Inigo Fraser Jenkins
[email protected]
+44 20 7102 4658
Nomura Global Quantitative Equity Conference in London
+44 20 7102 4658
Quantitative Equity Strategy
May 2011
© Nomura International plc © Nomura International plc
C
t
t
Contents
P f
f
t
d l i th
t
i
t
Performance of quant models in the current environment
The factor outlook
Four views on current areas of quant development:
1 Is there a case for minimum variance?
1. Is there a case for minimum variance?
2. Improving performance through timing of screening
3. Factor rotation
4 Non-linear interaction factors have been working recently
4. Non linear interaction factors have been working recently
1
N
Gl b l Q
t B
h
k i d
l ti
t
k t
Nomura Global Quant Benchmark index relative to market
2 Jan 06 = 100
111
113
Inferno
Purgatorio
Paradiso??
107
109
equal-weighted relative asset weighted relative103
105
99
101
95
97
06
06
06
06
07
07
07
07
08
08
08
08
09
09
09
09
10
10
10
10
-11
-11
-11
Figure shows the performance of equal-weighted and asset-weighted quant fund indices relative to the global market and also the absolute return of the Hedge Fund Research Equity Market Neutral index. Source: Hedge Fund Research, Bloomberg, Nomura Equity Strategy research
Jan-
Apr-
Jul-
Oct-
Jan-
Apr-
Jul-
Oct-
Jan-
Apr-
Jul-
Oct-
Jan-
Apr-
Jul-
Oct-
Jan-
Apr-
Jul-
Oct-
Jan
-Apr
-Jul
-2M t M d l
Meta Model
t
t
t
t
t
t
Correl
Value
Growth
Momentum
r
:
1
2
3
4
Where Correl is the average pairwise correlation between stocks. Value, Growth and Momentum all refer to
th di
i
f th
f
t
th
k t
β
s.e.
t-statistic
C
l i
48 2
11 3
4 3
the dispersion of these factors across the market.
Correlation
-48.2
11.3
-4.3
Value
-17.1
9.6
-1.8
Growth
2.8
0.4
6.7
Growth
2.8
0.4
6.7
Momentum
-0.6
0.5
-1.2
Constant
-10.6
8.1
-1.3
R
20.7
3P di t d
t
t
35
Predicted quant returns
% 20 25 30 Realised 12m FWD Return Predicted 12m FWD Return 5 10 15 -5 0 5 -15 -10 a y-00 a y-01 a y-02 a y-03 a y-04 a y-05 a y-06 a y-07 a y-08 a y-09 a y-10 M a M a M a M a M a M a M a M a M a M a M a
The quant proxy for which we are forecasting 12 -month forward return (r) is the Nomura Quant Benchmark . Before 2006, we use a sector-neutral blend of P/E and 12 -month price momentum combined using a simple average approach for the quant proxy. We use a regression based approach to forecast the 12 -month forward returns using average
pairwise correlation between global stocks and dispersion of value, growth and momentum as detailed in Quant and Existentialism , 9 October 2010.
Source: Nomura Strategy Research
ANY AUTHORS NAMED ON THIS REPORT ARE RESEARCH ANALYSTS UNLESS OTHERWISE INDICATED.
A
l ti
f t
k Gl b l
d US
Average correlation of stocks Global and US
0.45
0.9
Correlation
Correlation
0.35
0.40
0.7
0.8
0.25
0.30
0.35
0.5
0.6
0.7
0.15
0.20
0.25
0.3
0.4
0.5
0 05
0.10
0.15
0 1
0.2
0.3
75 day mean pair wise Median 63-day correlation of S&P 500
stocks to the S&P 500 index (LHS)
0.00
0.05
0
0.1
p
r-72
p
r-74
p
r-76
p
r-78
p
r-80
p
r-82
p
r-84
p
r-86
p
r-88
p
r-90
p
r-92
p
r-94
p
r-96
p
r-98
p
r-00
p
r-02
p
r-04
p
r-06
p
r-08
p
r-10
75 day mean pair-wise correlation (Global) (RHS)
Global correlation is defined as the mean of all the pairwise correlations between stocks over the prior 75 days. Universe is the 500 largest stocks in the FTSE World index. US correlation is the median 63-Day correlation of S&P 500 Stocks to the S&P 500 Index.
