The machine in the market: Computers and the infrastructure of price at the New York Stock Exchange,1965–1975

Introduction

On May 31, 1966, electronic stock quotes began to emanate from the floor of the New York Stock Exchange (NYSE). The Market Data System (MDS), an amalgam of computers and electronic input devices, had taken control of the stock market’s most famous medium, the tickertape. Hidden away from the bustle of the trading floor, on the third floor of the Wall Street exchange stood four computers. Together, they ran a network of optical card readers, printers and ticker-tape machines that enmeshed the exchange like a web. Employees called ‘reporters’, functionaries once responsible for recording trades and passing slips of paper to clerks for conveyance from the trading floor to ticker-tape typists, now placed specially formatted paper cards into electronic readers located at the Exchange’s trading posts. Trade data entered in this way hit the tape in a matter of seconds, circumventing a several-minutes-long process that had involved byzantine mazes of pneumatic tubing and teams of human hands. ¹ The speed of information in the new electronic system was described as both ‘real-time’ and ‘up-to-the-market’ but these categories were largely aspirational. ² ‘Tape lateness’ – when information lagged behind the action of the market – persisted through the 1960s and remained one among many points of contention between two groups active in early efforts to computerize market data: the Exchange itself and the US stock market regulation agency, the Securities and Exchange Commission (SEC). ³

The MDS reflected the technical aspirations of both groups. To Exchange executives, the system represented the possibility of complete managerial control over the human processes of information mediation. To regulators, the computerization of market data was a mechanism for digital disclosure: a tool for the surveillance of traders and the erasure of informational asymmetries that separated market insiders on the floors of exchanges and ticker-reading investors. Both groups pressed for the expansion of the machine’s responsibilities once it was installed, and by the end of the decade the MDS was a critical information hub for the Exchange. Its multitasking computers ran the tickertape, checked the work of human workers on the fly, and delivered price data to ‘mechanized’ brokerage-house computers in Lower Manhattan via electronic relay.

Market data technology from this early period of financial computerization embodied a complex political and legal arrangement between regulators and the NYSE. Although the rules of securities trading are largely determined by self-regulating organizations like exchanges themselves or the National Association of Securities Dealers (NASD, later FINRA), the flow and content of information about securities is subject to SEC oversight. As computerization blurred the line between market and media, the SEC sought to expand its oversight powers by pressing for new informational capabilities in stock exchange data-processing systems. Under the chairmanship of Manuel Cohen (1964–69) and with the guidance of the computer engineer Frank Verzuh, the SEC actively shaped computerization at the NYSE.

Although electronic data systems entered the market in part as tools of active regulation, they challenged spatial and institutional categories on which securities regulation and market rules depended. As early as 1966, for instance, NYSE lawyers and member firms wondered if orders delivered via an external broker computer would be considered ‘on the floor’ or ‘in the exchange’, or if an exchange itself could be (in) a computer. ⁴ Their most consequential effect on regulation occurred in the early 1970s. In the wake of an industry crisis (1968–69) and with the ascendance of computerized stock markets in Cincinnati, San Francisco and in the over-the-counter sector through the National Association of Broker Dealers Automated Quotation System (NASDAQ), economic and social problems in the market were increasingly answered with technical solutions. The perception that computers faithfully materialized rules-based governance within the market undercut efforts by regulators to shape technology in the public interest. Although a 1975 securities law promoted a ‘National Market System’ of data sharing, consolidated price information and harmonized regulation, the SEC took a backseat to powerful industry institutions in its technical actualization. The sociotechnical order of electronic market data infrastructure in the United States is a legacy of this period.

The technologies of financial markets have commanded significant attention from sociologists of finance. A number of researchers have documented the social environment of financial markets in their electronic form. ⁵ Far from enabling frictionless, fair, and geography-neutralizing markets, computerized finance and electronic telecommunications reproduce material politics of access and social asymmetries of information. By opening black boxes of market technology, these studies have demonstrated the persistent importance of space, physicality and materiality within seemingly virtual modern markets. In his recent study of high-frequency traders, MacKenzie (2017a) has offered the concept of ‘material political economy’ to describe some of these persistent effects. Microscale arrangements of technology resulting from ‘political-economic struggles’ (who gets data a microsecond earlier, who is physically located closer to the market’s servers, who owns the fiber optic cable between New Jersey and Chicago) structure macroscale advantages (MacKenzie, 2017a: 190). Traders well-positioned within the infrastructure of market data can occasionally predict short-term market behavior and generate risk-free profit.

This article examines the forces shaping early technologies for managing and disseminating electronic market data at the most influential stock market in the United States, the NYSE. It begins by describing a real-time electronic data system called the MDS and its simultaneous positioning by NYSE executives and federal regulators as both a check on human work and an agent of disclosure. A partial alignment of regulatory and managerial views on automation was supported by a common anxiety about the human intermediaries of financial market data: the teams of clerks, typists and reporters who recorded and managed information about trades.

Market automation did not come to all human tasks at the same time. Regulators pressed the NYSE to use computers as a form of digital disinfectant for a wider range of human problems facing the Exchange: not only to avoid clerical errors and manual slowness, but also to facilitate the surveillance of destabilizing trading patterns by speculators and floor traders. By emphasizing the historical relation between the emergence of market computerization and a regulatory logic of disclosure, this article offers context to the study of recent technologies for financial surveillance by Williams (2009). But in tracing why initial regulator-backed efforts to surveil traders were unsuccessful at the NYSE, it shows how asymmetries in the application and uptake of computer systems reflect the distribution of power within market institutions, especially between professional members of the Exchange and clerical staff. This analysis affirms that information technologies not only undergird social and epistemic orders of the market, but that infrastructures are also artifacts of historical struggles and indices of paths cut short. As such, it amplifies and builds from recent efforts to document the shaping of financial knowledge by powerful institutional users (Poon, 2009) and information vendors (Poon, 2007) as well as practitioner cultures (MacKenzie and Spears, 2014).

The second part of the article examines the expanding responsibilities of the computer. Market operators, drawing on the MDS’s timesharing ability, repositioned it as a general-purpose information hub for a variety of technical and human processes at the Exchange. Media and market became increasingly entwined as technical systems amplified the Exchange’s role as a switch for the filtering and movement of information, rather than as a venue for trading. Two projects are examined: the development of an electronic ‘specialist book’ and the computerization of order routing for odd-lot trades. Bradford and Miranti (2014) have provided a more comprehensive account of the odd-lot automation program at the NYSE and its effects on the brokerage business. Odd-lot routing appears here as part of a discussion of the changing legal ontologies and media imaginaries of stock exchanges as they began to transition into their modern electronic form. This analysis of market-media hybridity parallels Pardo-Guerra’s (2010) study of automation at the London Stock Exchange, and is motivated by a similar effort to bring to bear the materials and media of market practice on the history of the conceptualization of financial markets as information systems (Mirowski, 2002, 2007)

The third part describes the consequences of the marriage of machine and market in debates about the architecture of national financial infrastructure in the early 1970s. As exchanges came to be seen as things materializable in technical systems, reforms to a securities industry in crisis were captured by technological solutions. A perfectible market seemed increasingly achievable through a perfectible machine. Nonetheless, different views about electronic stock market infrastructure proliferated. Should there be a single venue for the entire nation, virtualized in a computer? Or should trading venues simply be linked together in a weaker network federalism; one that would largely preserve the concentration of information and liquidity in traditional market centers? These debates and their resolution were structured by changing institutional dynamics: on the one hand, the waning of direct involvement by the SEC in the details of market technology, and on the other, the increasing power of the NYSE as a developer and standard-setter in market data technology. The transformation in the early 1970s of a range of market problems including venue fragmentation into challenges with technical solutions supported this dynamic. The machine itself, rather than the human institutions that instrumentalized it, began to embody market governance.

