The Next Generation of Austrian Economics: Essays in Honor of Joseph T. Salerno

4. High-Frequency Trading: A Note on Spot vs. Future Trades, Property Rights, and Settlement Risk by Guillaume Vuillemey

Textbook* descriptions of financial markets draw a clear and seemingly unambiguous distinction between spot and future transactions. Whereas future transactions are often confined to derivatives markets, everyday trades on stocks, bonds or other assets are said to be spot. Furthermore, common descriptions of spot transactions usually do not distinguish between (i) the time a trade is agreed upon and (ii) the time it is paid for and delivered, as both are assumed, by definition, to take place virtually at the same point in time.

This chapter provides a theoretical investigation of high-frequency trading (HFT), which arises from the lag existing — even for seemingly spot transactions — between steps (i) and (ii). To this end, I shall redefine the dichotomy between spot and future transactions when the settlement of trades does not occur in real time but with a lag, and when this lag can be exploited by algorithms, computerized techniques or human decisions.

High-frequency trading consists of trade exposures opened and closed between settlement dates by market participants ensuring that their net open exposure at the settlement time is zero (implying that none of the trades performed intraday are either paid for or delivered). HFT transactions are not akin, for conceptual understanding, to usual trades that would merely be executed “faster” or to positions being liquidated after a shorter period of time. One distinguishing characteristic of HFT activities is that they can be performed with virtually zero cash or securities’ holdings in the first place, as the trader ensures a zero net position at the settlement date.

This chapter investigates two questions. First, does HFT imply that intraday buy and sell trades are performed using temporarily ex nihilo created fiat money? Second, can the case where securities are agreed-upon but never delivered create multiple (therefore conflicting) but valid property rights on particular assets? The issue at hand resembles those raised by fractional reserve banking. Importantly, this chapter does not comment on the status of high-frequency trading under various legal systems or jurisdictions — this is left for future research — and instead focuses only on the theoretical conditions under which the above-mentioned consequences may occur.

If the above questions are to be given a positive answer, then serious consequences follow as regards intraday liquidity management in payment and settlement systems. An example is that of “failures to deliver” arising from high-frequency trading from naked short selling, whereby a trading institution is not able to deliver at settlement date securities it has been selling during the day.1 Other consequences may relate to intraday collateral management, for instance in the case where securities are bought and delivered as collateral before the settlement of the initial purchase. Besides economics, ethical and legal issues raised by the potential over-issuance of property rights through high-frequency trading activities are akin to those raised by Mises (1996) or Huerta de Soto (2011) in the case of fractional reserve banking. An overview of Mises’s views on fractional reserve banking and monetary theory can be found in Salerno (1994).

Answering the above questions requires a careful analysis of the consequences of the lag between the time trades are agreed and the time they are paid for and delivered. I will show that, when clearing and settlement do not occur in real time, trades that are usually — theoretically and/or legally — described as spot must be treated as futures if a careful economic analysis is to be conducted. I also provide a criterion to distinguish between spot and future trades. Finally, I show that the over-issuance of property rights arising from HFT exists when transactions which should be treated as futures are legally or factually treated as spot.

The remainder of the chapter is structured as follows. First, high-frequency trading is described and is shown to be merely the exploitation of the lag between the time trades are agreed upon and the time they are settled. Its fundamental difference with other (“usual”) trading activities is also highlighted. Then, the distinction between spot and future transactions is refined. Trades on financial markets where settlement is delayed are shown to be meaningfully understood as futures. Finally, I define the conditions under which certain legal treatments of high-frequency trades as spot or as future transactions may lead to the over-issuance of property rights, thus give rise to liquidity risk in payment and settlement systems.

