DRAFT
A Sociology of Arbitrage: Market Instruments in a Trading
Room
Daniel
Beunza
Stern
School of Business
New
York University
44
West 4th Street
New
York, NY 10012
212-998-0224
(phone)
dbeunza@stern.nyu.edu
and
David Stark
Columbia
University
Department
of Sociology
1180
Amsterdam Ave
New
York, NY 10027
212-854-3972
(phone)
212-854-2963
(fax)
Prepared for the New York Conference on
Social Studies of Finance, Columbia University and the Social Science Research
Council, May 3-4, 2002, www.coi.columbia/ssf
A Sociology of Arbitrage:
Market Instruments in a Trading Room
Daniel
Beunza and David Stark
I. Introduction
In this paper we develop a sociological
approach to the market instruments that make up modern arbitrage. In contrast to value and momentum investing,
arbitrage involves an art of association – the construction of equivalence
(comparability) of properties across different assets. In place of essential or relational
characteristics, the peculiar valuation that takes place in arbitrage is based
on an operation that makes something the measure of something else –
associating securities to each other. We examine how these associations come about
by studying their material basis in the deployment of persons and things in the
Wall Street trading room of a major international investment bank where we have
been conducting ethnographic field research.
In the analysis that follows we argue that the cognitive work behind
these market instruments involves a situated cognition. A trading room is an engine for generating
equivalencies. Making associations among securities takes place in situ, that is, in a particular place
where formulas are formed by associations among persons. Innovation – making novel and unanticipated
associations – involves the organization of diversity (not fully ordered, but
not random noise) among heterogeneous rationalities.
But the story we tell is not simply another example of social
embeddedness (Granovetter 1985). That approach has rejuvenated economic
sociology; yet social network analysis in its American application is limited
by its obsessive sociologism that addresses only ties among persons (Stark
2001b). Drawing on the work of Michel
Callon (ref.) and Bruno Latour (ref.), we examine socio-technical processes in
which the work of making market instruments is distributed across a network of
persons and artifacts. Because the
social consists of humans and their non-humans (objects, things, artifacts), in
place of studying “society” we must construct a science of associations – an
analysis that examines not only links among persons but also among persons and
things.[1] Interwoven among the mathematics are the
machines.
.............................................................................................................................................................
[1] Although American sociologists have not yet incorporated the insight
that network analysis should include artifacts as well as persons, other social
scientists in this country have been working with similar concepts. Hutchins (1994), for example, argues that
cognition is distributed across a network of persons and instruments, and documents
this dramatically and painstakingly in the case of a US Navy cruiser that is
navigated into port after a power system failure. Suchman’s (1987) pathbreaking
work on human-machine interaction similarly resonates with the work of Callon
and Latour and provides the basis for further studies on distributed
design.
..............................................................................................................................................................
Thus, in studying the contemporary
practices of arbitrage, we examine the socio-cognitive and the socio-technical
processes from which derivates are derived.
In modern finance, the social
meets the calculative, the cognitive meets the technical, and formulas meet
judgment.
Our task is to analyze the organization of trading in the wake of a
technological revolution that has taken place on Wall Street, the so-called
quantitative revolution. Modern finance
presents an analytically privileged setting for examining the co-evolution of
organization and technology: as an
industry, not only has it been an “early adopter” of the tools of information
technology but also it has been a site where powerful combinations of tools
from different fields have been especially acute. With the creation of the
NASDAQ in 1971,Wall Street had an electronic market long before any other
industry. With the development of Bloomberg data terminals in 1980, traders in
investment banks were connected to each other in an all-inclusive computer
network long before other professionals. With the development of formulas for
pricing derivatives such as the Black-Scholes formula in 1973, traders gained
powerful mathematical tools. And with
the dramatic growth in computing power and the declining costs of such, traders
were able to combine these equations with powerful computational engines. This
mix of formulas, data to plug into them, computers to calculate them, and
electronic networks to connect them was explosive, leading to a mass shift to
“quantitative finance.”
To explore the socio-cognitive, socio-technical practices of
arbitrage, we conducted ethnographic field research in the Wall Street trading
room of a major international investment bank, following trades, observing interactions,
and interviewing traders and managers throughout the room. Pseudonymous International Securities is a
global bank with headquarters in Japan.
It has a large office in New York, located in a financial complex in
Lower Manhattan that includes the offices of Merrill Lynch and other major
investment banks. If a retired trader of
the bank were to visit these days the trading room of his old firm, he would
find it changed beyond recognition. To appreciate the changes, consider the
following description of a typical Wall Street trading room in the 1980s
(Wolfe, 1987, p. 58):
No
sooner did you pass the fake fireplace that you heard an ungodly roar, like the
roar of a mob... the bond trading room of Pierce & Pierce. It was a vast
space, perhaps sixty by eighty feet, but with the same eight-foot ceiling
bearing down on your head. It was an oppressive space with a ferocious glare,
writhing silhouettes… the arms and torsos of young men… moving in an agitated
manner and sweating early in the morning and shouting, which created the roar.
This boiler-room imagery is absent from
the trading room of International Securities. Entering the trading room is like
entering the lobby of a luxury hotel. Instead of a low ceiling, the observer
finds high ceilings and a huge open space occupying almost the entire 22nd
floor of a skyscraper in Lower Manhattan filled with rows of desks, computers
and traders. Instead of a roar, the observer hears a hushed buzz among the
traders in a background of numbers flickering in hundreds of flat-panel
screens. Instead of an oppressive space, the observer finds generous corridors,
elegant watercolors on the walls, and a dramatic view of Manhattan. Instead of
agitated employees, the observer finds relaxed traders in business-casual wear
standing up, walking around and even having coffee among themselves. Instead of a fake fireplace, the room is
further populated by non-human “intelligent agents,” the computer programs
executing automated trades, referred to by the traders themselves as “robots.”
Initially, we approached this research setting as providing an
almost pure case of risk and calculation.
Among these sophisticated derivatives traders, we thought, there would
be a premium on information, a single metric of performance, and an emphasis on
risk and calculability. But after months
of field work, we realized that as increasingly more information is almost
instantaneously available to nearly every market actor, the more strategic
advantage shifts from economies of information to socio-cognitive
process of interpretation. This
particular trading room makes profits (considerably higher than
industry-average profits) not by access to better or more timely information
but by producing a community of interpretation.
Similarly, the more we studied the
actual dynamics, the more we became aware that, alongside the shared metric
that a trader’s value could be measured by the profitability of his “book,”
there were contending performance criteria for measuring the value of a model
of the market. That is, your model
(perhaps even one codified into an algorithm for automated trading) might be
losing money in the short run but is, nonetheless, a valid representation
capable of performing well in the long run.
Although theories, models, and formulas are, indeed, performatives
(representations that do not mirror social reality so much as constitute it by
structuring expectations, see Callon 19xx and MacKenzie 2002), our attention
here is on the performances (Hennion 19xx) and the multiple criteria and
metrics of assessing their worth.
In analyzing the actual processes whereby traders construct these
models, “test” them, and repair them, our attention turned to the importance of
the spatial configuration of the trading room.
That is, in trying to understand the modus operandi of the
trading room, we came to see that its locus operandi was so
important. These findings, elaborated
below, are particularly salient in light of the work of Bruegger and Knorr
Cetina (2002). That work is pathbreaking
for the insight that the numbers on the screens of the electronic traders do
not represent a market that is elsewhere; instead, the market is
“appresented.” Just as the eyes of
traders in a commodities pit are glued to the gestures of other traders, so Bruegger
and Knorr Cetina found that the eyes of their currency traders are glued to the
screen – because in both cases that is where the market is. But whereas Bruegger and Knorr Cetina
discovered “global microstructures” in cyberspace, we found that trading
practices are intimately tied to the spatial deployment of traders and things in
the room.[2] If the “marketplace” is online, nonetheless,
the work of making and using market instruments is situated in a material
locale. While attentive to the screen, arbitrage traders are attentive to and
interact with other traders in the room.
As we shall see, the cognitive practices of creating models are
situated, the interpretation or sense-making of information online occurs in
situ, and even the “interactions” among robots are situated by their deployment
in the room.
[2] Note that Bruegger and Knorr
Cetina studied currency traders. As we
shall see, the practices of arbitrage traders are less conducive to
localization online.