This account of market technology argues that the arrangement of financial data infrastructure in the US was influenced by institutions whose ability to shape technology in practice and meaning oscillated historically. Alignments among these institutions was partial and mediated by common interests, including the common suspicion of human work in information processing. Although federal regulators were critical actors in early efforts to computerize market data systems, this research, like that of Bradford and Miranti (2014), affirms the importance of the NYSE as an influential agent of technological development and standardization. However, unlike the case of the odd-lot program, where automation’s effects on the securities industry appear ‘unanticipated’ (Bradford and Miranti, 2014: 850), this article frames NYSE market data systems in the context of contests over the sociotechnical order of market regulation. In the early 1970s, regulators, bankers, and economists openly discussed the centralization of stock trading venues through computer networks run by the government itself or by self-regulating organizations. Viewed in this light, efforts by the NYSE’s technology company SIAC to generate and commercialize market data automation services, and to proliferate alternative technical standards, appear like strategic choices to preserve institutional power.

Human reporters, electronic disclosures: 1965–1966

Before the Market Data System went on-line in 1965, the process for reporting a trade at the New York Stock Exchange involved a tight web of human practices and analog technologies of mediation. After a trade was made, an employee of the NYSE called a ‘reporter’ wrote the details on a paper slip. Recording the name or ticker symbol, the quantity of shares and price, the reporter then handed this slip to a ‘tubeman’ who sat inside the trading post sharing space with papers, time-stamps and a set of pneumatic tubes, through which the slip was routed to the ticker room. There, it was copied onto the ticker tape queue by a team of typists and editors. Clerks used the extensive tube communication system to send details of the trade into the separate but equally important clearing processes, where trades were actualized, accounts credited and certificates delivered signifying ownership. There was a normal lag of up to several minutes from transaction to tape. ⁶

Efforts to speed up this process in the 1960s were part of broad managerial reforms of the increasingly crowded Exchange floor, its technologies and its regimes of human work. The Floor Automation Committee of the NYSE took aim at everything: the square footage of desk space and the size of chairs in high traffic areas, the trial of experimental technologies like pocket-sized radio-signaling devices, and, critically, the labor efficiency of floor staff. ⁷ The ticker itself had undergone a revolution in speed in 1965. A new system developed by the Teletype Corporation could print between 500–900 characters per minute. This upgrade from the 300 characters per minute rhythm in place since 1930 required the development of a faster ‘big board’ for projecting quotes inside the Exchange. The board showed quotes on ‘luminescent plastic discs’ bearing numbers that could be rapidly actuated by electrically signaled air jets. (NYSE, 1966b: 5) The Exchange touted the new ticker as marking a new epoch in securities trading, with President Keith Funston noting that it operated so fast as to be ‘almost at the limit of human readability’. All the same, he stressed, the ‘Exchange’s planners recognize that the day may come when even this remarkable instrument … may require some form of augmentation’ (NYSE, 1965).

Clerical and technological reforms were in part mandated by a changing stock market. The 1965 NYSE annual report offers ample evidence of the coevolution of a new market and a new technical regime. The cover featured an image of an integrated circuit described in the table of contents as an ‘unusual photographic collage of computer parts [that] symbolizes the growing importance of automation at the Exchange’ (NYSE, 1965: 1). In his annual note, President Funston celebrated the success of the MDS, which was undergoing floor tests, and introduced two new NYSE divisions that emerged the same year: Electronic Systems and Institutional Investor Services (NYSE, 1965: 3). The former dealt with computers, the latter with a new breed of market actors: institutions like mutual funds, pensions and insurance companies. These institutional investors increasingly flocked to the stock market to generate reliable returns for their large capital holdings; while their business was welcome, they traded in novel scales and strategies that put pressure on the mechanics of trading. Large block trades – in excess of 10,000 units – proliferated, and so too did basket trades – simultaneous trades on a range of securities to achieve risk-hedging strategies. ⁸

Nonetheless, the NYSE’s push for market automation was not motivated only by novel economic conditions and new technical possibilities. It was also the product of managerial efforts to mechanize and disempower human labor. Changes in trading strategy and in the potential speed of electronic market data put new pressures on the humans that made the stock market function in the 1960s. Managerial anxieties about the reliability of human intermediaries in the production and mediation of market data offer an important context for the uptake of computerized data systems at the NYSE.

Fallible humans, market technology, and trading speed converged during the Crucible Case of 1966. In January, the NYSE detected a significant fraud on the ticker tape. Volume figures in stock for the Crucible Steel Company had been inflated by as many as 141,000 shares. The number of shares reported on the ticker as having changed hands was far higher than what had actually traded. Beyond the embarrassment of inaccuracy, such a fraud had market implications: inflating perceived activity in the stock could drive investors to the market and move the price. (A related strategy for manipulating the ticker feed to generate a false sense of trade activity is called ‘painting the tape’.) Two investigations followed, one by a former Assistant US District Attorney, and a second by the SEC. ⁹ The investigations found that the reporter assigned to Crucible had not tried to manipulate the market, but had nonetheless ‘deliberately’ overwritten the volume of trades reported to the ticker. He had done so for social and financial benefit: out of ‘a desire for prestige in handling the most active’ stock and with ‘some consideration for the volume bonus’ given to busy reporters. ¹⁰

Beyond the failing of an individual tempted by social prestige and monetary reward, investigators diagnosed something rotten in the human-technical system of market data mediation itself. This crucible revealed general anxieties about the judgment of human mediators in the analog system of market data dissemination of the time. The investigators centered on an NYSE-sanctioned practice of ‘digit deletion’ in the reporting of trades to the tape. Occasionally, trading activity on the floor would become so busy that the ticker tape lagged behind the market by many minutes. To catch up, the Exchange would instruct ticker operators to cut volume numbers from outgoing market data. Temporarily omitting information about the volume of shares in a trade reported on the ticker saved precious characters from the feed, allowing more quotes to pass onto the tape faster. The Crucible investigation concluded that digit deletion practices ‘may have led to the feeling [that] volume accuracy is not regarded as essential’. The instantiation of an information-processing procedure in human workers had a side effect of changing the way workers actually viewed data. This prompted the Chairman of the NYSE to meet with reporters to stress the ‘importance of accuracy’, all the while reiterating the ‘vital’ need to maintain the practice to ‘avoid tape lateness’. ¹¹

It was in this context of strict rules for market data handling and a perceived crisis in human judgment that computerization hit the NYSE floor. Between the Crucible event (January, 1966) and an internal memo reflecting on the recommendations of the investigations (February, 1967) the MDS had come online as a key tool not only for ticker tape control, but also for employee surveillance. As an internal summary of the Crucible affair underscored in February 1966, not only would the MDS systematize decisions about digit deletion, avoiding human judgment, it would automatically check for suspicious data entries by workers. ¹²