High-Frequency Trading as the Exploitation of Delayed Settlement

I shall start by examining the nature of high-frequency trading and the conditions under which it arises. High-frequency trading on an exchange platform consists of trades usually performed by computer algorithms so as to benefit from private information regarding the order flow or from small price variations over short horizons (ranging from a few milliseconds to a few hours). The major characteristic of high-frequency trading algorithms is that they ensure a virtually zero net open exposure at the end of each trading day, so that no cash or securities have to be physically delivered. High-frequency trading has recently become a sizeable phenomenon on financial markets, as it represents up to 70 percent of all trades on some organized stock exchanges (see Swinburne 2010).

I do not propose an extensive review of the literature (which can be found in Gomber et al., 2011). Most of the academic work revolves around the consequences of high-frequency trading on particular aspects of the price system, typically on the price formation mechanisms (bid-ask spreads, “price discovery” mechanisms, etc.).2 For instance, one oft-mentioned concern relates to the fact that high-frequency trading may amplify price volatility to the extent of triggering “flash crashes.”3 Among the main findings documented in the empirical literature are a reduction in trading costs and bid-ask spreads (see Brogaard 2010; Hasbrouck and Saar 2010) and a decline in short-term volatility (see Jarnecic and Snape 2014 or Brogaard 2011). Contrasting with the existing literature, this chapter focuses on an issue of a completely different order, largely neglected up to now. I do not focus on the empirical or theoretical consequences of high-frequency trading on particular aspects of the price system, but instead provide a theoretical analysis of high-frequency trading as regards property rights on cash and on traded securities. More precisely, do HFT activities lead to the over-issuance of property rights or to the ex nihilo creation of money?

An essential preliminary to be mentioned is a key institutional feature of present-day financial systems, namely the lag that exists on financial markets between the time trades are agreed (prices and quantities are decided upon) and the time payment and delivery actually take place. Whereas trade orders can be executed at any point in time during the trading day, clearing and settlement occur at one point only during the day, usually at the end of the trading session or up to T+72 hours. It is of utmost importance to highlight that such a time lag for so-called spot transactions is essentially institutional, i.e., that it does not primarily exist as a consequence of any physical or operational constraint. With the advent of computerized technologies at all stages of post-trade processing, real-time settlement (or quasi real-time settlement, as several actors have to be coordinated) could be a perfectly valid and implementable contractual or legal framework. For instance, real-time gross settlement systems (abbreviated RTGS4 ) exist for interbank payments — such as Fedwire in the United States and TARGET2 in Europe.

As a preliminary, I shall examine the extent to which high-frequency trades differ from other (“usual”) trades and show that high-frequency trading primarily exists as a consequence of delayed settlement. One key theoretical question for my purposes is actually whether high-frequency trades are akin to “usual” trades that are performed faster (an asset being bought at some date and sold a short moment — from microseconds to several hours — later), i.e., trades that could be fully described in theoretical terms by the canonical description of exchange phenomena (see Mises, 1996, for example). I aim to show that high-frequency buy-and-sell trades cannot be understood theoretically as a combination of spot buy and sell transactions.

I shall begin with a mere description of the steps involved in any combination of spot buy and sell transactions. For trader A, a usual buy-and-sell transaction amounts to (i) agreeing with B on prices and quantities, (ii) paying the agreed-upon monetary units to B in exchange for the agreed-upon good, and at a later date (iii) agreeing with C on prices and quantities and finally (iv) delivering the agreed-upon good to C in exchange for the agreed-upon monetary units.

On the contrary, high-frequency buy-and-sell operations do not imply, at any time, either any disbursement of cash or any physical delivery of a security or good. This is due to the fact that steps (i) and (iii) occur between two settlement dates, so that the buy and sell transactions never have to be paid for or delivered. If a buy-and-sell operation is performed within a few seconds, or even within a few hours, it will never have to be physically settled. One characteristic of high-frequency trading is indeed that investment positions are held for short periods of time so that net exposures are virtually zero at the end of each trading day, when clearing and settlement occur. As a result, high-frequency trading activities can virtually be performed with zero initial cash and zero initial securities (neglecting trading fees or initial cash balances to be maintained at the exchange platform). One may thus move in and out of investment positions thousands of times a day without having either to pay for the securities it buys or to physically deliver the securities it sells. A trader who consistently ensures a zero net open exposure at the end of the trading day can perform his activities without any holding of either cash or securities in the first place.