Arbitrage trading can be seen as an economy of information and
speed. So is flying a fighter aircraft
in warfare. Without the requisite
information and the requisite speed neither trader nor pilot could do the
job. But manuevering in the uncertain
environment of markets, like manuevering in the fog of battle, requires
situated awareness. As we shall see, the
cognitive practices of the trading room entail pattern recognition
(e.g., making associations, matching data to models); but they also involve practices of
re-cognition (making unanticipated associations, breaking out of lock-in,
reconceptualizing the situation). Arbitrage trade involves codification
(sometimes quite literally as when a model is codified in software code) and
de-codification (trading can be automated, but knowing when to turn off the
robot cannot). Expressed differently,
arbitrage traders exploit knowledge, but they are continuously exploring. They engage in that most interesting kind of
search – unlike the search where you call up “information” for a phone number –
the search where you don’t know what you’re looking for but will recognize it
when you find it. This process of
locating occurs in a locale.
Place becomes a means of organizing diversity among heterogeneous
communities of interpretation; and the trading room pulses with the tensions
between risk versus uncertainty, information versus interpretation, and
calculation versus judgment.
II. A sociological approach to arbitrage
Arbitrage hinges on the
possibility to interpret stocks in multiple ways. By associating one security
to another, the trader highlights different properties of the property he is
dealing with. Hence, for example, associating the stock of Boeing with that of
Microsoft, with Northrop, with Disney or with the S&P 500 index implies
categorizing it respectively as (1) a technological stock; (2) an aviation
stock; (3) a consumer-travel stock; or (4) an American stock. Derivatives play
an important role when two firms cannot be directly compared. The point of
arbitrage is finding two stocks with symmetrical exposure so that selling is a
way to hedge buying the other. If the exposures of two stocks are not the same,
a trader can modify them by, e.g., buying an option on the price of one of the
two. Traders use derivatives such as swaps and options to slice and dice the
exposure. These are like a surgeon’s tools: scalpels, scissors, proteases,
which give the patient (the stock) the desired resemblance.
Our interest in arbitrage lies in
that it constitutes the central trading strategy used in Wall Street to move
from economics of information to socio-technical processes of interpretation.
According Jon Corzine, then-CEO of Goldman Sachs: “information,” Corzine said,
“is transparent today. No one really generates a long-term competitive edge
just because they know something that someone else doesn’t.”[3]
Yet, despite such lack of informational advantage, the returns from arbitrage
strategies are impressive. While the average return from investing in a
diversified index of stocks from 1987 to 2001 is 13.6%, returns of
arbitrage-based strategies at hedge funds is 18%.[4] In
the case of International Securities, in the last five years -- when it has
been engaged in arbitrage -- it has attained average returns of 15-20%.
Arbitrage is defined in finance textbooks as “locking in a profit by
simultaneously entering into transactions in two or more markets” (Hull, 1996,
p. 4). The archetypal example of classic arbitrage is connecting two markets
that are geographically separate: if the prices of gold in New York and London
differ by more than the transportation costs, a trader can realize a profit by
buying where gold is cheap and shipping and selling it to city where it is
expensive (Ross, Westerfeld and Jaffe, 1986). Profits are “locked in” because
once the arbitrageur buys and sells gold simultaneously, no change in the price
of gold can affect the profits: the trader has hedged his exposure. As such,
however, classic arbitrage lacks sociological as well as economic interest-- it
relates markets that are the same in every dimension except for an obvious one
such as the geographical, and as a result there are very few such obvious
opportunities for profit (Soros, 2000).
It is modern arbitrage that sparked our interest. Instead of
“locking in” a certain profit by connecting markets for the same security,
modern arbitrage connects markets that are similar along some relevant
dimension-- as opposed to the same thing. The securities involved have to be
similar enough as to hedge exposure, but different enough so that other traders
have not seen the resemblance before or cannot adequately gauge the extent of
the similarity. Unlike simpler forms of association such as brokerage,
arbitrage associates across, and benefits from the information-generating power
of markets.[5]
To grasp the complexity of arbitrage, consider the trading
strategies that it replaced, value and
momentum investing. All three of them
involve the identification of an opportunity, that is, a mismatch between the
purported value of a security and the value that the stock market designates to
it, its price. Value investing is the traditional “buy low, sell high” approach
in which investors look for opportunities by identifying companies whose
“intrinsic” value differs from its market value. Investors do so by studying
companies’ annual reports, financial results, products and comparing the
intrinsic value that emerges from those with the market price (Graham and Dodd,
1934). Value investors are essentialists; they believe that value has some
essential or intrinsic property independent from other investors, and that they
can attain a superior grasp of that value through careful perusal of the
information about a company. They proceed with the belief that the mispricing
will be eventually corrected -- that is, that the rest of investors will
eventually “catch up” with the intrinsic value and drive the price towards it,
producing a profit for those who saw if first.
On the other hand momentum strategies, also called chartism, turn
away from focussing on the company and towards the rest of investors (Malkiel,
1973). As with value investing, the goal is to find a trading opportunity.
However, investors do not pretend to know the intrinsic value of a stock.
Momentum investors focus instead on whether other market actors are bidding up
or down the value of a security and assume that the trend will continue for
some time, i.e., that there is “momentum,” a self-sustaining social process
amenable to be discovered by the study of patterns in the time series of the
stock -- its chart. Momentum investors are relational. Like fashion victims or
night-life socialites, they derive their strength from obsessively asking themselves,
“what is everyone else doing?” in terms of choice of stocks, clothes, or
cocktail bars -- and finding an better answer than most. They believe that
opinions about value do not diffuse among investors instantly but slowly
instead, producing an inertia that creates the visual patterns they look for in
the charts. In contrast with value investing, a momentum strategy can involve
buying when the price is extremely high, as long as the patters in the chart
suggest that it is getting higher.
As with value and momentum investors,
the arbitrageur also needs to find an opportunity -- an instance of
disagreement with the market valuation of a stock. This opportunity is found by
association. Instead of claiming superior abilities to process and aggregate
information (as value investors do) or better knowledge of what other investors
are doing (as momentum traders do), the arbitrageur seeks some other security
-- another stock, or group of stocks, or bond that is somehow related to it,
and values one in terms of the other by establishing a value equivalency
between the two.
Thus, whereas value trading is
essentialist and momentum trading is relational, arbitrage is associational.
Like a striking literary metaphor, a new arbitrage trade reaches out and
associates the value of a stock to some other, previously unidentified
security, or one that is tenuously related to it.
Modern arbitrage differs from classic arbitrage in that the
equivalencies are new or uncertain. This reduces the scope of competing
arbitrageurs, increasing the profitability. To see an uncertain equivalency at
work, consider now an example of the most common case of modern arbitrage,
so-called merger arbitrage. On August 13th 2001, Cendant Corporation and Cheap
Tickets Inc. announced an agreement for Cendant to acquire all of the stock of
Cheap Tickets, making them worth $17 per share if the deal went through. As the
announcement came out, the price of Cheap Tickets shares jumped from $12.00 to
$16.50. The merger offered traders an opportunity for arbitrage: if the merger
was finalized, the traders would be able to obtain $17 for stocks that cost
them only $16.50, realizing a profit of $0.50 per share. The purchase price
that Cedant was willing to pay for Cheap Tickets became an alternative valuation
mechanism. The merger created a possibility to establish equivalencies -- in
this case, between the value of Cheap Tickets with and without merger with
Cedant. Other types of arbitrage use other equivalencies. Convertible bond
arbitrage, for example, is based on legal “convertibility provisions” of the
stocks that stipulate the transformation of stocks into bonds at a fixed,
non-market rate. In index arbitrage, the value equivalency is based on the
premise that a stock will move in lock step with the rest of companies
belonging to the index.
The tenuous or uncertain strength of the equivalency reduces the
number of traders that can play a trade increasing its profitability. In the
case of arbitrage, assessing the strength of the association amounts to finding
out the probabilities of the merger taking place. In the example above,
arbitrageurs had to take the acquisition price of Cheap Tickets and discount it
to reflect the possibility that the merger might not take place at all. If it
did not, the price of Cheap Tickets would presumably fall back to the original
pre-merger price of $12.00, losing the arbitrageurs a ruinous $4.50 per share.
Those traders that do not have competencies in assessing the likelihood of a
merger taking place refrain from entering into the trade, or end up being
selected out of the market anyway by force of their losses. Conversely, traders
at International Securities possess a stock of historical information on
various mergers, and are able to use data about the identity of the lead
bankers, the degree of anti-trust risk, legal risk, or material adversity
clauses risk to assess and refine the probability of the merger taking place.
Financial instruments such as options, swaps and futures play a
crucial, surgical role in arbitrage. Arbitrage presents a challenge of
dissociation similar to the requisites for constructing a neoclassical market
identified by Callon (1998). Callon (1998) argues that creating a logic of
abstract, anonymous calculability in markets requires “formatting” a framing
process that extricates commodities from the social relationships that produced
them. But, he adds, this always finds the problem of overflow. Thus, while car
companies attempt to circumscribe their relationship with car buyers to the
moment of exchange, the car transfers with it to the buyer the company’s
know-how (who can cash it in by reverse engineering), potential tort
liabilities, etc.
