High-frequency trading, relative tick size, and expected volatility

Research proposal for fulfilmentof the requirements for Doctor of Philosophy in Finance

School of Economics and FinanceMassey University

Khairul Zharif Zaharudin

26 March 2018

Supervisors:Professor Martin Young

Dr Wendy Hsu

CHAPTER ONE: INTRODUCTION.................................................................................................41.1 Introduction.........................................................................................................................4

1.2 Overview of the essays........................................................................................................6CHAPTER TWO: ESSAY ONE.........................................................................................................8

2.1 Chapter overview................................................................................................................82.2 Defining HFT......................................................................................................................8

2.3 Definition of terms............................................................................................................132.4 HFT mechanics and strategy.............................................................................................15

2.5 Beneficial HFT strategies..................................................................................................182.5.1 Market-making.....................................................................................................18

2.5.2 Statistical arbitrage...............................................................................................192.5.3 Directional trading................................................................................................20

2.6 Harmful HFT strategies.....................................................................................................212.6.1 Front-running, order-anticipation, and quote-matching.......................................21

2.6.2 Spoofing and layering..........................................................................................222.6.3 Quote-stuffing......................................................................................................24

2.7 The effect of HFT on market quality................................................................................242.8 Controversies on HFT.......................................................................................................30

2.8.1 The Flash Crash of May 6, 2010..........................................................................302.8.2 HFT arms race and welfare issues........................................................................33

2.8.3 Market-making obligations..................................................................................34CHAPTER THREE: ESSAY TWO..................................................................................................36

3.1 Chapter overview..............................................................................................................363.2 Introduction.......................................................................................................................36

3.3 Research objectives...........................................................................................................423.4 Hypotheses development..................................................................................................42

3.5 Expected contribution of the Study...................................................................................443.6 Methodology.....................................................................................................................45

3.6.1 Data and Sample...................................................................................................453.6.2 Measures of HFT activity.....................................................................................46

3.6.3 Method to address RO1........................................................................................483.6.4 Method to address RO2........................................................................................49

3.6.5 Method to address RO3........................................................................................50CHAPTER FOUR: ESSAY THREE................................................................................................51

4.1 Chapter overview..............................................................................................................514.2 Introduction.......................................................................................................................51

4.3 Research objectives...........................................................................................................534.4 Hypotheses development..................................................................................................53

4.5 Expected contribution of the Study...................................................................................54

2

4.6 Methodology.....................................................................................................................544.6.1 Data, Sample, and HFT measures........................................................................54

4.6.2 Event selection.....................................................................................................554.6.3 Method to address RO1........................................................................................58

4.6.4 Method to address RO2........................................................................................584.6.5 Method to address RO3........................................................................................58

BIBLIOGRAPHY................................................................................................................................59PROPOSED TIMELINE FOR THE COMPLETION OF DISSERTATION...............................65

3

CHAPTER ONE:

INTRODUCTION

1.1 Introduction

Technological advancement has shaped the financial world. Prior to the invention of the telegraph in

1844 and the telephone in 1876, communication in securities markets had been primitive – using

humans and carrier pigeons to transmit information across the markets (Markham, 2002). For nearly a

century, telegraph and telephone are used as the main channel for financial communication – data is

received via telegraphic stock ticker, and orders are transmitted via phone calls. However, in recent

years, fiber-optic cables and microwave towers are used as the medium to transfer trading

information, traveling at lightning speed. A group of traders, armed with complex algorithms, are

willing to spend a large amount of money to gain access to these state-of-the-art facilities and pay to

collocate their server within stock exchanges, as those services give them the speed advantage that

they need for their trading strategies that banks on being the fastest. In addition, they hire

mathematicians and statisticians to work as quantitative analysts or “quants”, to develop the various

trading algorithms. This unique group of traders is commonly referred to as “high-frequency traders”,

or HFT in short.

In the U.S., HFT’s market share in total equity trading peaked at around 60% in 2009, from

around 20% in 2005. The percentage gradually decrease to approximately 50% in 2013 and has been

stable until 2016 (Avramovic, Lin, & Krishnan, 2017). In Europe, HFT’s contribution to total equity

trading was almost 0% back in 2005, before reaching its highest point at around 40% in 2010. The

value slightly decreased since then, settling at around 35% in 2014 (Kaya, 2016). As for Australia,

HFT accounts for approximately 27% of all equity market turnover in S&P/ASX 200 securities, based

on an estimate conducted over a nine-month period, from January to September 2012 (ASIC, 2013).

The figure remains stable even after three years, until March 2015. However, ASIC (2015) notes that

there is more concentration in the HFT-driven volume – 10 largest HFT account for 21% of all trading

turnover in 2015, compared to 17% three years earlier.

SEC (2010, 2014) states that there is no standard or clear definition on HFT, and thus,

regulators, researchers, and market participants have different ways to describe HFT. The absence of a

universal definition of HFT also makes it difficult to classify and identify HFT, which might lead to

other problems such as inaccurate estimation of HFT's market shares, and inability to estimate the

reach and influence of HFT in a market (AFM, 2010). Regardless, even with an accurate definition of

4

HFT, it is still not possible to properly distinguish HFT from other forms of algorithmic trading (AT).

To properly the problem, AFM (2010) suggest that trading platforms should take the initiative to

catalog and estimate the market share of the various trading strategies that employed AT, which is

currently impossible. In a similar note, NASDAQ has taken an initiative to identify the firms

submitting orders using its access to order-level information on its market and identified 26 of the

firms as HFT. This dataset was supplied to and used extensively by academic researchers for the past

decade, attempting to understand the new phenomena.