It must be clear at this stage that the latter feature — the absence of any physical payment or delivery — exists only because of the delayed settlement of all trades. If trades were to be cleared and settled in real time, or in approximately real time, then high-frequency trading would essentially disappear as it would become impossible to trade without virtually any cash or securities initial endowment. What would remain would eventually be buy-and-sell trades that are executed “quickly,” but not high-frequency trades. In order to further understand high-frequency trading, the legal consequences of delayed settlement have to be clearly grasped.

Spot vs. Futures and the Status of Financial Trades

Given delayed settlement, can trades on financial markets be regarded as spot transactions? A clear understanding of the distinction between spot and future transactions is of utmost importance for my purposes, as each of these transactions implies different consequences regarding the property rights at stake. What is usually referred to as a spot transaction is a transaction where both (i) the agreement between two parties on prices and quantities and (ii) the payment on one side, the delivery of the agreed-upon goods on the other side (or clearing and settlement) occur virtually at the same time, meaning that the time span between steps (i) and (ii) is insignificant for human action and for economic theory. One can see that what is crucial to the definition of a spot transaction is whether settlement is delayed or not.

The dichotomy, however, is not as clear-cut as it seems. Strictly speaking, agreement on prices and quantities on one side, and payment and delivery on the other side, are very unlikely to occur at the exact same time in everyday exchanges. Think of a baker who gives a piece of bread to a customer and receives cash only a few seconds after both parties agreed on prices and quantities. Clearly, considering physical time, there is a lag between the agreement between the parties and the process of payment and delivery. Does this imply that this transaction should not be considered as spot but as future? Considering physical constraints, what lag is low enough so that a transaction can be considered spot and not future? One hour? Ten seconds? One microsecond? Phrased this way, the question is misleading and the distinction between spot and future transactions has to be rephrased. The relevant time to be considered is not the physical time but the time of human action. More precisely, one is faced with the problem of continuums in human action and economic behavior. Rothbard (2001, pp. 264–65) argues:

The human being cannot see the infinitely small step; it therefore has no meaning to him and no relevance to his action. Thus, if one ounce of a good is the smallest unit that human beings will bother distinguishing, then the ounce is the basic unit. … If it is a matter of indifference for a man whether he uses 5.1 or 5.2 oz. of butter, for example, because the unit is too small for him to take into consideration, then there will be no occasion for him to act on this alternative.

Similarly, if the lag between the time a trade is agreed and the time it is paid for and delivered has no relevance for human action, then it does not make sense to label as future a transaction where such lag is, say, of 10 seconds. Asserting that it is irrelevant for human action means that the buyer of the agreed-upon good does not and cannot engage in any other transaction or operation involving property rights on the good between the time prices and quantities are decided upon and the time payment and delivery take place. For example, the good bought cannot be pledged as collateral once its purchase is agreed but before it has actually been received. What fundamentally distinguishes a spot from a future transaction is not the physical time lag that virtually always exists (even if very short) between the time a trade is agreed and the time it is paid for and delivered, but whether this time lag is relevant and meaningful for human action. A similar argument has recently been made by Bagus and Howden (2012), who distinguish between demand and term deposits in the debate on fractional reserve banking.

Consider a trading platform with a low level of computerized automation, a relatively low speed of order execution (as compared to present-day speeds) and an end-of-day clearing and settlement. This is roughly akin to what used to exist about fifteen years ago before the tremendous technological improvements underwent by trading platforms. On such an exchange, a lag between clearing and settlement exists but it is essentially irrelevant for human action, as it cannot be exploited — or possibly very marginally. Thus, everyday transactions on such a platform can, without any major theoretical difficulty,5 be treated legally and conceptually as spot.

