The EconomistOctober 5th 2019 BriefingAutomatic investing 192
1rises) or yield (paying high dividends). Ini-
tially only a few money-managers had the
technology to crunch the numbers. Now
everybody does.
Increasingly, the strategies of “rules-
based” machine-run investors—those us-
ing algorithms to execute portfolio deci-
sions—are changing. Some quant funds,
like Bridgewater, use algorithms to per-
form data analysis, but call on humans to
select trades. However, many quant funds,
such as Two Sigma and Renaissance Tech-
nologies, are pushing automation even
further, by using machine learning and ar-
tificial intelligence (ai) to enable the ma-
chines to pick which stocks to buy and sell.
This raises the prospect of the comput-
ers taking over human investors’ final task:
analysing information in order to design
investment strategies. If so, that could lead
to a better understanding of how markets
work, and what companies are worth.
The execution of orders on the stock-
market is now dominated by algorithmic
traders. Fewer trades are conducted on the
rowdy floor of the nyseand more on quiet-
ly purring computer servers in New Jersey.
According to Deutsche Bank, 90% of equ-
ity-futures trades and 80% of cash-equity
trades are executed by algorithms without
any human input. Equity-derivative mar-
kets are also dominated by electronic exe-
cution according to Larry Tabb of the Tabb
Group, a research firm.This must be the place
Each day around 7bn shares worth $320bn
change hands on America’s stockmarket.
Much of that volume is high-frequency
trading, in which stocks are flipped at
speed in order to capture fleeting gains.
High-frequency traders, acting as middle-
men, are involved in half of the daily trad-
ing volumes. Even excluding traders,
though, and looking just at investors,
rules-based investors now make the major-
ity of trades.
Three years ago quant funds became the
largest source of institutional trading vol-
ume in the American stockmarket (see
chart 2). They account for 36% of institu-
tional volume so far this year, up from just
18% in 2010, according to the Tabb Group.
Just 10% of institutional trading is done by
traditional equity fund managers, says Du-
bravko Lakos-Bujas of JPMorgan Chase.
Machines are increasingly buying to
hold, too. The total value of American pub-
lic equities is $31tn, as measured by the
Russell 3000, an index. The three types of
computer-managed funds—index funds,
etfs and quant funds—run around 35% of
this (see chart 3). Human managers, such
as traditional hedge funds and other mutu-
al funds, manage just 24%. (The rest, some
40%, is harder to measure and consists of
other kinds of owners, such as companies
which hold lots of their own shares.)Of the $18trn to $19trn of managed as-
sets accounted for, most are looked after by
machines. Index funds manage half of that
pot, around $9trn. Bernstein, a research
firm, says other quantitative equity man-
agers look after another 10-15%, roughly
$2trn. The remaining 35-40%, worth $7 to
$8trn, is overseen by humans.
A prism by which to see the progress of
algorithmic investing is hedge funds. Four
of the world’s five largest—Bridgewater,
aqr, Two Sigma and Renaissance—were
founded specifically to use quantitative
methods. The sole exception, Man Group, a
British hedge fund, bought Numeric, a
quantitative equity manager based in Bos-
ton, in 2014. More than half of Man Group’s
assets under management are now run
quantitatively. A decade ago a quarter of to-
tal hedge-fund assets under management
were in quant funds; now it is 30%, accord-
ing to hfr, a research group. This figure
probably understates the shift given that
traditional funds, like Point72, have adopt-
ed a partly quantitative approach.
The result is that the stockmarket is
now extremely efficient. The new robo-
markets bring much lower costs. Passive
funds charge 0.03-0.09% of assets under
management each year. Active managers
often charge 20 times as much. Hedgefunds, which use leverage and derivatives
to try to boost returns further, take 20% of
returns on top as a performance fee.
The lower cost of executing a trade
means that new information about a com-
pany is instantly reflected in its price. Ac-
cording to Mr Dalio “order execution is
phenomenally better.” Commissions for
trading shares at exchanges are tiny:
$0.0001 per share for both buyer and seller,
according to academics at Chicago Univer-
sity. Rock-bottom fees are being passed on,
too. On October 1st Charles Schwab, a lead-
ing consumer brokerage site, and tdAme-
ritrade, a rival, both announced that they
will cut trading fees to zero.
Cheaper fees have added to liquidity—
which determines how much a trader can
buy or sell before he moves the price of a
share. More liquidity means a lower spread
between the price a trader can buy a share
and the price he can sell one.
But many critics argue that this is mis-
leading, as the liquidity provided by high-
frequency traders is unreliable compared
with that provided by banks. It disappears
in crises, the argument goes. A recent paper
published by Citadel, a hedge fund, refutes
this view. It shows that the spread for exe-
cuting a small trade—of, say $10,000—in a
single company’s stock has fallen dramati-
cally over the past decade and is consis-
tently low. Those for larger trades, of up to
$10m, have, at worst, remained the same
and in most cases improved.Grandmaster flash
The machines’ market dominance is sure
to extend further. The strategy of factors
that humans devised when technology was
more basic is now widely available through
etfs. Some etfs seek out stocks with more
than one factor. Others follow a “risk parity
strategy”, an approach pioneered by Mr Da-
lio which balances the volatility of assets in
different classes. Each added level of com-
plexity leaves less for human stockpickers
to do. “Thirty years ago the best fund man-
ager was the one with the best intuition,”
says David Siegel, co-chairman of Two Sig-
ma. Now those who take a “scientific ap-
proach”, using machines, data and ai, can
have an edge.
To understand the coming develop-
ments in the market, chess offers an in-
structive example. In 1997 Deep Blue, an
ibm supercomputer, beat Garry Kasparov,
the reigning world champion. It was a tri-
umph of machine over man—up to a point.
Deep Blue had been programmed using
rules written by human players. It played
in a human style, but better and more
quickly than any human could.
Jump to 2017, when Google unveiled
AlphaZero, a computer that had been given
the rules of chess and then taught itself
how to play. It took four hours of training to
be able to beat Stockfish, the best chessGoodbye, Gordon Gekko^2Source:
TABB Group*Excludingretailandhigh-frequencytradingfirms
†Institutionsincludingpensionfunds,mutualfunds
and other money managers ‡EstimateUnited States, share of institutional trading
volumeofshares*,%0102030402010 11 12 13 14 15 16 17 18 19‡Assetmanagers†Hedge funds Quant fundsBanksPassive aggression^1*August31stSource:JPMorganChaseUSEquityStrategy&
GlobalQuantResearch,EPFRAssets tracking an index
% of measured equity assets under passive management010203040502003 05 07 09 11 13 15 17 19*United StatesRest of worldTo t a l