of FUM, they become less aggressive in the implementation of their investment style and this is particularly true if they are inferior funds Paul Woolley Centre for the Study of Capital 17

18 The Strategic Implementation of an Investment Process We build a model based upon the following assumptions: the objective of the fund is to maximise the PV of its future fee income each fund s investment process yields an active portfolio that has an inherent excess return (α a ) and volatility (σ a ) each fund s actual portfolio is a combination of this active long/short portfolio and its benchmark portfolio where the proportion (w) invested in the active portfolio is a measure of the aggressiveness with which it implements it strategy each fund benefits (g) to differing degrees through growth in FUM following a sustained period of large outperformance but equally suffers (c) to a loss of FUM following a sustained period of large underperformance 18

19 Simple Representation of Our Model We assume that g (c) is obtained by achieving an upper (lower) threshold performance over a specific time period (t) Excess returns k Payoff = g t Time k Payoff = -c 19

20 We built a simple model to explain these stylised facts with the following basic assumptions: Funds are managed to maximise their market value Funds are rewarded on the basis of past performance Model and Findings Paul Woolley Centre for the Study of Capital 20

21 The model tells us that... interpretation... A rational fund will behave in a way consistent with the stylised facts Boutiques and better managers will be the most aggressive Managers will become less aggressive as FUM grow Eventually it will be optimal to become a closet indexer It is an industry where a lack of ability should not discurage wanting to have a go (and again, and again.. ) When a manager hits pay dirt, then it is (very) sensible for them to shut up shop The less than talented managers should do this earlier One of the interesting things to contemplate is how to be passive but appear active Whatever way, it is bad news for the client Paul Woolley Centre for the Study of Capital 21

22 ... and implications... There is the potential for significant principle/agent issues between The fund and its clients The probability of the client benefiting from the skills of the manager will become significantly less likely as the fund grows The fund management organisation and the fund managers This has significant implications for how fund manager are rewarded and explains why better managers leave and set up boutiques One very interesting finding is that funds evaluated relative to an index will be much less aggressive than fund evaluated on the basis of absolute returns The introduction of restrictive mandates has very likely been to the detriment of the client Paul Woolley Centre for the Study of Capital 22

23 Paper 3: Performance Implications of Active Management of Institutional Mutual Funds Paul Woolley Centre for the Study of Capital 23

24 Do managers have any special skills? In this piece we evaluated the association between active positions taken by managers and their performance Two papers of particular interest: Cremers and Petajisto (2009)identified that more active managers achieve better performance Wermers (2012) found that style drift added to the performance of a fund We extend the analysis undertaken in these two papers in order identify the areas where managers do actually add value Paul Woolley Centre for the Study of Capital 24

25 Regression results for effectiveness of active management Dep. Var. Excess Returns Tracking Error Information Ratio Variable Active Pos *** *** *** V/G Div. + (Value) *** (Growth) *** 0.002*** *** Mom. Div. + (Winners) *** *** *** - (Losers) *** *** *** Size Div. + (Large cap) ** *** ** - (Small cap) *** *** ** Beta *** -0.01*** *** Turnover *** *** ** Fees *** *** Log (TNA) E (Log(TNA)) E Age -3.25E *** Qtr inflows -8.06E-06* 3.94E-06*** Prev. returns *** ** *** Index returns *** *** *** Adjusted R Paul Woolley Centre for the Study of Capital 25

26 Effectiveness of active management by investment styles and market conditions Part A: Excess Returns (Dependent Variable) Fund Type Growth Manager Value Manager Core Manager Mkt. Conds. Weak Strong Weak Strong Weak Strong Variable Active Pos * *** *** *** *** V/G Div *** *** * *** *** * Mom. Div *** ** *** 0.012*** *** ** *** *** *** Size Div * ** *** *** ** *** ** * Adjusted R Part B: Active Position and Style Divergence Combinations with Average Weights Active position only -0.82% 1.40% 1.84% 1.99% 0.70% 0.23% Value/winners/large -0.25% 1.52% 1.67% 1.13% 1.05% 0.37%.Value/winners/small -0.50% 1.65% 1.21% 1.40% 1.07% 0.42% Value/losers/large -1.11% 0.91% 0.51% 1.12% 0.48% 0.23% Value/losers/small -1.37% 1.04% 0.05% 1.39% 0.49% 0.28% Growth/winners/large 0.43% 1.61% 2.18% 1.37% 1.27% 0.40% Growth/winners/small 0.18% 1.74% 1.72% 1.63% 1.28% 0.45% Growth/losers/large -0.43% Paul 1.00% Woolley Centre for 1.01% the Study of Capital 1.35% 0.69% 0.27% Growth/losers/small -0.69% 1.13% 0.55% 1.62% 0.70% 0.32% 26

