Kenneth Winther - Tryg [Compatibility Mode]

Transcription

Kenneth Winther - Tryg [Compatibility Mode]
Smart Beta…
…or
Smart Alpha?
Kenneth Winther
Senior Vice President, [email protected], Tryg
External lecturer, [email protected], Copenhagen Business School
1
26. november 2015
Smart beta in a nutshell
Smart beta is the new black!
Many names for same “topic”: Advanced beta, scientific beta, risk premia
investing, risk factors etc.
Concept: Overweight (usually academic) proven premia instead of market cap
Great long-term smart beta excess returns
Index returns (1998M12=100)
MSCI World
MSCI World Minimum Volatility
MSCI World Value Weighted
MSCI World Momentum
MSCI World Equal Weighted
350
300
250
200
150
100
50
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
2
Smart beta intuition
3
Factor
Risk-based explanation
Structural & behavioral explanation
Value
Costly reversibility of assets in
place: High sensitivity to
economic shocks in bad
times
Overreaction to bad news and
extrapolation of the recent past leads to
under-pricing
Size
Low liquidity, high distress
and downside risk is
compensated by higher
returns
Limited investor attention to smaller cap
stocks
Low risk
Liquidity-constrained
investors have to sell
leveraged positions in low-risk
assets in bad times when
liquidity constraints become
binding
Disagreement about high-risk stocks leads to
overpricing due to short-sales constraints.
Capital usage within Solvency II and many
rating models do not distinguish riskiness
making low risk stocks capital inefficient.
Managers worry about low risk’s tracking
error-risk which is not compensated
Momentum
High-expected-growth firms
are more sensitive to
disappointments to expected
growth trends
Investor overconfidence, initial
underreaction and self-attribution bias
leads to returns continuation in the short
term
Smart beta in the active vs. passive debate
Smart beta: The new box to dominate active and passive?
Active management
Passive management
Smart beta
Smart beta dominates due to:
• Average active fund do not
beat market cap
Winther & Steenstrup, 2016 Spring:
Smart Beta or Smart Alpha?
(The Journal of Investing)
4
Smart beta dominates due to:
• Transparent, easy
implementable static long term
investment rules which
outperform through academic
proven weighting schemes
( )
Winther & Steenstrup, 2015:
Kan ambitionsniveauet
hæves højere end smart
beta? (Finans/Invest)
Smart beta or smart alpha?
Another revolution: From 3 to 4 boxes
Manager
Focus (Universe)
Active
Active
management
Market
cap
5
Smart (beta)
Other
active funds
Smart alpha
Passive
Classic ETFs
Passive
management
&
index funds
Smart beta
strategies
Smart beta strategies
Smart beta or smart alpha?
Comparing apples with apples
Manager
Focus (Universe)
Active
6
Passive
Market cap
Smart beta
Smart (beta)
True smart alpha
Intuition behind smart alpha
How can active managers potentially create value?
Smart focus
Active manager create value through
Value
Ability avoid stocks that are cheap for a
reason (i.e. value traps) & assess individual
company risks better
Size
Smaller stocks get less attention (# of
analysts and quality of analysis) and more
inefficient pricing
Low volatility
Ability to avoid overpriced low volatility
stocks & understand intrinsic stability
rather than just price stability
Momentum
[No clear intuition?]
+ Smart active can adjust with news flow instead of waiting for
the smart beta rebalancing point and avoid crowding
7
Constructing data set
Groups of active smart beta funds
Step 1: Focus on professional investors (separate account/CIT-database*)
Step 2: Select managers that focus on smart beta
Smart beta focus
Active manager selection criteria
Value
All “Value”-groups: All Cap Value, Foreign Large Value,
Giant Value, Large Cap Core Value, Large Deep Value etc.
Size
All “Mid cap”-groups: Mid Core, Mid Core Growth, Mid
Core Value, Mid Core Deep Value, Mid-relative Value etc.
Low volatility
Ranking all fund according to 3yr trailing standard
deviation and select 20% with lowest risk
Momentum
Funds with “technical investment analysis”
1.880 smart active funds in data set
8
*Morningstar database
Active
Passive
Market cap
Smart active vs. traditional passive
Smart
Smart alpha?