Source: Nomura Equity Strategy, Ned Davis
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
A
p
5A
l ti
f t
k (E
)
Average correlation of stocks (Europe)
Correlation
0.5
0.6
0.4
0.5
0.3
0 1
0.2
0.0
0.1
0
0
0
0
0
1
0
1
0
2
0
2
0
3
0
3
0
4
0
4
0
5
0
5
0
6
0
6
0
7
0
7
0
8
0
8
0
9
0
9
1
0
1
0
1
1
Correlation is defined as the mean of all the pairwise correlations between stocks over the prior 75 days. Universe is the stocks that are in or ever have been in the FTSE World index. Source: Nomura Equity Strategy
Apr-0
Oct-0
Apr-0
Oct-0
Apr-0
Oct-0
Apr-0
Oct-0
Apr-0
Oct-0
Apr-0
Oct-0
Apr-0
Oct-0
Apr-0
Oct-0
Apr-0
Oct-0
Apr-0
Oct-0
Apr-1
Oct-1
Apr-1
6Di
i
f
t d
th
t
Dispersion of expected growth rates
22
Gap in expected
growth, % points
16
18
20
Europe
W ld
12
14
16
World
8
10
12
4
6
2
Jan-90
Jan-91
Jan-92
Jan-93
Jan-94
Jan-95
Jan-96
Jan-97
Jan-98
Jan-99
Jan-00
Jan-01
Jan-02
Jan-03
Jan-04
Jan-05
Jan-06
Jan-07
Jan-08
Jan-09
Jan-10
Jan-1
1
Chart shows the cross-sectional dispersion of expected growth rate for each region. This is defined as the median long-term expected growth of high-expected-growth names less the median for low-high-expected-growth names. The portfolios are defined as the top/bottom quartiles of stocks screened on long-term expected growth and FY0-FY3 expected growth. The universes used are the 300 largest stocks in the FTSE World Europe and the 500 largest companies from the FTSE World index. Portfolios are rebalanced every quarter.
P/E V l
ti
f
t d
th
P/E Valuation of expected growth
3.5
Ratio
3.0
Europe
World
2.5
World
1.5
2.0
1.0
0.5
a
n-90
a
n-91
a
n-92
a
n-93
a
n-94
a
n-95
a
n-96
a
n-97
a
n-98
a
n-99
a
n-00
a
n-01
a
n-02
a
n-03
a
n-04
a
n-05
a
n-06
a
n-07
a
n-08
a
n-09
a
n-10
an-1
1
Chart shows the P/E ratio of the top/bottom quartiles of stocks screened on long-term expected growth and FY0-FY3 expected growth with stocks selected from the 300 largest stocks in the FTSE World Europe and the 500 largest companies from the FTSE World index every quarter.
Source: IBES, FTSE, Exshare, Nomura Strategy research
J
a
J
a
J
a
J
a
J
a
a
J
J
a
J
a
J
a
J
a
J
a
J
a
J
a
J
a
J
a
J
a
a
J
J
a
J
a
J
a
J
a
J
P i /B
k
l
ti
f
t d
th
Price/Book valuation of expected growth
5.0
Ratio
4.0
4.5
Europe
3.0
3.5
4.0
World
2.0
2.5
3.0
1.0
1.5
2.0
0.5
1.0
a
n-90
a
n-91
a
n-92
a
n-93
a
n-94
a
n-95
a
n-96
a
n-97
a
n-98
a
n-99
a
n-00
a
n-01
a
n-02
a
n-03
a
n-04
a
n-05
a
n-06
a
n-07
a
n-08
a
n-09
a
n-10
a
n-1
1
J
a
J
a
J
a
J
a
J
a
a
J
J
a
J
a
J
a
J
a
J
a
J
a
J
a
J
a
J
a
J
a
a
J
J
a
J
a
J
a
J
a
J
a
Chart shows the price/book ratio of the top/bottom quartiles of stocks screened on long-term expected growth and FY0-FY3 expected growth with stocks selected from the 300 largest stocks in the FTSE World Europe and the 500 largest companies from the FTSE World index every quarter.
H
d
th
t
t
ll
di ti
How good are growth measures at actually predicting
growth?