Arbitrageurs need to engage in this process of dissociation because
they do not commit capital to a company but to a property of a company such as
the merger it takes part on. In the Cedant-Cheap Tickets example above,
arbitrageurs who buy Cheap Tickets to bet on the merger will be exposed not
only to the possibility that the merger might be cancelled, but also to other
changes in the price of Cheap Tickets stemming from its other properties such
as, e.g., changes in the price of small capitalization stocks, of American
companies, or of travel companies.
Arbitrageurs reject exposure to a whole company: they associate
markets by disassociating the companies they long and short from those
properties not involved in the equivalency principle. To eliminate from their
exposure all the non-merger properties of Cheap Tickets, arbitrageurs “sell
short” the acquirer, Cedant Corporation -- that is, they bet that its price
will fall. This second leg of the trade is the hedge. Traders hedge by adopting
opposite exposures between acquirer and target. They simultaneously bet that
the economy will improve and that it will worsen, that low-capitalization
companies will rise in value and will fall, or that air travel will be more
profitable and also less profitable. Whatever happens in all these respects,
the gains from one leg of the trade will offset the losses from the other,
leaving traders “market-neutral,” “industry-neutral,” and neutral with respect
to all the properties of the stocks but one -- the merger.
In the example above, hedging seems like a relatively trivial task.
But hedging is a centerpiece of what arbitrage is also about: associating
markets by isolating the stocks from all properties that are not the
equivalency principle. Consider a more complex -- and more lucrative --
operation that the trade presented above. Imagine two firms have announced
their intention to merge, but the acquirer belongs to an index, say the Dow
Jones biotech index, while the target does not. As a consequence, the stock
price of the acquirer stock moves up and down with wide movements of the DJ
biotech index. This would not necessarily be a bad thing for the arbitrageurs,
who, as we know, bet on the merger by simultaneously shorting the acquirer and
longing the target. DJ biotech membership could benefit their short position on
the acquirer if that index fell. But the biotech index could also rise and in
any case, that is not the point; the point is that arbitrageurs tailor their
exposure to what matters to them -- in this case, their bet that the merger
will take place. To avoid exposure to
those properties of the stock that are not the deal, traders hedge their
exposure to them.
Arbitrageurs slice and dice their exposure with the help of
derivatives such as options, futures or swaps, or other financial instruments
such as indexes. In the simple example of Cedant and Cheap Tickets, the close
similarity between the two merging stocks -- both were American,
low-capitalization, airline stocks -- made hedging almost trivial. In the
second case, for example, they offset their exposure to the DJ biotech index by
engineering a separate long exposure to the index.
Arbitrage, then, is not about reducing or eliminating the trader’s
exposure -- it is not a trading strategy for cowards. Rather, arbitrage is
about tailoring the traders’ exposure to their position vis-à-vis the market,
biting what they can chew, betting on what they know best and avoiding risking
their money on what they don’t. Traders expose themselves profusely, but
precisely because their exposure is custom-tailored to the relevant deal. Their
sharp focus and specialized instruments gives them a clearer picture on the
deals they examine than the rest of the market. Thus, the more the traders
hedge, the more solidly they can position themselves.
This points out to the irony of arbitrage. Arbitrage entails
establishing multiple associations using several instruments and drawing from
various formulas... but in the end it can boil down to a bet on a single
event. Such events either happen or not
such as, will these two companies merge or not?
Their simplicity is increased because they carry with them a date in
time that will resolve the uncertainties. We normally think of events as
something different from companies, but arbitrageurs manage to equate the two
by stripping the companies’ stocks of all properties that are not the event
itself. For all their sophisticated use
of derivatives, arbitrageurs can end up engaging in bets about the outcome of a
discrete event. Given the
associational task of the arbitrageur it should be clear how existing network
theory is not sufficient to account for it fully. Network theory is ideally suited for the
study of ties between people and the identification of structural holes (Burt,
1992). But, as noted above, arbitrage requires financial instruments to slice
and dice exposure, formulas, a degree in finance, and monitors, computers,
robots, cables and other material tools to find new associations and execute
them. A depiction of who talks to whom would miss the crucial information of
what talks to what.
This last point is underscored by a
comparison of market making and arbitrage. Market making, the process of
matching buyers and sellers of securities, can be thought of a particular case
of brokerage -- that is, the identification and exploitation of structural
holes in a network of social ties (Burt, 1993). Before electronic markets,
market-making was responsible for the cacophony of traditional Wall Street
trading rooms as salesmen and traders matched deals internally in frenetic
speed by shouting to each other. With the arrival of electronic markets, as
Bruegger and Knorr Cetina (1999, p. 4) documented, the rooms became much more
silent and the action moved to the screens.
But both rooms and screens are doing the same function -- supporting
conversations between people.
Arbitrageurs do not make markets -- they link markets. In doing so,
they use prices and the wealth of information contained in them (Hayek, 1944)
to profit from mispricings. This clever use of prices can be observed in the
Cedant-Cheap Tickets. Merger arbitrageurs take the stock price of both
companies, calculate the difference between them to come up with the “spread”
between the value of the two securities being linked. Then they read the spread
backward to gauge the probability that other market actors assign to the deal.
For example, a price difference of $ 0.50 means total gain of $0.50. If the
possible loss if $4.50 ($16.50 minus $12.00), this means that a in order to
break even the probability of the merger would have to be at least 90%. To decide whether to play the deal,
arbitrageurs then only need to determine whether they are more or less
confident than 90% that the merger will go through. This spread changes every
day, and arbitrageurs follow its ups and downs to better determine what their
positioning should be. By reading
probabilities back from spreads, the arbitrageurs tap into the collective
knowledge that two markets have aggregated (the markets for Cheap Tickets and
Cedant) and use it in their favor. Arbitrageurs make value investors work for
them.
How are these unprecedented and uncertain associations
accomplished? What resources,
organization, technology, would a bank need to establish them? Brokerage, filling structural holes, is about
maintaining and expanding a social network of buyers and sellers. Modern
arbitrage is also about networks, but networks that include not just people but
also ideas, artifacts, material objects and securities -- all of them thrown in
for generative recombination. In the next section we turn to a trading rooms to
examine how it works in practice.
III. The trading room as
an ecology of evaluative principles
To examine the organization of arbitrage we conducted field research
in the trading room of pseudonymous International Securities, a global
investment bank with offices on Wall Street. Over the past two years, we have
been regularly observing interactions in the trading room, interviewing traders
and managers on several desks, taking bus rides with clerical workers, and
sharing lunch at the trading desk with traders.
Our traders face a conundrum presented to Wall Street by the
quantitative revolution: the more pervasive and immediate information becomes,
the more ideas diffuse as traders become connected; and the more technology
facilitates the execution of derivatives trading, the lower the profits.
Bruegger and Knorr Cetina (1999) have argued that the rise of electronic
markets has brought the marketplace to the trader’s screen, prompting the
traders to shift from a “face-to-face” to “face-to-screen/ world,” and bringing
about a “diminishing relevance of the physical setting” (p. 23). Following a detailed examination of a
market-making trading room of currency traders, the authors noted how the most
relevant interaction takes place by means of email-like electronic conversations
across trading rooms spanning organizations, continents, and time-zones.
Participants, although physically located in the trading room “appear
viscerally plugged into the screen reality of the global sphere” (p. 15).
Heath et al.’s (1995) account of traditional trading rooms (i.e.,
pre-electronic markets), by contrast, argues that co-location makes optimal use
of traders’ limited attention and bounded rationality at the local and global
level. At the local level, grouping
traders by desk lets them collaborate with each other with minimal intrusion
into each other’s activity. For example,
the authors document how co-location allowed a trader to “time, with precision,
an utterance which engenders collaboration, so that it coincides with a
colleague… swallowing a mouthful of lunch” (1995, p. 6). At the global level of the room, the authors
noted that clustering different desks together allows traders to monitor
markets beyond the one in which they are trading. This is facilitated by the
institutionalized practice of the “outcry,” the habit of “when necessary,
shouting names and calling numbers without either waiting for, or apparently,
expecting a reply” (1995, p. 10). These
are, according to Heath et al., an economical way to reach everyone in the trading
room, and to deliver information in an un-intrusive manner as the recipient is
not asked to respond. However, because
Heath et al remain within the economics of information (for them, the point of
a trading room is to reduce the costs of transmitting information), their
account would fail to explain the persistence of trading rooms in the age of
electronic markets with its conditions of more comprehensive, nearly
frictionless exchange of information.