Human fallibility and labor speed in the mediation of data had long formed the background of computerization efforts at the NYSE. The initial impetus for MDS had come from a series of contracted studies by Bell, the New York Telephone Corporation, and IBM from 1951–52 that also recommended the automation of two other labor-intensive information services (NYSE, 1966b: 5). One was the NYSE’s Quotation Bureau, an organization founded in 1928 where operators provided on-demand last sale prices to member firms by phone. The other was the work of attaching market price from the floor to small trades called odd-lots (NYSE, 1966b: 11). The MDS’s responsibilities were partly forged from these managerial initiatives. In its initial deployment, MDS ran the market data process from report to wire, answered phone calls and played back electronically produced vocalizations of stock quotes, and kept a chronological record of transactions. ¹³

Human reporters stayed in the loop of transaction reporting when the MDS took over, but now faced a mechanical disciplinarian that checked their work and demanded perfection in transcription: the coding, marking, and legibility of cards submitted to the system. The first press release introducing the MDS emphasized not only its improvements to quotation speed but also its disciplining function, describing the system as a ‘school marm’ and a ‘teaching machine’. Reporters were disciplined by reports on a ‘supervisory’ printer that would spit out an error note ‘as plainly as a teacher’s red pencil scores an examination’. Submitting the wrong security would return in all caps: ‘IDENTIFICATION’. The machine chastised reporters for writing outside the highly structured card’s boxes and diagnosed suspicious trades. If ‘the difference in price between latest trade and the previous one is too large’, a message was returned: ‘SPREAD’. The MDS embodied a model of how markets normally operated and used this in order to check the work of Exchange workers. ¹⁴ For the early months of operation at Post 17, for instance, the MDS was entirely a teaching machine: Reporters dutifully submitted trades to the computer in order to practice on the system, but the computer was not connected to the ticker. ¹⁵

Although the language of ‘supervision’ was pervasive in the discourse surrounding the MDS’s early applications, precisely who was not surveilled reveals much about the contours and contests behind early market computerization. Surveillance was initially directed at NYSE employees like tubemen and reporters, unionized clerical workers, and less so at professional traders and brokers whose firms owned portions of the Exchange as members. ¹⁶ For the NYSE, the pursuit of efficient electronic markets was tied up with a mistrust of the human work of intermediation. But it was not financial intermediation (the work of brokers on behalf of clients) that concerned them, but informational mediation: the work of human clerks, typists, and reporters in the faithful representation and management of market data.

The decision to surveil workers but not traders was indeed a choice. Moreover, it was one made by the NYSE despite lobbying by the Securities and Exchange Commission, who pressed the Exchange to take advantage of computerization to monitor traders in real-time. Although the NYSE did pursue ‘real-time stock watching’ in the MDS – the monitoring of suspicious trading activity as it happened on the floor – they claimed that a system for actually individuating traders or brokerage clients – to know who was behind a suspicious trade in real-time – was technically infeasible. ¹⁷ But how was feasibility constructed and performed to the SEC in this early market system?

The NYSE noted internally that the SEC’s technological proposals involved ‘a great deal of fancy’. ¹⁸ Yet they tried to build the systems all the same. The NYSE made at least one effort to develop a system for monitoring individual traders from 1966 to early 1967. Each trader was given a unique, punched ‘calling card’ that the reporter would use to associate the transaction with an individual party and permit real-time electronic ‘identification’ and surveillance of market actors. The tests seem to have been a failure. In one memo to the NYSE President, a market automation manager described the problem as stemming from ‘adapting [the MDS] to do something it was not designed to do’. ¹⁹ An IBM psychologist and human factors consultant put the blame on the human reporters who remained the critical entry point of market data into the electronic world. Additional data to individuate traders could be handled by reporters only at the cost of accuracy and speed. ²⁰

Traders had something to do with the failure of the calling card system as well. The 1966 tests put some responsibility on the trader to self-identify to the reporter. But they had no real incentive to do so. ²¹ Nearly one third failed to self-identify during one test. ²² Floor traders, who bought and sold stocks for their own accounts, and brokers, who traded on behalf of clients external to the floor, were members of the Exchange, not employees of it. As a result, they were relatively independent of the kind of control the NYSE exerted on reporters. During the test of the calling card system, even those traders with the very worst performance were only disciplined with a talking-to. ²³

The trader surveillance program shows how responsibility for failure in a technical system is constructed by the priorities and social politics of a community. It was unfeasible because union labor (represented by the reporters) was inefficient and ineffective, or because changing the machine to a new goal was too expensive, or because professional work could not be effectively disciplined. That the SEC partly trusted the judgment of the NYSE reflects some aspects of the broader politics of markets, technology, and regulation in the United States during this period.

Created by the 1934 Securities Exchange Act, the SEC had focused not on intervention into the rules of the market (which are determined largely by self-regulating organizations like exchanges themselves or by industry consortia like NASD or its successor FINRA), but on the imposition of disclosure upon the institutions facilitating securities trading and the companies traded. By dictating that financial information about traded companies move beyond the walls of the boardroom or the exchange floor and by dictating that exchanges publicize rule infractions, the markets could become more transparent, less ripe for insider trading at the public’s expense and more efficiently self-regulating. Justice Louis Brandeis’s (1914: 92) famous argument for transparency (‘Sunlight is said to be the best of disinfectants; electric light the most efficient policeman’) explicitly guided the SEC’s regulatory mission for much of the twentieth century (Levy, 2011: 1–7).

But by 1960 the SEC was to many observers a ‘toothless tiger and a training ground for Wall Street Securities lawyers’ (Phillips and Zecher, 1981: 13). A stock manipulation scandal on the American Stock Exchange (AMEX) prompted President-Elect John F. Kennedy to request a report on the work of executive branch regulatory agencies throughout the government, including the SEC, which was followed by a highly influential 1963 SEC ‘special study’ (Phillips and Zecher, 1981: 14). This, in turn, led to legislation (the Securities Act Amendments of 1964) and a revitalized SEC empowered with resources and newly expanded disclosure requirements to enforce (Bayot, 2004; Seligman, 2003: Ch. 10).

Disclosure was a regulatory motif that meant more than just registering market members, requiring reports and mandating the publicity of certain market events (see Seligman, 2003: 39–40 on the disclosure philosophy). It guided the SEC’s 1960s attack on certain structural asymmetries in the arrangement of trading that gave some market participants informational advantages over public investors. ‘Floor trading’ came under the thumb of SEC Chairman William Cary (1961–64) and his successor, Manuel Cohen. Floor traders inevitably came across information that was unavailable to those outside the exchanges. They acted more quickly and faced lower costs in the execution of trades than the wider public: The fundamental structure of floor trading was, in Cary’s words, a ‘license to steal’ (Seligman, 2003: 308; for an overview of the critique of floor trading see Eiteman, 1966; Phillips and Zecher, 1981: 327). The SEC would come to see emerging market computerization projects as an opportunity to curtail the power of floor traders.