HFT can be categorized into three groups based on the strategy that they use, which are (1)

market-making; (2) statistical arbitrage; and (3) directional trading. HFT market-maker holds

positions on both sides of the order book, in which limit buy (sell) orders are placed just below

(above) the market price, providing liquidity to the market. HFT market-maker is also exposed to

typical risk associated with market-making activities, i.e. inventory risk and adverse-selection risk

(Aldridge, 2013). To compensate the HFT for the risk, as well as attracting more trading volume,

some electronic exchanges use maker-taker pricing model to price their order-matching service

(Harris, 2015). Statistical arbitrage HFT formulates their strategy to make a profit from temporary

deviations between securities that have statistically significant relationships. HFT that use this

strategy will hunt for temporary price divergence that creates profitable windows and exploits them

before the phenomenon disappears – i.e. when their relative price converged again (Moosa & Ramiah,

2015). Directional strategies are based on the theory that the price movement has directions and they

are predictable, which might be following a trend (momentum strategies) or reversal of a trend (mean

reversion strategies). HFT that employ directional trading strategies are time-sensitive, as they need to

forecast and exploit the opportunity that may arise at any time (Aldridge, 2013).

In general, empirical studies on HFT and automated trading find that their activities have a

positive influence on the market quality – evidenced by the reduction in bid-ask spreads, greater

market liquidity, and increase stock prices efficiency (Jones, 2013). Hasbrouck and Saar (2013) study

the effect of HFT on market quality using the NASDAQ HFT dataset and find that an increase in HFT

activities reduce quoted spreads, reduce price impact, increase depth, and lowers short-term volatility.

HFT activity is also claimed to promote liquidity through rapid price adjustments, allowing for

narrower bid-ask spreads within a market, strengthening the inter-market linkage and activity and

lowering the cost of intermediation (Goldstein, Kumar, & Graves, 2014; Jones, 2013). In addition,

since the market is not inherently efficient, trading activities by informed traders will move the stock

prices towards efficiency, either through market or limit orders – which will incorporate their private

information about a security on its price (Cao, Hansch, & Wang, 2009; Cooper, Davis, & Vliet, 2016).

Regardless, even with the many empirical studies support HFT participation due to their positive

effects on the market, the possibility that HFT, in theory, may harm the market through their speed

advantage cannot be ruled out (Manahov, Hudson, & Viktor, 2014).

5

1.2 Overview of the essays

The first essay is a survey of literature on HFT, which cover the various HFT definitions from

both regulatory and academic perspectives. The essay then presents how HFT works, and what makes

HFT different than other groups of investors, and proceeds with a discussion on beneficial HFT

strategies (e.g. market-making, directional trading, and statistical arbitrage) and harmful HFT

strategies (e.g. front-running, spoofing, and quote-stuffing). The essay continues with an argument on

the effects of HFT activities on market quality (e.g. liquidity, price discovery, transaction cost, and

volatility), and wrapped with a discussion on several critical issues associated with HFT (the Flash

Crash, the arms race, and market-making obligation).

The second and third essays are empirical research on HFT, using order-level data time-

stamped at milliseconds, provided by Securities Industry Research Centre of Asia-Pacific (SIRCA).

For both essays, the study will use information from securities listed on the Australian equity market

as a sample. There are two reasons for the selection: (1) most of the studies on HFT have been using

information from the U.S. or European markets, and there are only several academic papers on HFT

that use data from Australian market; (2) the Australian market has different market microstructure

design than the U.S. market such as different tick size structure, smaller minimum trading unit, less

market fragmentation, and different stocks composition, which in theory, might directly affect HFT

strategy; and (3) the Australian equity market is a pure order-driven market, while the U.S. market is a

quote-driven market that uses brokers and specialists as market-makers. Since there is no dedicated

market-maker in the former market, HFT might face less competition to make the market, which

might affect their strategy, profit, and market share (as a percentage of total equity trading volume) in

Australia.

The second essay will investigate the effect of relative tick size on HFT activity. Angel (1997,

2012) asserts that that tick size preserves the price and time priority in an order book which

incentivizes traders to supply liquidity by posting limit orders, and it creates a floor for the quoted

bid-ask spread which motivates dealers to make markets. Stocks with smaller tick size will have

narrower bid-ask spreads, and thus, smaller minimum trading costs, which is favorable for HFT

(Comerton-Forde, 2012; Harris, 1994). Relative tick size, on one hand, represents HFT’s profit

potential from one tick of price movement. O’Hara, Saar, and Zhong (2016) find evidence that HFT

tends to leave their orders longer, trade more aggressively, and have higher profit margins in stocks

with larger relative tick size, indicating that different relative tick size may influence HFT’s

willingness to participate. The study has identified the following objectives for Essay Two: (1) to

determine the effect of tick size borders on HFT’s activity, using stocks priced within the proximity of

the borders; (2) to determine the effect of crossing tick size borders on HFT’s activity; and (3) to

6

determine the effect of relative tick size on HFT’s activity, using stocks within a similar tick size

structure.

The third essay examines how HFT activity is influenced by the different level of expected

volatility in the market, measured by the VIX index. The index, which is also known as “investor fear

gauge”, represents the expected stock market volatility over the next 30 calendar days. The index

serves as a benchmark for short-term market volatility, as it is implied by the current prices of options

of its underlying market. In Australia, the VIX index tracks the S&P/ASX 200 (RIX: AXJO) and

known as S&P/ASX 200 VIX (RIC: AXVI). Market-making is one of the strategies commonly used

by HFT. However, they lack the commitment to make market, and thus, raise the issue of whether

they should be obliged to stay active in volatile markets (Goldstein et al., 2014). In practice, the level

of volatility may influence traders’ risk management policies and their decision to supply liquidity in

the market (Hagstromer & Norden, 2013). Jarnecic and Snape’s (2014) evidence shows that HFT is

found to participate in the best quote when price volatility is high, which provide additional liquidity

in the limit order book, and thus offers valuable service during periods of market uncertainty. The

objectives for Essay Three are as follow: (1) to examine the effect of different level of expected

volatility on HFT activity and preference; (2) to determine the effect of HFT activity on liquidity

during the period of low and high level of expected volatility; and (3) to investigate whether there is a

causal relationship between HFT activity and transaction costs in different level of expected volatility.