We approach the tension between electronic markets and physical
rooms as a particular case of what can perhaps be referred as the “Castells
axiom.” How, Castells (2000) asks, has
the role of space changed in a network society structured by the Internet and
information technology? Castells
distinguishes between spaces of place, that is, locales “whose form, function
and meaning are self-contained within the boundaries of physical contiguity,”
and spaces of flow, which he defines as “the organization of purposeful,
repetitive, programmable sequences of exchange between physically disjointed
positions.” Contrast, for example, a street and an airport. According to
Castells, as information technology creates spaces for repetitive,
preprogrammed, machine-like interactions (spaces of flows), the original, spontaneous,
unexpected interactions found in physical spaces (spaces of place) can provide
an ever-increasing source of competitive advantage. Thus, for example, as the
technology for remote surgery develops and surgeons gain the possibility to
intervene from a remote location, they are clustering more and more in two or
three neighborhoods of Manhattan. From
the perspective of arbitrage as association, trading rooms can be seen as the
“space of place” where novel associations emerge.
One exemplary passage from our fieldnotes finds Bob, the manager of
the trading room, restating the
Castells’ paradox:
It
is hard to say what percentage of time people spend on the phone vs. talking to
others on the room. But I can tell you the more electronic the market goes, the
more time people spend communicating with others inside the room.
The response of International Securities to the challenge posed by
information abundance has been to step up the level of complexity of its
arbitrage trades. To be sure, traders must
have access to the most timely and complete array of information; but this is
not enough. In addition to being a nexus
of data flows the trading room also constitutes a community of interpretation. In this section, we examine the room first as
a kind of “primate society” of 160 traders and anciliary personnel, exploring
some features of their interaction;
second, we examine how these traders are grouped into desks, exploring
the specialized functions by which each recognizes patterns through distinctive
financial instruments; next, we examine
the trading room as an ensemble of desks, exploring how this ecology of
evaluative principles facilitates practices of re-cognition; and finally, we
examine the room as an assemblage of material tools, exploring how the
socio-cognitive and the socio-technical are intertwined.
Facilitating sociability
The architecture of the trading room at International Securities is
designed to create an atmosphere conducive to association. Consider, for
example, Gladwell’s description of the layout and distribution of “a classic
big-city office tower,”
The center part of every floor is given over to the guts of
the building: elevators, bathrooms, electrical and plumbing systems. Around the
core are cubicles and interior offices, for support staff and lower management.
And around the edges of the floor, against the windows, are rows of offices for
senior staff… The executive in one corner office will seldom bump into any
other executive in a corner office. Indeed, stringing the exterior offices out
along the windows guarantees that there will be very few people within the
critical sixty-foot radius of those offices (Gladwell, 2000, p. 64).
The contrast between the classic corporate offices and International
Securities is striking. Turn right from the reception area on the 22nd
floor of the skyscraper where the bank is located, and the trading room opens
up in all its magnificence: almost an entire floor filled with multi-colored Bloomberg screens,
moving images in the CNBC monitors and relaxed traders clad in elegant business
casual wear. No cubicles. No partitions to block the view. There is even a
strict “low-monitor” policy enforced by Bob, the manager of the room, since, as
he notes, “people are insecure on the floor and build themselves a nest,” that
is, they stack their Bloomberg monitors two- or three-high, to produce a sense
of privacy. Bob prevents that. “We try,” he says, “ to keep the PCs at a low
level so that they can see the rest of the room.”
Moreover, the composition of the room promotes association among
disparate communities of practice: the
room not only accommodates traders and their assistants, but also salesmen,
analysts, operation officers, and computer programmers. Flouting an
industry-wide trend of relegating these latter employees to a back-office,
International Securities has kept programmers and operations officers in its
money-making core. They not only stay in the trading room but are given desks
as large as the traders’, and their area of the room has the same privileged
feel as the rest. The objective, Bob states, is to prevent differences in
professional status from undermining communication among these groups. If
placed in a different building, says Bob, “they might as well be in a different
planet.”
The size of the trading room is designed to maximize communication.
At 160 people, it has a small size by
current Wall Street standards. But its
dimension was purposely chosen by Bob to create a space where communication
could be facilitated. Bob says,
“managers, they’ll tell you, ‘communication, communication,’ but you wonder.”
For example, he pointed us to a trading room of another international bank
located in Connecticut:
It is the size of three aircraft carriers.
And the reason for it is that it is a source of pride to the manager. It is
difficult to see how can traders communicate shouting at each other across two
aircraft carriers. At [name of bank], what you’ll find is chaos that looks
grand.
Instead, at the trading room of International Securities,
The key is [to avoid] social awkwardness. Two
traders are talking to each other. A third needs a piece of information. He has
to interrupt. ‘Can I interrupt? Can I interrupt?’ The key there is the social
cost of the interruption [emphasis by Bob].
Promoting sociability among traders is not an easy task. In the age
of mathematical finance, arbitrageurs, unlike Tom Wolfe’s image of traders as
Master of the Universe, are all very qualified, very competitive, and fiercely
individualistic. They are intellectually
over-confident, but socially inept:
A trader is like an engineer type. Difficult
when they think they’re right. Abrasive. And not very social. Not socially
adept. I can easily find you ten traders in the room who would be miserable at
a cocktail party.
This results in territoriality in the trading room. For example,
back in the early 80s, Bob recalls, in the bank where he began his career,
There were areas of the trading floor I would
never venture onto in years. People I never, absolutely never, talked to. There
was no reason why I should go there, since we traded completely different
things. Being there felt strange. There were these cold looks.
International Securities has avoided this territoriality in the
trading room by moving traders around. “I rotate people as much as I can.” Bob
says, “because sitting near each other is the best rule of thumb to predict
that they will talk to each other.”
However, Bob is careful not to displace them excessively. He describes
his approach as “not really shifting, more like drifting,” as with those
puzzles that have only one hole and in which only one piece can be moved at a
time.
Once two traders have been sitting together,
even if they don’t like each other they’ll cooperate, like roommates. Everyone
gets moved every six months on average. But not everyone at a time.
This emphasis on communication underscores
that the cognitive tasks of the arbitrage trader are not those of some isolated
contemplative, pondering mathematical equations and connected only to a
screen-world. Cognition at International
Securities is a distributed cognition.
The formulas of new trading patterns are formulated in association with
other traders. Truly innovative ideas,
as one senior trader observed, are slowly developed through successions of
discreet one-to-one conversations within the desks,
First you talk to others. You tell someone
else, ‘I’ve got this great idea,’ and if he tells you ‘I read it yesterday in
Barron’s,’ you say, ‘Oh, I did too.’
An idea is given form by trying it out, testing it on others,
talking about it with the “math guys,” who, significantly are not kept apart
(as in some other trading rooms), and discussing its technical intricacies with
the programmers (also immediately present). Because they have been stirred up by the
subtle churning of the room, they can test the ideas on those with whom they
were once “like roommates” but who might now be sitting in different parts of
the room. Appropriately, the end of this
process of formulation (and the beginning of the next stage of material
instrumentation, see below) is known as
a “victory lap” – a movement
around the room in and through which the idea was generated.
And where is Bob, the trading room manager? He sits in the middle of the room despite the
fact that he has a very well-appointed office in one corner, complete with
designer furniture, a small conference table and a home cinema-sized Bloomberg
screen to watch the markets that can be controlled from a wireless mouse and
keyboard. But he prefers to sit in a
regular trader’s desk in the middle of the room.
I have that office over there – you just saw it. But I like this place better
[referring to his desk]. Here, I am more connected. No one would come to tell
me stories if they had to come into my office. Also, here I get a feel for how
the market is doing. I have to know this, because the atmosphere definitely
influences the way these guys trade.
“The phone and on-line communication are inefficient,” said Bob. “It
takes longer for people to tell each other what they want. Also, you miss body
language.” As he emphasizes,
Body language and facial expressions are
really important. You’re not conscious
of body language and so it’s another channel of communication, and it’s one
that’s not deliberate. So it’s a good source for what’s happening. I don’t try to get too conscious of how I’m
reading body language and facial expressions.
I just let it work its way to where its useful.
Bob’s observations (and those of many other
traders with whom we spoke) highlight that cognition in the trading room is not
simply distributed. It is also a situated cognition. A trader needs tools – the financial
instruments of derivatives and the material instruments to execute a
trade. But in addition to these
calculative instruments, the trader also needs a “sense of the market.” Knowing how to use the tools combines with
knowing how to read the situation. This
situated awareness is provided, in large part, by the room. As the passage quoted above about
deliberately not being too conscious of how one reads body language nicely
illustrates (points emphasized in different terms by other traders), the
activities of the trading room involve pre-calculative components.