27 Portfolios of Active Positions and Style Divergences Style divergence Small Medium Large Large combination Active Active Active All Small Position Position Position Sample Active (%p.a.) (%p.a.) (%p.a.) (%p.a.) Position (%p.a.) Value/winners/large ** *** Value/winners/small ** 4.14*** * Value/losers/large Value/losers/small Growth/winners/large * Growth/winners/small 2.99** 2.79* 5.01*** 3.94** 2.02* Growth/losers/large Growth/losers/small All Sample * * Paul Woolley Centre for the Study of Capital 27

28 Some Important Findings The more aggressive managers do add most value The major exception being growth managers during weak markets However, the superiority of aggressive managers (i.e. large active positions) is very much conditioned by the manager s style divergences The most aggressive funds actually detract value (to a greater extent than the least aggressive) when making large bets on losers The least aggressive managers actually do well when making large bets on growth, winners The biggest outperformance comes from the most aggressive managers tilting towards small cap, growth, winners The biggest losers are also the most aggressive managers when tilting towards value/losers 28

29 An Interesting Finding : Style Choice or Implementation Added value (%pa): Total Full Sample Weak Markets Strong Markets Style choice Total Style choice Total Implementation Implementation Implementation Style choice All Funds Growth Value Core Overall managers add value by their process design with their day-to-day implementation of the process adding nothing The value style adds significant value (all of which comes in weak markets) while their active implementation actually destroys value (especially in weak markets) The growth style still add value (this time in strong markets) with the day-to-day implementation also being slightly value-adding managers Best combination would be an ( active / passive?) growth manager in strong markets and a passive value manager in weak markets

30 Paper 4: What are the traits of a superior fund manager? 30

31 Recent Literature Recent literature has identified a number of factors that tend to be identified with fund performance: Active (Cremers & Petajisto, Bird et al) Style tilts (Wermers, Bird et al) Trend followers (Jiang and Verardo) Industry concentration (Kacperczyk et al) R squared (Amihud and Goyenko) The return gap (Dyikov and Verbeet) Public information (Kacpercyzk and Seru) Several control variables: Beta, fees, turnover, age, size, funds flow, past performance, benchmark performance In this study (WIP) is to examine how we can combine these insights to identify the super manager 31

32 Results Key Variable Market Conditions Strong Weak 1 - Herding *** Active *** *** Value tilt *** Growth tilt *** *** Large cap tilt *** Small cap tilt ** *** Winners tilt **** *** Loser tilt *** *** Industry *** * concentration 1 r-squared *** Gap return *** *** Control Variables Age * Funds Flow Turnover *** Fees *** TNA ** Benchmark *** return Past return *** Adjusted R

33 The Findings Managers on average have quite strong stock selection skills The reasons why these do not show up in performance is because of self-imposed and externally imposed constraints Investors need to change the way that they: Select managers Choose young managers and especially those that act as though they are the better managers Utilise managers Extensively use their stock selection skills only and contract to avoid agency costs Fire managers Be more aggressive in firing, especially with larger managers 33

34 Additional Introductory Slides 34

35 Choice of benchmark The benchmark for each fund is that index which minimises the active position during the greatest number of quarters Table A1 Index benchmark allocations and index returns This table shows the distribution of funds to the allotted benchmark indices. We determine each quarter the index that most closely tracks each fund s actual portfolio (i.e. the index that gives it the smallest active position). As a consequence over the life of each fund we have a benchmark index assigned each quarter. The index that was chosen in the greatest number of quarters will be allocated as the fund s benchmark. Returns are each index s average annualised index returns over the sample period. Index Name Number Of Funds Return (%PA) Russell % Russell 1000 Growth % Russell 1000 Value % Russell % Russell 2000 Growth % Russell 2000 Value % Russell Mid % Russell Mid Growth % Russell Mid Value % S&P % S&P 400 Growth % S&P 400 Value % S&P % S&P 500 Growth % S&P 500 Value % S&P % S&P 600 Growth % S&P 600 Value % 35

