[Marxism] Harvesting data

Louis Proyect lnp3 at panix.com
Mon Apr 22 06:27:02 MDT 2019


The Data Race; Fund managers are paying handsomely to sift through 
oceans of data on everything from flight patterns to parking lots for 
clues on deals and trends that move markets. Tim Shufelt looks at why 
sophisticated investors are increasingly turning to Big Data to gain an edge

The Globe and Mail (Canada), April 20, 2019 Saturday, Ontario Edition
Byline: Tim Shufelt, Staff

Last summer, a corporate jet owned by Encana Corp. embarked on a strange 
pattern of flights.

Over the course of a couple of months, the Calgary company's aircraft 
was tracked landing close to oil fields in Oklahoma, Utah and Montana. 
Encana had operations in none of those places. Houston-based Newfield 
Exploration Co., however, owned assets in all three locations.

And it turns out that Newfield's jet was also on the move around the 
same time. Its executives seemed to take a particular interest in flying 
to Denver, which happens to be home to Encana's chief executive officer, 
Doug Suttles.

The mutual visits hinted that something was in the works. And in early 
November, Encana struck the largest deal in its history when it agreed 
to buy Newfield for US$5.5-billion. An aggressive expansion into U.S. 
shale represented a dramatic change of course for the Canadian driller, 
and the deal caught the market by surprise. When trading started the 
next morning, Encana's stock plummeted by 17 per cent.

A select group of investors, however, may not have been quite so 
shocked. They had access to the movements of Encana's jet, giving them 
strong clues about what the company's executives were up to in the 
months before the acquisition.

Quandl Inc., a Toronto-based startup, allows investors to peek in on the 
travel habits of companies such as Encana. By stitching together data 
from aircraft registries, corporate filings and flight communications, 
Quandl can track the movements of thousands of corporate jets around the 
world, giving investors a new advance signal of potential marketmoving 
corporate deals.

Known as "alternative data," this kind of insight into publicly traded 
companies is proliferating, as investors and fund managers look to data 
science for an edge that will help them beat the market. Jet activity 
can foreshadow corporate deals; aggregated credit-card tallies can 
reveal consumer trends; satellite imaging can track oil inventories; 
information "scraped" from job sites can indicate who's hiring; data 
collected by auto insurers can give clues on future car-sales figures. 
The vast pools of information being generated by the digital economy 
hold the power to better predict what companies will do and how their 
stocks will perform.

The computer sophistication and machine learning needed to make sense of 
all that information, meanwhile, is quickly evolving. Until recently the 
purview of quantitative hedge funds, alternative data methods are 
spreading to mainstream investing. In December, Quandl was acquired by 
Nasdaq Inc., which runs a data products business in addition to 
operating stock exchanges. The deal amounted to an "inflection point for 
the industry," wrote Richard Johnson, a vice-president in Greenwich 
Associates' market structure and technology group.

Competition among data providers is heating up. In February, Bloomberg 
launched a site featuring 20 alternative data sets. "The race to take 
alternative data mainstream has now begun in earnest," Mr. Johnson said.

It's a race that Canadian investors have been reluctant to join.

Despite its Toronto address, Quandl has zero Canadian names on the list 
of major clients subscribing to its 50-odd data products, which range 
from estimates of Tesla sales to industrial-auction results. Outside of 
the big pension funds, in fact, it appears few Canadian investors are 
dabbling in alternative data.

There are reasons to be cautious. There is a growing discomfort over the 
capacity for Big Data analytics to observe the intimate details of 
people's lives. Meanwhile, the legalities over data collection and 
distribution can be murky, raising concerns over who owns particular 
information and who has the right to sell it. There are tough questions 
regulators are just starting to grapple with, including whether 
sophisticated investors gain an unfair advantage when they have access 
to data that is effectively unavailable to the masses.

And yet, wary Canadian investors run the risk of being left behind if 
they wait too long. Alternative data will soon be essential to 
generating competitive returns, says Tammer Kamel, Quandl's CEO. "It 
will become unacceptable to be basing your investment decisions on what 
happened a few months ago."

For decades, standard financial data has been the lifeblood of 
fundamental investing. Investors glean what they can from whatever 
public companies are required to disclose through regulatory filings and 
quarterly financial statements. A handful of data providers, including 
Bloomberg, Refinitiv and FactSet, have come to dominate the distribution 
of that information. (Refinitiv is partially owned by Thomson Reuters 
Corp., which is controlled by Woodbridge Co. Ltd., owner of The Globe 
and Mail.) In 2018 alone, investors spent in excess of US$30-billion 
globally for access to market data and analysis, according to an 
estimate by BurtonTaylor International Consulting.