Mar 93 = 100
3000 3500
Composite Growth FY0-FY3 Growth Trailing Earnings Trailing Revenues Internal Growth Long Term Growth
Mar 93 = 100
2000 2500
Expected Growth Price/Book
1500 2000
500 1000
0
Figure shows the ex post one year forward growth for a wide variety of growth factors. This is done by constructing high/low growth portfolios every December on these factors and track the growth in 12 month forward consensus EPS over the subsequent year. We chart the spread in cumulative earnings growth between the high and low growth portfolio which is also the data highlighted in yellow to the right of each sheet. Portfolios are selected from the 500 largest stocks in the FTSE World.
V l
ti
f l
/
ll
Valuation of large/small caps
Ratio
3.00
3.50
PE
price/book
2 00
2.50
price/book
1.50
2.00
0.50
1.00
0.00
a
n-90
e
p-90
ay-91
a
n-92
e
p-92
ay-93
a
n-94
e
p-94
ay-95
a
n-96
e
p-96
ay-97
a
n-98
e
p-98
ay-99
a
n-00
e
p-00
ay-01
a
n-02
e
p-02
ay-03
a
n-04
e
p-04
ay-05
a
n-06
e
p-06
ay-07
a
n-08
e
p-08
ay-09
a
n-10
e
p-10
Figure shows the relative P/E and price/book of the largest/smallest quartile of stocks selected from the FTSE World Europe index and rebalanced every quarter. Source: FTSE, WorldScope, Exshare, IBES, Nomura Equity Strategy research
J
a
S
e
M
J
a
S
e
M
J
a
S
e
M
a
J
S
e
M
J
a
S
e
M
J
a
S
e
M
J
a
S
e
M
J
a
S
e
M
J
a
S
e
M
J
a
S
e
M
J
a
S
e
11C
t T
d i Q
t A
t M
t
Current Trends in Quant Asset Management
Quant Managers are offering wider range of models:
Quant Managers are offering wider range of models:
This includes many non-benchmark strategies eg minimum variance
Strategies applied to broader universe: Emerging markets, small caps
Discretion in quant models
Discretion in quant models
Looking at alternative ways of using the data that we have already: timing
and interaction of factors
Factor rotation approaches
H
Mi i
V i
b
i d?
Has Minimum Variance been re-priced?
Cross-sectional distribution of P/E by volatility decile
18 0
PE Ratio
16 0
17.0
18.0
PE Ratio
Mar-11
Average
Average ex 99-01
14 0
15.0
16.0
Average ex 99 01
13.0
14.0
11.0
12.0
9.0
10.0
1
2
3
4
5
6
7
8
9
10
Figure shows the valuation of volatility deciles defined as the median 12 month forward PE of stocks in each decile screened on trailing 12 months volatility of returns. The universe is the 300 largest stocks in the FTSE World Europe index. The portfolios have been rebalanced every quarter. The average valuation is calculated over the period January 1990 – March 2011.
H
Mi i
V i
b
i d?
Has Minimum Variance been re-priced?
Slope of cross-sectional regression of P/E for volatility deciles
1.0
Regression Coefficient
0.0
0.5
Valuation downward
sloping with volatility
-1.0
-0.5
-2.0
-1.5
-3.0
-2.5
sloping with volatility
Valuation upward
-3.5
Jan-90
Jan-91
Jan-92
Jan-93
Jan-94
Jan-95
Jan-96
Jan-97
Jan-98
Jan-99
Jan-00
Jan-01
Jan-02
Jan-03
Jan-04
Jan-05
Jan-06
Jan-07
Jan-08
Jan-09
Jan-10
Jan-1
1
Figure shows the slope coefficient from a regression of the valuation of volatility deciles on a cross-sectional dummy variable. Valuation is measured as the median observation within the portfolio of consensus 12 month forward PE. Portfolios have been created by sorting on trailing 12 month realised volatility. A negative coefficient implies that valuation is upward sloping with volatility and vice versa. The universe is the 300 largest stocks in the FTSE World Europe index. The portfolios have been rebalanced every quarter.