Pattern Recognition at the
Desk
We now move from the individuals that compose the trading
room as a “baboon society” to the teams that compose the trading room as
a more complex organization of diversity. This organization of diversity begins
by demarcating specialized functions.
The basic organizational unit, “team,” has a specific equipment,
“desk.” The term “desk” not only denotes
the actual piece of furniture where traders sit, but also the actual team of
traders – as in “Tim from the equity loan desk.” Such identification of the animate with the
inanimate is due to the fact that a team is never scattered across different
desks. In this localization, the
different traders in the room are divided into teams according to the financial
instrument they use to create equivalencies in arbitrage: the merger arbitrage
team, options arbitrage team, futures arbitrage team, convertible bond
arbitrage team, etc. (see fig. 1 below).
The desk is an intensely social place.
The extreme proximity of the work setting enables traders to talk to
each other without lifting their eyes from the screen or interrupting their
work. Lunch is at the desk, even if the
sandwich comes from a high-end specialty deli.
Jokes are at the desk, a never-ending undercurrent of camaraderie that
resurfaces as soon as the market gives a respite.
Each desk has developed its own way to look at the market, based on
the principle of equivalence that it to uses to calculate value and the
financial instrument that it enacts in its particular style of arbitrage
trade. For example, traders at the
merger arbitrage desk value companies that are being acquired in terms of the
price of the acquiring firm and specialize in asking themselves, “how solid is
company X’s commitment to merge?” Analytical and calculating, for them the
companies in the S&P 500 index are little more than a set of potential
acquirers and acquisition targets. In contrast, traders at the convertible bond
arbitrage desk look at stocks as bonds, and specialize in information about
stocks that would typically interest bondholders such as their liquidity and
likelihood of default. Traders at the index arbitrage desk value market indexes
according to the weighted average of the prices of the companies that
constitute them, and specialize in executing high-volume high-speed trades that
exchange indexes for their baskets and in finding companies that upset the
accuracy of indexes when they split into two or disappear. The traders at the customer sales desk, meanwhile,
take and propose orders to customers outside the confines of the room. Although
not specialized in a distinct financial instrument, this most sociable bunch in
the room provide a window on the anxiety level of their customers and thus of
the market at large by the sound of their voices on the phone and the bangs of
the headsets against the desks in frustration.
Each of these desks organizes a specific form of pattern
recognition. For example, merger
arbitrage traders, keen on finding out the degree of commitment of two merging
companies, look for patterns of progressive approximation in stock prices. As
noted in the previous section, they probe into commitment to a merger by
plotting the “spread” (difference in price) between acquiring and target
company over time. As with marriages between humans, mergers between companies
are scattered with regular rituals of engagement intended to persuade others of
the seriousness of their intent. As time passes, arbitrage traders look for a
pattern gradual decay in the spread as corporate bride and groom come together.
Such joint focus on visual and economic
patterns creates, at each desk, a distinctive community of practice within a
principle of equivalence with its own tacit knowledge and its own denunciations
of the others. Traders on a desk develop a common sense of purpose, a real need
to know what each other knows, a highly specialized language, and idiosyncratic
ways of signaling to each other. This
sense of joint membership translates into denunciations of others. A senior arbitrage trader, for example,
confided in us that, to him, statistical arbitrage is “like playing video
games. If you figure out what the other
guy’s program is you can destroy him.
That’s why we don’t program trades,” he added, referring to his own
desk. Tod, one of the statistical arbitrage traders, told us that the more he
looked at his data (as opposed to letting his robot trade), the more biased he
becomes.
These functional teams are thus markedly
different from the cross-functional teams that make up “projects” in fields
such as film, new media, construction, etc., where heterogeneous evaluative
principles combine in a single team (Girard and Stark, 2001). Within a new
media team, different members may find the value of a web site in metrics as
varied as its speed, ease of use, profit-making abilities, or beauty and
elegance. Within an arbitrage desk, in contrast, all traders find value along
the same equivalency principal. Nonetheless, the complex trades that are
characteristic of our trading room seldom involve a single desk/team in
isolation from others.
Connecting for cutting, co-location for disassociating
The desk, in our view, is a unit organized around a dominant
evaluative principle and its arrayed financial instruments (devices for
measuring, testing, probing, cutting).
This principle is its coin; if you like, its specie. But the trading room is composed of multiple
species. It is an ecology of evaluative
principles. Complex trades take
advantage of the interaction among these species. To be able to commit to what counts, to be
true to your principle of evaluation, each desk must take into account the
principles and tools of other desks.
Recall that shaping a trade involves disassociating some properties in
order to give salience to the ones to which your desk is attached. Recall
also that cutting and slicing (dis-associating) requires making
associations (identifying the relevant categories along which exposure will be
limited). It follows that shaping a
trade involves associations among desks.
Co-location, the proximity of desks, facilitates the connections to do
the cutting.
While in most textbook examples of arbitrage the
equivalence-creating property is easy to isolate, in practice this isolation is
difficult to accomplish completely. For example, merger arbitrage traders lend
and borrow stock as if they were going to be able to reverse the operation at
any moment of time. However, “if the company is small and not traded often, it
may be difficult to borrow,” and traders may find themselves unable to hedge,
notes Carl. Similarly, some companies have “A-” and “K-class” stocks, depending
on the voting rights they carry. “Arbs hedge with the As, because they are
easier to borrow,” according to Carl, “but actually get the Ks,” creating the subsequent
challenge to get transform the K into A stocks. In other cases, “one of the
parties may have a convert provision [that is, its bonds can be converted into
stocks if there is a merger] to protect the bondholder. Then, the question is how does that affect
the deal.” Finally, traders will
occasionally resort to financial engineering to create synthetic arbitrage
products. As a result traders end up exposed to properties of the companies
different from that which constitute the equivalency.
In the case of arbitrage – linking,
as opposed to creating markets – traders
face a similar problem. Hedging
their exposure to overall market movements may leave them with exposure to
something else through convertibility provisions. Interestingly, traders reintroduce
the overflow exposure in their calculations in the same way as they achieve
association: through co-location.
Physical proximity in the room allows traders to survey the financial
instruments around them and assess what additional variables they should take
into account in their calculations. The
stock loan desk, according to Carl, “help[s] us by telling us how difficult it
is to borrow a certain stock.” The index
arbitrage desk helps with regard to A and K stocks; since it profits from the
fact that A stocks are included in indexes while K stocks are not -- and
therefore sometimes the two classes of stocks move out of lockstep. The convertible bond arbitrage desk helps
merger arbitrage traders clarify the ways in which a convertibility provision
can affect the deal. “The market in
converts is not organized,” says Carl. There is no single screen representation
of the prices of the converts. For this reason, says Carl,
So
we don’t know how the prices are fluctuating, but it would be useful to know it
because the price movements in converts impacts mergers. Being near the
converts desk gives us useful information.
In any case, according to Carl, “even if you don’t learn anything,
you learn there’s nothing major to know about.” It’s important because, as he
says, “what matters is having a degree of confidence.” Or, as one senior trader
puts it, a hedge entails multiple components, and as time passes “the component
pieces of the hedge are moving.”
In the previous section we noted that the teams/desks of a trading
room are very different from the teams that make up “projects” in other
industries. But if a given desk is organized around a relatively homogeneous
principle of evaluation, a given trade is not.
Because it involves hedging exposure across different properties along
different principles of evaluation, any given trade can involve heterogeneous
principles and heterogeneous actors across desks. If a desk involves simple teamwork, a
(complex) trade involves something more like collaboration. This collaboration
might be as primitive as an un-directed expletive from the stock loan desk
which, overheard, is read as a signal by the merger arbitrage desk that there
might be problems with a given deal. Or
it can be as formalized as a meeting (extraordinarily rare at International
Securities) that brings together actors from the different desks. And much in between. A trade is a project.
Re-cognition in an ecology of equivalences
By putting close the teams that trade in the different financial
instruments involved in a deal, the bank is able to associate different markets
into a single trade. As a senior trader
observed,
While
the routine work is done within teams, most of the value we add comes from the
exchange of information between teams.
This is necessary in events that are unique and non-routine,
transactions that cross markets, and when information is time-sensitive.
How to use the creativity, vitality and serendipity stemming from
the trading room to make new interpretations?
By interpretation we mean processes of categorization, as when traders
answer the question, “what is this a case of?” but also processes of
re-categorization such as making a case for. Both work by association -- of people to
people, but also of people to things, things to things, things to ideas, etc.