The 1963 special study offers a window into both the SEC’s imagined computer future for the securities industry and an associated regulatory effort to shape it. ‘Developments in electronic data processing’, it read, ‘have foreshadowed the emergence of new and perhaps revolutionary quotation systems’ (SEC, 1963: 6). But despite ‘vital significance … they should not be allowed to emerge without due regard to the welfare of the market and to the public interest’. Technology, in other words, should be checked by regulators mindful of its effects on market conditions. One direct implication of this view was to impose registration and disclosure requirements upon electronic quotation providers. This approach to market technology reflected the long-term political arrangement of the SEC and the self-regulating organizations as well as the legal-regulatory philosophy of ‘cooperative regulation’ (Seligman, 2003: 112). This philosophy, dating to the first SEC Chairman Joseph P. Kennedy, stressed bilateral consultation and coordination between regulators and securities industry institutions (Cohen, 1964).

One critical site of cooperation in the 1960s was in coordination of automation efforts for the operation of stock markets. At this early stage of financial computerization, the SEC saw malleable infrastructure – computer systems in the making – as sites for more active regulatory intervention that nonetheless left the boundary between Wall Street institutions and government regulators intact. For instance, the SEC used the legal structure of self-regulation to induce the NYSE to undertake electronic surveillance. In a 1963 analysis, the Chief Economist of the SEC’s Office for Program Planning argued that the adaptability of computer devices essentially expanded the responsibilities of self-regulating entities like exchanges to enforce compliance with rules among their members ‘as far as within its powers’. ‘Given the capabilities of the computer equipment’, he noted, the NYSE would not be meeting this responsibility if it did not use the MDS to perform trader surveillance. ²⁴

Building on this expanded view of self-regulatory responsibility, regulators at the Securities and Exchange Commission actively shaped the MDS as a new vector for regulatory oversight of market practices during a multi-year program of meetings and communications, structured by the advice of SEC economists and a hired consultant, Frank Verzuh, an MIT electrical engineering researcher and computer pioneer who had worked on the Whirlwind computer program of the 1940s. ²⁵ These meetings brought together executives at the vice-president level of the NYSE and a number of SEC staff from the Trading and Markets division. Through meetings and correspondence, Cohen’s SEC pushed for computer functions that would enable a more transparent market, where brokers were surveilled and real-time data curtailed the insider power of floor trading by spiriting up-to-market data away from the exchanges to the general public faster. ²⁶ These meetings discussed both matters of policy and law, and technical details of the NYSE computing machines, including how the NYSE planned to transition to the IBM System/360 computer. When, during a May 17, 1966 meeting, the NYSE reported that the broker identification program had proven unsuccessful, the SEC alleged that this was because computing memory had been over-allocated to ticker tape processes within the MDS. The SEC had studied the technical and organizational hurdles to trader surveillance independently since 1963. ²⁷ In a 1967 a speech honoring the President of the NYSE on the occasion of his retirement, Cohen lambasted the Exchange, calling the recently installed ticker technology ‘already obsolete’, and describing an environment stymied by ‘hesitant and limited use of computers and recent communications developments’. ²⁸

The NYSE was not the only target of SEC pressure to automate. Meetings and communications with the American and Midwest stock exchanges reveal a similar, if less highly charged, effort to direct computerization in the securities industry, including in the creation of a national certificate clearing house for the management of the paper material of stock trading. ²⁹

In addition to real-time, identifiable trader surveillance, Cohen’s SEC pressed for computer-to-computer ‘tie-ins’ between the MDS and computers outside the exchange. This was aimed at guaranteeing investor access to timely and complete market data. In practice, the completeness of market data at the NYSE was often in question. Digit and volume deletion rules, including those implicated in the Crucible affair, were part of the problem. Also troubling to the SEC were exchange practices of omitting trades at ‘zero tick’ (sequential trades for the same stock, temporally separate, made at the same price) and continued tape lateness, even after the MDS came online. ³⁰

In 1967, the SEC lobbied the NYSE to allow computer data vendors to obtain complete electronic market data directly from the MDS rather than via the tape. ³¹ The NYSE pushed back, and in the same year reiterated the importance of practices to remove data from the ticker tape to avoid tape lateness. The exchange planned to eliminate zero tick quotes in order to make the 900-ticker capable of keeping up with a 15 million share trading day. The NYSE claimed that they had found no demand for computer-delivered quotations and argued that such a service would bifurcate and confuse the streams of trading data emanating from the market and their differing temporalities: ‘one up-to-date, and the other (the ticker) behind’. ³²

The SEC countered, threatening more explicit regulatory arm-wringing in a manner typical of their strategy during the period. In one meeting between the two groups, an SEC staff member cited the legal basis of their authority over the ticker and argued that this would include its computerized form, threatening that the NYSE pursue the technology the SEC suggested or be forced to do so under more explicit legal command. ³³ The NYSE complied.

The market-machine, 1966–1969

In May 1966, the computer took full control over the generation of the ticker feed. ³⁴ In a public relations move aimed at solidifying the Exchange’s status as forward-looking, the NYSE set up a demonstration of the MDS in their Wall Street visitor’s center, flaunting the system’s ‘split-second reporting of stock transactions’ (NYSE, 1966a: 20). But the machine was no sideshow. The MDS immediately played a critical role in market practice. Its expanding importance within the Exchange is visible from the proliferation of fallback procedures that accompanied its deployment. Historical plans for system failure index a technology’s value to its community in relief. Specifically, MDS’s failure conditions show how some infrastructure and information was deemed critical while others were merely supplemental.

Not only did the NYSE have a second pair of IBM computers running in parallel to take over if one set failed; they funded an entire, separate computer system called the ‘ticker standalone’. ³⁵ This would be used in the worst-case scenario of complete failure in both pairs of MDS computers to keep trade reports flowing from the floor to the ticker tape. ³⁶ Electronic speed had become so central to the work of trade reporting that only computers could provide adequate backup for computers. ‘Fallback’ to paper slips would happen only if computer systems could not be restarted within two minutes. ³⁷ Trading and the ticker tape would continue in the event of a complete failure of the MDS, but a set of new functions would be lost, including the doomed test system for ‘registered trader surveillance’, ³⁸ quote-by-phone, and other automatically calculated metrics of market activity generated by the MDS, including the NYSE index (a stock market index like the Dow or the S&P) and hourly volume statistics. ³⁹ Reporting trades to the ticker was essential; statistics and market surveillance were less so.

The presence of these backup systems also reveals a transformation within MDS from a ticker tape control device into a central information hub for market data. This transformation hinged on what to its promoters was MDS’s most cutting-edge technology: Timesharing. Its computers multitasked. In 1968, for instance, the MDS did the following: Stored records of trades from the floor, checked the work of exchange employees, statistically analyzed trading patterns to detect some forms of suspicious trading, calculated ‘indices’ and responded automatically to various quote-request systems. Some of these responsibilities happened in real-time and concurrently, others happened on-demand or on fixed schedules, and others only happened when the machine had downtime. ⁴⁰ As a 1965 press release explained, timesharing was a ‘relatively new technique [that] permits spare computer time to be used by a number of independent programs simultaneously’. Although timesharing had ‘wide potential for the computer industry’, the release proclaimed, ‘the exchange is among the first organizations to put it into actual operation’. ⁴¹

Computerization efforts on the NYSE expanded almost immediately from the moment MDS came online, extending into two other significant market systems. These systems moved computerization from the mediating work of data processing and information dissemination into the heart of trading itself. One was the specialists’ ‘electronic book’, the other was odd-lot order routing. The development of these new technologies pushed a new market ontology to its legal and regulatory limit. Where did a market end and its media begin?

