“流動(dòng)性因素”：評(píng)估使用流動(dòng)性數(shù)據(jù)的優(yōu)點(diǎn)

1、Global Quantitative & Derivatives Strategy 26 May 2020The Flow FactorAssessing the merits of using flow dataThis note explores whether sentiment affects stock prices and whether there is any predictive power in using particular definitions of flows as a measure of sentiment. Specifically, we assess

2、whether flow is a factor, how it interacts with traditional factors, and its role in multi-factor investing.Is Flow a Factor?Many discretionary investors use flow data, yet there is little research on its effectiveness from either a commercial or academic basis. Our analysis suggests, there is indee

3、d evidence that flow data may be used as a factor and our back-testing suggests above average annualized returns (5.8%), low draw-downs (-6.8%), a high hit rate (65.0%) and a Sharpe ratio near to 1.0 over the past decade.Whats the most effective definition of Flows?Using flow Momentum, Trend, Positi

4、oning and Positioning Delta as key flow definitions, we observe that in the US flow Positioning is an effective strategy (3.0% L/S rtn), yet in Europe it performs poorly (-2.3% L/S rtn), and there are similar findings in Asia Pac (-2.7% L/S rtn). By contrast, flow Momentum delivers the best returns

5、in Europe (+7.8% L/S rtn) compared to other regions. In an attempt to find one universal definition across the regions, we find the combination of ST flow Momentum coupled with flow Positioning is the most effective return metric. We call this the Flow Factor.Flows Predictive or Coincident?Investors

6、 typically perceive flows to be a leading indicator. Our analysis suggests it may just inform you about what is happening right now and the momentum in the signal may persist for up to a quarter. However, beyond 4- 5 months, the alpha decay frequently turns negative. Moreover, among the average bask

7、et of stocks with the most attractive flow data, we often find relatively cheaper valuations than those stocks that have the least attractive flow data.How does flow data impact Factor/Style returns?Using the definition of the Flow Factor described above, we learn that investment style returns can,

8、on average, be enhanced by the addition of flow data. In general, the benefit for long-only Styles is observed through lower volatility and increased Sharpe ratios, alongside lower drawdowns. The results are further boosted for long/short portfolios. It is noteworthy, that the Flow Factor shows a po

9、sitive returns correlation to Quality, Momentum and Low Risk Styles; however, the cross sectional rank correlations suggests little to no correlation among traditional risk factors.Introducing flow data into a multi-factor modelWe highlight two concepts to introduce flow data into a multi-factor mod

10、el (Value, Growth, Momentum, & Quality). 1) Enhance the model with flows; turning a 4 factor model into a 5 factor model, and 2) introduce flows by substituting it for one factor at a time; maintain a 4 factor model. The results for L/S strategies show higher returns and Sharpe ratios for the 5-fact

11、or model, adding flows and eliminating Growth leads to a better 4-factor model.Global Quantitative Strategy Khuram Chaudhry AC(44-20) 7134-6297 HYPERLINK mailto:khuram.chaudhry khuram.chaudhryBloomberg JPMA CHAUDHRY J.P. Morgan Securities plcAyub Hanif, PhD AC(44-20) 7742-5620 HYPERLINK mailto:ayub.

12、hanif ay HYPERLINK mailto:ub.hanif ub.hanifBloomberg JPMA HANIF J.P. Morgan Securities plcWilliam Summer AC(44-20) 7742-6079 HYPERLINK mailto:william.summer william.summerJ.P. Morgan Securities plcGlobal Head of Quantitative & Derivatives StrategyMarko Kolanovic, PhD(1-212) 622-3677 HYPERLINK mailto

13、:marko.kolanovic marko.kolanovicJ.P. Morgan Securities LLCGlobal Quantitative and Derivatives StrategyMarko KolanovicGlobal Head Dubravko Lakos-BujasGlobal/Americas Narendra SinghAmericasArun JainAmericasKhuram ChaudhryEMEAAyub HanifEMEAWilliam SummerEMEARobert SmithAsia PacificBerowne HlavatyAsia P

14、acificMixo DasAsia PacificSee page 35 for analyst certification and important disclosures, including non-US analyst disclosures.J.P. Morgan does and seeks to do business with companies covered in its research reports. As a result, investors should be aware that the firm may have a conflict of intere

15、st that could affect the objectivity of this report. Investors should consider this report as only a single factor in making their investment decision. HYPERLINK / Table of Contents HYPERLINK l _bookmark0 The Flow Factor One Page Summary 3 HYPERLINK l _bookmark1 Introduction to Flow, Positioning and

