the prattle primer
TRANSCRIPT
Contents
Introduction
The Prattle Process, Part 1: Why Robots Are Reading
The Prattle Process, Part 2: How Central Banks Move Markets
The Prattle Process, Part 3: Under the Hood
Conclusion
Introduction
What does Prattle do—and why? We put together this primer
to give short, clear answers to those two questions. A tale
littered with reading robots, incomprehensible bureaucrats, and
a financial crisis, the following pages will take the reader on an
interesting—and hopefully helpful—journey towards an
understanding of our company, its vision, and the technological
approach we believe is part of the future of finance.
The Prattle Process, Part 1:
Why Robots Are Reading
In large part, the economy is shaped by words. The attitudes
of investors, bankers, politicians, and the general public drive
the market—a linguistic and monetary reality that has only
become more apparent since the ascendency of the internet in
global trade.
Because of this phenomenon, a methodology of systematic
language interpretation was developed. This process, known as
sentiment analysis, had several forms. Basic sentiment analysis
used a fairly straightforward method: first, two separate
dictionaries of positive and negative buzzwords were created;
then, the sum total of the negative buzzwords in a given
document was subtracted from the sum total of the positive
words in that document; and this simple mathematical
operation produced the document’s score.
Including positive and negative phrases into their term
dictionaries, the most sophisticated iterations of sentiment
analysis bolstered accuracy by taking this higher degree of
complexity into account.
The advantages of such an automated process were numerous:
programs could consume far more information far more quickly
than a human analyst—or even a team of analysts—ever could;
programs could produce evaluations nearly instantaneously;
and, perhaps most importantly, programs did not fall prey to the
various biases that plague human interpretation.
Yet for all its advantages, classic sentiment analysis could not
reproduce the sophistication of human evaluation. To truly
assess a piece of communication, an interpretation must
account for far more than individual words and phrases:
meaning develops over sentences, paragraphs, and a piece as
a whole, and credible evaluations take such breadth into
account.
In addition, these techniques often operated with dictionaries of
words assembled by programmers—not analysts with domain
expertise. Since the accuracy and inclusiveness of the term
dictionaries is vital to the credibility of the automated
evaluations, the failure of programmers to incorporate domain
expertise into the foundation of the automated process counted
as a significant strike against classic sentiment analysis.
Such shortcomings have inspired the founders of Prattle to
develop a new generation of sentiment analysis. Leveraging the
inclusivity, speed, and objectivity of classic sentiment analysis
while benefiting from the sophistication and depth of domain
expertise, this propriety process puts the best of both under
one hood.
The Prattle Process, Part 2:
How Central Banks Move Markets
While Prattle’s language interpreting tech has broad financial
applications, our process was originally built to deal with a
very specific set of institutions: central banks. Because of this,
the story of how Prattle built its machine doesn’t begin with the
particulars of our innovations…it begins with the history of these
institutions.
Many modern central banks are practically engines of content.
The Federal Reserve, for instance, produces press releases and
meeting minutes, gives speeches and congressional testimony,
and arranges press conferences and interviews. Outside of the
People’s Bank of China, this level of dedication is standard
among the world’s leading central banks. While transparency
may be the status quo these days, turn back the pages of
central banking history 30 years, and you’ll find institutions with
an altogether different approach to communication. Back in the
early nineties, central banks were the black boxes of finance.
When it came to opaque communication, former Fed Chairman
Alan Greenspan was a particularly gifted practitioner. He and his
predecessor, Paul Volcker, had developed a rhetoric so
renowned for its incomprehensibility that it even earned its own
moniker: Fedspeak. For Greenspan, Fed communications
should serve as a buffer between the delicate inner workings of
the Fed and the prying eyes and (perhaps) contaminating
influence of those outside, and so, when Greenspan did speak,
his language was so esoteric, so convoluted that there seemed
little point to even listen.
Slowly, Greenspan began to see Fed communications in a
different light, and, by the end of his time as chairman,
Fedspeak was no longer used as a tool of obfuscation. Instead,
Greenspan used such opportunities to give politicians and the
public insight into the Fed’s thinking—a policy of open
communication that has not only blossomed under the reign of
Bernanke and Yellen, but in central banks all over the world.1
These days central banks use communication not only as a
means of explaining their policy decisions, but as policy in its
own right. As transparency began to be embraced, it became
clear that communication may have more utility than was first
thought. While undoubtedly useful for encouraging trust in
central banks, communications actually powerfully influence
markets, and, given the power central banks wield over the
respective markets they oversee, this makes perfect sense.
And this power has only grown since the financial crisis.
Despite this development, the connection between central
bank communications and the market has yet to receive the
1 It should be noted that the Fed did not introduce transparency as a central banking tool. That distinction is the Bank of England’s, who began implementing transparency measures in the eighties. Their forward-thinking strategies eventually influenced the rest of Europe and America to adopt the tool.
kind of attention due its significance: academic literature and
private research publications have hardly touched the topic.
Why? We’ll save the answer to that question and (now that the
stage is set) finally dive into the gears of Prattle’s machine in the
third and final section of this book.
The Prattle Process, Part 3:
Under the Hood
Why have the connections between central bank
communications and the market been largely untouched? 2
Because much of central bank watching has been absorbed in
an interpretive practice lifted straight from your university
English classes.
2 The communications we're referring to here are the text (not data) releases.
As we touched on earlier, central banks produce a wide variety
of content. But, rather than being treated as the financial data
that they are, these communications are approached with a
very different mindset—almost as if they were poetry.