7

CHAPTER TWO:

ESSAY ONE

A SURVEY OF LITERATURE ON HIGH-FREQUENCY TRADING

2.1 Chapter overview

This chapter presents a survey that highlights the key theoretical and empirical research papers on

high-frequency trading (HFT). This essay discusses the existing literature on HFT with regards to its

definitions, characteristics, types, operations, activities, and surrounding issues. Section 2.2 provides

definitions on HFT from the regulatory and academic perspective, and Section 2.3 defines the terms

commonly found in HFT related studies. Section 2.4 explains the mechanics and general strategies

applied by HFT, followed by in-depth discussions on beneficial and detrimental HFT strategies in

Section 2.5 and Section 2.6 respectively. The effect of HFT on market quality is discussed in detail in

Section 2.7. Lastly, Section 2.8 presents selected controversies related to HFT.

2.2 Defining HFT

High-frequency trading, or HFT, as noted by the U.S. Securities and Exchange Commission (SEC)

has no clear definition (SEC, 2010, 2014). As there is no standard definition of HFT to date,

regulators, researchers, and market participants have different ways to describe HFT. The term “high-

frequency trading” is typically associated with "trading that utilizes computer technology", "the use of

technology to execute orders", "electronic trading", “electronic markets", or “automated trading”.

While the terms are indeed closely related to HFT, they are not the same thing, and only portray an

incomplete picture of HFT.

The absence of a unanimous definition of HFT also makes classification difficult (AFM,

2010), which leads to other problems such as inaccurate estimation of HFT' market shares, and

inability to estimate the reach and influence of HFT in their markets. This lack of consensus on HFT

definition complicates research conducted in this area and contributes to the various conclusion on the

effect of HFT’s activity in the market. The inexistence of precise definition of HFT also leads to

confusion of HFT with other forms of activities, and consequently, blamed for things that have

nothing to do with it (Moosa & Ramiah, 2014).

8

The incomplete definition of who or what is HFT is a problem to both HFT-existed and HFT-

free markets alike. Financial authorities need to meticulously analyze and consider the costs and

benefits of having HFT in their market. However, before they can effectively tackle the issue, first and

foremost, they need to have a sound definition of HFT. An inaccurate definition would be too costly

to the market – any microstructural changes introduced will likely involve a huge sum of money and

may affect all class of market participants, from the smallest individual investors to the mutual fund

giants.

Zhang (2010) broadly defines HFT as all short-term trading activities by hedge funds and

other institutional traders not captured in the 13f database. Kirilenko, Kyle, Samadi, and Tuzun (2017)

describe HFT as traders with high volume and low inventory, and Baron et al. (2012) added low

overnight inventory to the list. Moosa and Ramiah (2014) define HFT based on its characteristics i.e.

data-intensive, latency-sensitive, high-volume, low-margin activity, extremely short holding periods,

and rarely held positions overnight. Other scholars define HFT as a large number of small-quantities

orders, high-speed order cancellations, and have short position-holding periods (Aldridge, 2009;

Brogaard, 2010; Gomber et al., 2011).

SEC (2010) refers to HFT as "...professional traders acting in a proprietary capacity that

engage in strategies that generate a large number of trades on a daily basis" (p. 45). SEC (2010) also

lists down five characteristics commonly attributed to HFT: (1) use of extraordinarily high-speed and

sophisticated computer programs for generating, routing, and executing orders; (2) use of co-location

services and individual data feeds offered by exchanges and others to minimize network and other

types of latencies; (3) very short time-frames for establishing and liquidating positions; (4) the

submission of numerous orders that are cancelled shortly after submission; and (5) ending the trading

day in as close to a flat position as possible. Regardless, SEC never suggests that all of the

aforementioned characteristics should be met for a firm to be categorized as HFT. By doing so, a

broader range of proprietary firms can be classified as HFT (SEC, 2014).

Netherlands' Authority for the Financial Markets (AFM) produced a report on HFT to shed

some lights on the new phenomenon. AFM (2010) defines HFT as a form of automated trading based

on mathematical algorithms that implement certain short-term trading strategies by utilizing advanced

technology, and not as a separate trading strategy by itself. The main characteristics of HFT according

to AFM (2010) are: (1) use trading strategy that involves rapid calculation and execution speeds; (2)

use sophisticated systems and efficient infrastructures; (3) use earnings model with very small profit

margins in very large volumes; (4) usually take market-neutral (non-directional) and delta neutral

(hedged) position, thus in many cases close out their positions with flat position at the end of the day;

(5) have a really short average holding period, ranging from seconds to several minutes; and (6) have

a very high order-to-transaction ratio.

9

Australian Securities and Investments Commission (ASIC) also agrees to the notion that there

is no unanimous definition of HFT. ASIC (2010) characterised HFT as a specialised form of

algorithmic trading that (1) generate large numbers of small size orders with high rate of amendment

and cancellation; (2) typically have to hold positions with very short time horizons; and (3) use

variety trading strategies, but the most common strategy is electronic liquidity provision. HFT also

employ high-speed, low-latency technology infrastructures which requires them to: (1) process direct

market feeds to have access to the fastest market information available; (2) co-locate their servers in

the data centres with the exchange market’s matching engine to reduce access times; (3) develop their

own sophisticated trading strategies to trade on a short-term basis; and (4) typically end the trading

day with no carry-over positions that use capital (ASIC, 2010).

In 2010, the Committee of European Securities Regulators (CESR) conducted a survey to call

for evidence on micro-structural issues of the European equity markets. The survey is intended to

assess the impact of technology-driven developments such as HFT, sponsored access, and co-location

services that have intensified following the implementation of the Markets in Financial Instruments

Directive (MiFID) on November 1, 2007. In the survey, CESR (2010) describes HFT: (1) as a form

of automated trading that uses sophisticated computers and IT programs; (2) execute trades in matters

of milliseconds; (3) hold new equity positions possibly down to a “sub-second”; (4) ends their day

with a flat position; (5) use their own capital and do not act on behalf of clients; and (6) and employ

trading strategies that are generally geared towards extracting very small margins from hyper fast

speed trading. In a response to the survey, London Stock Exchange Group (LSE) refers to HFT as a

wide variety of different strategies utilizing ultra-fast technology to conduct electronic market-making

and/or to seek arbitrage opportunities. LSE (2010) also noted that HFT is very fast and requires low-

latency connection to exchanges' trading systems.

