The Failure Of Liability In Modern Markets

Yesha Yadav

Vanderbilt University – Law School

August 28, 2015

Virginia Law Review, No. 106, 2016, Forthcoming

Vanderbilt Law and Economics Research Paper No. 15-21


This Article argues that the liability framework governing securities trading is unable to effectively deter and compensate harms in algorithmic markets. Theory underscores the significance of robust laws to safeguard information flows and the trading process. Without this assurance, investors internalize the costs of privately policing markets and will rationally discount the capital they invest. A detailed body of regulation seeks to ensure that markets function safely, benchmarking compliance using the three familiar standards grounding liability: (i) intent; (ii) negligence; and (iii) strict liability. This Article shows that this framework is ineffective in markets that rely on algorithms – or pre-programmed computerized instructions – for trading. It makes two claims. First, a basic level of error is endemic to the operation of algorithmic markets. Especially when designed to trade in fractions of a second, algorithms must be programmed in advance of trading and anticipate how markets are likely to behave. This predictive dynamic means that error and imprecision are inevitable, irrespective of constraints that liability imposes. Secondly, liability standards fare poorly in high-speed algorithmic markets where errors can spread rapidly across multiple exchanges and security types. Even small, “reasonable,” risk taking can give rise to outsized harms, diminishing the protection provided under the negligence standard. Strict liability also fails. With error inextricably a part of predictive, pre-set algorithms, liability can arise too frequently to function as an informative signal of bad behavior. Further, small errors can create large-scale losses that may be too high for any single firm to pay. Finally, punishing only intentional bad actors leaves a swath of the market unsanctioned for careless behavior. With each standard falling short, the current design of the liability framework can leave markets facing pervasive costs of mistake, manipulation and disruption. In concluding, this weakening of laws points to a need for structural solutions in automated markets. This Article explores avenues for reform to institutionalize better behavior and fill the gaps left by the law.

The Failure Of Liability In Modern Markets – Introduction

In April 2015, the Department of Justice charged Navinder Sarao for his role in causing the Flash Crash – the near 1000-point drop-andrebound in the Dow Jones Index that roiled markets in May 2010. Sarao, a small-time, British trader operating out of his parents’ suburban basement, stood accused of putting together a string of illusory, fake orders that fooled markets enough to spark the largest single day drop in the Index’s history. Commentators rightly contest whether a bit-player like Sarao could have unleashed a near-catastrophe on U.S. securities markets singlehandedly. Yet, the complaint – and its causal account – point to a troubling dilemma facing scholars and policymakers today. This Article shows that the longstanding liability framework undergirding securities regulation looks increasingly fragile in the face of modern market design. With trading growing ever more automated – characterized by complex algorithms, a proliferation of specialist traders and interconnections between markets – single weak links can create outsize costs. This evolution in market design poses a profound challenge for well-established liability regimes governing fraud, negligence and mistakes. Trading firms are easily capable of creating far larger risks than they can either provision for ex ante or pay for ex post. In decoupling the riskiness of trading firms from their capacity to realistically bear the cost of their conduct, market structure casts doubt on the law’s ability to credibly constrain as well as punish mistakes and misbehavior in trading.

While scholars have vigorously debated the design of liability regimes in securities regulation, few dispute the underlying need for a guiding framework in this context. Securities amount to little more than simple claims on the future value of a company’s cash flows. Without credible, trustworthy information to substantiate these claims, investors face deep uncertainties in valuing them and in determining how much of their capital to invest. The risk of fraud, mistakes or manipulation in presenting information can dissuade investors from bringing their money to markets or force them to rationally discount for the risks and the costs of verification. A regulatory framework that punishes misinformation constrains those whose expressive conduct and communication matter to investors. Given this importance, scholars have devoted extensive attention to the regime underlying fraud and misrepresentation in securities regulation, generating a vast literature studying its effectiveness. But, investors face a much broader set of risks than just the harm of misinformation. In particular, they depend on the operation of trading mechanisms to buy and sell their securities and to permit timely entry into and exit from investments. Without such mechanisms, investors face being left holding sticky securities or missing out on profitable trading windows. Exchanges, brokers, and other intermediaries operationalize the trading process and allow investors to interact fluidly within the marketplace. Though attracting much less scholarly attention, regulation also controls these “execution” and “liquidity” risks through an intricate system of rules, regulations and best practices to deliver robust trading systems.

The dense volume of rules comprising the liability regime in securities regulation obscures the observation that, for the most part, it measures compliance according to three well-established legal standards: (i) intent; (ii) negligence; or (iii) strict liability. Certain infractions, notably fraud or manipulation, demand that authorities show that defendants acted with intent to deceive and disrupt trading. In other cases, defendants face sanction when they act negligently by failing to abide by a standard of reasonable care. Finally, for certain harms, particularly for more technical breaches, regulation can punish using strict liability. These standards of liability are familiar to lawyers and their application is well established under jurisprudence and scholarship. By choosing one or other head of liability for a particular offence, regulation calibrates the costs that defendants – as well as authorities – face in maintaining order in the marketplace. With a robust disciplinary system in place, an effective liability framework should prevent undue discounting by investors.

These familiar standards of liability, however, are rapidly losing relevance in modern automated markets. Recent years have been marked by a shift towards a near-fully automated marketplace, with algorithms – pre-programmed electronic instructions – driving almost all aspects of trading. Instead of relying on human beings to perform the task of submitting orders, routing them to exchanges, concluding and completing trades, these functions are instead undertaken by algorithms. Unlike human traders, computers can transact in microseconds, at high volumes and deploy an enormous reserve of data and quantitative input to inform trading. Algorithmic trading – as measured by the subset of hyper-fast, data-driven high frequency trading – is responsible for around 50-70% of equity volume and an estimated 60% of all trading in futures markets in the U.S. When added to trading by (relatively) slower algorithms, these figures reach considerably higher.

Flash Crash Failure Of Liability

See full PDF below.