The whole picture changes with technological improvements when high-frequency trading arises, i.e., when the lag between the time trades are agreed upon and the time they are paid for and settled can be meaningfully exploited. More precisely, a security that has been bought at some point during the day can then be re-sold before being first physically received. Faced with the above-outlined continuum problem, I explained that the distinction between spot and future transactions is to be expressed not in terms of the physical time between agreement and settlement but in terms of time meaningful for human action. Therefore, if high-frequency trades are to be understood as trades that are agreed upon but never paid for and delivered, they can no longer be understood as spot transactions and can conceptually be defined more meaningfully as future transactions. Future transactions differ from spot transactions in that they are agreed in the present but paid for and delivered at a future date, so that the time lag between the agreement on prices and quantities on one side, and the clearing and settlement on the other, is no longer irrelevant for economic and legal theory. In terms of property rights, spot and future transactions are different in esse. Spot transactions are the exchange of property rights over present goods, whereas future transactions are the exchange of claims on property rights on future goods.

If it is clear that high-frequency trades are to be considered as futures, what about trading positions that are kept open until the settlement date, i.e., transactions that will indeed be paid for and delivered? An important issue to highlight is that nothing makes it possible to distinguish ex ante a high-frequency trade from any other trade. When a buy or sell order is executed on the market (“execution” here referring not to the fact that a trade is paid for and delivered, but merely to the fact that a buyer is matched with a seller, i.e., that an agreement on prices and quantities is reached), nothing makes it possible to identify trades of two different types as there cannot exist prescience, at least for an external observer, about whether the position will be liquidated or not before the settlement date. All trades are potentially high-frequency trades ex ante. When there is no real-time settlement, all trades must therefore be regarded as futures in the first place, so as to account for the institutional lag between the time of order execution and the time of clearing and settlement. Indeed, the possibility that a particular trade be high-frequency always exists before the settlement time. In this context, trading positions that are left open over at least one settlement date can be considered similar to future contracts that are kept until maturity, whereas trading positions that are liquidated before settlement date are akin to future contracts that are never delivered.

Legal Treatment and Consequences for Property Rights

All transactions that are usually regarded as spot in economic analysis have been shown to be better understood as futures. Moreover, I explained how different are the implications of spot and future transactions in terms of property rights. Following the above analysis, one needs now to investigate how various legal or contractual arrangements may result or not in the over-issuance of property rights or in the ex nihilo creation of fiat money. Can one think of cases where such over-issuances from high-frequency trades exist because of the lag between the time trades are agreed and the time they are cleared and settled?

First, if all trades on financial markets are to be seen as futures, it must be emphasized that future transactions do not entail any over-issuance of property rights. When one sells at some date a security to be delivered in the future, it does not matter at all whether he actually owns the security in the first place. To understand this, the distinction between a present good and a future good must be restated. What is exchanged in a future transaction is a claim on a future good against a claim on future money. One must emphasize that only claims are exchanged, so that no property rights on present money or securities are exchanged (or involved in any way). Therefore a future transaction, if properly dealt with contractually and legally, is not and cannot imply any over-issuance of property rights. The only point in time where property rights on actual physical securities and on money matter is at the maturity date, i.e., when the future transaction has to be settled. The same reasoning applies for any trade (including high-frequency trades) correctly understood as a future trade. When a security “is bought” during a trading session, what is actually bought is a claim on a future security to be delivered at the settlement time (say, the end of the trading day). Similarly, what is sold in such a transaction is not present money but a claim on future money. If all trades on financial markets are to be treated legally and contractually as future transactions in this precise sense, then high frequency trading does not imply any over-issuance of property rights. A high-frequency trader would then be perfectly akin6 to a trader on futures markets who buys and sells contracts on oil, currencies or whatever securities but consistently unwinds his positions before the maturity date (i.e., never gets delivered with the underlying assets nor pays for any of these assets). Such traders consistently trade claims on future goods but never wait for the maturity of the future contract. This cannot lead to the over-issuance of property rights. In such a case, it is likely beneficial to market liquidity, similar to dealers in futures markets providing liquidity to end-user investors.