In its analog form, the specialist book was a physical record of offers to buy or sell stocks. A ‘specialist’ keeps track of the two-way auction for a set of securities and matches buyers and sellers; the specialist keeps the record of bids and offers and can buy on his or her own account. The computerization of the specialist book was an important pursuit in Wall Street automation for two reasons. First, specialist computerization represented a turning point in the spread of automation from clerical to professional roles. Second, the specialist’s work in market-making was viewed as a bottleneck in the mechanics of trading and a barrier to the achievement of market efficiency. In 1971, the mathematician and market theorist Fischer Black (of options-pricing fame) pointed specifically to the electronic specialist book as a lynchpin in the process of making securities markets as efficient in practice as they had come to appear in financial theory. (Black, 1971a, 1971b)

The stakes of specialist automation were particularly high at the NYSE, because specialists were a powerful constituency, establishing the core of the Exchange’s trademark auction style. One argument voiced against the mechanization of the specialist role was the idea that human specialists, but not machines, would take positions unfavorable to their own pocketbooks to slow volatile markets. In the famous ‘Black Monday’ crash of 1987, for instance, there is evidence that this happened: Specialists at the NYSE ‘leaned against the wind’ and bought to keep their markets afloat (Carlson, 2007: 9). However, in the late 1960s, analysis by the SEC demonstrated that specialists exacerbated volatile markets as much as 30% of the time by buying in rising markets and selling in declining ones (Seligman, 2003: 343). The automation of specialist work was therefore a complex process that involved not just technical problems and the familiar politics of professional automation, but also economic theory, institutional contests and regulatory prerogatives.

As a result, specialist mechanization promoted by NYSE executives in the late 1960s was tame in its ambitions. An electronic device, automators envisioned, would help specialists keep up with the new demand for trading and clear the market space of more pneumatic tubes, all while keeping human specialists on the floor. Automators imagined a specialist using a ‘possibly … hand-held …computer-driven … TV-like display’ that would ‘replace pneumatic tubes and pencils’ and ‘manual operations of report generation’. It would ‘mechanize’ the work of specialists, but not replace it. Electronic displays were critical to the system because the pace of trading in 1968 had outpaced even the 900-ticker. ⁴² A keyboard and electronic display would also enable adaptability in the future. A system proposed in 1966 would allow the specialist to run the book like its paper analog (with a ‘page turning feature’), while giving the user digital advantages: the ability to sort entries, perform bookkeeping, monitor electronic orders from ‘machine firms’ and perform computer analyses of the market. ‘[U]p-to-date computer techniques such as … trend projection, probability calculation [would] assist the Specialist … in handling his broker/dealer obligations’. The work of market analysis and the care of market-making would be supplemented by computer aids. ⁴³ It would have, in short, general purposes beyond the maintenance of the order book itself. ⁴⁴

From 1966–68, ‘specialist mechanization’ was automation in two senses: On the one hand, the NYSE pursued a short-term goal of specialist assistance to double the capacity of trading handled daily. On the other, they tentatively imagined a long-term future in which brokerage firms would send orders from their own computers directly to the electronic specialist book, bypassing floor brokers. ⁴⁵ Nonetheless, NYSE automation plans of the period reflected a desire to maintain the status quo professional sociality of the pre-electronic floor. In one memo about specialist automation, a system designer catalogued a number of member concerns about the program, including disruptions to the ‘traditional relationship’ between specialists and brokers. ⁴⁶

While specialist computerization seemed poised to change the social dynamics of the floor, a parallel computer project threatened to redefine the Exchange itself from a trading venue into an electronic switch for directing computer-delivered orders. In 1963, the NYSE began a program to automate the movement and pricing of ‘odd-lot’ stock trades. The odd-lot brokerage sector (trades in quantities of stock between 1–99) had consolidated early in the 1960s and by 1968 there were only a few brokerage firms left on the NYSE handling the lion’s share. ⁴⁷

At the center of NYSE’s odd-lot automation program was the Electronic Switching Unit (ESU), a computer that read orders from brokerage firms arriving via teletype, and directed them electronically to the correct odd-lot firm for pricing, matching and execution. The ESU replaced the human work and pneumatic tubes of internal market communication. The Exchange ESU computer ‘would behave in the manner of a communications switchboard … taking over some functions now performed by the Exchange’s tube men and tube system’. Then, so-called ‘machine firms’ or ‘mechanized’ brokers would sort and fill orders on their own computers based on a price provided by the MDS. They then responded to the order’s originating brokerage firm via the ESU with a report of the completed transaction. In this system, computers routed trades, priced and filled orders, and closed the loop with automated transaction reports. ⁴⁸

Just as in trade reporting, clerical accuracy was both the anxiety surrounding and the precondition for automation. Before the market could become an information switch, routing orders from broker to specialist, communications had to be transformed into a machine-readable format. Detailed manuals disciplined broker communication on the NYSE’s teletype infrastructure in 1967. Although these orders would still be read by clerks until 1969, the manual explicitly noted that exacting standards ‘anticipate[d] the eventual computer-pricing [of orders] and automatic generation’ of reports back to brokerage firms. At forty-seven pages, the manual attempted to codify every conceivable trade-related communication from brokerage house to odd-lot firm: From a simple ‘market order’ (to buy at the current price) to an update to ‘change a limit order to a market order’. ⁴⁹

The brokerage firm DeCoppet and Doremus was the first to mechanize its odd-lot business. In an era of timeshared computers with distant mainframes tethered to local terminals, the materiality of computing rattled the ontology of the market. In 1968, DeCoppet planned to use ‘repeating relays to send orders over low-speed communication circuits’ from the Exchange to their computers elsewhere in Manhattan for processing, and then back to the NYSE to complete the order. Their competitor Carlisle & Jacquelin contested DeCoppet’s plans for a test of their live odd-lot system in Fall 1968, arguing that computer processing offsite would violate Exchange rules prohibiting ‘off-floor’ trades. ⁵⁰ If DeCoppet’s computer terminals were in the Exchange, but their mainframes were afield, where was trading actually happening?

Studies of the implications of odd-lot automation on SEC regulations and Exchange rules began in 1966 when one external legal advisor noted that ‘if the exchange owns the computer which executes odd-lots, it might be deemed to fall within the [statutory] definition of broker’, representing a ‘departure from its historical role in merely providing market facilities for members’. Furthermore, the advisor stressed that ‘the Exchange may be liable if it has been negligent in its operations’, pointing out that this was new territory: ‘[N]o cases involving failure of computer equipment in contract or tort have been discovered’ and an odd-lot customer whose order is misrouted by computer and erroneously executed might have claims against the Exchange as well as the builders and maintainers of the equipment. ⁵¹ By contrast, the lawyer continued, the MDS would be sufficiently like older media systems like the ticker or quote-by-phone services that it ‘probably will not change the degree of legal responsibility of the Exchange for erroneous price quotations’. ⁵² The MDS largely sidestepped the troubling hybridity that odd-lot routing had brought to the NYSE. Computerized order routing, however, threatened the media/market, venue/broker and facility/execution binaries that distinguished categories of securities-market participants and responsibilities.