The following instance illustrates a case of re-cognition across two
desks, the customer trading and the special situations desk. Jay L., senior
trader at the customer trading desk, takes orders from clients and suggests new
trades to them. Across the table sits Josh P., head of the special situations
desk, close enough to overhear Jay’s conversations with his clients at any
given point in time. On one occasion, a
client phoned Jay L. with an order to buy company X, sell company Y and short
Z. Jay recollects:
I looked at it and it didn’t make any
sense. Why does he do that? I talked to
Josh [gestures over there at the nearby desk]. ’That guy’s crazy,’ Josh
said. That was the tipoff.
This
fleeting exchange was the tipoff that set Jay to work on the customer’s trade
to find an opportunity for a genuinely good trade. Following a quick
give-and-take between the two of them about what was wrong and what was right
with the trade, Jay says,
We
structured what we thought was a better trade. Then I phoned my client. ‘This
is the trade you should be doing. And this is why.’ I operate in this way. And
he might say, ‘You’re an idiot, that’s never going to happen.’ Then I’ll say,
‘Great. Do you want to take the other side?’”
In doing so, Jay and Josh use the
client’s trades as the starting point for their own brainstorming. It is also a
useful way for the traders to put fresh thinking into their strategies –
even if it’s wrong.
Re-cognition can be more dramatic as in the following example in
which the proximity of desks lead to a new kind of trade, a hybrid that
resulted from recombination across equivalences. The case involved an announced merger between
two financial firms. On January 25th,
2001, Investors Group announced its intention to acquire MacKenzie Financial.
The announcement of the merger immediately set off a rush of deals from merger
arbitrage desks in trading rooms all over Wall Street. Close contact with the
merger arbitrage desk and the equity loan desk allowed the special situations
desk at International Securities to conceive a new arbitrage trade, which they
called “election trade,” that recombines in an innovative way two previously
existing strategies, merger arbitrage and equity loan strategies.
The facts of the merger were as follows: following established
practice, the acquiring company, Investors Group offered the stockholders of
the target company to buy their shares and offered them an alternative of cash
or stock in Investors Group. The offer favored the cash option. Despite this,
Josh and his traders reasoned that a few investors would never be able to take
the cash. For example, board members and upper management of the target company
are paid stocks in order to have an incentive, so it would look wrong if they
sold them -- they had “symbolic” or “hierarchical” rationalities, as opposed to
a purely financial profit-maximizing approach.
The presence of such symbolic investors in effect created two
markets of sorts -- one made up of profit maximizers, and one made up of
captive investors in MacKenzie. As with any other situation of disconnected
markets with diverging local valuations, this could open up an opportunity for
arbitrage. But, how to connect them? The genesis of the idea lied precisely in
co-location in the room. The “special situations” desk (its name denotes a
purposive attempt to cut through the existing categories of financial
instruments in the room) is located in between stock loan and merger arbitrage. Stock loan makes its margin by a lending and
borrowing operation. Just as the merger arbitrage desk uses mergers to create
equivalencies, the possibility to lend and borrow stocks for election day
became the equivalency principle that tied together two disparate options, cash
and stock, faced by investors in MacKenzie. According to Josh P., head of the
desk,
[The idea was generated by] looking at
the existing business out there and looking at them [the trades] in a new way.
Are there different ways of looking at merger arb?
The traders used physical proximity to the other desks and the
strong acquaintance that it creates to envision opportunities for arbitrage:
We
imagined ourselves sitting in the stock loan desk, and then in the merger
arbitrage desk. We asked, is there a way to arbitrage the two choices, to put
one choice in terms of another?
The traders found one. What if International Securities were to
borrow from “symbolic investors” their shares in the target company at the
market price, exchange them on election in the more favorable terms (i.e.,
taking the cash instead), then return the stock to symbolic investors? That way
these investors would be able to bridge the divide that separated them from the
most favorable option on election day. The special situations desk was linking
these investors to the cash by engineering a financial bridge between the two,
not unlike the classic operation of shipping gold from London to New York.
Physical proximity and social similarity helped further in assessing
the risk-rewards of the trade. As with merger arbitrage, the possibilities for
a new equivalency imagined by Josh and his traders were not cast in stone, but
tenuous and uncertain (and that was what made them so profitable -- the fact
that no one had done them before). Depending on how many investors chose cash
over stocks, International Securities might find itself with large losses. The
reason was as follows: IG, the acquiring company, intended to devote a limited
amount of cash to the operation; if, as planned, on election day most investors
elected cash, the acquiring company would prorate its available cash (i.e.,
distribute it equally) and use some of its shares to pay even those
stockholders that elected “cash.” This was the preferred scenario for
International Securities, for then it could use these shares to return the
stock it owed to those symbolic investors it borrowed from. But if, in an
alternative scenario, most investors were to opt for stock, then the acquirer
would not run out of cash on election day, investors who elected cash such as
International Securities would not get stocks back, and Josh and his traders
would find themselves without stock in IG to return that which they had
borrowed. They would then be forced to buy the stock of IG on the market,
perhaps at prices high enough to turn the profits from the deal into
losses.
The profitability of the trade, then, hinged on a question: would
investors elect cash over stock? Uncertainty about what investors would do on
election day posed a problem for the traders. Finding out about their plans,
note, is not a trivial challenge in a capitalistic economy that is based on the
notion of dispersed stock ownership across various people in different places
and approaches to investment. The traders overcame this infinite-search problem
with the help of co-location and technology. As a first step, Josh pulled up a
list with the names of the twenty majority holders in the target company in the
screen of their Bloomberg terminal. Then he went down the list with his team.
According to his own account,
What
we did is, we [would] meet together and try to determine what they’re going to
do. Are they rational, in the sense that they maximize the money the get?
But
knowing the identity of the owner of stock was only half of the equation. What
would this company or that individual do? Josh and his crowd resorted to social
similarity.
See...
the major owner is Fidelity, with 13%. They will take cash, since they have a
fiduciary obligation to maximize the returns to their shareholders. [Some others are] hedge funds doing merger
arbitrage... so I shouted across the question to the merger arbitrage team here
at International Securities... who were doing the same and work right across
me.
To summarize, co-location helped develop an opportunity for
arbitrage that took advantage of the existence of multiple rationalities among
investors. Traders were then able to gauge the prevailing rationality in the
market from their close knowledge of the rationality of one of the desks. This
process of going from the local to the global the trading room was transformed
into a tool for probing the uncertainties left in electronic markets.
IV. The trading room as an assemblage of market
instruments
To this point, we have emphasized the role of abstract market
instruments. But there is another, equally important set of instruments that
includes cables, phone lines, computers, modems, electrical power, algorithms,
TVs and buildings -- the material tools of trading. The formers are considered
worthy of study in The Journal of Finance, while the latter supposedly
belong to the province of handymen, contractors or electricians. But
overlooking this last group would mean missing another important way in which
the room makes up for more than the sum of the parts. We refer to both
financial instruments and the material tools as market instruments.
Paradoxically, while the former have a reputation for being more interesting,
we would argue that the latter are the least understood of the two. While
derivatives have a clear purposive nature, the tangible tools have a diffuse,
complex use, and their effect on socio-technical networks is multifaceted.
We know from the previous section that traders use financial
instruments to associate and dissociate and reshape the trader’s exposure from
a stock into the desired form. Similarly, material market instruments associate
data, models and ideas. Instruments, according to Latour (1986), are actively
inscribed to translate other people’s interests into those of the instrument
designer. The key weight that is customarily found in traditional hotels is
inscribed to persuade the guests to leave their keys in reception as they go on
the street. In the context of science,
Latour (1987) defines an instrument as an inscription device that shapes the
reader’s view, that is “any setup that provides a visual display of any sort,”
as long as it “is used as the final layer in a scientific text.” Examples of
instruments include radio telescopes, Geiger counters, or petri dishes.
Perhaps the most visible of all instrument are the Bloomberg
terminals that traders use, three extra-wide high-contrast flat panel screens
that make a manifest difference between a trading room now and twenty years
ago. The screens are not mere transporters of data. They are heavily inscribed.
They select, modify and present data in ways that shape what the trader sees.
As we show below, some of the data is arranged to be a “magnifying glass,”
other data constitutes trading “baskets,” and yet other data is tied among
itself in the form of “active links.”
Steve P. at the customer trading desk devotes a lot of care to the
conscious inscription of his screens. He
arrives to the trading room at 8:30, but the markets do not open until 9:30. An
important part of the preparation for the day is preparing the screen set-up.
One by one, Steve opens each of his windows and places it in its customary
place. Steve’s screens are divided into several areas that reflect the many
types of data he faces. On the extreme left, a series of windows give him
general information on the market: a Bloomberg window, charts for the Dow
Industrials and NASDAQ. The indexes
provide a general view. A more
personalized perspective is provided by another window further to the right,
that Steve calls his “magnifying glass.” These are 60 additional stocks that he
considers representative of what happens in different sectors such as chips,
oil, or broadband. They give more information than just the general direction
of the market indexes. Visually, the numbers of this window momentarily
increase in size when an order is received, resembling a pulsating meter of
live market activity.