The introduction of Directive 2014/65/EU of the European Parliament and of The Council of

May 15, 2014, on markets in financial instruments, sees the amendment of the MiFID. The new

directive (commonly referred to as MiFID II) provides a legal definition of HFT. Article 4(1)(40) of

MiFID II describes a HFT technique as “an algorithmic trading technique characterised by: (a)

infrastructure intended to minimise network and other types of latencies, including at least one of the

following facilities for algorithmic order entry: co-location, proximity hosting or high-speed direct

electronic access; (b) system determination of order initiation, generation, routing or execution

without human intervention for individual trades or orders; and (c) high message intraday rates which

constitute orders, quotes or cancellations”.

Brogaard, Hendershott, and Riordan (2014) state “one of the difficulties in empirically

studying HFT is the availability of data identifying HFT. Markets and regulators are the only sources

of these and HFT and other traders often oppose releasing identifying data” (p. 2270). An example of

10

such dataset is the one provided by NASDAQ, which covers 120 U.S. equities over the 2008-2009

period, timestamped to the milliseconds. NASDAQ used its access to order-level information on

its market to identify the firms submitting orders, and manually classified 26 of the firms as HFT.

The dataset also categorizes whether the execution is either aggressive (liquidity taking) or

passive (liquidity providing), and further grouped them into either HFT or non-HFT, resulting in

four types of order execution: “HH”: HFT take liquidity from other HFT; “HN”: HFT take liquidity

from non-HFT; “NH”: non-HFT take liquidity from HFT; and “NN”: non-HFT take liquidity from

other non-HFT.

Even so, the data has its limitations. NASDAQ cannot identify all HFT in the market and

possibly has excluded HFT firms that also act as brokers while engaging in proprietary lower-

frequency trading strategies (e.g. Goldman Sachs, Morgan Stanley). Thus, the orders from HFT firms

routed through those large integrated firms might be excluded as well (Brogaard et al., 2014). In

similar note, Hagstromer and Norden (2013) assert that the use of mediation trading services such as

sponsored access and/or trading desks of banks, which consist a mixture of HFT and non-HFT, makes

it difficult to distinguish the origins of the trading activity, and to interpret the results obtained from

this group. According to AFM (2010), even with an accurate definition of HFT, trading platforms

would still be unable to properly distinguish HFT from other forms of AT. To do so, the trading

platforms need to establish market share of the various trading strategies that employed AT, in which

based on today's technology, is not yet possible.

Albeit not having a conclusive definition of HFT, certain characteristics distinguishing HFT

from other forms of trading can be specified. In general, the majority of the regulatory body agree that

HFT is: (1) a specialised form of algorithmic trading; (2) use high-speed, sophisticated computer

programs and systems; (3) have a very high order-to-transaction ratio; (4) have extremely short

average holding periods; and (5) end their trading day with flat positions. Table 1 summarizes the

definition of HFT from regulatory perspectives:

11

Table 2.1: Summary of HFT definition from regulatory perspectives

Definition/characteristics of HFTSEC

(2010)AFM(2010)

ASIC(2010)

CESR(2010)

MiFID II(2014)

Use efficient, high-speed, low-latency infrastructures such as co-location services

Use extraordinarily high-speed and sophisticated computer programs and systems

Have a very high order-to-transaction ratio, generated by large numbers of small size orders with high rate of quotes, amendments, and cancellations

Have a really short average holding period, possibly down to a “sub-second", and execute trades in matters of milliseconds

Close out their positions at the end of the day with no carry-over positions (flat positions)

A specialized form of automated trading based on mathematical algorithms/ an algorithmic trading technique

Use earnings model that is generally geared towards extracting very small margins from hyper-fast speed trading

Use a trading strategy that involves rapid calculation and execution speeds

Use a variety of sophisticated short-term trading strategies, but the most common strategy is electronic liquidity provision

Use their own capital and do not act on behalf of clients

Use system for the determination of order initiation, generation, routing or execution without human intervention for individual trades or orders

12

2.3 Definition of terms

The following are the terms commonly found in HFT-related studies. The definition is taken directly

from either academic articles, regulatory bodies, financial exchanges, or online definition.

Table 2.2: Definition of terms commonly found in HFT studies

Terms Definitions

Adverse selection In exchange market trading, there is a risk that the person you trade with is more informed than you are. If this is so, or you fear that it is so, you may respond by becoming more risk-averse, reducing the price at which you are willing to buy or increasing the price at which you are willing to sell. The consequence of such adverse selection is a widening of spreads. In an extreme, investors might decline to participate in trades or to not post limit orders. If many participants in a market act according to the principles of adverse selection, trading becomes encumbered and inefficient (ASIC, 2010, p. 15).

Co-location A practice by securities firms to house their computers in the same location as electronic exchange computers. By doing so, firms can move closer to a market’s central computer, increase trading speed, and gain an edge in highly competitive and fast-moving securities markets (Garvey & Wu, 2010, p. 368).

Direct market access

An arrangement through which an investment firm that is a member/participant or user of a trading platform permits specified clients (including eligible counterparties) to transmit orders electronically to the investment firm’s internal electronic trading systems for automatic onward transmission under the investment firm’s trading ID to a specified trading platform (ESMA, 2011, p. 32).

Effective spread Effective spreads measure the difference between a trade’s execution price and the pre-trade midpoint. Effective spreads compensate liquidity providers for adverse selection costs when trading with informed traders and are expected to contain an additional component that covers inventory risk, order processing costs, and market-maker rents (Carrion, 2013, p. 696). Effective spread is an ex-post measure of liquidity and is calculated as the difference between the trade execution price and the midpoint of the best bid and offer at trade time (SEC, 2014, p. 23). The effective spread can be decomposed into the realized spread, i.e. liquidity suppliers’ revenue, and the price impact after time x (Zhang & Riordan, 2011).

Latency The time that elapses between an investor making a trading decision and the execution and confirmation of the desired trade (Garvey & Wu, 2010, p. 369). An expression of how much time it takes for data to get from one point to another (ASIC, 2010, p. 109).