Against MSCI World
Value
Annualized excess return (smart alpha)
Sharpe ratio (difference)
Information ratio
Size
Low vol
Momentum
3,07%
2,34%
2,11%
Numbers5,02%
to be disclosed
in
0,10
0,17
0,18 20160,06
Journal
of
Investing
Spring
issue
0,46
0,58
0,30
0,27
Outperformance probability*
Against MSCI World
Normal probability
Value
Size
Low vol
Momentum
1yr
62,1%
69,3%
53,5%
54,8%
3yr
70,2%
82,8%
67,8%
66,9%
Numbers
to be disclosed
in
5yr
75,6%
1yr
Beta adj. probability
3yr
5yr
88,5%
78,4%
74,4%
99,6%
70,8%
99,8%
78,5%
Journal
of
Investing
201667,0%
80,8%
68,4%
96,1%
84,8%
73,2%
Spring
issue
86,4%
82,7%
Note: Returns after fees are used in the calculations.
9
*Average outperformance probability when choosing a random fund at a random point in time and
hold for a given horizon.
Active
Passive
Market cap
Smart active vs. smart beta
Smart
The ultimate skill test - True smart alpha?
Against smart beta benchmarks
Annualized excess return (true smart alpha)
Sharpe ratio (difference)
Information ratio
Value
Size
Low vol
1,77%
0,78%
-1,00%
Numbers1,31%
to be disclosed
in
0,12
0,07
0,03 2016
-0,06
Journal
of
Investing
Spring
issue
0,26
0,14
0,12
-0,09
True smart alpha exists!
• Active Value, Size and Low vol beat smart beta
• Active Momentum is lagging
Note: Realized active performance vs. backstated benchmark
10
Momentum
Note: Returns after fees are used in the calculations. Source: Morningstar, MSCI and Bloomberg.
Active
Passive
Market cap
Smart active vs. smart beta
Smart
Academic smart alpha?
Percentage of funds with positive estimated alphas
Against Carhart 4-factor model
Alpha > 0
Value
Size
Low vol
Momentum
86,8%
93,2%
77,9%
Numbers89,0%
to be disclosed
in
Above 90% significance level
Above 95% significance level
49,1%
50,5%
59,1%
30,6%
37,4%
39,3%
50,9%
20,2%
29,5%
7,8%
Journal of Investing 2016
Spring issue
Above 99% significance level
20,3%
20,5%
Results equivalent to true smart alpha
11
Note: Returns after fees are used in the calculations. Values state percentage of funds with positive estimated alphas and different statistical significance.
Significance level derived from regression of Carhart 4-factor model: Rfund = rf + b1*(Rm – rf)+ b2*HML +b3*SMB +b4*MOM +α, where Rfund is the return for the
individual fund, rf is the risk free rate, b1…n are factor loadings, Rm is the market return measured by MSCI World Index and HML, SMB and MOM are the CarhartFama-French factors and α is the alpha. Source: Kenneth French Data Library, Morningstar, MSCI and Bloomberg.
Smart beta for tactical purposes
•
Fluctuations in performance for active mandates poses risk in short term even
when long term outperformance is expected
•
Theoretical probability of an active fund outperforming over a given horizon:
N([Exp. Alpha – Trading costs]/Tracking Error)
•
+ Overlay of non-smart alpha mandates (just like on sectors/countries)
Use smart beta instruments on short investment horizons
Example of cumulative probability of outperformance for an active mandate*
60%
Tactical horizon
55%
50%
45%
40%
1
12
2
3
4
5
6
7
8
9
10
11 12 13
Months
*Exp perf = 100bp, trading cost = (2x)20bp and Tracking error = 600bp.
14
15
16
17
18
19
20
21
22
23
24
Dynamic smart beta…
…impossible?
