|
 That’s the upshot of two new investment studies which seek to use
Google search and Twitter to inform trading strategies.
The first, from MIT, showed that measuring tweet sentiment on the
day of Federal Open Market Committee meetings yields data which can
be profitably used in trading. (http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2756815)
Another, from Norway, using Google trends data, shows that high
search interest in companies actually predicts low future returns in
the following week, a finding in contrast to earlier similar papers.
(ttp://papers.ssrn.com/sol3/papers.cfm?abstract_id=2756370)
It is, of course, not quite as simple as this, but the studies, part
of a huge rush to invest profitably based on social media and
Internet data, do tend to confirm Twitter’s reputation as a sort of
insiders’ clubhouse, as opposed to Google’s more mass appeal.
Ironically, this view has held back Twitter stock due to fears it
will have difficulty growing much beyond its base of fierce
adherents. Google parent Alphabet stock, in contrast, with its place
embedded at the heart of everything online, is up 38 percent over
one year, as against a 67 percent loss for Twitter.
Twitter does seem to produce good information about sentiment on
those most market-moving of events: Federal Reserve interest rate
decisions.
The MIT study looked at English-language tweets in 2007-14 about the
Fed, measuring sentiment and adjusting for the reach of the tweeter.
“We exploit a new dataset of tweets referencing the Federal Reserve
and show that the content of tweets can be used to predict future
returns, even after controlling for common asset pricing factors,”
the authors, Andrew Lo, a professor at MIT, and Pablo Azar, a PhD
student there, write.
They further “find that a tweet-based asset-allocation strategy
outperforms several benchmarks, including a strategy that buys and
holds a market index as well as a comparable dynamic asset
allocation strategy that does not use Twitter information.”
On days the FOMC meets, a one-standard deviation increase in tweet
sentiment will increase returns by 0.62 percent, the study shows, a
significant outperformance over a short period.
CHICKEN VS EGG VS QE
Importantly, during the 2010-2014 period a simple strategy which
buys stocks on the day before FOMC decisions did not perform well,
in contrast to historical norms, but the Twitter-informed strategy
did. That indicates that the results are not simply a product of the
well known pre-FOMC drift effect of rising stocks.
There are plenty of other caveats though, mostly indicating the need
for further research. Twitter, the Fed and financial markets were
all in unusual or unprecedented situations from 2007-14. Markets
were generally rising, supported in part by totally unprecedented
monetary policy.
[to top of second column] |
At the same time Twitter was coming into its own and became, for
many market participants, a key part of how they monitored events
and one another on FOMC day. While Twitter shareholders might want
lots of new users, it is very possible that tweet-derived data
becomes less useful over time if more of the tweets are pictures of
people’s cute cats batting at Janet Yellen on the television.
A 2015 Johns Hopkins study used Twitter to help generate sentiment
data on company earnings in a predictive way.
On to Google, where earlier studies of search volume showed that
rising interest led to positive one- or two-week returns. This
study, by academics at the Norwegian University of Science and
Technology in Trondheim, did not confirm those findings. Looking at
Google search volume data from 2008 to 2013, they tried to relate
search queries to above-market returns in individual stocks.
From a trader’s point of view the results, while interesting, were
sort of the worst of all worlds. Growing search interest in
companies led to negative returns over subsequent days. Sadly,
though, the effect was small enough that it couldn’t overcome the
drag from transaction costs when turned into a trading strategy.
As with all such studies it is hard to know exactly where this
leaves us.
As a contrarian you might argue that the companies people are
searching for are typically hot momentum stocks, and thus short
candidates. On the other hand, as the authors point out, it may
simply be that the market is getting better and faster at
incorporating information from Internet search results into market
prices. Certainly there are algo-based traders out there trying to.
For now, though, the lesson seems to be: use Twitter to learn what
is going on and Google to find tire shops near you.
(At the time of publication James Saft did not own any direct
investments in securities mentioned in this article. He may be an
owner indirectly as an investor in a fund. You can email him at
jamessaft@jamessaft.com and find more columns at http://blogs.reuters.com/james-saft)
(Editing by James Dalgleish)
[© 2016 Thomson Reuters. All rights
reserved.] Copyright 2016 Reuters. All rights reserved. This material may not be published,
broadcast, rewritten or redistributed. |