36 Additional Slides on Paper 1 36

37 Objective of this Study Determine how best to use manager skills The method that we employ is to: identify a manager s preferred stocks This is done on the basis of the active bets that they take on individual stocks (as a measure of their level of conviction) build concentrated portfolios based on these preferred stocks using conviction weights Compare the performance of these concentrated portfolios with the manager s actual performance We then compare two strategies: Build a diversified portfolio of funds that hold concentrated portfolios Build a concentrated portfolio of managers who hold diversified portfolios Paul Woolley Centre for the Study of Capital 37

38 Results Best Bets (Long Only) Total Returns - Full Sample Conviction Weights Against Own Index Portfolios Total Returns Standard Deviation (annualised) (annualised) Top %*** 26.33% Top %*** 23.40% Top %*** 21.83% Top %*** 20.65% Top %*** 19.79% Top %*** 19.13% All Funds 6.30%* 19.51% Own Index 5.05% 19.96% Sharpe Ratio The performance above is based on an equal investment in each of funds available in each quarter The numbers suggest that managers do have good stock selection skills which are dissipated in the delivery process Paul Woolley Centre for the Study of Capital 38

39 Factor-adjusted Performance Conc. Port. Four Factor Model : Excess Returns based on Weighting against Own Index Alpha P-Value Rm-Rf P-Value SMB P-Value HML P-Value Momentum P-Value Top Top Top Top Top Top The adjusted alpha varies between in excess of 8%pa and 4% pa There is a big bet towards momentum stocks There is a smaller (and decreasing) bet towards high market beta Paul Woolley Centre for the Study of Capital 39

40 Best Bets: Incremental Performance Portfolios Total Returns (annualised) Total Returns - Full Sample Conviction Weights Against Own Index Standard Deviation (annualised) Sharpe Ratio Top 1 to Top % 26.61% Top 6 to Top % 19.84% Top 11 to Top % 18.51% Top 16 to Top % 17.42% Top 21 to Top % 17.02% Top 26 to Top % 16.75% Top 31 to Top % 16.41% Top 36 to Top % 16.25% All Funds 6.30% 19.51% Own Index 5.05% 19.96% This provides insights into the extent of the managers ability to add value 40

41 Incremental Returns a/transaction costs Total and Net Returns - Full Sample Conviction Weights Against Own Index Portfolios Annualised Returns Transaction Costs Annualised Net Returns Incremental Portfolios Annualised Returns Transaction Costs Annualised Net Returns Top % 1.60% 9.16% Top 1 to % 1.60% 9.16% Top % 1.45% 7.94% Top 6 to % 1.30% 6.04% Top % 1.35% 7.32% Top 11 to % 1.04% 5.44% Top % 1.28% 6.84% Top 16 to % 0.98% 4.89% Top % 1.24% 6.54% Top 21 to % 0.94% 4.82% Top % 1.18% 6.26% Top 26 to % 0.89% 4.31% All Funds 6.30% 0.00% 6.30% All Funds 6.30% 0.00% 6.30% Own Index 5.05% 0.20% 4.85% Own Index 5.05% 0.20% 4.85% We estimated transaction costs using a method developed by As you see almost all of the added value comes from the five favoured stocks with no added value beyond the top 20 stocks This has significant implications for how to benefit from the skills of active managers 41

42 A Feasible Strategy Rebalancing every three years Concentrated Portfolios Returns StdDev Sharpe Ratio Top %*** Fund of Each type Top %*** Top %*** Top %*** Funds of Each Type Top %*** Top %** Any 1 Fund 6.17%* Any 5 Fund 6.28%** One Fund of each Cap Fund 7.29%** One Fund of each Style Fund 6.28%** One Fund of Each Type of Fund 7.07%** Funds are divided by nine types by growth and market cap. Strategies are then simulated of running concentrated versus diversified portfolios The concentrated portfolios would appear to win out by between 4% pa and 5%pa. 42

