Last month BlackRock said it plans to pour resources into
data science finding investment ideas in everything from
social media data to satellite images of busy shipping ports
as it shifts to offering more quantitative funds. One
quant investment shop is going in a different direction.
QMA, the $116 billion multi-asset manager owned by
Prudential Financial that has managed U.S. equities for
institutions since 1975, is somewhat
skeptical of the big-data revolution, even as it is being
touted as the savior of active managers. Many asset managers
will only be able to pull short-term signals out of the vast
data that has become available in recent years, including
on social media platforms, says Joshua Livnat, a managing
director who focuses on global accounting research at QMA.
Instead, QMA is using its data and computer-based expertise to
incorporate insights that have historically been associated
with fundamental researchers into its quant process.
To be considered useful to the firm, the data it uncovers
has to meet a high bar. Among other things, signals or patterns
have to work all over the world, and they have to tie directly
back to an underlying economic rationale, executives at the
Gavin Smith, a vice president and researcher for the firm,
says QMA has had success in analyzing earnings call transcripts
once he began to look at them differently.
Smith explains that he started off looking at the sentiment
of the call, even assessing the ratio of positive to negative
words. But I was mostly intrigued by changes in tone
relative to past calls, he says. Once he realized that
changes in the way that management would talk about results
were predictive of stock returns, he dug in to find out why and
to tie it to an economic reason. If you cant
establish that, then youre falling into the trap of data
mining, he says.
What Smith found was usable: Once the change in sentiment
was detected, he could look back four quarters and see a
decline in sales, profitability or margins. For four quarters
after the change in tone was detected, the companys
The change in tone was giving us insight into the
future. And it wasnt reflected in the hard numbers at the
time of the call, says Smith. Management was seeing
light at the end of the tunnel.
Using unstructured data also decreases the likelihood that
others will stumble across identical insights, reducing the
ability to squeeze extra returns out of expensive big data
work, the firm says. Thats important given the asset
management industrys obsession with data science.
Traditional quants looking at financial statement
data, say, will all uncover the same signal, says Smith,
who emphasizes that hes always monitoring models to make
sure they arent becoming less effective over time.
Next up for QMA is looking at reports from research analysts
at brokerage firms to go beyond over-used measures such as
earnings forecasts and revisions. Analysts are often slow to
change their forecasts, especially when they are negative,
because of concerns about hurting corporate relationships. QMA
wants to see if it can find clues in these texts that could
predict a shift in analyst sentiment on a stock and get ahead
of a forecast.
By gleaning useable pearls from unstructured data, quant
managers are accessing information once the province of
fundamental managers who rely on talking to management,
suppliers, and other qualitative methods. Unstructured data is
nothing more, in some cases, than a word-by-word account of a
conversation with a companys senior executives or a
discussion with investors at a brokerage conference.
Smith says this type of information, including comments on
sites where employees can provide opinions about their
employers or the fairness of their salaries, can give a much
truer picture of a companys culture and reputation. All
of these factors can figure into a stock valuation.
Theres a blurred line between information that
is exploitable by a fundamental manager and what a quant
investor can now take advantage of, says Smith. Its
something QMA hopes will differentiate it in an increasingly