Steve complements the magnifying glass with the “footprints” of
competitors. These carry information about the stocks that he is trading each
day, contained in about six individual windows that vary in appearance
depending on whether the stock trades in the NYSE or the NASDAQ. In addition to
magnifying sectors and tracing his rivals’ footprints, Steve’s screens double
as workbench for his operations. These are displayed on several windows that
show the trades that Steve undertakes, called “trading baskets.” An additional
workbench window shows pending work. This is contained in a spreadsheet in
which Steve introduces entries with “active links” to stock prices, so that
they are automatically updated. That allows him to program in the cells next to
the links the conditions that the clients give them (e.g., if the order is “set
the spread at 80,”) and an additional cell that changes color depending on
whether the conditions are met for him trade or not (cyan equals good, dark
green not good). The window is a traffic light for operations.
Instead of reducing the nature and importance of social interaction
in the room, the screens give a new rationale for it. As a trader said, “thanks
to technology, you don’t need to hear from your neighbor on price action. All
that information, we can find in Bloomberg”.
But there is a sense in which a screen reproduces the rigid categorical
structure created by bureaus and bureau-based organizations, bureaucracies.
Screens can be “bureau-tronic.” According to the trader referred above,
“Bloomberg shows the prices of normal stocks. But sometimes, normal stocks
morph into new ones.” Whenever stocks morph, that is, change their properties
and defy the categorization system, electronic information is not enough. Furthermore, Steve comments, “no two screens
are the same.” “We all have different information, so I sometimes check with
them,” he adds, referring to other traders. How often does this happen? “All
the time.”
Just as Latour (1987) defined a
laboratory as “a place that gathers one or several instruments together,”
trading rooms can be understood as places that gather market instruments
together. Seen in this light, the move from traditional to modern finance can
be contemplated as an enlargement in the number of instruments in the room,
from one to several. The best scientific laboratories maximize
cross-pollination across instruments. For example, the Radar Lab of MIT in the
1940s made breakthroughs by merging the epistemologies of physicists and
engineers (Galison, 1996). Similarly, the best trading rooms encourage
different desks to relate to each other.
Monitoring the price mechanism
Another example of inscripted instrument are the “robots,” computer
programs used by statistical arbitrage traders that automate the process of
buying and selling stocks. Robots are inscribed in the sense that they execute
only one possible trading strategy -- the one they were programmed to perform.
For example, the mean reversion robot will only buy and sell stocks based on
whether their prices are close or distant from their historic average price.
The earnings robot will only buy and sell based on companies’ earnings. In this
sense, robots contain in them a complex set of assumptions about the market and
undertake an active selection of the available data that is consistent with it.
Robots are models and representations.
The room plays a crucial part in the robots from the moment of their
inception, the process of codifying, literally turning tacit knowledge into
algorithms and code to make it usable by robots. This takes place at the whiteboard, in
meetings that include for example, a trader specialized in indexes, and a
programmer from the computer department.
At the whiteboard, an idea for arbitrage trade mutates in medium from a
trader’s utterances, to graphs, to models, to equations, and finally to
computer code. The robot is quite
literally codified knowledge.
Once codified into a computer program, the robot goes to traders
specialized in implementing computer programs such as the statistical arbitrage
desk. End of the story? Not really. Piloting a robot requires inputs from a kind
of emergent traffic control – cues and signals from other parts of the
room.
Consider the case of Tom, a trader at
the statistical arbitrage desk. Instead
of trading directly, Tom uses, programs and maintains a robot. Automated
trading poses the same challenge as driving a car at a high speed: any mistake
can lead to disaster. “I have,” Tom says, “a coin that comes heads 55% of the
time.” With margins as low as 0.05, the only route to high returns is trading a
very high volume or, as Tom says of the coin, “the point is to flip it a lot.”
As with formula 1 car or speed-boat racing, traders need a very good
instrumentation Indeed, they have navigation instruments as complex as an
airplane cockpit -- a price
mechanism. Yet, as it turns out, these
are not enough. The price mechanism has to be monitored, and calibrated, and
for that purpose Tom obtains crucial cues from the social interaction in the
desks around him.
To illustrate how bad things can go when instrumentation fails, Tom
recounts an instance in 1997. Price
movements were large, the index was going up very quickly, and other banks were
getting delayed information because of problems in the Reuters server. The
result of delay, in a rising market, was that these competitors consistently
saw the index below its real level. But while the index data were delayed, they
were getting timely price feed on the futures index, so what seemed to them
cheap was in fact expensive. Tom and others at International Securities, on the
other hand, were getting timely information on both (see fig. 2). Tom explained
to us,
While they were buying, we were
selling... the traders here were writing tickets until their fingers were
bleeding. We made $2 million in an hour, until they realized what was happening.
More generally, the episode not only illustrates the challenges of
execution but also the dangers of representation. Given the advanced automation
of the market, what traders see – the
numbers on their screen – is a representation of the market. If they
take this to be reality but it turns out not to be accurate, there can be huge
problems. The key to good execution, then, is an accurate representation. How
does Tom know that the robot’s representation of the market is accurate?
The most immediate solution to the challenge of execution is more
technology. Tom’s robot provides him with as many dials as a cockpit in an
airplane. He trades with three screens in front of him. Two of them correspond
to two powerful Unix workstations and the third one is a Bloomberg terminal.
The first of the two Unix terminals has real-time information about his trades.
Across the top of one of them there is a slash sign that rotates and moves from
side to side. It is a “pulse meter” to gauge the speed with which information
on prices is arriving to him, or “price feed”. The character stops moving when
prices stop arriving. It is very important to realize when this happens,
because then the “price robot” gets confused: it thinks that prices don’t
change and imagines false opportunities, while in reality prices are moving but
not arriving to his terminal.
On the right hand-side corner of his second Unix station Tom has
five squares; each of them is a kind of traffic light that indicates how
quickly the orders are getting through the servers of the specialists or
electronic communication networks. If they are green, everything is fine. If
they are yellow, the network is congested and deals will get through slowly. If
they are red, their servers are clogged. The clocks in the Unix workstations
are synchronized everyday to the Atomic Clock. In addition to a large display
of an analog clock in his computer, Tom has two “CPU-meters” which measure how
busy is the database that deals with the order flow of International Securities.
When it is busy for long periods of time, orders take longer to execute. Thus, to monitor prices in the market,
traders must monitor the price mechanism – literally, they must monitor the
machines that bring and make prices.
However, technology is not the only answer to the problem of
execution, for the dials that measure the accuracy of the technology are a
representation themselves. Technology, in other words, answers one question,
“is the robot getting the data?” but raises another one, “is the robot right in
what it says?” We call this infinite-regress problem the “calibration” problem.
The calibration problem became notable following the nuclear
accident at Chernobyl. An unfortunate circumstance that reportedly made the
damage much worse is that radiation was so high that the dials of the Geiger
counters of the control room of the Soviet nuclear power station did not
register any abnormal level of radiation even at the peak of the escape. The
dials had been calibrated for registering nuances, so the sharp increase went
unnoticed. Technology, then, helps in execution of automated tasks, but needs
calibration.
How to solve the calibration problem? Tom solves it by the use of
space and sociability in the trading room. He sits in between the merger arbitrage
desk and the systems desk. There he hears what the system people tell others
through their microphones, getting a sense of how well the systems are running.
According to Tom,
When
you hear screams of agony around you, it indicates that perhaps it is not a
good time to trade. If I hear more screams, maybe I should not use the system
even if it’s green.
Similarly, price feed in stocks and futures has to come at the same
speed. Hence sitting near the futures arbitrage desk is helpful in answering the
question of whether there is something anomalous in the data feed. Tom’s
solution to the calibration problem suggests that when technology and the
existing representations come under doubt, traders resort to the social
relations that spawned them. Thus, the calibration problems is relevant to our
discussion so far because it takes from the screen and its economics of
information back to the room
The final challenge that Tom can encounter is that the assumptions
underlying his arbitrage robot may not apply, or, as he puts it, that “the
world may change in a way that is not envisioned in the system.” In these
cases, he will take a stock out of the robot. How does Tom know when to do so?
Listening to the media is only partly helpful:
Rumors in CNBC--should I listen to them?
I need to find out when a stock is going to go through exceptional
circumstances, takeover or restructuring. In principle, it would be useful to
anticipate what is going to happen. But there are ten rumors for every
takeover. So I just turn off the volume [As he said this, he turned off the
volume of CNBC and ignored it.]
According to Tom, there is an additional problem with rumors.