Nonetheless, systems for electronic delivery of market data still demanded careful management of the Exchange’s legal liability. Beginning in 1967, and in part motivated by the SEC pressure described above, the Exchange sold data from the MDS to a limited number of ‘private wire networks’ outside the Exchange. This was on a small scale: In 1968, requests to expand the number of supported terminal lines beyond twenty was met with concern by Electronic Systems Division executives who framed the expansion of interrogation lines as a trade-off against the fidelity of the machine in keeping pace with the market – its ‘thru-put’. ⁵³ Unlike the ticker, computer data served outside the exchange was not authoritative, as the Exchange did not ‘guarantee [its] sequence, accuracy, or completeness’. ⁵⁴ In fact, the bulk of the Scantlin ‘interrogation machines’ for delivery of MDS data were placed within the physical space of the Exchange, where the accuracy of data could be checked. ⁵⁵

The careful management of legal and regulatory categories of market and media mattered little in the crisis to come. In April 1968, the paper practices and media mechanisms of the Exchange and Wall Street brokerage firms began to come apart at the seams, threatening public, trader and regulatory confidence in the stock market. Amidst rising levels of trading activity, back-office clerks working in the paper-intensive business of trade clearance for member firms suddenly found themselves in an acute crisis: They were no longer able to clear a day’s worth of transactions before the next day’s trading began. The number of trades that ‘failed to deliver’ piled up. The situation snowballed until the Exchange closed the floor for trading one day a week for six months, and curtailed hours over an additional six, to give the paper and its people time to catch up (Blume, 1993; Wells, 2000). As the summer arrived, clerks struggled to complete the Friday work before the weekend began; some firms invested in liquor they sardonically called ‘delivery glue’ to keep clerks at their desks into the summer afternoons (Blume, 1993: 118; Cortada, 2006: 165–167).

The paperwork crisis made public the continuing human and operational challenges facing the securities industry. The MDS was facing its own crises of capacity. Since 1967, core memory limits were being reached with alarming frequency. ⁵⁶ A ‘drum wrap’ would occur in the MDS memory when enough trades were recorded such that the drum memory device had done a full loop, and returned to where it had started that day. ⁵⁷ The last trade recorded would find itself written next to the day’s very first, and if the drum were not stopped and replaced, the machine would become a digital ouroboros: An electronic snake eating its own tail. The day’s memory would be overwritten by incoming trades. This situation wasn’t catastrophic: it took staff just 45 seconds to swap out a drum for a new one. Nonetheless, drum wrap and ‘double wrap’ days were an alarming reminder of the computer’s material limits in an age of rising trade activity. ⁵⁸

Market federalisms, computer architectures, 1970–1975

The crises of the late 1960s brought heightened interest to the material and technological practices of securities trading in the US. ⁵⁹ A 1979 article in the Wall Street Journal intended as a retrospective on the fiftieth anniversary of the 1929 stock market crash instead focused attention to the 1969 ‘crunch’. The 1968 crisis of high volume was followed by one of low. Firms that had invested in manpower and automation to keep up with the pace of trading found themselves in 1969 with a drought of trades to execute. As many as two-thirds of NYSE member firms operated their brokerage businesses at a loss. Many were liquidated; others, including the fifth largest in the nation, Goodbody and Co., were absorbed into larger firms (NYSE, 1971; Rustin, 1979). ‘Brokers agree’, Richard Rustin (1979) noted, that 1969 ‘was even rougher on the securities industry than the 1929 stock market crash, which accompanied the onset of the Great Depression’.

The crunch brought a spate of government interest in the securities industry, which culminated in legislation and new market-regulation efforts, including the end of fixed commission rates, efforts to harmonize national markets and a continued push towards computerization in order routing and trade execution (Rustin, 1979). In addition to clerical inefficiency, by 1970 market fragmentation had become a major concern to economists, elected officials and the SEC. Fragmentation was the condition whereby the same good (like a share of common stock) was traded in multiple locations for potentially diverging prices. Individual investors might not care where a trade was executed, as long as the price was roughly right. But to large institutions the costs of market fragmentation were high. Institutional investors took large positions and brokerage was a contributor to the costs of managing funds. A fund manager might go to brokers on the floor of the NYSE or the Pacific Stock Exchange, or pursue a bilateral trade with another institutional investor through the NASDAQ OTC system, depending, in part, on the costs of execution – the amount charged to actualize a trade. Adding to the complexity was the fact that the prices might actually differ by small amounts between exchanges, but these differences would not be visible because of different media for reporting trades between various exchanges. Furthermore, some exchanges did not have enough trading activity to support large trades. Although 1970 was the year that Eugene Fama published his paper on the efficient markets hypothesis, in practice, capital markets were hardly efficient processors of information.

Regulatory studies and congressional hearings following the crisis years focused intensely on these problems (United States Congress House Committee on Interstate and Foreign Commerce, Subcommittee on Commerce and Finance, 1972; United States Congress Senate Committee on Banking Housing and Urban Development, Subcommittee on Securities, 1971, 1973). Supplemented by testimony and stacks of reports by consultants, investment banks and economists (Arthur D. Little Inc. [ADL], 1974; NYSE, 1971; Petruschell et al., 1970), the consensus conclusion of these studies culminated in the 1975 Securities Acts Amendments, which stressed the need for the nation’s various venues for trading stocks to become electronically harmonized to facilitate the reform, unification, and automation of clerical and administrative practices. Information technology featured centrally in these deliberations, and yet the specifics of market technologies were largely framed out of legislative discussion, left to regulators or self-regulating institutions to concretize. As a result, what was discussed as a ‘central market system’ in 1971 was ensconced in law in 1975 as an anomalously defined ‘National Market System’.

In fact, in the intervening years two trajectories had emerged for national securities industry harmonization. On one side was a coalition of economists, legislators and investors in favor of a centralized national market, virtualized in a computer system that would feature an electronic order book for all trading activity in a given stock. On the other side were the NYSE and allied exchanges and a technology called the Intermarket Trading System. This system preserved the separation of geographically and physically distinct trading floors. It also maintained the status quo market domination of the NYSE (Mendelson and Peake, 1979). The divergence of these two visions reflected technological as well as political positions. Their resolution shaped modern market infrastructure and governance.

The NYSE’s own position on financial computerization was shaped by the crisis years. Some brokerage firms, the NYSE President Robert Haack stressed in 1970, had simply given ‘little or no attention to long-range planning’. Not only are those ‘the most seriously affected by the widely publicized paper-work problem’, they are also gumming up the works for the rest of the market. The ‘Big Board Chief’ pledged ‘the highest priority’ toward aiding member firms in their own long-range planning’, a double effort to keep member firms from going under and to expand the NYSE’s role as the dominant voice in financial computerization (in Seligman, 2003: 353).

In 1972, Haack’s NYSE started a joint venture with the American Stock Exchange (AMEX), called the Securities Industry Automation Corporation (SIAC), to economize on financial technology between the two exchanges by consolidating ‘planning, design, and operation of automation services’. ⁶⁰ SIAC would bring under one roof the management and development of stock quotation, clearance and trading infrastructure of lower Manhattan stock trading, taking responsibility for every piece of technology owned by the exchanges, from time stamps to computer programs. ⁶¹ Over the following years, the NYSE was able to make a claim to being an automation authority not just for lower Manhattan, but throughout the country.

