A Material Sociology of High-Frequency Trading

30年前，几乎所有的金融交易都是通过电话或面对面直接在人与人之间进行的。在20世纪80年代末和90年代，电子交易势头强劲，但最初几乎所有交易仍是由人类通过电脑屏幕和键盘进行的。然而，自2000年以来，交易越来越完全自动化。特别突出的是超快速的算法高频交易（HFT）。在几个重要的市场中，高频交易现在占所有交易的四分之一到一半以上（例如，Baron等人，2012年，Bouveret等人，2014年，Brogaard等人，2014年）。我们的研究将经济社会学和科学技术研究（STS）联系起来，将调查：1)高频交易是如何进行的；2)它是如何以及为什么在一些市场站稳脚跟，而在另一些市场却未能进入。从经济社会学和STS交叉的角度来看，高频交易是一个理论上的关键话题。经济社会学家将市场作为有争议的、历史上的“领域”进行了富有成效的分析，重点关注市场文化、社会秩序和治理以及围绕这些领域的冲突等主题。相比之下，研究市场的STS学者通常关注市场的物质性：技术设备在其中的作用；物质程序的；有形体的人（而不是无形体的认知，无论理性与否）；数学模型不是抽象的，而是实际的计算例程。每种方法的支持者经常认为另一种方法是不完整的，甚至是不充分的，而本研究的理论目标是发展一种我们称之为“物质社会学”的综合分析。高频交易是一个适合发展物质社会学的实证主题，原因有二：首先，高频交易的物质方面是至关重要的。例如，即使是光速也是一个限制。高频交易中显著的时间单位是一微秒——百万分之一秒——而在一微秒内，即使光纤电缆中的光也只能传播200米左右。因此，例如，计算机系统的确切位置和通信链路的精确路线非常重要。其次，高频交易的兴起引发了有关市场结构、交易规则等方面的冲突：经济社会学家感兴趣的那种冲突。几乎所有的高频交易公司都是新近进入市场的小公司。他们面对的是大型银行等老牌参与者（按照高频交易标准，它们的技术系统通常较慢）。到目前为止，由此产生的冲突的结果差别很大。在一个极端（例如，美国股票交易），传统的制度结构几乎完全被侵蚀，取而代之的是一种促进高频交易的制度/技术结构（除了一些有趣的例外）。在另一个极端（例如英国和欧元区政府债券市场），高频交易公司根本无法获得任何准入。我们的实证研究将以我们的“物质社会学”理论目标为依据。例如，当我们研究高频交易算法如何预测价格变化时，我们将研究它们这样做的技术程序实际上依赖于历史位置的程度，市场的竞争特征。当我们研究高频交易和市场结构之间的相互作用时，我们不仅要研究社会角色和规则，还要研究物质基础设施。我们将采用半结构化访谈（高频交易员，为他们提供技术服务的人，以及交易所员工、监管机构和现有公司成员等其他人），辅以文献来源和观察性研究（例如在高频交易行业会议上）。广泛的前期研究为我们提供了必要的人脉，使我们能够在不侵犯“商业秘密”的情况下深入调查受访者，并为我们的初步数据分析框架提供信息。

Thirty years ago, nearly all financial trading took place directly among human beings, over the telephone or face-to-face. In the late 1980s and 1990s, electronic trading gained momentum, but initially nearly all of it was still by human beings, using computer screens and keyboards. Since 2000, however, trading has increasingly become entirely automated. Particularly prominent has been ultra-fast, algorithmic high-frequency trading or HFT. In several important markets, HFT now makes up between around a quarter and just over a half of all trading (see, e.g., Baron et al. 2012, Bouveret et al. 2014, Brogaard et al. 2014). Our research, which links economic sociology and science and technology studies (STS), will investigate: 1) how HFT is conducted; 2) how and why it has become firmly established in some markets while failing to gain entry to others. HFT is a theoretically pivotal topic from the viewpoint of the intersection of economic sociology and STS. Economic sociologists have productively analysed markets as contested, historically situated 'fields', focusing on topics such as the cultures of markets, their social ordering and governance, and the conflicts surrounding these. STS scholars who study markets have in contrast typically focused on the materiality of markets: the role in them of technological devices; of material procedures; of embodied human beings (rather than disembodied cognition, whether rational or not); of mathematical models not as abstractions but as material computational routines. Proponents of each approach have often seen the other approach as incomplete, even inadequate, and the theoretical objective of this research is to develop an integrated analysis that we call a 'material sociology'. HFT is a suitable empirical topic for the development of a material sociology for two reasons: First, HFT's material aspects are crucial. For example, even the speed of light is a constraint. The salient time unit in HFT is a microsecond - a millionth of a second - and in a microsecond even light in a fibre-optic cable travels only around 200 metres. So, e.g., the exact locations of computer systems and precise routes of communications links matter hugely.Second, the rise of HFT has triggered conflicts about the structures of markets, the rules of trading, and so on: conflicts of the kind that interest economic sociologists. Nearly all HFT firms are small, recent entrants. They confront established players such as big banks (whose technical systems are typically slow by HFT standards). The outcomes so far of the resultant conflicts vary considerably. At one extreme (e.g., US share trading), the traditional institutional structure has eroded almost completely, to be replaced by an institutional/technical structure that - with some interesting exceptions - facilitates HFT. At the other extreme (e.g. the markets for UK and Eurozone government bonds), HFT firms have simply failed to gain any access. Our empirical research will be informed by our 'material sociology' theoretical objective. E.g. when we examine how HFT algorithms predict price changes, we will investigate the extent to which the technical procedures by which they do so actually depend on historically-situated, contested features of markets. When we examine the interaction between HFT and the structure of markets, we will investigate not simply social roles and rules, but also material infrastructure. We will employ semi-structured interviewing (of high-frequency traders, those who supply them with technical services, and others such as exchange staff, regulators and members of incumbent firms), supplemented by documentary sources and observational research (e.g. at HFT industry meetings). Extensive preparatory research has given us the necessary contacts, allowed us to discover how deeply we can probe interviewees without intruding on 'trade secrets', and informed our initial data-analysis framework.