13
26. november 2015
Tactical make sense even if you are passive
Passive is another way of being active: Style drift
10-2-2015
9-11-2015
8-21-2015
7-31-2015
7-10-2015
6-19-2015
5-29-2015
5-8-2015
4-17-2015
3-27-2015
3-6-2015
2-13-2015
1-23-2015
1-2-2015
12-12-2014
11-21-2014
10-31-2014
10-10-2014
9-19-2014
8-29-2014
8-8-2014
7-18-2014
6-27-2014
6-6-2014
5-16-2014
4-25-2014
4-4-2014
3-14-2014
2-21-2014
1-31-2014
1-10-2014
12-20-2013
11-29-2013
11-8-2013
10-18-2013
9-27-2013
9-6-2013
8-16-2013
7-26-2013
7-5-2013
6-14-2013
5-24-2013
5-3-2013
4-12-2013
3-22-2013
3-1-2013
2-8-2013
Note: MSCI World exposures are found by optimizing with MSCI smart beta benchmarks and finding
weightings that explain MSCI World’s return best. Calculated on 1yr rolling basis.
14
Ot he r
M in V ol
M ome nt um
Size
V alue
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Smart beta popularity
Low volatility is especially popular among professional investors
Let’s focus on low volatility premium
Methodology and data
The survey allowed us to collect the opinions of 128 respondent, largely representative of alternative equity beta users. The survey covers different parts of the
world: however, European respondent were predominant as they represent two-thirds of the sample, while 16% of respondents were from North America and 17%
from other parts of the world, including Asia Pacific, the Middle East, Africa and Latin America. The respondents to the survey were mainly asset managers (64%)
and institutional investors (20%). The majority of respondents were also key decision-makers, including board members and CEOs (12%), CIOs, CROs, heads of
asset allocation or heads of portfolio management (31%), and portfolio or fund managers (27%). Respondents were mainly from large firms having over 10 bEUR
in assets under management (51%) or medium-sized companies with assets under management of between 100 mEUR and 10 bEUR (39%)
15
Source: EDHEC
Approach
Potential performance drivers for low vol
Horizon
Operational
Implied volatility (VIX)
Tactical
Leading indicators (ISM Man.)
Strategic
Valuation (P/E, P/B & P/CF)
Investment alternative
16
Bond yield (US 10yr yield)
Low volatility premium history
Low vol premium has outperformed historically
MSCI World Minimum Volatility index relative to MSCI World
140
130
Low volatility outperforms
broad equity market
120
110
100
90
Low volatility underperforms
broad equity market
80
1988
17
1992
1996
2000
2004
2008
2012
Operational horizon: Implied volatility
VIX vs. low vol excess return
1M excess return
10%
3M excess return
15,0%
5%
7,5%
0%
0,0%
-5%
-7,5%
-10%
-25,0
-15,0%
-12,5
0,0
12,5
1M change in VIX (% -point)
25,0
-25
0
25
3M change in VIX (% -point)
6M excess return
30%
12M excess return
30%
15%
15%
0%
0%
-15%
-15%
-30%
Horizon
Slope
Constant
2
R
50
-30%
-50
18
-50
-25
0
25
6M change in VIX (% -point)
50
-50
-25
0
25
12M change in VIX (% -point)
50
1M
3M
6M
12M
0,14***
0,09***
0,24***
0,29***
0,01
0,09
0,34
0,66*
15,3%
8,8%
22,1%
13,5%
Note: Data from 1990M1-2015M1. *, ** and *** indicate statistical significance at 10%, 5% and 1%
level, respectively.
Tactical horizon: Leading indicator
ISM Man. vs. low vol excess return
3M excess return
20%
1M excess return
10%
5%
10%
0%
0%
-5%
-10%
-10%
-20%
-10
-5
0
1M change in ISM
5
10
6M excess return
30%
15%
0%
0%
-15%
-15%
-30%
-30%
Horizon
Slope
Constant
R
2
-10
-5
0
5
3M change in ISM
10
15
12M excess return
30%
15%
-30
19
-15
-15
0
6M change in ISM
15
30
-30
-15
0
15
12M change in ISM
1M
3M
6M
12M
-0,10*
-0,35***
-0,50***
-0,52***
0,09
0,30
0,61**
1,03***
1,1%
13,0%
25,4%
26,6%
Note: Data from 1990M1-2015M1. *, ** and *** indicate statistical significance at 10%, 5% and 1%
level, respectively.