13

Limit order An order to buy a stock at or below a specified price, or to sell a stock at or above a specified price. For instance, you could tell a broker "buy me 100 shares of XYZ Corp at $8 or less" or "sell 100 shares of XYZ at $10 or better" The customer specifies a price, and the order can be executed only if the market reaches or betters that price. A conditional trading order designed to avoid the danger of adverse unexpected price changes (NASDAQ glossary).

Market order Used in the context of general equities. Order to buy or sell a stated amount of a security at the most advantageous price obtainable after the order is represented in the trading crowd. You cannot specify special restrictions such as all or none (AON) or good 'til canceled order (GTC) on market orders (NASDAQ glossary).

Price impact The price impact is the effective spread minus the realized spread and measures the information content of a trade. It approximates the permanent impact of a trade under the assumption that information impacts are permanent and realized at the five-minute horizon, whereas other effects, such as inventory and explicit trading costs, are temporary (Riordan & Storkenmaier, 2012, p. 424). If price impact (an adverse price move from the standpoint of the liquidity supplier) exceeds the effective spread, the realized spread will be negative (SEC, 2014, p. 23).

Proprietary trading Proprietary trading occurs when a firm or bank invests for its own direct gain instead of earning commission dollars by trading on behalf of its clients. This type of trading occurs when a firm decides to profit from the market rather than from the thin-margin commissions it makes from processing trades. Firms or banks that engage in proprietary trading believe that they have a competitive advantage that will enable them to earn excess returns (Investopedia).1

Realized spread Realized spreads is the difference between the trade price and the midpoint of the spread five minutes later (O’Hara, 2015, p. 267). Realized spread is the portion of transaction costs that can be attributed to liquidity provider revenues (Malinova & Park, 2015, p. 530). Realized spread measures the potential for a liquidity supplier to profit from a trade by liquidating the position at some specified point in the future (SEC, 2014, p. 23). Realized spread can also be viewed as the losses of the market marker to better-informed traders (Zhang & Riordan, 2011).

Sponsored access An arrangement through which an investment firm that is a member/participant or user of a trading platform permits specified clients (including eligible counterparties) to transmit orders electronically and directly to a specified trading platform under the investment firm’s trading ID without the orders being routed through the investment firm’s internal electronic trading systems (ESMA, 2011, p. 32)

1 Investopedia is wholly owned by IAC (NASDAQ: IAC), and it is the largest financial education website in the world. See https://www.investopedia.com/corp/about.aspx.

14

2.4 HFT mechanics and strategy

There is nothing new in the way HFT works. The short-term trading strategies employed by HFT has

long existed (AFM, 2010). The way HFT profit from the market, in general, is similar to other traders’

strategy. For instance, they will buy stocks at a lower price, then sell them at a higher price. For stocks

with short selling option, they have more choices ‒ they are able to make money from either bearish

or bullish stocks. For stocks with options, they could make a profit from price disparities between the

parent stocks and their option securities. They might as well assume the role of a market-maker, in

which they stand ready to buy and sell securities, and profit from the market-making spread.

Moreover, market-making HFT might receive rebates from certain trading venues for providing

liquidity to their market.

What makes HFT a unique class of investor lies in their speed; to observe for profitable

trading opportunities, to quickly process new information and execute the appropriate action, to

analysis textual context of news flow, at a much higher-frequencies and shorter time-frames than a

human being capable to. This is also an important advantage that the machines (i.e. HFT) have over

humans (Menkveld, 2014). The infrastructural and technological advantages that they possess allow

for the optimization of a wide-array of complex trading strategies, from the beneficial market-making

strategies to the harmful and devious strategies such as quote stuffing (AFM, 2010; O’Hara, 2015).

According to Angel (2014), the trading speed nowadays is so fast that it almost reached the theoretical

speed of light ‒ approximately 300,000 km/s. This superhuman speed also makes certain trading

strategies exclusive to HFT, especially the ones that rely on speed, as other types of market

participants unable to replicate such strategies (Harris, 2013), which further stressing HFT’s need for

superior calculation and execution speeds (AFM, 2010).

Hasbrouck and Saar (2013) find that some algorithms are so fast that the time it takes to

complete a trading cycle starting from the detection of a market event, analyses it, and send an order

appears to be 2‒3 milliseconds. This intense activity within the “millisecond environment” is also

where computer algorithms react to each other (Hasbrouck & Saar, 2013). O'Hara (2015) states that

order latencies are now measured in milliseconds (one-thousandth of a second), microseconds (one-

millionth of a second), and even nanoseconds (one-billionth of a second). For comparison purpose, it

takes the human eye 400–500 milliseconds to respond to visual stimuli, and human reaction times are

generally thought to be around 200 milliseconds, which in both cases is far behind the HFT's speed

(Kosinski, 2013; O'Hara, 2015). At such speeds, human traders cannot accurately follow the low-

latency activity on their trading screen, and the market dynamics that may be driven by the

interactions between algorithms (Chordia, Goyal, Lehmann, & Saar, 2013; Hasbrouck & Saar, 2013).

Due to HFT strategies depends heavily on speed, latency issues such as the speed of cross-market

15

information flow, and transmission speed across geographical locations are of their concern.

Therefore, to utilize their trading strategies, many HFT would have multiple locations across several

cities such as New York, Chicago, and London.

The earnings model for HFT consists of executing many transactions with very small profit

margins in very large volumes (AFM, 2010). Using fully automated trading strategies, HFT attempt to

identify and profit from short-term irregularities, and earn small amounts of money from every trade.

Even though the profit per trade is often as small as a few basis points only, it is amplified by the high

trading volume (Zhang, 2010). The ability to trade at low latency allows HFT to profit from the

trading environment itself, rather than from investing in financial securities (Hasbrouck & Saar,

2013). Budish, Cramton, and Shim (2015) state two common characteristics used by HFT in their

trading strategies, which are (1) often cancel their orders soon after placing them, and (2) high-ratio of

messages to completed trades.2

HFT exhibit variability in their trading strategies by documenting differences in liquidity

provision, end-of-day and maximum intra-day positions, trading revenues, etc. The variability in

strategies also translates into different sensitivities of HFT' position changes to inventory levels and to

recent price changes (Benos & Sagade, 2016). Brogaard et al. (2014) find that the direction of HFT

trading is correlated with publicly available information, such as macroeconomic news

announcements and limit-order book imbalance. They also find that HFT followed contrarian trading

strategies, evidenced by the negative correlation between HFT overall trading with past returns.