Alternative theoretical cases shall nevertheless be considered. Up to now, I have explained without further explanation that high frequency trading does not imply the over-issuance of property rights if trades are “treated legally and contractually as future transactions.” Such a proviso is of the utmost importance. Confusion may indeed come from the fact that what has been here described as future transactions is usually, in textbook explanations of the phenomenon, described as spot transactions. What if trades that are factually futures (as they are paid for and delivered only at an end-of-day settlement date) were to be treated legally and contractually as spot? Or, in other terms, what if an inconsistency in the legal framework exists, so that delayed settlement is the norm for transactions legally treated as spot? Once again, I shall make clear that the issue whether trades are treated as future or as spot under various legal systems or jurisdictions is complex and is not discussed in the present chapter, as my focus is on economic theory only.

In this case, a high-frequency trader buying a security during the day (to be delivered at the end of the trading day) could possibly engage in other operations involving property rights on a present security — not only claims on property rights on future securities — for example by pledging this security as collateral. Until either the settlement date or the date the position is liquidated, there would be two seemingly legitimate owners of the exact same security. This case would clearly result in an over-issuance of property rights that are not backed by actual physical securities. This is reminiscent of “circulation credit” or “inflation” in Mises’s sense (Mises 1981; Salerno 2000). Similarly, assume that a seller is able to use intraday the cash he is supposed to be delivered only at the settlement date — for example to repay a maturing debt — then such cash must be considered as ex nihilo created fiat money, as no one renounced yet to this quantity of money in the present. Once again, this would merely be an over-issuance of fiat money, which may have serious implications for liquidity risk in payment and settlement systems in a stressed environment.

Conclusion

This chapter provided a theoretical examination of high-frequency trading, focusing on whether it creates either additional property rights that are not backed by physical securities or ex nihilo created money. This is likely to occur as high-frequency traders can buy and sell large amounts of securities without virtually any cash or securities endowment in the first place. One key feature for a theoretical understanding of high-frequency trading is that it exploits the lag between the time trades are agreed and the time they are paid for and settled. In turn, high-frequency trading as it is currently practiced would essentially disappear if clearing and settlement were to be implemented in real time.

Whereas the time lag between the execution of a trade (i.e., the matching of a buyer and a seller) and its settlement has long been virtually irrelevant for human action as it could not be exploited — or only to a very limited extent — the advent of electronic trading platforms and of computerized trading algorithms enabled exploiting this lag to a greater extent. What used to be considered as spot transactions without any major conceptual difficulty can no longer fit the stylized description of a spot transaction, i.e., a transaction where payment and delivery occur virtually at the same time as the agreement on prices and quantities. Given that powerful computer techniques enable exploiting smaller and smaller lags (nowadays a few microseconds), the dichotomy between spot and future transactions has to be re-thought. Faced with the continuum problem, I argue that the distinctive criterion which ultimately matters is not the physical time lag that almost necessarily exists between trade agreement and delivery, but whether this lag is meaningful for human action — or, eventually, for algorithms executing models designed by humans. In that regard, all transactions usually regarded as spot have to be treated conceptually as futures with the advent of high-frequency trading techniques (of course, as long as the institutional lag between trade execution and delivery is maintained).

Turning to a legal analysis of high-frequency trading, I show that — in a system where settlement is delayed — the issue whether an over-issuance of property rights exists ultimately depends on whether it is treated legally as spot or future. If high-frequency trades are properly dealt with as futures — i.e., not as an exchange of property rights on goods, but as claims on property rights on goods — then no such consequences follow. This implies, for example, that traded securities cannot be pledged as collateral before they are physically delivered. On the contrary, if high-frequency trades are treated legally, contractually or factually as spot, then there exists over-issuance of property rights, even though it is for short time periods. This gives rise to liquidity risk in payment and settlement systems.