You
never know how they’re going to be interpreted. Take Motorola, the big news
today. They have done worse than expected. But the stock went up. Why? Because
all tech stocks went up, apparently because some trader at Salomon decided that
tech’s so low that we should buy anyway.
But free interpretation is dangerous, according to Tom.
I
find that the more that I can articulate simple rules for myself, the more I
can be consistent in my own interpretation of events. Otherwise, if I start to
interpret events freely, I’m using a 50% coin... to go against a 55% one. The
algorithm in computer trades gives you the necessary discipline to stick to
strategies.
For that reason, Tom limits himself to reading the second column in
the cover of the Wall Street Journal and avoids media that is closer to the
rumors. He finds it very useful to listen to activity in the merger arbitrage
desk.
While promoting association through proximity, the trading room also
uses distance to preserve the requiste measure of variety amongst traders -- a
form of organized diversity. Instead of the work of cleansing differences that
produces “one right way” to calculate, the trading room actively promotes
diversity in itself. Thus, for example, in the case of statistical arbitrage
traders, the notion of the desk has been abandoned. For example, of the four
statistical arbitrage robots, a trader said,
We
don’t encourage the four traders in statistical arb to talk to each other. They
sit apart in the room. The reason is we want some diversity.” The diversity of
these stat arb units is ensured by making sure that they have different P &
L [profit and loss] patterns, and different risk profile.
The stat arbs are organized according to
the push-pull mechanism. What pushes them apart is the the need to avoid the
danger that they become similar to each other.
They might evolve closer because robots are partly “alive” -- they
change as they are re-tooled and re-fitted to changes in the market.[6] They are kept separated to reduce the
possibility that their evolution will converge (thereby resulting in a loss of
diversity in the room). On the other
hand, robots are pushed closer (or at least kept inside the room) because a
given stat arb unit cannot be too far from the other types of arbitrage
desks – proximity to which provides the
cues about when to turn off the robots.
REFERENCES
Abolafia, M. 1997. Making
Markets. Cambridge, MA: Harvard University Press
Bloomberg, M., 1997. Bloomberg by Bloomberg. New York: Wiley.
Bruegger, U and K. Knorr Cetina. 2000. Global Microstructures: The
Interactional Order of Financial Markets.
Paper presented at the Heterarchies Seminar, Center on Organizational
Innovation, Columbia University.
Callon, M. 1998. The
Laws of the markets. Oxford:
Blackwell Publishers.
Castells, M. 1996. The rise of
the network society. Cambridge, MA: Blackwell Publishers,
Clippinger, John. 1999.
“Tags: The Power of Labels in Shaping Markets and Organizations.” Pp. 67-88
in The Biology of Business:
Decoding the Natural Laws of Enterprise edited by John Clippinger. San Francisco: Jossey-Bass.
Fama, E. 1970. Efficient Capital Markets: A Review of Theory and
Empirical Work. Journal of Finance,
49, 283-306.
Galison, Peter. 1997. Image and Logic: A material Culture of
Microphysics, University of Chicago Press.
Garud, Raghu and Peter Karnoe.
2001. Garud, R. and P. Karnoe. 2001. Path Creation as a Process of
Mindful Deviation, in Garud and Karnoe, eds., Path Dependence and Creation. Mahwah, NJ: Lawrence Erlbaum
Associates.
Girard, M. and D. Stark. forthcoming. Distributing Intelligence and
Organizing Diversity in New Media Projects.
Environment and Planning A.
Graham, B. and D. Dodd, 1934. Security
analysis; Principles and Techniques. New York: McGraw-Hill.
Granovetter, M. 1985. Economic action and social structure: The
problem of embeddedness, American Journal
of Sociology 19(3): 481-510.
Heath, Christian, Marina Jirotka, Paul Luff
and Jon Hindmarsh (1995). “Unpacking Collaboration: the Interactional
Organization of Trading in a City Dealing Room.” Journal of Computer Supported Cooperative Work, vol. 3, no. 1.
Hutchins, E. and Klausen, T.
(1991) Distributed cognition in the cockpit. In Engestrm, Y. and Middleton, D.
Cognition and communication at work. Cambridge University Press.
Hutchins, E. 1995. Cognition in the wild. Cambridge, MA: MIT Press.
Kahneman, Daniel, and Amos Tversky. 1979. Prospect Theory: An
Analysis of Decision under Risk. Econometrica,
vol. 46, no. 2 (March):171-185
Knight, F. 1921. Risk,
uncertainty and profit. Boston: Houghton Mifflin Company.
Malkiel, B. 1973. A random walk down Wall Street. New York: Norton
Lane, David, and Robert Maxfield. 1996. “Strategy under complexity:
Fostering generative relationships.” Long
Range Planning vol. 29(2):215-31.
Latour, Bruno. 1986. “Powers of Association.” Pp. 264-280 in John
Law, ed., Power, Action, and Belief: A
New Sociology of Knowledge.
Routledge.
-- 1991.
“Technology is Society Made Durable.”
Pp.103-131 in John Law, ed., A Sociology of Monsters: Essays on Power, Technology, and Domination. Routledge.
Lepinay, V. and F. Rousseau, 2000. Les trolls sont-ils incompétents?
Enquête sur les financiers amateurs. Politix,
vol. 13, n°52, pp. 121-154. p. 73-97.
MacKenzie, D 2000. Long-Term Capital Management and the Sociology of
Finance. The London Review of Books,
13 April 2000.
Markowitz, H. 1991 Portfolio selection : efficient diversification
of investment / Harry M. Markowitz. Cambridge, MA.: B. Blackwell.
Muniesa F. 2000, Un robot walrasien. Cotation électronique et
justesse de la découverte des prix, Politix,
vol. 13, n°52, pp. 121-154.
Soros G., 2000. Open Society : Reforming Global Capitalism.
New York: Public Affairs.
Suchman, L. 1987. Plans and
situated actions : the problem of human-machine communication. New York :
Cambridge University Press, 1987.
Podolny, Joel M. and Toby E. Stuart.
1995. “A role-based ecology of
technological change.” AJS 100(5):1224-1260.
Shiller, R.J. (1998). Human behavior and the efficiency of the
financial system. NBER Working Paper Series. Downloaded September 1, 2001.
Stark, D. 2001a. “Ambiguous Assets for Uncertain Environments:
Heterarchy in Postsocialist Firms,” In The
Twenty-First-Century Firm: Changing Economic Organization in International
Perspective, edited by Paul DiMaggio.
Princeton, NJ: Princeton University Press, 2001, pp. 69-104.
--
2001b. “Have Theory, Will
Travel.” Review of Actor Network
Theory and Beyond, edited by John Law and John Hassard. Contemporary Sociology, vol. 30, no. 1, January 2001, pp. 96-7.
Thaler, R.
1993. Advances in Behavioral Finance, editor. Russell Sage Foundation.
Wyser-Pratte, G. 1982. Risk Arbitrage II. Monograph Series in Finance and Economics, New York: Graduate
School of Business Administration, New York University.
 |
|||||||
 |
|||||||
 |
|||||||
|
|||||||
Real-time futures index
$
Real-time S&P 500
S&P 500 as seen by late traders
delay
Time
elapsed in the day
Fig.
2: Note distance between futures and spot-market indexes: a slight time delay
can make it look as if there is an opportunity for arbitrage.
[1] Although American sociologists have not yet incorporated the
insight that network analysis should include artifacts as well as persons,
other social scientists in this country have been working with similar
concepts. Hutchins (1994), for example,
argues that cognition is distributed across a network of persons and
instruments, and documents this dramatically and painstakingly in the case of a
US Navy cruiser that is navigated into port after a power system failure.
Suchman’s (1987) pathbreaking work on human-machine interaction similarly
resonates with the work of Callon and Latour and provides the basis for further
studies on distributed design.
[2] Note that Bruegger and Knorr
Cetina studied currency traders. As we
shall see, the practices of arbitrage traders are less conducive to
localization online.
[3] Joshua Cooper Ramo, “Welcome to the Wired World,” Time Magazine (February 3rd,
1997).
[4]
Anonymous, “Hennessee Releases 8th Annual Hennessee Hedge Fund Manager
Survey(R) Findings; 2001 Marks Record Capital Inflows and Record Number of
Survey Respondents,” Yahoo Finance, Thursday March 7, 8:32 am.
[5] For example, in the case of market making, association and
disassociation predate the market, and calculation follows disentanglement. In
the case of arbitrage, local calculation feeds into global association.
[6] For example, according to one trader, if one of the stat arb
traders left the bank, “it is unlikely that we would retain the profitability
in the medium term... they [the robots] would loose fit with the market, and
who would adjust them?”
沒有留言:
張貼留言