Goldstein, Kumar, and Graves (2014) state that naturally, the HFT strategies are employed by

proprietary firms, in which the majority are either broker-dealer proprietary trading desks,3 hedge

funds,4 and proprietary trading groups.5 This is only logical due to the high cost involved in

employing sophisticated technology and obtaining the big data to execute HFT strategy (Moosa &

Ramiah, 2014; Kauffman, Hu, & Ma, 2015).

Aldridge (2013) generally categorized HFT trading strategies into three groups, which are (1)

statistical arbitrage, also known as value-motivated strategies; (2) directional strategies, also known as

informed trading; and (3) market-making, also known as liquidity trading. The algorithms employed

by HFT may determine their order execution style, such as either being aggressive or passive or to

2 Regulatory bodies intend to introduce minimum resting time and impose maximum order-to-trade ratio, in which each rule is aimed to address the aforementioned characteristics respectively. 3 Proprietary trading desks is a trading unit in a firm such as banks, that trade using the firms’ own money to make profit, instead of relying on commissions from their clients. In the U.S., the Volcker Rule prohibits banks from engaging in high-risk, speculative trading activity on their own account, such as the short-term proprietary trading.4 Example of hedge funds that employ HFT strategies are Renaissance Technologies, Worldquant, DE Shaw, and Millennium.5 Example of proprietary trading groups that use HFT strategies are Getco LLC, Allston Trading LLC, Infinium Capital Management LLC, Hudson River Trading LLC, Quantlab Financial LLC.

16

send the orders in either one trade or split them into smaller trades.6 Similarly, AFM (2010) also

divides HFT strategies into market-making, statistical arbitrage, and low-latency. While the first two

groups are similar to Aldridge’s (2013), the third group classification, i.e. low-latency, has a broader

scope. AFM (2010) states that the success factor of the latter group is determined by the sheer speed

of the users, hence, creating the need to have the fastest systems and the best connection to trading

venues. Harris (2013) on the other hand, grouped HFT trading strategies into three (3) groups based

on their effect on the market. The first group, Valuable, is a group of trading strategies that are

acceptable to the market in general, such as market-making and statistical arbitrage. The other two

groups, namely Harmful and Very Harmful, are a group of trading strategies that is intolerable, with

the latter worse than the former. The strategies belong to these groups benefits the HFT at the cost of

other market participants, such as front-running and quote stuffing.

Most HFT-based strategies such as market-making promote market liquidity, while the

arbitrage strategies have a positive contribution to price discovery and market efficiency. Therefore,

the action to prevent or hamper these strategies by inadequate regulation, or imposing specific

constraints for this group of strategies, may trigger counterproductive effects to market quality.

Regardless, regulatory bodies should always combat any predatory strategies that go against market

integrity or create disruptive or confusing effects on other market participants (Gomber, Arndt, Lutat,

& Uhle, 2011). Harris (2013) highlights that the financial authorities should be meticulous in

regulating the market, to avoid from unintentionally harming friendly HFT strategies. Cooper et al.

(2016) examine regulatory efforts related to HFT, particularly on the issue of HFT’s deception in the

market. They conclude that the action to treat a deception, or even an intentional deception, as a

misconduct in a financial market, is a mistake. They outlined three (3) acceptable criteria for

algorithm trading strategies, which are; (1) the trading strategy should be prudent, in which it would

not be harmful to the market should they behave unexpectedly; (2) the trading strategy should not

block price discovery, i.e. it should not interfere with the ability of other market participants to reflect

their private information on the price; and (3) the trading strategy should not circumvent transparent

price discovery, and therefore, strategies that conceal information from being discovered, such as

using dark pools or hidden orders, should be prohibited.

6 An aggressive order is an order that is placed on the current market price, a.k.a. market order, or a limit-order with price near to the current market price. A passive order on the hand is a limit-price placed far from the current market price.

17

2.5 Beneficial HFT strategies

The following sections provide a brief a discussion on acceptable HFT trading strategies, namely

statistical arbitrage, directional trading strategies, and market-making. These strategies are deemed as

acceptable as they do not harm the market and have positive effects on market quality.

2.5.1 Statistical arbitrage

Statistical arbitrage, also commonly known as “stat arb”, is a trading strategy that is based on the

theory that two similar instruments should share similar behavior, and therefore, any short-term

divergence between their relative prices are likely to converge again. The temporary divergence is

more likely to be driven by momentary order imbalance in the market, rather than by any meaningful

fundamental change (Narang, 2013).7 This trading strategy is designed to make a profit from price

disparity, and temporary deviations of statistically significant relationships,8 while considering tens or

hundreds of stocks to utilize this strategy (Lhabitant & Gregoriou, 2015; Golub, Dupuis, & Olsen,

2013; Moosa & Ramiah, 2015). Accordingly, HFT will hunt for the opportunities that arise during the

temporary deviations period and exploit them before the phenomenon disappears (Moosa & Ramiah,

2014).

Wissner-Gross and Freer (2010) highlight the importance of minimizing information

transmission delay in modern-day securities trading. In their paper on relativistic statistical arbitrage,

they demonstrated that there exist optimal intermediate locations between trading centers that host

cointegrated securities, which minimizes transmission delays and maximizes profit potential. As

traders continue to aim at being the fastest, the importance of having optimal locations is even more

pronounced (Donner, 2010; Wissner-Gross & Freer, 2010). Regardless, Kozhan and Tham (2012 .)

argue that while competition is commonly associated with improved price discovery, competition

among arbitrageurs might inflict negative externalities on each other due to the crowding effect,

which in turn will limit efficiency.