16、 Investment insights 4 HYPERLINK l _bookmark2 Flow Definitions 5 HYPERLINK l _bookmark4 What flows best? What is the right definition to use? 6 HYPERLINK l _bookmark5 Europe 7 HYPERLINK l _bookmark6 US 8 HYPERLINK l _bookmark7 Developed Markets (GDM) 9 HYPERLINK l _bookmark8 Japan 10 HYPERLINK l _bo

17、okmark10 Emerging Markets (GEM) 11 HYPERLINK l _bookmark11 Asia ex-Japan 12 HYPERLINK l _bookmark12 Introducing the Flow Factor 13 HYPERLINK l _bookmark18 Flow factor composition comparison 14 HYPERLINK l _bookmark19 Alpha Persistence and Alpha Decay 15 HYPERLINK l _bookmark20 Flow Factor in Active

18、vs. Passive universe 17 HYPERLINK l _bookmark23 Flows, positioning and valuations 20 HYPERLINK l _bookmark25 Comparing High/Low Flow and Positioning 21 HYPERLINK l _bookmark26 Investment Style and the Flow Factor 22 HYPERLINK l _bookmark29 Can Flow enhance existing Style? 24 HYPERLINK l _bookmark32

19、Building a Multi-Factor Model 26Appendices HYPERLINK l _bookmark34 Appendix I: Characteristics of Flow Factors 28 HYPERLINK l _bookmark35 Appendix II: Investment Style and Flow 29 HYPERLINK l _bookmark36 Appendix III: Enhancing Investment Style with Flows 30 HYPERLINK l _bookmark37 Appendix IV: Mult

20、i-Factor Models + Flow 32 HYPERLINK l _bookmark38 Appendix V: Statistical Definitions 33 HYPERLINK l _bookmark39 Appendix VI: Simple Definitions Style & Factor 34The Flow Factor One Page SummaryIs Flow a Factor?Many discretionary investors use flow data, yet there is little research on its effective

21、ness from either a commercial or academic basis. Our analysis suggests, there is indeed evidence that flow data may be used as a factor and our back-testing suggests above-average annualized returns (5.8%), low draw-downs (-6.8%), a high hit rate (65.0%) and a Sharpe ratio near to 1.0 over the past

22、decade.Whats the most effective definition of Flows? Using flow Momentum, Trend, Positioning and Positioning Delta as key flow definitions, we observethat, in the US flow, Positioning is an effectiveHow does Flow data impact Factor/Style returns? Using the definition of the Flow Factor as described

23、above, we learn that investment style returns can, on average, be enhanced by the addition of flow data. Ingeneral, the benefit for long-only Styles is observed through lower volatility, and increased Sharpe ratios alongside lower drawdowns. The results are further boosted for long/short portfolios.

24、 It is noteworthy, that the Flow Factor shows a positive returns correlation to Quality, Momentum and Low Risk Styles; however, the cross sectional rank correlations suggests little to no correlation among traditional risk factors.Table 1: Enhancing Investment Style with Flow (EW) L/Sstrategy (3.0%

25、L/S rtn), yet, in Europe, it performspoorly (-2.3% L/S rtn), and there are similar findings inAnn. RetSharpeMax DDAsia Pac (-2.7% L/S rtn). By contrast, flowFlow Factor5.80%0.99-6.80%Momentum delivers the best returns in EuropeValue-0.70%-0.06-22.70%(+7.8% L/S rtn) compared to other regions. In anVa

26、lue + Flow Factor4.00%0.47-10.60%attempt to find one universal definition across theGrowth-2.00%-0.27-32.20%regions, we find the combination of ST flowGrowth + Flow Factor2.90%0.42-12.50%Momentum with flow Positioning to be the mostQuality6.90%0.73-12.90%effective return metric. We call this the Flo

27、w Factor.Quality + Flow Factor7.80%0.97-9.20%Flows Predictive or CoincidentMomentum9.90%0.8-11.70%Investors typically perceive flows to be a leadingMomentum + Flow Factor13.40%1.37-9.50%indicator. Our analysis suggests it may just inform you about what is happening right now and the momentum in the

28、signal may persist for up to a quarter. However, beyond 4-5 months, the alpha decay frequently turns negative. Moreover, among the average basket of stocks with the most attractive flow data, we often find relatively cheaper valuations than those stocks that have the least attractive flow data.Figur

