Quantitative Equity Strategy

Transcription

1 Inigo Fraser Jenkins Nomura Global Quantitative Equity Conference in London Quantitative Equity Strategy May 2011 Nomura International plc

2 Contents 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? 2. Improving performance through timing of screening 3. Factor rotation 4. Non-linear interaction factors have been working recently Source: Nomura Equity Strategy 1

3 Nomura Global l Quant Benchmark index relative to market 2 Jan 06 = Inferno Purgatorio Paradiso?? 109 equal-weighted relative 107 asset weighted relative Jan- r- Jul- Oct- Jan- r- Jul- Oct- Jan- r- Jul- Oct- Jan- r- Jul- Oct- Jan- r- Jul- Oct- Jan- r- Jul 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 2

4 Meta Model rt : t 1Correlt 2Valuet 3Growtht 4Momentumt Where Correl is the average pairwise correlation between stocks. Value, Growth and Momentum all refer to the dispersioni of fthese factors across the market. β s.e. t-statistic Correlation Value Growth Momentum Constant R Source: Nomura Equity Strategy 3

5 Predicted d quant returns % Realised 12m FWD Return Predicted 12m FWD Return Ma ay-00 Ma ay-01 Ma ay-02 Ma ay-03 Ma ay-04 Ma ay-05 Ma ay-06 Ma ay-07 Ma ay-08 Ma ay-09 Ma ay-10 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 Source: Nomura Strategy Research ANY AUTHORS NAMED ON THIS REPORT ARE RESEARCH ANALYSTS UNLESS OTHERWISE INDICATED. PLEASE SEE ANALYST(S) CERTIFICATION(S) ON SLIDE 2 AND IMPORTANT DISCLOSURES BEGINNING ON PAGE 4. 4

6 A Average correlation l ti off stocks t k Gl Global b l and d US Correlation Correlation Median 63-day correlation of S&P 500 stocks to the S&P 500 index (LHS) day mean pair pair-wise wise correlation (Global) (RHS) pr-10 pr-08 pr-06 pr-04 pr-02 pr-00 pr-98 pr-96 pr-94 pr-92 pr-90 pr-88 pr-86 pr-84 pr-82 pr-80 pr-78 pr-76 pr pr-72 0 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 5

7 Average correlation of stocks (Europe) Correlation r-00 Oct-00 r-01 Oct-01 r-02 Oct-02 r-03 Oct-03 r-04 Oct-04 r-05 Oct-05 r-06 Oct-06 r-07 Oct-07 r-08 Oct-08 r-09 Oct-09 r-10 Oct-10 6r-11 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

8 Dispersion i of expected growth rates Gap in expected growth, % points 18 Europe 16 World Jan-90 Jan-91 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Chart shows the cross-sectional dispersion of expected growth rate for each region. This is defined as the median long-term expected growth of highexpected-growth names less the median for low-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. Source: IBES, WorldScope, FTSE, Exshare, Nomura Equity Strategy research Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11 7

9 P/E Valuation of expected growth 3.5 Ratio Europe World Ja an-90 Ja an-91 Ja an-92 Ja an-93 Ja an-94 Ja an-95 Ja an-96 Ja an-97 Ja an-98 Ja an-99 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 Ja an-00 Ja an-01 Ja an-02 Ja an-03 Ja an-04 Ja an-05 Ja an-06 Ja an-07 Ja an-08 Ja an-09 Ja an-10 an-11 8J

10 Pi Price/Book valuation of expected growth 5.0 Ratio Europe World Ja an-90 Ja an-91 Ja an-92 Ja an-93 Ja an-94 Ja an-95 Ja an-96 Ja an-97 Ja an-98 Ja an-99 Ja an-00 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. Source: IBES, WorldScope, FTSE, Exshare, Nomura Equity Strategy research 9 Ja an-01 Ja an-02 Ja an-03 Ja an-04 Ja an-05 Ja an-06 Ja an-07 Ja an-08 Ja an-09 Ja an-10 Ja an-11