The new cooperative ownership of computer hardware and software, as well as proprietary communications cables and switches, produced small frictions between the NYSE and AMEX. These frictions reveal much about the emerging legal ontology of proprietary electronic infrastructure. The text of early service agreements between SIAC, AMEX and the NYSE offer clues concerning contextual anxieties over the ownership of data and material infrastructure. Physical access to hardware and wires, and their vulnerability to tampering, were clear concerns to both firms. Extensive provisions enumerated rights to access physical switches, computers, and cables by one company or the other. The NYSE owned much of the physical property where SIAC was housed, and AMEX officials were anxious about getting locked out of their own market data feed. Not only was their access guaranteed, but SIAC would prevent anyone from ‘attach[ing] … to the wires, apparatus or equipment by which AMEX sales data or the bid-asked quotations are received’, and would ensure against the ‘tapping’ of lines and ‘circuits’. SIAC further agreed to prevent access to the content of market data unless approved by AMEX – that is, they were responsible for data security. ⁶² These efforts reflect an effort to maintain control over the content and revenue of the market data feed while handing off to SIAC the management of materials and technicians. ⁶³ This was electronic infrastructure in the making: A capitalization of the data feed through legal and technical black boxing that simultaneously outsourced responsibility over data’s materiality to a third party and solidified the owner’s control over the feed as a revenue and content stream.

SIAC’s work is demonstrative of the early 1970s business model of a computer services company in the timesharing era; indeed, many of SIAC’s executives came from that industry. ⁶⁴ In this model, companies owned the capital of data processing (computers) and rented time to subscribers who connected to their mainframes via terminals. SIAC maintained and covered the costs of installing terminals provided to brokerage firms in Manhattan, ⁶⁵ centralized ownership of hardware whose use it then sold back to its clients, and even sold spare computer-time as a general-purpose computer-time vendor. This practice of spare computer-time vending was so widespread in the 1970s that SIAC found itself in direct competition with NYSE member firms also selling time in the Manhattan computer market. ⁶⁶

Beyond the administration of existing technologies and infrastructure, SIAC also took responsibility for the development of new technical systems for moving stock quotations and orders. These were largely of two types: Telecommunications systems and market switches – computer technology for routing information within exchanges.

The Common Market Switch (CMS), one part of an ambitious market computerization program called Project CENTAUR, materialized the dream that had propelled the twin odd-lot order routing and MDS projects of the 1960s: A stock market computer as a universal hub for exchange information flows. Operating as central ‘hub’ from 90 West Street in Manhattan’s Financial District, the SIAC CMS linked more than thirty different computers within and beyond AMEX and the NYSE. It was also the heart of the ‘Designated Order Turnaround’ system, on-line in 1976, which routed electronically submitted orders from outside the exchange to specialists on the NYSE floor. CMS routed orders and administrative traffic ⁶⁷ and was a central node in an expanding web of market data technology that stretched not only to the market’s traditional output media – the ticker tape – but increasingly to computer data vendors. ⁶⁸

In 1974, SIAC began to bid on contracts to provide its technologies and data processing services to other exchanges in the US, a move that extended its reach from Manhattan to the national financial system. Its first contract with an outside firm was the Chicago Board Options Exchange, but SIAC also bid for stock clearing services and other data processing services to the Midwest, Philadelphia, and Pacific Stock Exchanges, and the National Clearing Corporation. ⁶⁹

This expanding portfolio of financial technology and data-processing services set the stage for the Intermarket Trading System (ITS), as a technical and political project by SIAC. ITS connected stock exchanges through terminal devices located on trading floors that showed the price of stocks at other exchanges, and allowed brokers to send orders from the floor of one exchange to the specialist of another. ITS was the NYSE’s model of how to connect markets in the era of the proposed ‘central market system’. Notably, it preserved the primacy of the vocal auction market and the centrality of human specialists. At a time when the industry faced increasing pressure to unify and harmonize, ITS offered a form of connection that largely preserved the status quo: Geographically distinct floor trading venues where specialists controlled the market-making role for individual stocks. ITS was the product of a consortium of exchanges including NYSE, AMEX, and the Philadelphia-Baltimore-Washington Stock Exchange. SIAC would be its data processor.

Opponents of ITS argued that it only nominally allowed tele-traders to get in on the action of venues like the NYSE. ‘Being on the exchange floor still conveys certain advantages over being on a secondary exchange or at a brokerage firm’, noted Morris Mendelson and Julius Peake (1979: 34), supporters of a competing national market infrastructure called the Composite Limit Order Book (CLOB). Peake and his allies pushed for a different technical and capitalist order than what ITS would provide. It was one that, in contrast to the loose federalism of market switches and interconnections promoted by NYSE, emphasized centralization in trading, governance, and technology. Pressure for such centralization had begun in earnest in 1971 during Federal legislative hearings into the troubles of the securities industry. New Jersey Senator Harrison Williams Jr. opened one hearing by castigating the securities industry, describing the ‘near debacle of 1970’ and calling the state of the industry ‘a cause of serious concern not only to the investing public, but also to the Nation at large’. Operational problems, including failures of trades to execute, had rattled public confidence in stock markets. Because operational failures threatened perceptions of the stock market’s liquidity, the efficient functioning of the markets’ people, practices and technologies were critical to its economic security. This sentiment was echoed by several witnesses, including the chairman of the investment firm Merrill Lynch, who stressed the link between market efficiency, transparency and centralization: ‘When an investor comes into the market to buy or sell a stock, he should be able to come to the central place where all bids and offers for that particular stock are congregated’. But the situation was not like this in practice: Instead ‘there is one market in New York, another market on the Pacific coast, a third market that is nebulous – except in terms of wires that connect brokers’ trading departments together … the investor does not know what is going on in the market’. ⁷⁰ But this was not only a problem for well-heeled traders. The rise of institutional investing in the stock market had made Main Street and Wall Street increasingly connected. Inefficiencies in the stock market risked damaging retiree pensions and mutual funds. The structural importance of the exchanges to the national economy was becoming increasingly apparent as well. As Williams stressed, foreign investments in the US were highly concentrated in stocks, and a significant portion of the US trade balance was tied up in the stock markets. ⁷¹

A solution to fragmentation and operational problems began to solidify around the time SEC published a policy statement in February 1972. ⁷² In it, the Commission discussed the pursuit of a ‘central market system’, reflecting a new technical and regulatory ambition whereby, ‘because of modern communication and data processing facilities, it is possible to preserve geographically separated trading markets while at the same time tying them together on a national basis’. New technology, the SEC (1972) stressed, made it possible to virtually represent ‘all interests of investors … in a central market’, without creating monopoly power in any particular exchange or firm. Orders entered in Cincinnati, San Francisco or New York would have the same access to the auction market. But as to the technical qualities of such a system, the SEC (1972) wanted leave it to the industry, and to ‘observe, and if necessary, to modify the structure which evolves through the ingenuity and response of the marketplace to the extent changes occur that appear inconsistent with the public interest’ (p. 5286).