Alternative data means anything considered to be outside the realm of 
traditional financial information, but that can yield valuable market or 
company insight.

And it's by no means a newly invented category. Investors have long 
hunted for tradable information outside the bounds of financial 
reporting. It used to be said that the thickness of U.S. Federal Reserve 
chair Alan Greenspan's briefcase could portend monetary policy 
announcements. (A big haul meant he was carrying the documentation to 
support a rate cut, or so the theory went.) Hedge fund managers have 
also been known to directly observe retail foot traffic or cross-border 
shipping or executives appearing in certain airports - anything to get a 
read on what, or how well, a company is doing at that very moment.

What has changed is the sheer volume of data now being produced, 
everywhere. The internet has more than 1.5 billion live sites.

Facebook users create about 3.3 million posts a minute. The Internet of 
Things is connecting everything from cars to household appliances, and 
smartphones are constantly tracking their users' locations. By next 
year, roughly 1.7 megabytes of data will be generated each second for 
every person in the world.

Most commercial information is simply "exhaust" - a byproduct of a 
company's main business, Mr. Kamel says. But there is an active market 
for those companies to turn their data into revenue.

Many telecommunications companies give third parties access to user 
location data for a fee; financial intermediaries will compile 
credit-card transaction data; and policy information from auto insurers 
can reveal which models of cars are selling best. "You can find out 
almost anything you want to know about a stock or a commodity or a 
consumer, if you connect to the right database," Mr. Kamel says. 
"Somebody's taking that measurement."

That's where alternative data providers come in, typically licensing 
that information and turning it into data sets marketed to big hedge 
funds and asset managers. There are currently more than 400 providers 
like Quandl, up from around 100 a decade ago, according to 
alternativedata.org.

"Web scraping," or data extracted from websites, is the largest 
subcategory. Using data scraped from Best Buy Co. Inc.'s website, for 
example, New York-based startup Thinknum showed robust sales in Amazon 
products, such as its Alexa-powered smart speakers, starting around 
Black Friday last year - nearly one month before Amazon.com Inc. 
announced record holiday sales for its devices.

Research firm Opimas estimates that hedge funds and asset managers 
scraping sites for investment purposes accounted for 5 per cent of all 
web traffic last year.

So-called sentiment data scraped from social media, financial news and 
online forums are among the more established alt data products. 
Toronto-based Buzz Indexes built a model that scours sites like 
StockTwits and Twitter for insight into how investors feel about 
individual stocks.

A natural language processing algorithm looks for signs of investor 
positivity toward U.S. large-cap stocks and calculates a sentiment score 
for each name. The 75 stocks with the highest scores are included in the 
Buzz NextGen AI US Sentiment Leaders Index, which, back-tested to the 
start of 2013, has returned an average of 17 per cent annually, compared 
to 11 per cent for the S&P 500 index. Not too long ago, the idea that 
there might be wisdom in the online conversations of investors was met 
with cynicism, says Buzz Indexes founder Jamie Wise. "Today, there's 
probably not a CIO at any major asset manager on the continent that 
isn't thinking about an alt data strategy."

Many have progressed well beyond the thinking stage. BlackRock uses an 
active quantitative approach in its Advantage funds, which search for 
investment signals from a range of data sources, including weather 
patterns, travel-site bookings and employee reviews from sites like 
Glassdoor.com. "Combining millions of responses can indicate a company's 
state of health, as those with happy employees tend to outperform their 
competitors," BlackRock said in a recent brochure for its Advantage 
funds. Meanwhile, Franklin Templeton Investments recently signed a deal 
with platform company Elsen, giving traditional portfolio managers 
easier access to big sets of data.

For large investors and asset managers, getting access to market and 
company intelligence that gets as close as possible to real-time data is 
worth paying good money for. Quandl's data sets range in price from 
US$25,000 to US$250,000 a year. Other products on the market, like 
specialized satellite intelligence, can cost upwards of US$1-million a year.

A pair of professors from the University of California at Berkeley 
recently demonstrated how satellite images' predictive power can justify 
such an extravagant price tag. They looked at daily images of the 
parking lots of major U.S. retailers, including Walmart and Target, over 
a six-year period to identify whether counting car traffic could help 
predict earnings and stock movements. A trading strategy built on buying 
shares in retailers with abnormally high parking lot traffic, and 
shorting those with low traffic, would have paid off handsomely once 
earnings were announced, the analysis found. Compared to a buy-andhold 
approach, that satellite-informed portfolio generated average excess 
returns of 4.6 per cent.

