finding alpha in esg data - truvalue labs...2019/05/06  · he has an extensive track record of...

Post on 27-Jun-2020

0 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Finding Alpha in ESG Data

May 2019 Alpha Strategies

Dozens of Strategies for Achieving Outperformance

May 2019

Dr. Stephen MalinakChief Data and Analytics OfficerTruvalue Labs

Stephen leads Truvalue Labs’ quantitative research team in applying artificial intelligence and machine learning techniques to create new financial signals from unstructured data. An industry leader in quantitative analytics, Stephen joined Truvalue Labs from Thomson Reuters, where he spearheaded the company’s quantitative analytics offering, StarMine, and developed over 15 quantitative models. He has an extensive track record of successful predictive models using widely varying techniques across numerous domains.

He attended college at the Massachusetts Institute of Technology where he received his S.B. in Electrical Engineering and Computer Science. Stephen went on to receive his Masters and Ph.D. in Engineering-Economic Systems from Stanford University.

Alpha Strategies

3

Executive Summary

4

• AI-technology mines unstructured data to uncover hidden risks and opportunities

• Customizable platform provides proprietary and customizable alternative datasets

• Next-gen solution provides real-time, objective, transparent signals

• Focused on actual ESG performance vs. company policy

Truvalue Labs brings structure to unstructured data to uncover the intangible factors that drive company value.

5

Truvalue Labs produces four key scores from text data.

Pulse Score• Capture day-to-day variation• Responds to news as it happens

Insight Score• Applies EWMA to pulse score • Provides rating equivalents for

longer-term investors

Momentum Score• Slope of Insight Score over TTM• Identifies companies with

improving or deteriorating ESG

TTM Volume Score• # of articles tagged to SASB

categories in past 12 months

• Measure of information flow

1) Negative exclusion (screening out ESG controversies)

2) Positive ESG screening (seeking positive alpha from ESG topics)

3) Spotting future ESG leaders (active engagement with future leaders)

4) Linear behavioral quant factor: adding ESG to long-only quant factors

5) Long-short performance: adding ESG to long-short quant factors

6) ESG exposures as a risk factor (TVL ESG scores vs. common risk factors)

7) Broad ESG indices (smarter beta)

8) Thematic ESG indices (e.g. Data Security, Consumer Welfare)

9) Monitoring what ESG issues are material over time

10) Combining TVL data with traditional ESG data: Serafeim / Harvard paper

11) Using pulse to alert on spotlight events (alerting for fundamental analysts)

12) Short-term trading of biggest spotlight events (short-term event studies)

Truvalue Labs ESG data can drive over a dozen strategies for achieving quantitative outperformance and gaining ESG research insight.

6

Equity Screening

ESG ActivityQuant Factor

ESG Indices

ESG Research

ESG Events

7

Equity Screening Strategies

Traditional ESG funds, based on narrow exclusion strategies, tend to hug the benchmark.

8

For R1K stocks with high data volume, positive ESG companies significantly outperform negative ESG companies.

9

0-33 33-67 67-100

67-100

33-67

17-33

0-17

Insight

-2.10%

3.52%

PercentilesMomentum

TVL Insight and Momentum can be combined to identify likely future ESG leader companies.

10

0-33 33-67 67-100 0-33 33-67 67-100

67-100 2.68% 2.26% 0.15% 67-

100

33-67 1.96% 4.39% 5.34% 33-67

0-33 1.11% 2.13% 1.36% 0-33

PercentilesMomentum

Insight

PercentilesMomentum

Insight

4.94%

Each new piece of TVL ESG information adds to potential portfolio performance.

11

TVL Strategies

12

ESG ActivityQuant Factor

Smaller companies outperform larger ones, but heavily discussed small companies perform best of all (on ACWX).

13

Both high Activity and high Insight are good.What if we multiply Activity Rank by Insight Rank?

14

Insight Rank x (Data Volume / $ Trading Volume) = ESG Activity Signal (backtest for ACWX benchmark)

1515

ESG Activity Signal greatly outperforms ACWX.

16

ESG Activity Signal adds to multifactor long-shortSmart Beta performance.

17

5-Factor Smart Beta = Size + PriceMo + MinVol + Value + Quality

TVL ESG factors show consistent negative exposure to traditional value factors.

18

19

ESG Indices

ESG, index, and quantitative experts collaborate to create indices customized to specific needs.

20

TVL has partnered with customers to develop custom index solutions.