11 How good are growth measures at actually predicting growth? Mar 93 = Composite Growth Trailing Earnings Internal Growth Expected Growth FY0-FY3 Growth Trailing Revenues Long Term Growth Price/Book 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. 10 Source: IBES, FTSE, WorldScope, Exshare, Nomura Equity Strategy research

12 Valuation of large/small ll caps 3.50 Ratio PE price/book Jan-90 Sep-90 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 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 11

13 Current Trends in Quant Asset Management 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 Looking at alternative ways of using the data that we have already: timing and interaction of factors Factor rotation approaches 12

14 Has Minimum i Variance been re-priced? Cross-sectional distribution of P/E by volatility decile 18.0 PE Ratio Mar-11 Average Average ex 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 Source: Nomura Equity Strategy research 13

15 Has Minimum i Variance been re-priced? Slope of cross-sectional regression of P/E for volatility deciles 1.0 Regression Coefficient 0.5 Valuation downward sloping with volatility Valuation upward sloping with volatility 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-11 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. Source: Nomura Equity Strategy research 14

16 Factor efficacy over the month 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 construction ti and portfolio implementation ti less effective In some cases implementing just before the end of the month can lift performance Source: Nomura Equity Strategy 15

17 PE and PBK become less effective at month end Efficacy of factors for 3 month forward stock selection when implemented on day: Regres ssion coef fficient Price/book PE Day of month Chart 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 for the 300 largest stocks in the FTSE World Europe index. Source: Nomura Equity Strategy 16

18 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: Regre ession Coe efficient Day of the month 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 for the 300 largest stocks in the FTSE World Europe index. Source: Nomura Equity Strategy 17

19 V + M interaction ti also becomes less effective at month end Efficacy of factors for 3 month forward stock selection when implemented on day: 2.50 PBK+12m mtm (3m holding) 2.30 PE+12m mtm (3m holding) PE+1m reversal (1m holding) Day of Month Chart 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 for the 300 largest stocks in the FTSE World Europe index. Source: Nomura Equity Strategy 18

20 Section Header (used to create Tab s and Table of Contents) Style selector Style Ranking 1/2 1/2 Value (cheap) Value (expensive) High FCF yield Low FCF yield High h Growth Low Growth High Profitability Low Profitability High Gearing Low Gearing High Risk Low Risk High Momentum Low Momentum Mid/Small Cap Measure of Long term over reaction 1/4 1/4 P/E of Style relative to history Price/Book of Style relative to history Measure of Short term under reaction Style Momentum (12M) 19

21 Performance of alternative ti style selector model (Europe) May 91= persistence only Value + persistence 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, FTSE, Exshare 20

22 Model recommendations for May This updates our ranking of European styles. We use a value and momentum approach as detailed in European Style Selector, 5 June The model recommends going long styles highlighted at the top of the table and short the styles highlighted at the bottom. Average Price/Book Momentum Rank on PE Rank on Rank on Overall Rank : Current PE Average PE PE (z score) Current PBK Price/Book Momentum 1/4 A + 1/4 B Final Rank PBK (z score) (12m) (A) (B) (C) + 1/2 C Low Profitability High FCF Yield High Gearing Value (Expensive) High Growth Low Growth Low Gearing Large Cap High Risk Low Momentum Low FCF Yield High Profitability High Momentum Value (Cheap) Low Risk Mid/Small Cap Source: Nomura Strategy Research 21

23 Interaction ti factor A normal multifactor model may select a stock as attractive if it is the cheapest in the 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 momentum. An example of this would be: r i, t : t i, t. ( V i, t M i, t 1. 1 V i, t 2. 2 M ) If we make the interaction function non-linear, this will give a disproportionate p weight 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 22

24 Value added d from non-linear interaction ti factor 125 Dec 91 = 100 Out of sample Source: Nomura Equity Strategy research 23

25 Performance of Nomura Global l Multifactor t Model Dec 91 = 100 Return, % pa Annualised monthly volatility Annualised monthly IR 550 Out of sample Whole Period Jun 04 - Present Dec 92 - Jun Live period 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