30
Strategic horizon: Valuation
Low vol increasingly more expensive
P/E-spread (right)
Low vol-equity
MSCI World
P/E
P/CF-spread (right)
Spread
Low vol-equity
MSCI World
P/CF
40
30
20
30
20
15
20
10
10
10
0
5
-10
0
1999
Spread
10
5
0
0
1999
2001
2003
2005
P/B-spread (right)
2007
2009
2011
Low vol-equity
2013
-5
2001
2003
2005
2007
2009
MSCI World
P/B
Spread
5
3
4
2
3
1
2
0
1
0
1999
20
-1
2001
2003
2005
2007
2009
2011
2013
Note: Data from 1999M12-2015M1. Due to data quality in the P/B-serie values have been interpolated
between 2000M2-2000M11 and 2003M2-2003M11.
2011
2013
Strategic horizon: Valuation
Valuation-spread explains future low vol excess return
Slope
Horizon (yrs)
1
2
3
4
5
6
7
8
9
10
P/E-spread
0,49***
0,29***
0,20***
0,04
0,01
0,04*
0,07***
0,05***
0,11***
0,12***
P/B-spread
9,53***
7,37***
5,52***
3,16***
2,74***
1,99***
1,61***
1,78***
2,58***
1,78***
P/CF-spread
2,24***
2,01***
1,15***
0,91***
0,99***
0,52***
0,40***
0,47***
0,63***
0,51***
R2
Horizon (yrs)
1
2
3
4
5
6
7
8
9
10
P/E-spread
10,5%
9,9%
12,7%
1,4%
0,1%
2,8%
15,6%
7,3%
26,7%
37,6%
P/B-spread
25,6%
38,2%
51,3%
31,4%
28,7%
36,5%
46,8%
50,2%
87,5%
75,4%
P/CF-spread
28,8%
53,3%
36,4%
38,0%
54,0%
33,3%
34,7%
46,2%
63,9%
67,7%
5 yrs excess return (left)
Excess return
Spread
20%
10,0
15%
7,5
10%
5,0
5%
2,5
0%
0,0
-5%
-2,5
-10%
-5,0
2004
21
P/CF-spread (right, 5 yrs lead)
2006
2008
2010
2012
2014
2016
Note: Data from 1999M12-2015M1. *, ** and *** indicate statistical significance at 10%, 5% and 1%
level, respectively.
2018
Investment alternative: Bond yield
US 10yr yield vs. low vol excess return (in-/outflow)
3M excess return
20%
1M excess return
10%
5%
10%
0%
0%
-5%
-10%
-10%
-120
-80
-40
40
80
120
1M change in US 10yr (bp)
6M excess return
30%
-20%
-200
15%
0%
0%
-15%
-15%
Horizon
-100
-100
0
100
3M change in US 10yr (bp)
200
12M excess return
30%
15%
-30%
-200
0
100
6M change in 10yr (bp)
200
-30%
-300
-200
-100
0
100
12M change in US 10yr (bp)
200
1M
3M
6M
12M
-0,024***
-0,033***
-0,042***
-0,043***
Constant
0,034
0,043
0,007
-0,073
R2
11,6%
22,7%
30,2%
28,1%
Slope
22
0
Note: Data from 1988M5-2015M1. *, ** and *** indicate statistical significance at 10%, 5% and 1%
level, respectively.
300
Investment alternative: Bond yield
Not explained by downtrend in interest rates
1M excess return
3M excess return
Falling rate
10%
20%
5%
10%
0%
0%
-5%
-10%
-10%
-120
-80
6M excess return
-40
0
40
80
1M change in US 10yr interest rate (bp)
Rising rate
120
-20%
-200
12M excess return
Falling rate
30%
15%
15%
0%
0%
-15%
-15%
-150
-100
-50
0
50
100
6M change in US 10yr interest rate (bp)
Note: Data from 1988M6-2015M1.
-150
Rising rate
-100
-50
Falling rate
0
50
100
150
200
3M change in US 10yr interest rate (bp)
30%
-30%
- 200
23
Rising rate
150
200
-30%
-300
-200
Rising rate
Falling rate
-100
0
100
200
12M change in US 10yr interest rate (bp)
300
Putting it all together
Go for the smart alpha!
Smart alpha for Value, Size and Low volatility
Smart beta is still useful!
Smart beta for Momentum (consider insourcing)
Use for tactical purposes on both return and risk (like sectors)
Dynamic indicators seems possible
Manager consequences
Focus on rewarded styles (use smart alpha)
Evaluate and monitor active managers against smart beta
Use cheap smart beta for fee discussions
24
?