29、e 1: Flow factor Alpha persistence lagged signal L/S ReturnsDownward trending Returns profile Europe US GDM EMJapanAverageNegative Returns6.0%4.5%L/S Returns3.0%1.5%0.0%-1.5%-3.0%-4.5%Source: Source: J.P. Morgan Quantitative and Derivatives Strategy, EPFR Global Test Period 2011 2019Introducing flow

30、 data into a multi-factor modelWe highlight two concepts to introduce flow data into a multi-factor model (Value, Growth, Momentum, & Quality). 1) Enhance the model with flows; turning a 4- factor model into a 5-factor model, and 2) introduce flows by substituting it for one factor at a time; mainta

31、in a 4-factor model. The results for L/S strategies show a higher returns and Sharpe ratios for the 5-factor model, adding flows and eliminating Growth leads to a better 4-factor model.Table 2: Flow incorporated multi-factor models (Europe L/S)0123456789101112Ann. RetSharpeMax DDV, G, Q, M9.60%1.14-

32、11.00%V, G, Q, M + Flow10.70%1.27-10.60%V, G, Q + Flow5.20%0.92-7.10%V, G, M + Flow8.20%1.2-7.80%x-months LaggedSource: J.P. Morgan Quantitative and Derivatives Strategies, EPFR Global.V, Q, M + Flow12.70%1.63-8.70%G, Q, M + Flow10.00%0.89-15.80%Source: J.P. Morgan Quantitative and Derivatives Strat

33、egy, EPFR Global Test Period 2011 2019Introduction to Flow, Positioning and Investment insightsWhy look at flow data?Our motivation to write a research note on flows, positioning, and the insights of how to use this type of data, stems from two key observations. 1) Many investors and market commenta

34、tors cite flow data, yet research from both an academic or commercial perspective is fairly scarce in assessing its efficacy or market impact. 2) Ultimately, these data are used to measure sentiment, but it is unclear whether the flowFlow data sourcesFor the purpose of our analysis into flow data, w

35、e have sourced data from EPFR global, and consider mutual fund flows across assets, and then within equities.The chart and table below illustrates that 31 trillion US$ is captured across Equities, Bonds, Money Markets, Balanced and Alternative market funds, where Equities make-up 52% of aggregate AU

36、M.Figure 3: Flow AUMmomentum is more useful than reversal in extreme investor positioning, and is this predictive or coincident to stock price moves?In this research note our objective is use various definitions of flow data, to assess its impact as a tool for addressing market sentiment, and then t

37、o understand the merits of using flow data as a standalone factor and/or its role alongside traditional factor & multi- factor models. In essence, this note tests whether sentiment affects stock prices, and whether there is any predictive power in using particular definitions of flowsCoverage by Ass

38、etsAUMUS$ MnsAlternative139,970Balanced2,450,537Bond7,567,962Equity16,438,063Money Market4,780,205 Total31,376,737 as a measure of sentiment.What is flow data aiming to capture?Investors (Institutional and Retail) actively allocate their funds/money across different mutual funds, ETFs or individual

39、stocks. This allocation intends to capture investors risk appetite on the basis of past returns for funds (with a view the trend will persist), or a shift to changing economic conditions through the selection of different assets, regions, sectors or investment styles. The flow data looks at active &

40、 passive funds which have inflows and outflows, and relate these positions to aggregate fund holdings or individual stock positions.Theory suggests net buying as a result of capital flows is highly correlated to the traditional definitions of price Momentum. Yet, does this still hold if rising senti

41、ment lifts equity prices above what is perceived to be the fundamental value?Moreover, if flows impact stock prices then we are trying to learn whether it is sensible for investors to build on this momentum (predictive), or if this is precisely the time to consider neglected areas within financial m

42、arkets (mispricing). We aim to analyse these data while attempting to find a flow-based rule that is also effective across different regions.Source: J.P. Morgan Quantitative and Derivatives Strategies, EPFR Global.Within the equity component, 59% is deemed to reflect active equity positions and 41%

43、passive equity positions. Additionally, ETFs represent 23% of the equity sample, and 77% are Non-ETFs.Figure 4: Flow data AUM coverage within EquitiesSource: J.P. Morgan Quantitative and Derivatives Strategies, EPFR Global.From a geographical perspective, the USA reflects 58% of the equity sample, a

44、nd the largest top 50 funds capture just below 30% of the total AuM.Figure 5: Flow data AUM split by size & geographySource: J.P. Morgan Quantitative and Derivatives Strategies, EPFR Global.Flow DefinitionsWe begin our assessment of flows by defining some common flow factors. These include magnitude