F t
ffi
th
th
Factor efficacy over the month
There are many managers who use month-end prices for
There are many managers who use month end prices for
screening and then implement factor portfolios early in the
month
There is evidence that this makes end-month factor
t
ti
d
tf li i
l
t ti
l
ff
ti
construction and portfolio implementation less effective
In some cases implementing just before the end of the month
In some cases implementing just before the end of the month
can lift performance
15
PE
d PBK b
l
ff ti
t
th
d
PE and PBK become less effective at month end
Efficacy of factors for 3 month forward stock selection when implemented on day:
2.10
2.30
Price/book
PE
1 70
1.90
2.10
fficient
PE
1 30
1.50
1.70
ssion coe
f
0 90
1.10
1.30
Regre
s
0.70
0.90
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
D
f
th
16Chart shows the regression coefficients of a regression run of 3 month forward returns on factor values screened on a given day of the month. Regressions run 2000-2011 for the 300 largest stocks in the FTSE World Europe index. Source: Nomura Equity Strategy
Sh t t
l l
l
ff ti
t
th
d
Short term reversal also less effective at month end
Efficacy of 1 month reversal factor for 1 month forward stock selection when
implemented on day:
p
y
0 80
1.00
0.60
0.80
efficient
0.20
0.40
ession Co
e
-0.20
0.00
Regr
e
-0.40
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of the month
17
y
Chart shows the regression coefficients of a regression run of 1 month forward returns on factor values screened on a given day of the month. Regressions run 2000-2011 for the 300 largest stocks in the FTSE World Europe index. Source: Nomura Equity Strategy
V
M i t
ti
l
b
l
ff ti
t
th
d
V + M interaction also becomes less effective at month end
Efficacy of factors for 3 month forward stock selection when implemented on day
:
2 10
2.30
2.50
PBK+12m mtm (3m holding)
PE+12m mtm (3m holding)
PE+1m reversal (1m holding)
1 70
1.90
2.10
PE+1m reversal (1m holding)
1.30
1.50
1.70
0.90
1.10
0.70
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
18Chart shows the regression coefficients of a regression run of 3 month forward returns on factor values screened on a given day of the month. Regressions run 2000-2011 for the 300 largest stocks in the FTSE World Europe index. Source: Nomura Equity Strategy
Section Header (used to create Tab s and Table of Contents)
Style selector
Style selector
Style Ranking
Measure of Long term
Measure of Short term
1/2
1/2
•Value (cheap)
Measure of Long term
over reaction
Measure of Short term
under reaction
1/4
1/4
Value (cheap) •Value (expensive) •High FCF yield •Low FCF yield Hi h G th •High Growth •Low Growth •High Profitability •Low ProfitabilityP/E of Style
relative to
history
Price/Book
of Style
relative to
history
Style Momentum
(12M)
•High Gearing •Low Gearing •High Risk •Low Risky
•High Momentum •Low Momentum •Mid/Small Cap 19P f
f lt
ti
t l
l t
d l (E
)
Performance of alternative style selector model (Europe)
450
May 91= 100
350
400
persistence
only
250
300
150
200
Value + persistence
50
100
Value only
50
May-91
Jan-92
Sep-92
May-93
Jan-94
Sep-94
May-95
Jan-96
Sep-96
May-97
Jan-98
Sep-98
May-99
Jan-00
Sep-00
May-01
Jan-02
Sep-02
May-03
Jan-04
Sep-04
May-05
Jan-06
Sep-06
May-07
Jan-08
Sep-08
May-09
Jan-10
Sep-10
Figure shows the relative performance of attractive and unattractive styles according to a range of switching models on a long-short basis. Portfolio holdings g p y g g g g g have been rebalanced each month. Stocks have been equally weighted within the style portfolio at the beginning of each holding period and styles are equally weighted. Style performance is on a total return, common currency basis. From 2000 a three-day implementation lag is introduced as each rebalancing point to allow for a conservative trading horizon. The investable universe is the 300 largest companies in the FTSE World Europe indexSource: Nomura, WorldScope,
M d l
d ti
f
M
This updates our ranking of European styles. We use a value and momentum approach as detailed in European Style
Selector 5 June 2009 The model recommends going long styles highlighted at the top of the table and short the styles
Model recommendations for May
Selector , 5 June 2009. The model recommends going long styles highlighted at the top of the table and short the styles
highlighted at the bottom.