In traditional literary studies, close reading is taught as a
standard means of interpreting texts. In short, the method uses
the subtle, granular details embedded within a given text as the
main source of interpreted meaning. While close reading is
adept at elaborating literary works, its merits as means of
evaluating central banking communications aren't nearly as
obvious.
Despite this, close reading seems to be the precise method
of choice for orthodox central bank watchers.
For example, a standard breakdown of a speech by Fed
Chairman Janet Yellen would likely spend a sizable chunk of its
print on a select few words and phrases. Such an analysis
would use the smallest details to form an assertion of the
central bank's mood: the Fed is hawkish (for instance) because
Yellen used "moderate" instead of "modest."
The issues with this methodology are numerous, but there are
at least four large problems worth discussing here:
1. Central banks produce too much material. If central banks
only published press releases updated incrementally through
track changes, then each minute change has a great deal of
importance, and close reading could have some merit. But
central banks produce a litany of content, and, therefore, proper
evaluation necessitates a more comprehensive approach.
2. Human interpretation is simply too prone to cognitive
errors. Each element of a standard close-reading analysis of
central bank communications, whether it's word choice or the
particularly history used to evaluate those words, is subject to
the innumerable biases and mistakes that plague human
evaluation. When it comes to poetry, such drawbacks may be
negligible, but, when it comes to producing objective analyses
of communications that move billions of dollars, "negligible" is
not a word that comes to mind.
3. The detail-centric nature of this approach—and the errors
come along with it—hide what could be the most important
question in central bank watching. While what central bank
communications reveal about monetary policy is undoubtedly
important, what is perhaps equally important is how
communications themselves act as policy. Words move the
markets, and the preoccupation with what a handful of
expressions could mean about future policy obscures the
immediate effects of the policy of words currently at work.
4. The type of conclusions this method produces are difficult
to operationalize. Investing has becoming a game of
quantitative models—even the discretionary macro space is
moving in that direction—and quantitative models need
quantitative data. Qualitative assessments, like those that
standard analyses produce, are hard to plug into multi-factor
financial models. "Fairly dovish" just isn't an ideal input.
It seems clear that close-reading is not a viable method of
interpreting central bank communications, but, as we covered
in an earlier section, classic automated interpretation (sentiment
analysis) has equally fatal flaws.
It is in this context that Prattle developed its method of analysis.
Taking the best from automated interpretation technology and
domain expertise, Prattle has produced the world's first
unbiased, comprehensive, and quantitative evaluations of these
institutional communications: the Prattle Central Bank
Sentiment Indexes.
Prattle's Central Bank Sentiment Indexes are rooted in
reference texts. These texts are central bank communications
that have led to particular, identifiable market reactions—the
type and level of which allow the texts to be expertly scored.
For central banks, this score is an indication of the
“hawkishness” or “dovishness” of a central bank’s position on the
economy. A hawkish central bank views the economy as strong
and growing and, because of this perception, will soon
implement contractionary monetary policy—raising interest
rates to ensure credit is less available—in an attempt to keep
the market from overheating. A dovish central bank believes the
economy is struggling and takes the corresponding strategy:
lowering interest rates to encourage growth through a climate
of easier credit. Directly connected to varying degrees of
market reaction, these reference documents are firmly rooted
in history—making them an excellent foundation of comparison.
Using these reference texts, Prattle has mathematically linked
specific words, phrases, sentences, paragraphs and whole
communications to specific market reactions. Expressions
linked to hawkish policy and the corresponding market reaction
are awarded positive numbers based on the level of the
response. Conversely, dovish terms are awarded negative
numbers. This lexicon of hawkish and dovish expressions is the
backbone of our methodology.
With a lexicon in place, it now becomes possible to accurately
evaluate current central bank communications.
Aggregating text from all the streams of a given central bank's
communication within whatever timeframe is desired, our
process then generates a score for the sample in light of the
pool of hawkish and dovish communications. The score
generated is the average rating of all the hawkish and dovish
expressions embedded within the selected documents for a
given time period.
This score represents the central bank’s “mood”—i.e. their
inflation expectations. We call each bank’s mood their “Index,”
and these signals are the only comprehensive, unbiased, and
quantitative data on the economic outlook of central banks in
existence.
These scores can be used by quantitative traders as a plug-in
for their multi-factor models or by portfolio managers looking
for a discrete, unbiased measure of economic performance.
There are numerous applications for Prattle's central bank data,
but decoding these institutional communications is only the first
application of our approach.
The future of our analytical technology is transforming all
manner of market-moving texts into tradable data. Corporate
communications, regulator documents, articles in the financial
news media—these are only a few major currents in the ocean
of content that informs price, and we at Prattle seek to
understand and map the economic influence of these flows.
In the today's global economy, that process begins with central
banks, but it certainly doesn't end there. The “meaning” of text in
the information age is only beginning to be understood, and
sentiment analysis technology is the key to progress in this
endeavor. Prattle is building systems trained to draw from a
flood of words the remarkable, untapped potential of these
data streams, helping lead the way towards the future of
finance.
Conclusion
What does Prattle do—and why? We hope this primer has
given short, clear answers to those two questions. If you’re
interesting in our solutions—or would simply like to learn
more—our team is always ready to help. Contacting us is easy.
Just visit prattle.co and click the “contact” button, and one of our
team members will reply as soon as possible. Thanks for your
time, and we hope to hear from you soon.
The Prattle Team