Smart beta in the future
Smart beta has a future…but not as rosy as backtests
•
Academia’s evidence is basically a
backtest and not realized returns
historical returns are a bit too
rosy
•
Increasing smart beta popularity
can bid up valuation
future premia will be lower
(but will not disappear)
•
Smart beta’s transparency makes
them pray for index arbitrageurs
future returns will be lower
•
“Hot and a lot of money” in and out
of smart beta increase risk
25
Factor zoo!
Number of new smart beta:
Source: “…and the Cross-Section of Expected Returns” (Harvey, Liu & Zhu, 2015) via SSRN
Appendix – Supporting numbers (Gross returns)
Percentage of funds beating benchmark
Smart alpha
Benchmark: MSCI World
Value
Size
Low Volatility
Momentum
Excess return
91,5%
94,6%
89,4%
80,5%
Sharpe ratio (difference)
84,5%
88,6%
91,2%
73,6%
Information ratio
91,8%
95,2%
89,4%
80,6%
Benchmark: Smart beta
Value
Size
Low Volatility
Momentum
Excess return
88,9%
78,8%
82,4%
49,2%
Sharpe ratio (difference)
89,4%
76,0%
76,4%
51,6%
Information ratio
89,1%
79,2%
82,4%
49,2%
True smart alpha
26
Appendix – Supporting numbers (Net returns)
Percentage of funds beating benchmark
Smart alpha
Benchmark: MSCI World
Value
Size
Low Volatility
Momentum
Excess return
85,3%
90,8%
81,4%
74,8%
Sharpe ratio (difference)
73,4%
81,5%
82,7%
61,6%
Information ratio
85,4%
91,1%
81,4%
74,8%
Benchmark: Smart beta
Value
Size
Low Volatility
Momentum
Excess return
77,4%
69,1%
63,3%
35,9%
Sharpe ratio (difference)
78,3%
64,6%
58,2%
37,2%
Information ratio
77,5%
69,4%
63,2%
35,7%
True smart alpha
27
Investment alternative: Bond yield
Monitoring low vol excess return
3M chg in US 10yr (6M chg, left)
Implied rate chg (3M chg, left)
Excess return (3M, right)
bp
%
150
-20%
Interest rate rises
Low vol underperforms
100
-15%
-10%
50
-5%
0
0%
5%
-50
10%
-100
Interest rate falls
-150
2006
28
Low vol outperforms
15%
20%
2007
2008
2009
2010
2011
Note: Implied rate change from forward rate curve
2012
2013
2014
2015
2016
Appendix - Measuring low risk
Low risk premium seems robust across definitions
29
Source: Deutsche Bank
MSCI World & MSCI USA from 1995 to 2011. Rebalanced quarterly, long only, max 5% in each stock and max 25% in each sector.
Appendix - Low risk on long horizons
Past 40 yrs may be exceptional in low vol returns
Low risk exposure towards size on the rise
30
Source: Dimensional Fund Advisors (Upper) and Deutsche Bank (Lower). Note to upper graph: US 1928-2012. Beta quintiles
rebalanced annually. Stocks in each quintile are weighted by market capitalization. Note to lower graph: Data: MSCI World from 1995
to 2011. Rebalanced quarterly, long only, max 5% in each stock and max 25% in each sector.
Appendix – Investment alternative: Bond yield
Steepnnes in falling and rising rate environment
Regression results†
Slope
Down
Constant
R
2
# Obs
Slope
Up
Constant
R
2
# Obs
1M
-0,031**
0,009
7,1%
165
-0,02***
0,130
3,7%
154
3M
-0,048***
-0,200
16,9%
168
-0,025***
0,277
5,6%
149
6M
-0,056***
-0,019
19,8%
179
-0,029***
0,281
6,1%
135
-0,03***
2,215
5,1%
203
-0,03***
0,297
14,7%
104
12M
Slope differences
Slope
Up - Down
Down
Up
Diff
Diff in %
Up rate steepness
1M
-0,031
-0,020
0,011
35,3%
Less steep
3M
-0,048
-0,025
0,023
47,3%
Less steep
6M
-0,056
-0,029
0,028
49,0%
Less steep
12M
-0,030
-0,030
0,000
-1,0%
More steep
Note: Data from 1988M6-2015M1.