43 Copycat Strategy Portfolios Total Returns (annualised) No Lags Lagged 2 Months Lagged 1 Quarter Sharpe Ratio Total Returns (annualised) Sharpe Ratio Total Returns (annualised) Sharpe Ratio Top %*** %*** %*** Top %*** %*** %*** Top %*** %*** %*** Top %*** %*** %*** Top %*** %*** %*** 0.31 Top %*** %*** %*** Problem is that the funds holding data does not become available until well into the quarter What happens is investment has to be delayed It seems that after two months the strategy would still add significant returns Other analysis we have conducted suggests that this is because funds enter mean-reverting stocks too early Paul Woolley Centre for the Study of Capital 43

44 What about the Worst Bets? Total Returns Conviction Weights Against Own Index Portfolios Total Returns Standard Deviation (annualised) (annualised) Sharpe Ratio Worst % Worst %*** Worst %*** Worst %*** Worst %*** Worst %*** All Funds 6.30%* Own Index 5.05% Except for the Worst 5, the concentrated portfolios of worst bets perform even better than do the best bets Managers have always been better at choosing stocks in which not invest Paul Woolley Centre for the Study of Capital 44

45 Portfolios.... and Long/Short? Total Returns - Full Sample Conviction Weights Against Own Index Standard Total Returns Deviation (annualised) (annualised) Sharpe Ratio Long Top 5/Short Worst %*** Long Top 10/Short Worst %*** Long Top 15/Short Worst %*** Long Top 20/Short Worst %*** Long Top 25/Short Worst %*** Long Top 30/Short Worst %*** Paul Woolley Centre for the Study of Capital 45

46 Implications and Extensions of Findings Managers stock selection skills are eroded by requiring them to hold diversified risk-controlled portfolios The implication being that it may well be better to acquire the best bets from a manager best bets and doing the diversification yourself The fees might be quite high but a viable option might be to wait until the bets become public knowledge As has been found elsewhere, the manager s worst bets perform even better than their best bets A risk-adjusted long/short portfolio based on the best/worst bets has returned in excess of 20% p.a. Paul Woolley Centre for the Study of Capital 46

47 Results by Manager Style Growth managers add by far the most by running concentrated portfolios but also introduce much more volatility into portfolios This is not surprising given that growth indices are made up on expensive stocks that do not well reflect the holdings of growth managers Paul Woolley Centre for the Study of Capital 47

48 Results by Fund Size Large funds have the best stock picking skills (5.8% v 4.8%) which is reassuring but for reasons that will become apparent later this does not translate into superior investment returns Paul Woolley Centre for the Study of Capital 48

concentrated portfolios of small cap managers would seem to

49 Results: small cap v large cap Funds The stock picking abilities of large cap and small cap managers would appear fairly similar but the concentrated portfolios of small cap managers would seem to Introduce much more volatility Paul Woolley Centre for the Study of Capital 49

50 Concentrated Portfolios Results: Retail v Institutional Funds Total Returns Conviction Weighted Portfolios Relative to Own Index Institutional Funds Standard Deviation Sharpe Ratio Total Returns Retail Fund Standard Deviation Sharpe Ratio Top % 27.69% % 28.36% 0.28 Top % 24.98% % 25.35% 0.25 Top % 23.52% % 23.80% 0.24 Top % 22.38% % 22.66% 0.22 Top % 21.57% % 21.82% 0.22 Top % 20.97% % 21.19% 0.21 All Institutional / All Retail Corresponding Index Return 6.43% 19.70% % 19.71% % 0.18% % 0.15% 0.07 There is very little difference between the performance of institutional and retail funds Paul Woolley Centre for the Study of Capital 50

51 Additional Slides on Paper 3 51

52 The Strategic Implementation of an Investment Process We build a model based upon the following assumptions: the objective of the fund is to maximise the PV of its future fee income each fund s investment process yields an active portfolio that has an inherent excess return (α a ) and volatility (σ a ) each fund s actual portfolio is a combination of this active long/short portfolio and its benchmark portfolio where the proportion (w) invested in the active portfolio is a measure of the aggressiveness with which it implements it strategy each fund benefits (g) to differing degrees through growth in FUM following a sustained period of large outperformance but equally suffers (c) to a loss of FUM following a sustained period of large underperformance 52

