Sie sind auf Seite 1von 7

E

q
u
i
t
y

R
e
s
e
a
r
c
h









I
n
t
e
r
n
a
t
i
o
n
a
l

T
o
p
i
c
s












May 30, 2013




Bulls, Bears, & Birds
Market Analysis Through Social Media

GAULTYGER
WATERSTONES
Author
Shine Zaw-Aung
Tel: (1) 516 730 6070


Gaultyger, Waterstones & Co. LLC
64 Wall Street
New York, NY 10005
United States of America
Tel: (1) 212 250 2500



Hedge funds selling points have changed dramatically in recent years.
Funds used to sell themselves as market-beaters. Now, they claim to provide
long-term stability in returns and low volatilities. Funds are now trying to
harness new developments in technology to get an investment edge.
Sentiment Analysis has been popular with investment funds recently. An
early mover was London-based Derwent Capital, which returned 1.85% in
February 2011 using a sentiment analysis algorithm. This return, on the
managed capital of US$100 million, beat a bear market, and the firm expected
an annual return on investment of between 15 pc and 20 pc on the new fund.
Sentiments are a form of signaling long used by tech companies.
Amazon, eBay, TripAdvisor and many more companies use rating systems to
convey to the consumers quality and reliability of the proffered product. Used
on the capital markets, sentiments signal a securitys attractiveness to current
and future holders of the security.
Sentiment and volumetric Twitter data are useful indicators of stocks
performance. Our licensed proprietary software which uses social media
analysis through Twitter and Facebook has managed to produce a portfolio
which outperformed base indices in 1Q13.
90.0
95.0
100.0
105.0
110.0
115.0
1/7 1/14 1/21 1/28 2/4 2/11 2/18 2/25 3/4 3/11 3/18 3/25
S&P 500 (GSPC Index) DJIA Periander
GW's Periander Fund outperforms benchmark indices in 1Q13
Birds, Bulls, and Bears
May 30, 2013 2
GAULTYGER
WATERSTONES
Introduction
For the hedge funds, it has been an uncertain and
disappointing decade. During the financial crisis, the HFRX
Global Fund Index -- an industry indicator of hedge-fund
performance posted calendar year decline for the first time
since 1998. After a brief rebound in 2009, the returns for
2010 were likewise disappointing. The index posted a return
greater than two percent in only one month (December
2010), and even this meagre improvement was cancelled
out soon by the political instability of 2011. In 2011, the
index posted another calendar year decline (See Chart 1).
During the last decade, the S&P 500 has outperformed its
hedge-fund rival for ten straight years, with the exception of
2008 when both fell sharply. A simple investment portfolio
(60-40 in equity and sovereign bonds) would have delivered
returns of more than 90% over the past decade, compared
to a meagre 17% (after management fees) in hedge funds.
To correct this trend, hedge fund has long tried to tap into
online barometers of public opinion to harness the crowd
data. As early as 2001, there are attempts to extract useful
market sentiments and investment advice from stock
message boards.
Simple search query volumetric analysis has been done
to forecast the performance for feature films, video
games, and songs. These analyses found the sheer
volume is highly predictive of future outcomes; in other
words, more online exposure a product possesses, more
successful it is. Search-based predictions are strongly
correlated with realized outcomes for movies (0.85) and
video games (0.76) and moderately correlated for music
(0.56).
In a separate experiment, 2.89 million tweets referring to 24
different movies released over a period of three months
were collected. Keywords present in the movie titles are
used as search argument to extract tweets. When the
tweets are compared with the Hollywood Stock Exchange
index -- a market where the prices for movie stocks can
accurately predict real box ofce results --- the resulting
regression model is an accurate prediction of movie
performances. (However, it should be noted that the HSX is
not an actual market. At the end of the rst weekend after
the movies debut, the HSX adjusts the price for a movie
stock to reect the actual box ofce gross. If the movie
performs better than expected, the price goes up). The
0
20
40
60
80
100
120
Jan-08 Jul-09 Jan-11 Jul-12
Hedge Funds' Dismal Recovery
HFRX Index
1 Source: GW Research
Birds, Bulls, and Bears
May 30, 2013 3
GAULTYGER
WATERSTONES
studys model using the tweet-rate time-series
outperformed the HSX-based model by returning more
accurate predictions.
A relationship between Twitter sentiment and Gallup
Poll of Consumer Confidence has also been identified.
A study was conducted by OConner et al. where the
analysts tallied over 1 billion Twitter messages posted over
the years 2008 and 2009 and compared it to polls, namely
the Gallup Organizations Economic Condence index.
The index asks the respondents to rate the overall economic
health of the country. They also used a second poll, Gallups
daily tracking poll for the presidential job approval rating for
Barack Obama over the course of 2009, which is reported
as 3-day rolling averages. For positive or negative
sentiments, they used a pre-defined lexicon from
OpinionFinder. It is a word-list that contains around 1,600
positive and 1,200 negative words. OConner et al. do not
use the lexicons distinctions between weak and strong
words. A message is dened as positive if it contains any
positive word, and negative if it contains any negative word,
but the study allows a tweet to be both positive and
negative. This relatively simple sentiment detector is shown
to replicate both consumer condence and presidential job
approval polls.