31
†Coefficients scaled by 100. *, ** and *** indicate statistical significance at a 10%, 5% and 1% level
using a two-sided t-test
Appendix - Low risk and interest rate risk
32
Source: Goldman Sachs
Appendix – Smart beta correlation
Example: EDHEC factor correlation of excess returns
EDHEC SciBeta US long-term track records (Dec 1974 – Dec 2014): Diversified multi-strategy
Mid cap
Momentum
Low volatility
Value
Low investment
Momentum
Low
volatility
Value
Low
investment
High
profitability
0,67
0,63
0,86
0,85
0,74
0,61
0,64
0,74
0,65
0,70
0,82
0,60
0,84
0,51
0,69
Note: All statistics are annualized and daily total returns from 31 December 1974 to 31 December 2014 are used for the US
Long-term universe. The universe contains 500 stocks. The full names of the indices used are: SciBeta United States LTTR
Mid-Cap Diversified Multi-Strategy, SciBeta United States LTTR High-Momentum Diversified Multi-Strategy, SciBeta United
States LTTR Low-Volatility Diversified Multi-Strategy, SciBeta United States LTTR Value Diversified Multi-Strategy, SciBeta
United States LTTR Low Investment Diversified Multi-Strategy and SciBeta United States LTTR High Profitability Diversified
Multi-Strategy.
Source: EDHEC
33
Appendix – Smart beta in single stocks or indices?
Momentum possible in sectors while you need stock pickers
on other factors
Source: AQR
34
Appendix - Smart arbitrage?
•
Some of smart beta’s key selling points are fixed investment rules and
transparency
•
…but is it smart to tell everyone what you buy in advance?
Smart index arbitrage can lower smart beta’s return
•
Already well-known effect from traditional passive investing in e.g. S&P
500 and Russell 2000
•
Negative effect rises as smart beta volume rises in the future
•
Especially the momuntum factor has high turnover and is easy to exploit
35
Appendix – Smart beta AuM ETP US
Appendix – Smart beta AuM ETP EU
Appendix – Passive sector drift
Passive is another way of being active: Sector drift
Market overweighing
the technology sector
in the early 2000
Index funds suffered
the loss from not
being able to get out
in time
Source: S&P, FactSet, ING U.S. Investment management
38
Performance persistence on active management
Past performance for active managers is simple and works (as starting point)
• Enhance value by selecting active fund managers who have a history of superior
performance, or avoiding managers with the worst track records (Harlow & Brown,
2006)
• Persistence is much more pronounced for the top and bottom performers (VidalGarcia, 2012)
• The best use of past relative performance information is to avoid persistently poor
performers (Aragon & Ferson, 2006)
• Performance persistence is even more pronounced for emerging equity funds (Huij &
Post, 2011)
39
Performance persistence (US)
3yr historical performance is simple starting point to chose managers
US equity funds ranked on 3yr past performance. “Worst funds” have
3yr average performance <-2,5%. “Best funds” are residual.
Median for groups
Distributions: Best
Distributions: Worst
Distributions: All Observations
6%
4%
2%
0%
-2%
-4%
-6%
-8%
-10%
2005
2006
2007
Source: Own construction. Data from Bloomberg.
40
2008
2009
2010
2011
2012
Performance persistence globally
Able to detect performance persistence on all markets
EU: Best vs. Worst 2005-2012
US: Best vs. Worst 2005-2012
Worst
Best
Best
14%
Worst
25%
12%
20%
10%
8%
15%
6%
10%
4%
5%
2%
0%
-25%
-20%
-15%
-10%
-5%
0%
5%
10%
15%
20%
25%
0%
-25% -20% -15% -10% -5%
EM: Best vs. Worst 2005-2012
Best
0%
5%
10%
15%
20%
25%
Average performance 2005-2012
EM
Worst
US
EU
18%
9%
16%
8%
14%
7%
12%
6%
10%
5%
8%
4%
6%
3%
2%
4%
1%
2%
0%
-25% -20% -15% -10% -5%
0%
5%
10%
15%
20%
25%
0%
-25% -20% -15% -10%
Source: own construction. Data from Bloomberg, 2005-2012.
41
-5%
0%
5%
10%
15%
20%
25%