The CLOB model reflected the highest ambitions of the early 1970s central market system aspirations: The centralization of market activity for each security on a national basis. CLOB connected trading floors to a single electronic order book that concentrated all orders in a single virtual market. Key to this system was strict price and time priority for orders and the automation of the specialist role. It hinged on the ability for a system to create a ‘National Best Bid-Offer’: A single price for each of the buy-and-sell sides of a given stock; no matter if you came to buy from New York or San Francisco. Such a system could eliminate inter-market arbitrage opportunities in which traders took advantage of price differences between markets and privileged locations within information networks to trade the same good risk-free between markets. Indeed, Peake (1978) argued: ‘One true test of a true national market system will be the impossibility of intermarket arbitrage’. (p. 31). He noted with concern the perils of a plan for harmonization that lacked strong central governance and pushed for a ‘Hard CLOB’ model of total market centralization, stressing that ‘the only way to build that system is with an integrated communications and computer environment’.

But who would run it? A self-regulatory organization like NASD or a governmental authority? Without strong governance, a central market system was a dream of standards without a standardizer; a set of terminals without a central computer or market processor. Despite advocacy by industry consultants and academics, the SEC did not promote a single entity as a market processor for the US. This weakened the push for the unification of American securities markets as proposed reforms of the market bumped up against the political assumptions underlying securities exchanges: That they should be locally managed and privately owned, and largely left to self-regulate and free to compete through the creation of innovative rules and market systems.

By contrast, and in light of their expanding national technical responsibilities over market clearance and data processing, SIAC made a good case for its position as de facto national data processor for the financial markets. ⁷³ As a result, ITS and the material political economy (MacKenzie, 2017a) that it concretized, of connected but decentralized markets, would slowly become the norm. Without a strong industry body or federal regulator to support centralization and its costs, the energy behind market harmonization regressed to the status quo. When the Securities Amendment Acts of 1975 became law, the language of market nationalization had changed significantly. What was once discussed as a ‘central’ market had become a ‘national’ one. This shift reflected a changing politics of market networking – from an ambition to centralize and unify the nation under a single market, to a weak federalism of market inter-connections. Although the law tasked the SEC with ‘facilitating the establishment of a national market system’ by ‘removing impediments to and perfect[ing] the mechanisms of a national market system for securities … and for the clearance and settlement of securities’, the specific meaning of this system, and the regulatory muscle to achieve it, were largely omitted. ⁷⁴ As a telling example of the vacuum of federal guidance, the SIAC 1976 long-term automation plan, which includes a glossary of technical computing and regulatory terms, includes the following entry for NMS: ‘National Market System. A system, which will treat equity trading on a national level. Not defined explicitly’. ⁷⁵

The SEC did not push aggressively for a particular model of market interconnection after 1975. As Seligman (2003) notes, ‘the SEC itself did not study the technological problems involved in implementing a national securities market system … instead … the SEC deferred to the congressionally required National Market Advisory Board’ (pp. 526–527). That group, made up of both a Pro-CLOB faction and NYSE veterans, released a series of indecisive reports, setting out a range of options, none of which received strong consensus advocacy.

Coordination among even allied groups to build a centralized system would have been a challenge without clear techno-regulatory guidance. Add to this NYSE’s work to promote an alternative standard, and we begin to see the shape of late-twentieth-century financial data infrastructure and regulation. A system of financial centralization, with all its ambitions of equal access and transparency, required central leadership and technical governance. A system that reproduced the status quo of market fragmentation demanded comparatively little.

Conclusion

This article has described the history of the development, deployment, and transformation of a critical material and epistemic infrastructure for electronic financial markets: Price information. It joins previous studies of market data systems in emphasizing the role of electronic media in the formation of consequential assumptions about where a financial market is (Knorr Cetina and Bruegger, 2002a) and what market data represents (Knorr Cetina and Preda, 2007). It likewise shares in broader efforts in the social studies of finance to analyze technological, organizational, and calculative ‘market devices’ as they distribute knowledge and agency to and within markets (Muniesa et al., 2007). Technologies of calculation and communication enmesh individual action within opaque technical systems and diffuse webs of distant trust. This is not unique to financial markets. At stake in the examination of finance’s technical orders is both the specific question of responsibility for the crises of modern finance but also the character of agency under conditions of mediated social life.

The history of financial technology presented here has emphasized the role of institutions in the construction and shaping of electronic market data systems. Initially, the NYSE and SEC were joined by aligned goals, including the oversight of clerical labor in the processes of market data mediation. But this alignment was only partial. As the SEC took advantage of a period of relative strength following the 1963 Special Study and the Securities Acts Amendments of 1964 to press for expanded financial market oversight functions through computerization, it was hampered by the Exchange, which made strategic choices to promote technologies that preserved the interests of its constituents: Traders, brokers, and specialists. Trader surveillance and the content of market data were early sites of friction between the two groups that culminated in a broader contest over the national unification of stock markets in the early 1970s.

The National Market System debate of the 1970s offers a historical antecedent to contemporary concerns around high-frequency trading strategies, such as those described by MacKenzie (2017a, 2017b) and Lewis (2014). In important ongoing work, Pardo-Guerra (2017) carries forward the history of market fragmentation from the 1970s to the electronic markets of today. The persistence of fragmentation appears attributable to a number of forces, including ongoing contests between regulatory and market institutions, the recasting of fragmentation as socially beneficial inter-market competition and the continuation of divergences in the political imaginaries underlying the organization of national markets. In a study of the technologies of algorithmic trading, MacKenzie (2017b) also notes the importance of the 1970s centralization debates for the fractured character of electronic markets today and their most famous consequence: The proliferation of high-frequency trading. These strategies depend upon an electronic market that is nominally virtual but is materially fragmented in a baroque network of physical infrastructures and geographically distributed venues, proximate access to which confers informational privileges within the market. As described in this article, market regulators of the 1960s saw computer communication and data-processing technologies as tools to curtail problems similar to those of the analog era: The concentration of information and power in the physical space of the Exchange floor. It is tempting to view this history in ironic terms, in that media brought into the market to diminish the effects of spatial and institutional privilege essentially reproduced that system of privilege among the owners of proprietary low-latency market connections. However, this was not some unforeseeable consequence of technological change. Instead, it was the predictable result of the decline of active governance of market infrastructure by federal regulators (Peake, 1978: 27).

Ultimately, the ability to shape market technology depended on historical oscillations in the relative power of groups including the NYSE and the SEC. Nonetheless, the diminishing role of the SEC in market technology in the early 1970s also reflects a broader transformation in the historical association between computers and governance. Computers took on a number of differing meanings as devices of governance in the twentieth century, including as tools for the consolidation of control over human organizations and technical systems (Edwards, 1996; Haigh, 2001; Medina, 2011; Noble, 1984) and as materializations of rules-based systems of political or economic self-governance (Agar, 2003; Chun, 2011). The former conception of computer governance supported the partial and temporary alignment of the NYSE and the SEC, as they shared the common view of computers as tools for the surveillance and control of human work. But in the wake of the paperwork crisis, the market appeared increasingly like a technical system in need of engineered solutions. It was primed, therefore, for a form of computer governance in which faith in technical systems outpaced faith in the institutions that instrumentalized them.

Beyond the history and social studies of finance, this work suggests that the wedding of the computer and the market represents an important chapter in the history of twentieth-century governance. Their entwining not only made possible the flash and Black Monday stock market crashes but undergirded a more general phenomenon, the bullish investment of public trust in markets as information systems and in computer media for the infrastructure of social life.