Twitter Sentiment Accurately Reflects Public Mood


Source: OConner et al (2010) 2
Birds, Bulls, and Bears
May 30, 2013 4
GAULTYGER
WATERSTONES
Periander Fund
To further survey the potential and possibility of a market-
beating sentimental analysis tool, Gaultyger Waterstones
has licensed a proprietary software. The patented software,
named Periander, is used for listed companies with biggest
online presence and attention. The software analyses a
companys presence, popularity, and positivity using the
following:

Social media presence (Facebook, Twitter, etc.)

Organic social media growth (rate of increase in likes
shares tweets and retweets associated with the
companys online presence)

Positivity (portion of favorable social media presence as
gauged by prop lexical software)
There are some difficulties. To analyze the data set,
academic researchers rely on thousands of workers from
the Amazon Mechanical Turk to label tweets, positive,
negative, or neutral. Each tweet was labeled at least three
times. This is a time-consuming, resource-consuming
practice impractical for real-time market predictions.
Instead, Periander relies on an automated lexicon and self-
learning algorithm to process sentiments as reflected by
tweets and posts. (For instance, the strong emotions such
as love or hate are assigned extreme positive and
negative values).
Periander is used on both most mentioned companies
online and on thirty companies that made up the Dow Jones
Industrial Average. However, not all companies are included
due to query limitations (company names such as 3M or
Johnson & Johnson are too noisy for crawlers). This
analysis is cross-referenced with base indices. Over the
study period, Dow Jones Industrial Average increased by
8% and S&P 500 increased by 6%. How does Perianders
stock pickings stack up?
Portfolio comprising most mentioned brands during the
period outperformed the market. Straight-forward
portfolio of most mentioned companies returned 8.8%, while
the portfolio of most mentioned companies whose positivity
is increasing (i.e., companies whose online approval
increased) returned 12%.
Relative popularity is more important than absolute
popularity. While most mentioned brands outperformed the
market, various weighed portfolios (i.e., where most
Top 12 Listed Companies
by weekly Twitter mentions

Source: GW Research
3
Birds, Bulls, and Bears
May 30, 2013 5
GAULTYGER
WATERSTONES
mentioned brands such as Apple or Facebook are weighed
according to their popularity) returns less than straight-
forward portfolios above.
Organic social media growth is a good indicator to
stock performance. Using the brands online presence and
popularity on January 1
st
2013 as a reference (Jan 1
st
=
100), an index is constructed. Firms dropping in social
media popularity (<100 at the end of the monitoring period)
or languishing (between 100-120 at the end of the period)
underperformed the market. Portfolios containing these
companies lost around 4% in value. Meanwhile companies
whose online profile grew aggressively, which was
observed in ~10% of the cases (index >180 at the end of
the period), their portfolio return is 17% -- double the DJIA
benchmark. (See Chart 4).

These portfolio are only updated weekly, to account for
liquidity in the markets, and to prevent high-frequency
trading biases. Periander is capable of returning similar
returns on daily trading basis. The software is also capable
of various portfolio combinations.
Various Periander Portfolios and Their Returns

Source: GW Research
4
Birds, Bulls, and Bears
May 30, 2013 6
GAULTYGER
WATERSTONES
One of the best performers is that of stocks which have
increased in positivity and decreased in negativity. At a
three-week trial run in late May 2013, this portfolio returned
3.08% -- slightly above the best performing index (which
returned 2.76%), and significantly above DJIA and S&P500,
which returned 0.2% and 0.6% respectively during the same
period.
An ideal portfolio can be constructed by taking both
long and short positions based on changes in Twitter.
By taking a long-position on companies gaining in
popularity/positivity and taking a short-position on
companies losing in popularity/positivity, Periander can
effectively hedge and multiply returns.



Ten Most Popular Companies Online
% Approval

... And Ten Least
% Disapproval

Source: GW Research
5
Birds, Bulls, and Bears
May 30, 2013 7
GAULTYGER
WATERSTONES
Appendix
List of Companies Included in Periander Analysis

Name Ticker Symbol
Adidas ADS.DE
Alcoa AA
Amazon AMZN
American Express AXP
Apple AAPL
AT&T T
Bank of America BAC
Blackberry BBRY
BMW BMW.DE
Boeing BA
Canon CAJ
Capital One COF
Caterpillar CAT
CBS CBS
Cisco CSCO
CNN TWC (Parent)
Coca-Cola KO
Daimler DAI.DE
Dell DELL
Disney DIS
Dupont DD
ebay EBAY
Exxon XOM
Facebook FB
Ferrari / Fiat FIATY
Ford F
Fox NWS
GE GE
Google GOOG
Name Ticker Symbol
Home Depot HD
Honda HMC
HP HPQ
HTC 2498.TW
IBM IBM
Intel INTC
JPMorgan JPM
KFC YUM (Parent)
LG 066570.KS
McDonalds MCD
Mercedes Benz DAI.DE (Parent)
Merck MRK
Microsoft MSFT
NBC GE (Parent)
Netflix NFLX
Nike NKE
Nokia NOK
Pepsi PEP
Pfizer PFE
Procter & Gamble PG
Samsung 005930.KS
Sony SNE
Starbucks SBUX
Taco Bell YUM (Parent)
Target TGT
United Health UNH
United Technologies
(UTC)
UTX
Verizon VZ
Wal-Mart WMT
Yahoo YHOO

Das könnte Ihnen auch gefallen