Current PE Average PE PE (z score) Current PBK Average Price/Book Momentum Rank on PE
Rank on Price/Book Rank on Momentum Overall Rank : 1/4 A + 1/4 B Final Rank Current PE Average PE PE (z score) Current PBK
PBK (z score) (12m) (A) Price/Book (B) Momentum (C) 1/4 A + 1/4 B + 1/2 C Final Rank Low Profitability 12.46 14.63 -0.87 1.42 1.65 -0.53 34.93 4 7 1 3.25 1 High FCF Yield 10.67 12.64 -0.84 1.75 2.09 -0.66 33.70 5 5 3 4.00 2 High Gearing 12.46 14.11 -0.56 2.18 2.15 0.05 33.84 8 14 2 6.50 3 Value (Expensive) 17.22 20.62 -0.53 3.95 4.59 -0.36 31.82 9 10 5 7.25 4 High Growth 13.40 16.50 -0.65 2.42 3.82 -1.06 28.42 7 1 11 7.50 5 Low Growth 11.21 12.89 -0.75 1.41 1.62 -0.56 28.63 6 6 9 7.50 5 Low Gearing 15.57 16.83 -0.31 3.39 3.75 -0.28 32.46 13 11 4 8.00 7 Large Cap 11.16 14.08 -0.90 1.80 2.38 -0.83 26.79 3 3 13 8.00 7 High Risk 9.34 15.14 -1.18 1.25 2.10 -0.83 26.47 1 2 15 8.25 9 Low Momentum 10 53 13 15 -0 92 1 36 1 85 -0 82 26 70 2 4 14 8 50 10 Low Momentum 10.53 13.15 -0.92 1.36 1.85 -0.82 26.70 2 4 14 8.50 10 Low FCF Yield 13.30 14.69 -0.52 2.45 2.12 0.53 30.74 10 16 6 9.50 11 High Profitability 13.90 15.14 -0.36 4.39 4.72 -0.21 29.94 11 13 7 9.50 11 High Momentum 14.51 15.54 -0.27 2.33 2.71 -0.41 28.82 14 9 8 9.75 13 Value (Cheap) 9.26 9.95 -0.35 1.17 1.35 -0.52 28.40 12 8 12 11.00 14 Low Risk 13.55 14.06 -0.23 2.84 2.67 0.32 28.44 15 15 10 12.50 15
Source: Nomura Strategy Research
Mid/Small Cap 12.48 12.73 -0.13 1.47 1.57 -0.22 25.07 16 12 16 15.00 16
I t
ti
f t
A normal multifactor model may select a stock as attractive if it is the cheapest in the
Interaction factor
universe but only attains a mediocre score on other factors.
This could be erroneous if what really matters is the joint score on value and
t
A
l
f thi
ld b
momentum. An example of this would be:
)
(
.
.
.
2
1
:
t
i
t
i
t
i
t
i
t
t
i
V
M
V
M
r
If we make the interaction function non-linear, this will give a disproportionate weight
)
(
,
,
,
2
,
1
:
,
t
t
i
t
i
t
i
t
i
t
i
,
g
p p
g
to stocks were the value and momentum signal are aligned.
May help in situations of overcrowding
We use a cubic interaction function to combine factor scores
V l
dd d f
li
i t
ti
f t
Value added from non-linear interaction factor
Dec 91 = 100
125
Out of sample
Dec 91 = 100
115
120
110
105
100
D
ec-91
D
ec-92
D
ec-93
D
ec-94
D
ec-95
D
ec-96
D
ec-97
D
ec-98
D
ec-99
D
ec-00
D
ec-01
D
ec-02
D
ec-03
D
ec-04
D
ec-05
D
ec-06
D
ec-07
D
ec-08
D
ec-09
D
ec-10
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
P f
f N
Gl b l M ltif t
M d l
Performance of Nomura Global Multifactor Model
Dec 91 = 100
550
600
650
O t f
l
Return, % pa Annualised monthly volatility Annualised monthly IRLive period
450
500
550
W hole Period 9.97 7.14 1.40Out of sample
Jun 04 - Present 4.67 3.96 1.18 Dec 92 - Jun 04 12.99 8.28 1.57
Live period
300
350
400
200
250
300
100
150
-91
-92
-93
-94
-95
-96
-97
-98
-99
-00
-01
-02
-03
-04
-05
-06
-07
-08
-09
-10
Out of sample is the period over which the factor coefficients have been constant. Live period is the period over which the stocks selected by the model have been published every month. Source: Nomura Equity Strategy research
24