The opportunities for statistical arbitrage might surge from long-term investors’ strategic

decision. For instance, their action to buy or to sell certain securities might create a price impact on

the securities’ price, which consequently create a ripple of price impact across the markets, especially

in correlated securities (Goldstein, Kumar, & Graves, 2014). The fastest trader who first notices such

opportunities and trades on them will make the most, if not take all, of the profits from the

7 This trading strategy is also commonly known as “pairs trading”. Although in theory it is possible to find directly comparable instruments, however, very few assets can be compared precisely with another instrument, rendering the potential benefits from this strategy to be infeasible (Narang, 2013).8 HFT might use statistical approach that measures the relationship between two or more instruments such as cointegration or correlation analysis.

18

phenomenon. Therefore, speed is essential to successfully execute this trading strategy, and HFT that

implement this strategy are willing to spend a lot to keep their technological capabilities up-to-date

(Chung & Lee, 2016; MacIntosh, 2015). This strategy plays a key role in the market in terms of

liquidity provision, as well as in price discovery and information dissemination process (Goldstein,

Kumar, & Graves, 2014). Nonetheless, Hasbrouck and Saar (2013) argue that even though HFT helps

in eliminating momentary price distortions but given that the improvement is only within millisecond

environment, the effect is deemed insubstantial.

2.5.2 Directional trading

Directional trading strategies is a group of high-frequency trading strategies based on the

identification of short-term trend or momentum, which includes event-driven strategies and short-term

price movements prediction strategies. Directional strategies are time-sensitive (Aldridge, 2013), as

they need to anticipate an intraday price movement, which involves taking un-hedged positions based

on forecasted price changes, such as exploiting the divergence between fundamental values and actual

market prices. Benos and Sagade (2016) find that HFT with neutral liquidity taking/making behavior

is trend chasers. They trade in the direction of short-term price changes, i.e. they buy when the price is

rising and sell when the price is falling, which is suggestive of momentum strategies.

As the name implies, directional strategies are based on the theory that the price movement

has directions and they are predictable, which might be following a trend (momentum strategies) or

reversal of a trend (mean reversion strategies). Under the momentum strategy, HFT will identify a

trend or a significant move, and bet that the price will continue to move in the same direction, driven

by the idea of there is a growing consensus among market participants (Narang, 2013). The mean

reversion strategy, on the other hand, is built on the notion that any deviations in price, such as a trend

or a consistent direction, may be temporary in nature. Thus, price movements do not persistently

move in one direction, and will eventually revert and bounce back (Easley, Prado, & O’Hara, 2012).

To be successful in implementing directional trading strategies, HFT needs to have superior

access to information (e.g.: information from paid-for news sources such Bloomberg) and able to

immediately assess and analyze market condition. Foucault, Hombert, and Rosu (2016) suggest that

the contribution of news trading to the directional trader’s profit increases with news informativeness,

and the fastest traders will gain the most profit. Furthermore, the competitive edge that directional

traders have from the early access to new information will not last long, as the information will soon

be available to the public. Thus, the directional traders are normally aggressive, as they use market

orders or post limit prices close to market (Aldridge, 2013).

19

2.5.3 Market-making

Market-making, in general, can be described as the placement of limit orders on both sides of the

market price, in which limit buy (sell) orders are placed just below (above) the market price, which

provides liquidity to the market. HFT market-making strategies help the market to be more efficient

and have stabilizing effects to the market as they (the HFT) provide buying power when others want

to sell and selling power when others want to buy (Angel, 2014). Despite the financial landscape has

developed so much as technology evolves, the general mechanics of market-making still hold even in

a high-frequency world. Goldstein, Kumar, and Graves (2014) states that market-making HFT uses

automated liquidity provision, a strategy which rapidly places, cancel, and replace bid (buy) and ask

(sell) limit orders, and profit from the resulting spreads. The high-frequency updating process

involves in the market-making process resulted in enormous orders volume and a high cancellation

rate of 90% or more (SEC, 2010).

Unlike HFT that uses directional trading strategies, HFT market-makers do not seek to make a

directional bet, but instead, take a position on both sides of the order book to maximize their inventory

turnover. HFT market-makers typically would turn over their inventory more than five times in a day,

which explains their high share of volume traded in the market. They also hold minimum or even zero

inventory positions at the end of a trading day. Since they have very small inventories and short

holding period, essentially they could perform their market-making activities with very low capital,

while using high-speed trading to control their position risk (Easley, Prado, & O’Hara, 2011). Benos

and Sagade (2016) find evidence that passive HFT is consistent with market-making activity, in which

they trade in the opposite direction (i.e. contrarian trading) of the most recent price changes, post limit

orders, and use aggressive trade to make quick inventory adjustments.9 Regardless, they also find that

passive HFT has a high information-to-volume ratio, suggesting that the HFT might use various

market-making strategies, rather than solely using aggressive orders to make the market.

Aldridge (2013) states that a market-maker is exposed to two types of risk once his market

limit orders are placed, which are (1) inventory risk and (2) adverse selection risk. Inventory risk is

the risk that the inventory that a market-maker is holding decline in value due to natural market

movements, while adverse selection risk is the possibility of the market-maker is trading against a

party that is better-informed about the true price of the stock. Thus, it is only natural that the market-

maker to be compensated not only for the liquidity-providing service that they provided, but also the

risk they have to bear from their role as a market-maker (Aldridge, 2013; Golub et al., 2013).

9 Benos and Sagade (2016, JFM) categorized HFT based on their liquidity taking/making behaviour. For computational details of the measure, kindly refer to their paper at https://doi.org/10.1016/j.finmar.2016.03.004

20

Some electronic exchanges use maker-taker pricing model to price their order-matching

service (Harris, 2015).10 Durbin (2010) defines the model as “a pricing policy of some exchanges

where active traders pay a fee, some of which is distributed to the associated passive trader” (p. 206).

The maker-taker pricing model is used to encourage market-making instead of market-taking activity

in the market through incentives in the form of rebates or reduced transaction costs to market-makers.

The rebate is indeed important for market-making HFT. The absence of rebate would put HFT in a

loss position (Hendershott & Riordan, 2013), and their revenue from supplying liquidity would be

negative (Brogaard, Hendershott, & Riordan, 2014), which in turn may discourage HFT’s liquidity

provision activities.