Following the above analysis, two research directions are to be outlined for future work. First, I set a theoretical framework indicating under which legal arrangements high-frequency trading may or not lead to the over-issuance of property rights. A survey of the existing legal frameworks in the United States or in Europe would be highly valuable as a complement. Second, from a theoretical perspective, the framework set out above could be extended to the study of another controversial market practice, namely naked short-selling. Naked short-selling occurs when a security is shorted before being first borrowed or located. A legal issue therefore is whether it is fraudulent in that one is selling something he does not own in the first place. This practice could be fruitfully analyzed not as the shorting of a security but as the shorting of a claim on a security, therefore as a future.

References

Bagus, Philipp, and David Howden. 2012. “The Continuing Continuum Problem of Deposits and Loans.” Journal of Business Ethics 106(3): 295–300.

Bank of International Settlements. (1997). Real-time gross settlement systems. Basle.

Brogaard, J. 2010. “High-Fraquency Trading and its Impact on Market Quality.” Northwestern University Working Paper.

——. 2011. “High-Frequency Trading and Volatility.” Northwestern University Working Paper.

Gomber, P., Arndt, B., Lutat, M., Uhle, T. 2011. High-Frequency Trading. Goethe Universität.

Hasbrouck, J., Saar, G. 2010. “Low-Latency Trading.” NYU Working Paper.

Huerta de Soto, Jesús. 2011. Money, Bank Credit and Business Cycles. Auburn, Ala.: Mises Institute.

Jarnecic, E., Snape, M. 2014. “The Provision of Liquidity by High-Frequency Participants.” Financial Review 49(2): 371–94.

Mises, Ludwig von. 1981. The Theory of Money and Credit. Indianapolis: Liberty Classics.

——. 1996. Human Action. Fox & Wilkes, San Francisco.

Rothbard, Murray N. 2001. Man, Economy and State. Auburn, Ala.: Mises Institute.

Salerno, Joseph T. 1994. “Ludwig von Mises’s Monetary Theory in Light of Modern Monetary Thought.” Review of Austrian Economics 8(1): 71–115.

——. 2000. “Inflation and Money: A Reply to Timberlake.” Money, Sound and Unsound, chap. 17. Auburn, Ala.: Mises Institute, 2010.

SEC (Securities and Exchange Commission). 2010. Findings Regarding the Market Events of May 6, 2010. Staff report.

Swinburne, K. 2010. Trading in Financial Instruments: Dark Pools and HFT. Brussels: Report to the European Commission.

  • *Guillaume Vuillemey is a PhD student in economics at Sciences Po, Department of Economics, Paris, France. I was a summer research fellow at the Mises Institute in 2009, under the guidance of Professor Joseph Salerno.
  • 1On the extent of failures to deliver in the United States, see SEC Fails-to-Deliver Data.
  • 2Another issue regarding high-frequency, which has been less dealt with in the literature, is the extent to which it is akin to insider trading, as some high-frequency traders benefit from their technological superiority to get market information (on incoming buy and sell orders especially) ahead of other market participants. This issue is not addressed in the present chapter.
  • 3The most prominent example of so-called “flash crash” occurred on May 6, 2010, when the Dow Jones Industrial Average plunged by about 9 percent before recovering in a few minutes. High-frequency trading algorithms have been shown to play a role in the amplification of the drop (see SEC, 2010).
  • 4A comprehensive overview of RTGS payment systems is provided by the Bank of International Settlements (1997).
  • 5In a world where the automation of stock exchanges through computer systems is low or inexistent, i.e., where high-frequency trading or multiple intraday transactions on the same security are virtually not possible, treating as spot a transaction that is technically future (with a maturity of a few hours up to 24 hours) may only matter in case of bankruptcy — for example, if bankruptcy is declared between the time a trade was agreed and the time it was supposed to be paid for and delivered.
  • 6One slight difference is that one party usually has to pay a present premium in order to enter a future transaction. This, however, is not a necessary element of a future contract. The only payment that a high-frequency trader has to make — like any other trader — is the trading fee to the exchange platform.