Silicon Valley-based Orbital Insight is one of the leaders in using 
satellite technology to spot tradable economic or company data in real 
time, mostly in the consumer and energy sectors. Last September, RBC 
Capital Markets announced a partnership with Orbital that provides the 
investment bank with geospatial data to include in its equity research. 
An RBC report from January said Orbital's images of storage tanks 
pointed to declining global crude oil inventories from their Decem-ber 
peak - one sign the market could be tightening and prices headed higher.

"Everybody's trying to get into the alternative data space," says 
Fardeen Khan, head of strategic initiatives at RBC Capital Markets.

But he adds that it's not a standalone investment approach. The idea 
behind RBC's arrangement with Orbital, as well as the bank's other data 
science endeavours, is to complement the fundamental and technical 
process. "When you look at alternative data as a standalone, the 
insights are not sufficient to say you should go fully long on this 
company or go short on a specific name," says Mr. Khan.

That sentiment is echoed by Ron Mock, CEO of Ontario Teachers' Pension 
Plan, which uses some data-driven trading strategies and is "leveraging 
the deep insights it's capable of bringing," he said during a discussion 
at the World Economic Forum in Davos, Switzerland, in January. "We have 
to be very, very mindful, that we can't push it so far that we turn our 
brains off."

Nearly a decade ago, hedge funds were the only ones most willing to take 
a chance on such an exotic, untested idea as using alternative data, and 
they have been the main driver behind the growth of that industry. 
Global spending on alternative data sets is currently about US$3-billion 
per year, according to JP Morgan, a small fraction of the size of the 
conventional data business.

For the industry to assume a larger profile, it will need to extend its 
appeal to more traditional asset managers. Now is a good time to do just 
that, Greenwich's Mr. Johnson says. "A lot of active managers are 
struggling to beat passive benchmarks, and they're looking for a new 
edge," he says.

The passive investing craze has made life difficult for traditional 
active managers. Franklin Templeton, for example, saw its global assets 
under management decline by 14 per cent last year.

The average active fund manager, however, has very different data needs 
than a giant U.S. quant fund. Without the infrastructure to analyze raw 
data, most fundamental investors require data that have been ingested, 
formatted and packaged, or fed into platforms they can incorporate into 
their own investment processes.

Quandl's corporate-jet-tracking app is one example of this.

The idea for an aviation-based investment tool came out of a hedge-fund 
trade from early 2017.

A trio of New York-based funds figured out how to track Johnson & 
Johnson's Gulfstream jet on the internet and found it sitting on the 
tarmac at a Swiss airport for more than a week, just a few kilometres 
away from the headquarters of pharmaceutical company Actelion Ltd. 
Convinced a major tie-up was being negotiated, the hedge funds loaded up 
on Actelion shares, which soared when J&J announced a US$30-billion deal 
to acquire the Swiss company a few days later. When Abraham Thomas, 
Quandl's chief data officer, read about that payday, he thought: "What 
if we could automate that process?" By combining flight location data 
and ownership information from several different sources, Quandl can now 
track the flight activity for a majority of the companies in the Russell 
1000 Index.

Most of those companies, however, would prefer to keep that information 
to themselves. They'll often try to conceal their own aircraft through 
subsidiaries and holding companies, or complex leasing arrangements. By 
poring over aircraft registrations, operator licences, public filings 
and corporate parent-subsidiary relationships, Quandl has built a 
database of 29,000 jets and counting.

Most subscribers are using the product as one part of an M&A investing 
strategy, to help shed light on rumours or suspected deals, Mr. Thomas 
says. Others use it for protecting their short positions.

Investors betting against a stock are vulnerable to that company being 
acquired, since such an announcement typically results in a big jump in 
share price - and big losses for short sellers. "On other occasions, 
activist hedge funds want to find out if the CEO is gallivanting around 
the world on the company dime," Mr. Thomas says.

Corporate jet data is one of dozens of data sets that Quandl says puts 
it in the alternative-data lead, Mr. Kamel says. Already, the company 
leads the industry in brand recognition, according to a recent Greenwich 
Associates study. And being acquired by Nasdaq represents a huge boost 
to the company's profile and credibility.

"When you hand someone a card that says Nasdaq - part of the fundamental 
structure of capital markets - that helps a lot," Mr. Kamel says. "Now 
we're standing on the shoulders of a giant."

So far, Canadian hedge funds have taken a pass on alternative data - 
almost all of them, in fact, according to Claire Van Wyk-Allan, head of 
the Canadian chapter of the Alternative Investment Management 
Association. Most Canadian players are just not big enough to justify 
the cost. The hedge fund industry here pales in comparison to behemoths 
on the other side of the border. Bridgewater Associates, for example, 
manages about US$160-billion, while only a handful of Canadian hedge 
funds surpass even the US$1-billion mark.