21

-1.51%

-0.74%

0.00%

-0.29%

-0.30%

2.25%

3.81%

-2.00% -1.00% 0.00% 1.00% 2.00% 3.00% 4.00% 5.00%

MSCI SUSA Benchmark

FTSE4Good Benchmark

S&P Equal-Weight Benchmark

S&P 500, equal weight, quarterly rebalance

S&P 500 with Categorical ESG Exclusions

TVL Insight Strategy (163)

TVL Momentum Strategy (78)

Annualized Alpha with Quarterly Rebalance and Transaction Costs

27 of the 30 SASB categories show positive alpha and can be combined to create thematic indices.

22

23

ESG Research

SASB has mapped materiality by industry.

24

TVL Data Volume can be used to track dynamic materiality by company. (Example: Facebook)

25

SASB material categories account for 80% of the overall alpha.

26

Truvalue labs provides a 3rd party ESG perspective that complements company self-reported data.

27To truly evaluate a company and eliminate ESG “blind spots” requires a comprehensive solution.

Company ESG Policy Disclosure as Communicated by the Company

Actual Company Behavior from an External Perspective

TruValue Labs

provides a 360° view

Prof. George Serafeim of Harvard demonstrated some ways that TVL data adds performance to MSCI data.

28

Major financial firms such as State Street are using TVL ESG factors for portfolio analytics.

29

TVL research also covers many private companies that have public debt or may be approaching IPO stage.

30

Third party perspective includes both positive and negative sentiment on specific ESG topics.

3131

32

ESG Events

Spotlight Events highlight the most important ESG stories over time for each company (e.g. Boeing).

33

Equifax: example of price overreaction to bad news; potential for short-term reversion strategy.

34

Spotlight events include metadata that enable evaluation of the importance of breaking events.

Breaking duration: how many days it takes to trigger spotlight event• Biggest events tend to break suddenly, in 1 or 2 days• Events can last for up to two weeks

Breaking volume magnitude: how many times threshold volume• Threshold = volume required to trigger any spotlight event, based on

typical data volume patterns• Biggest events trigger 3 to 4 or more times that threshold volume

Breaking SpotScore – Pstart: how much does the pulse jump• Compare pulse just before the event to average breaking pulse score• Stock price appears to react more to change in pulse than to level 35

Preliminary event studies show price responses lasting up to a month when volume > 3x and Δ pulse > 20.

36

37

Conclusions and Future Work

1) Negative exclusion (screening out ESG controversies)

2) Positive ESG screening (seeking positive alpha from ESG topics)

3) Spotting future ESG leaders (active engagement with future leaders)

4) Linear behavioral quant factor: adding ESG to long-only quant factors

5) Long-short performance: adding ESG to long-short quant factors

6) ESG exposures as a risk factor (TVL ESG scores vs. common risk factors)

7) Broad ESG indices (smarter beta)

8) Thematic ESG indices (e.g. Data Security, Consumer Welfare)

9) Monitoring what ESG issues are material over time

10) Combining TVL data with traditional ESG data: Serafeim / Harvard paper

11) Using pulse to alert on spotlight events (alerting for fundamental analysts)

12) Short-term trading of biggest spotlight events (short-term event studies)

Truvalue Labs ESG data can drive over a dozen strategies for achieving quantitative outperformance and gaining ESG research insight.

38

Equity Screening

ESG ActivityQuant Factor

ESG Indices

ESG Research

ESG Events

13) ESG predictive power for bankruptcies, defaults, and credit ratings

14) ESG impact on credit spreads and bond pricing

15) ESG driven fixed income indices

16) Predicting changes in ESG ratings

17) Mapping material factors to specific financial measures

18) Corporate engagement

19) Mapping SASB ESG topics to SDGs to score companies on SDGs

20) Creating outperformance while contributing to SDGs

21) Scoring industries, sectors, funds and ETFs on ESG topics

22) Monitoring external fund managers for asset owners

23) Score proprietary ESG data for custom signals and ESG reporting

24) Expanding NLP technology into related topics on intangible risk

Ongoing research projects look at additional applications.

39

ESG Impact onFixed Income

Deeper ESG Integration

SDGs

Fund Research

TVL Cloud

40

Appendix

TVL algorithms identify material issues, quantifythem, then produce a suite of data and analytics.

41

AGGREGATE EXTRACT ANALYZE GENERATE

Trade Blogs, Industry-Specific Publications, Twitter ESG Articles

NGO Sources (CDP, Echo), Watchdogs

Local, National, and International News

Dynamic Web Feeds

Category Scoring w/Freshness Weighting

Signal Categorization

Entity Detection

Content Parsing

Volatility / Confidence

Event Detection

Density Analysis

Salience Clustering

Peer, Sector, Benchmark Comparison

Portfolio Analysis

Alerts and Reports

Company Scorecardsand Trends

Terminal

SAAS

API

DELIVER

top related