53 Simple Representation of Our Model We assume that g (c) is obtained by achieving an upper (lower) threshold performance over a specific time period (t) Excess returns k Payoff = g t Time k Payoff = -c 53

54 Base Case Parameters µ 0.05 Avg return of the benchmark α 0.00 Avg excess return of the active process σ 0.10 Volatility of the benchmark b σ a 0.10 Volatility of the active process ρ 0.00 Correlation of active and benchmark T 3 Time horizon (years) k 0.05 Upper threshold k Lower threshold g 0.5 Gain if e T exceeds k at T c 0.05 Cost if e T is below k at T λ 0.3 Risk-aversion coefficient 54

55 The Base Case Outcome The decision variable on which we focus is the proportion of funds that are devoted to the active portfolio (w) The objective for the fund is to set this proportion at a level (w*) that maximises the expected value of its future fee income Under our base case scenario, the optimal strategy is to devote 74% of the funds to the active portfolio. This figure is in line with Cremers and Petajisto (2009) and Bird et al (2010) We then test the impact on w* varying in pairs the value for the various parameters in order to provide insights as to how a fund will behave as it passes through its life cycle 55

56 Optimal strategy under different pay-offs Remembering that α = 0, you see: Managers become more aggressive as g/c increases Manager becomes passive as g c (but before g = c as managers are risk-averse 56

57 Optimal strategy with different α s and payoffs It is optimal for the more talented managers to be more aggressive Even the below average managers will be active for relatively low payoffs (e.g. g/c = 2) 57

58 Optimal strategy with volatility of process 58

59 Implications of Our Findings 1. New funds with little or no reputation and low funds under management are most likely to take the most active positions 2. Funds that are more likely to outperform (i.e., with higher α) have a stronger incentive to be more aggressive in the implementation of their process than is the case with lesser funds (i.e., with lower α). 3. It is not only α that has implications for the optimum active positions but also the volatility of the managers' active process the impact that these positions have on the volatility 4. The attitude to risk of the fund management organization (and managers) will play an important role in determining how aggressive they are in implementing the fund s investment process. 5. When c g, k = k and α = 0, then w 0 (Index) 59

60 Strategy Through the Life Cycle Good manager Bad manager Good Bad Alpha (ά) Cost (c) Alpha (ά) Cost (c) Boutique Mid Cycle Mature Boutique Mid-Cycle Mature Embryonic bad managers will be quite active but less so than good managers All managers will become less aggressive as FUMs grow Eventually the optimal strategy will be to become passive (i.e. closet indexers) N.B. Given the level of noise, even bad managers can achieve 60 sustained good performance

61 Interesting Issues: Manager What are the implications for an organisation that has multiple funds? What are the implications of managers having equity in the firm? What are the implications for funds becoming closet indexers? What are the implications for funds being evaluated on the basis of absolute rather than relative returns? And so on 61

62 Interesting Issues: Clients What to look for when choosing managers? The good signs The bad signs How better to motivate managers? When to fire managers? Foster and Warren paper 62

63 What Else Can We Learn? We know that past performance is not a good way to chose managers but we use it anyway An alternative is to concentrate on manager characteristics that have been shown to be associated with good performance: Active (Cremers & Petajisto, Bird et al) Style tilts (Wermers, Bird et al) Trend followers (Jiang and Verardo) Industry concentration (Kacperczyk et al) Public information (Kacpercyzk and Seru) R squared (Amihud and Goyenko) The return gap (Dyikov and Verbeet) 63

64 .... and finally there is Berk and Green They provide a rational model to explain why fund managers are well remunerated when they do not seem to be able to outperform the market The implication of no persistence in performance is that it is random (luck ) and so why compensate for luck Their model has three elements: Competitive provision of capital to managers Some managers do have ability but diminishing returns to scale Learning about manager ability from past returns As a result funds chase performance to a point where future performance erodes and all of the rents are captured by the managers 64

65 Our Model and Findings We built a simple model to explain these stylised facts with the following basic assumptions: Funds are managed to maximise their market value Funds are rewarded on the basis of past performance The model tells us that A rational fund will behave in a way consistent with the stylised facts Boutiques and better managers will be the most aggressive Managers will become less aggressive as FUM grow Eventually it will be optimal to become a closet indexer It is quite rationale for a less than average manager to have a go (potentially time and time again) If such a manager hits pay dirt, then it will be (very) sensible for 65 them to shut up shop (early!)