"We are not able to spend $100,000 every month on all kinds of data. We 
aren't Bridgewater," says Ernest Chan, who runs QTS Capital Management 
in Niagara-on-the-Lake, Ont., and manages a small hedge fund. His own 
experience with alternative data suggests it generates one to two 
percentage points of "alpha," or excess returns. For Bridgewater, that 
would amount to a boost to annual returns of US$1.6-billion to 
US$3.2-billion. "But if you are $100-million fund, alternative data is 
not necessarily a must-have," Mr. Chan says.

The major Canadian pension funds, on the other hand, are certainly big 
enough to use algorithmic trading strategies and advanced data 
analytics. "If you go to a quantitative investment conference, it is 
dominated by pension plans," Mr. Chan says. Canada Pension Plan 
Investment Board, Ontario Teachers' Pension Plan and Alberta Investment 
Management Corp. all declined to comment on how they're using 
alternative data in their investment decisions.

Canada's big asset managers, meanwhile, appear to be on the sidelines 
when it comes to alternative data. While Quandl boasts of having 14 of 
the world's 15 largest asset managers as customers, the company has yet 
to land a big Canadian name. "It's a little frustrating that it's easier 
for me to sell in New York than in my own backyard," Mr. Kamel says. 
Though Toronto has emerged as a global fintech hub, Bay Street asset 
managers seem reluctant to evolve.

"Canadians just might be used to doing things the regular old way," Ms. 
Van Wyk-Allan says.

Part of the hold-up might be in how alternative data is typically 
marketed. "Looking at this data as a source of alpha is like entering 
into a nuclear arms race, where you're constantly in search of the next 
data set," says Ashby Monk, executive director of Stanford University's 
Global Projects Center, which studies, among other things, how 
technology can improve long-term investing. "If you're a patient 
investor, that's probably not the best use case for alternative data." 
Instead, he suggests it can be used to better understand risk in a 
portfolio, to assist in due diligence and to make better capital 
allocation decisions.

Alternative data is by no means risk-free, however. It's an unregulated 
space that lacks legal clarity.

While insider trading generally has a fairly narrow legal definition, 
some alternative data strategies are starting to look awfully similar. 
The common regulatory test for insider trading asks whether a piece of 
information is material and non-public. Alternative data certainly has 
the power to be material to a company's stock, by providing timely 
indicators of a company's health. And if distributed only to a limited 
number of investors, or even exclusively to a single hedge fund, it can 
be difficult to argue that data is public.

"The line between public and material non-public information is key 
here," says Kirsten Thompson, a partner at Dentons and national lead of 
the law firm's transformative technologies and data strategy group. 
"Securities regulation requires you to know which side of that line you 
are on, and alternative data makes that difficult." Some large hedge 
funds are known to avoid buying "exclusive" data sets for fear of legal 
risk.

There is growing alarm, meanwhile, around the sharing of personal data 
by companies that compile it. Last year, it was discovered that location 
information on U.S. cellphone users sold by telecom providers was ending 
up in the hands of bounty hunters. Canadian telcos also sell location 
data to third parties, which they say is only done with users' explicit 
consent.

Canadian privacy legislation regulates the personal information that 
corporations collect, limiting the ways it can be shared.

And in any case, investors are not interested in obtaining anyone's 
personally identifiable information, Quandl's Mr. Thomas says.

"We tell our vendors, if you have personal records, don't even send us 
the data. We don't even want it to touch our servers."

But it can be difficult to truly anonymize certain information.

"Our machine learning and our algorithms are now getting so 
sophisticated that you may have data that you think is not identifiable, 
but the algorithm can, in fact, identify somebody," Ms. Thompson says. 
Data science can also be used to approximate protected data. Access to 
Canadian credit scores is limited, for example, but a fairly accurate 
estimate can be computed from other sources, including social media.

Investors considering accessing alternative data need to conduct their 
due diligence, Ms.Thompson says. "There's a constellation of questions 
you should be asking, including the sources of data, appropriate 
consents, the genesis of the information." Data harvesting methods like 
web scraping can violate the terms and conditions set out on a website's 
fine print, in which case, data vendors would be prevented from selling 
that information.

Quandl says it is offered about 100 datasets each month by companies and 
data hunters looking to cash in on the data they have collected. In 
addition to evaluating the data quality and its potential value to 
investors, the company says it takes pains to trace each potential data 
set back to its original source. "Our customers understand we've vetted 
the data for personal information, insider information and ownership 
issues," Mr. Thomas says.

Outside of Canada, adoption of alternative data is growing fast, despite 
the regulatory limbo. Canadian investors won't have the luxury of taking 
a cautious approach much longer, Mr. Thomas says. "As it spreads wider 
and wider, if I don't have this data, I'm at a disadvantage. It becomes 
table stakes."




More information about the Marxism mailing list