45、 measures, momentum, mean-reversion among active managers and changes in these sentiments. The factors under consideration are summarised in HYPERLINK l _bookmark3 Table 2 below with detailed definitions thereafter.FactorDefinitionTable 2: Flow FactorsShort Term Flow1M (21 day) EWMA of daily stock f

46、lows Long Term Flow6M (130 day) EWMA of daily stock flowsFlow TrendMeasure of magnitude and persistence of 260D cumulative flowsPositioningMeasures active managers stock allocation relative to the benchmark weight Positioning Delta6M change in Positioning scoreSource: J.P. Morgan Quantitative and De

47、rivatives StrategyShort-Term Flow and Long- Term Flow are magnitude measures of flowShort/Long Term FlowThe first set of factors we define aim to capture short and long-term directional flows. Calculated as the daily flow percentage: the USD Flow amount for each stock on that day, divided by its lat

48、est AUM. Then we take an exponentially weighted moving average with a half-life of 1M (21 days) for the short-term signal and 6M (130 days) for the long-term signal.Flow Trend is a momentumFlow Trendflow pcti,t= $flowi,t AuMi,Tmeasure of flowNext, to capture longer-term flow dynamics, we calculate t

49、he cumulative Flow Trend a measure which is aimed at capturing the long-term magnitude and persistence of flows. Calculated as the t-statistic of the regression of the 260-day cumulative flow percentage time series against time (in days). A high beta of the univariate regression measures the magnitu

50、de of the flow, with the standard error of the beta measures how persistent the flows have been. Thus, the higher the ratio of the two (the t-stat) the stronger and more persistent the flows.Positioning measures the levelPositioningflow trendi,t=i,t se()i,Tof mean-reversion among active managersThe

51、Positioning score is designed to capture potential mean reversion in active manager allocations that are overextended relative to the market. Calculated by taking the weight of the stock based on reported AUMs and subtracting the benchmark weight of the respective stock, normalised by the benchmark

52、weight.AuMi,tFF MCapi,tposN= i=1AuMi,tNi=1FF MCapi,ti,tFF MCapi,tNi=1FF MCapi,tWhat flows best? What is the right definition to use?We start this exercise by testing the 5 candidate flow factor strategies: 1) Short-Term (ST) Flow, 2) Long- Term (LT) Flow, 3) Flow Trend, 4) Positioning, and 5) Positi

53、oning Delta.We run a series of analyses including strategy performance and risk measurement, and test for orthogonality among the factor using correlations (both rolling returns and cross-sectional ranks).Additionally, regional and global investors should also benefit from the comparisons of each of

54、 these flow strategies for the USA, Asia-Pac ex-Japan, Japan, GDM, GEM and European universes.Overall, we see a great degree of regional variability. For example, consider the following chart on performance of ST Flow and Positioning factors across regions. We find generally positive performance for

55、 ST Flow across regions, with Positioning showing positive risk-adjusted performance in the US and Japan.Figure 6: ST Flow and Positioning factor Return vs. Sharpe Ratio across Global regionsST FlowPositioingEuropeUSGEMJapanAsia x-JP USEuropeGDMAsia x-JPJapanGDMGEM1.201.000.80Sharpe Ratio0.600.400.2

56、00.00-0.20-0.40-0.60-2.0%-1.0%0.0%1.0%2.0%3.0%4.0%5.0%6.0%Annualised ReturnSource: J.P. Morgan Quantitative and Derivatives Strategy, EPFR GlobalBelow we provide a summary of the key observations from our analysis.EuropeInterestingly, most flow factors show stellar performance in Europe. This holds

57、up for both Long- only and Long/Short. From a portfolio construction perspective, we note all strategies suffer low Max Drawdowns and, alongside lower Volatility, could be a worthwhile inclusion into a broader equity portfolio.Unfortunately, bar a few instances, it would appear most flow factors are

58、 capturing the same underlying risk element as evidenced in both traditional returns correlations but also in cross-sectional rank correlations.USFlow strategy effectiveness is less conclusive in the US. From both an active strategy and a Long/Short perspective; magnitude-based measures and momentum

59、 measures both show mediocre to poor performance.There does, however, appear to be informational content in positioning data of active managers with both Positioning and Positioning Delta showing positive gains and positive risk-adjusted performance.Developed Markets (GDM)Most flow-based factors sho

60、w weak to negative performance in Developed Markets. LT magnitude, momentum and positioning all deliver negative risk- adjusted performance from an active strategy perspective with ST and LT Flow posting a positive Sharpe Long/Short.JapanOur backtesting suggests flow factors are not effective in Jap