66 Additional Slides on Paper 4 66

67 The Factors Active position: The aggregate of the bets against the benchmark portfolio Cremers and Petajisto (2009) found a strong positive association between active position and performance Bird et al. (2013) confirmed this findings but found it varied with fund style and market conditions Style divergence: The tilt relative to the position inherent in the benchmark for value/growth, winner/loser and large/small Wermers (2012) found the funds that take the largest fund bets outperform Bird et al. (2013) identified the tilts that worked and did not work 67

68 The Factors (cont.) Trend followers: The correlation between a fund s trades and the lagged trades across all institutional investors (controlled for past returns, size and book-to-market) Jiang and Verardo (2013) found a strong negative relationship between this correlation and fund returns indicators that leaders outperform followers Industry concentration: The difference between the industry weights of the fund s portfolio and that of the fund benchmark Kacperczyk et al. (2005) find that the funds that run concentrated industry portfolios outperform 68

69 The Factors (cont.) R-squared: The measure is the-squared from an equation that regresses the fund s return against a multi-factor model Amihud and Goyenko (2012) found that this r-squared to be negatively correlated with a fund s investment performance. The return gap: This is measured by the difference between a fund s actual return for a quarter and the return that it would have achieved if it had held its portfolio weights constant over the quarter Dyakov and Verbeek (2013) found that the return gap is positively correlated with the performance of the fund (i.e. those who benefit from adjusting their portfolio during the quarter do best) 69

70 The Factors (cont.) Public information: This measures the extent to which a fund reacts to a change in an analysts forecast (Kacperczyk and Seru (2007) found that the fund s that reacted least to the release of public information (e.g analyst s forecasts) achieved the best investment performance These is a common theme that develops from a consideration of these factors: Those fund s that behave as though they are superior realize the higher returns 70

71 The Factors Active position: The aggregate of the bets against the benchmark portfolio Cremers and Petajisto (2009) found a strong positive association between active position and performance Bird et al. (2013) confirmed this findings but found it varied with fund style and market conditions Style divergence: The tilt relative to the position inherent in the benchmark for value/growth, winner/loser and large/small Wermers (2012) found the funds that take the largest fund bets outperform Bird et al. (2013) identified the tilts that worked and did not work 71

72 The Factors (cont.) Trend followers: The correlation between a fund s trades and the lagged trades across all institutional investors (controlled for past returns, size and book-to-market) Jiang and Verardo (2013) found a strong negative relationship between this correlation and fund returns indicators that leaders outperform followers Industry concentration: The difference between the industry weights of the fund s portfolio and that of the fund benchmark Kacperczyk et al. (2005) find that the funds that run concentrated industry portfolios outperform 72

73 The Factors (cont.) R-squared: The measure is the-squared from an equation that regresses the fund s return against a multi-factor model Amihud and Goyenko (2012) found that this r-squared to be negatively correlated with a fund s investment performance. The return gap: This is measured by the difference between a fund s actual return for a quarter and the return that it would have achieved if it had held its portfolio weights constant over the quarter Dyakov and Verbeek (2013) found that the return gap is positively correlated with the performance of the fund (i.e. those who benefit from adjusting their portfolio during the quarter do best) 73

74 The Factors (cont.) Public information: This measures the extent to which a fund reacts to a change in an analysts forecast (Kacperczyk and Seru (2007) found that the fund s that reacted least to the release of public information (e.g analyst s forecasts) achieved the best investment performance These is a common theme that develops from a consideration of these factors: Those fund s that behave as though they are superior realize the higher returns 74

75 Discussion

Diversification versus Concentration... and the Winner is?

Diversification versus Concentration... and the Winner is? Danny Yeung Paolo Pellizzari Ron Bird Sazali Abidin Abstract: Diversification has its well-known benefits but its pursuit can involve a trade-off