2.6 Harmful HFT strategies

The controversial strategies are the strategies that profit at the expense of others through "dirty"

means such as front-running, order anticipation, quote-matching, quote-stuffing, spoofing, and

layering. Moreover, HFT’ ability to rapidly enter and cancel orders faster than other traders makes it

difficult to identify where liquidity exists across fragmented markets and this uncertainty creates even

more profitable opportunity for HFT (O’Hara, 2015).

2.6.1 Front-running, order-anticipation, and quote-matching

Harris (2013, 2015) describe front-running as “very harmful” trading strategies, and further

categorized them into “order-anticipating” and “quote-matching” strategy. Order-anticipation works

by examining trades and quotes to detect algorithms used by traders that intend to move large orders. 11

The HFT would then trade ahead of (i.e. front-run) the incoming large orders and profit from the

anticipated direction of the price changes. This will make the price higher (lower) for incoming large

buy (sell) orders, which increase the transaction costs for traders intended to execute a large order.

HFT that apply order anticipation strategy cleverly design their algorithms to play by the book,

without violation of a duty, misappropriation of information, or other misconduct (SEC, 2010).

Regardless, the strategy that they use is “parasitic” ‒ not only it does not contribute to price discovery

or liquidity, but it also preys on other traders and jeopardize the large traders the most (Harris, 2015).

10 The maker-taker pricing model is criticized for causing distortion in the market (Angel, 2014, FR), providing unfair advantages to high-speed traders. This issue is further discussed in the Section 2.7.4: Issues on maker-taker pricing model. 11 A trader will split their large orders into “smaller packages” to conceal their private information, and reduce the impact on the market. In this aspect, this is quite similar to the reason traders use dark pools to trade their large orders.

21

Some institutional investors even claim that the order-anticipation strategy may adversely affect their

trading strategy, which impacts costs for these institutional investors (Agarwal, 2012).

Quote matching, on the other hand, make profits by posting slightly better limit order, e.g.

one-tick higher (lower) than slow traders’ limit buy (sell) orders, which gives them price-priority. In

the case of the market is moving against their position, quote-matching HFT would trade with the

slower traders’ quotes (which has become the best quotes) to minimize their loss. The problem of

quote matching is not something new to the large buy-side traders. It was an important source of

profit for exchange specialists before the era of HFT. The main difference between now and then is

the identity of the quote matchers (Harris, 2013). Unlike the order anticipation strategy that requires

high-quality pattern-recognition algorithms, the success of quote matching strategy highly depends on

HFT’s low-latency communication. Speed is crucial to quote-matchers to get their orders be the first

to fill the large orders, also to revise their unexecuted orders should the large orders are canceled or

filled before they can be matched, thus, it is dominated by the faster HFT (Harris, 2013).

Nevertheless, both strategies unnecessarily increased the large traders’ transaction costs (Chung &

Lee, 2016), and may impede the process of impounding fundamental information into the price

(Jarnecic & Snape, 2014).

Aquilina and Ysusi (2016) empirically examine HFT order anticipation activity using data

from LSE and find no evidence that HFT systematically anticipate orders sent to different venues by

non-HFT, and try to front-run the orders. However, they do find trading patterns consistent with HFT

anticipate non-HFT’ order flow when analyzing longer time periods. Regardless, the result can also

mean that the HFT able to react faster to news and other public information than non-HFT. They

conclude that “HFT appear not to anticipate near-simultaneous orders…but they could be predicting

the flow over longer time periods” (p. 26).

2.6.2 Spoofing and layering

Spoofing and layering are defined as a strategy that:

Submitting multiple orders at different prices on one side of the order book slightly away

from the touch, submitting an order to the other side of the order book (which reflects the

true intention to trade) and, following the execution of the latter, rapidly removing the

multiple initial orders from the book (ESMA (2011, p. 27).

FINRA (2012) in general describe spoofing as a form of market manipulation intended for “triggering

another market participant(s) to join or improve the NBBO, followed by canceling the non-bona fide

order and entering an order on the opposite side of the market” (para. 5). Dodd-Frank Wall Street

22

Reform and Consumer Protection Act, on the other hand, outlined a broader definition of spoofing,

defined as a disruptive practice that involves “bidding or offering with the intent to cancel the bid or

offer before execution” (p. 364), which makes it unlawful to practice such strategy.12

Spoofing is executed with the intention to attract liquidity by posting fake market or limit

orders to mislead other investors, especially algorithms specialize in tape reading (Serbera &

Paumard, 2016), by forming an illusion that the market is moving soon due to a great demand in the

order book. For example, HFT may create such situation by posting large displayed limit orders just

below the best bid price, leaving others under the impression that the price will soon move upwards.

This situation encourages other traders to quickly buy the stock by quoting the stock at a higher bid

price or even execute market orders. In the meantime, the HFT might already own the stocks

beforehand, and can now sell them at a higher price in a bigger volume, thanks to the artificially

inflated price that was driven by the fake limit buy orders.

Layering is a form of spoofing, which involves placement of a large number of fake orders on

several different price limits on one side of the order book (AFM, 2010). This creates an appearance

of changing levels of supply and demand in the affected securities (FINRA, 2012). Others may falsely

interpret this pattern as a signal of an increasing directional pressure on the price and act accordingly.

The HFT will then make a profit from the price move they have initiated and cancel the fake limit

orders before they are executed. Both spoofing and layering convey an impression that a security is

more liquid than it actually is, or suggest that the security is currently under- or overpriced (Harris,

2015). Regardless, Cooper et al. (2016) claim that spoofing and layering is just another form of

bluffing, and just like poker, bluffing is allowed. They conclude that the regulators should not treat all

deception in the financial market as a misconduct and proposed a set of criteria in deciding which

trading strategy should be regulated, and which should not.13

2.6.3 Quote-stuffing

Quote-stuffing is another form of a market manipulation strategy that utilizes HFT’s ability to

rapidly send and cancel orders. Easley et al. (2012) describe quote stuffing as a strategy that “involves

sending and canceling massive numbers of orders with the intent of taking all available bandwidth and

thereby preventing other traders from being able to submit orders” (p. 228). In similar notes, The

Government Office for Science (2012, p. 168) defines quote-stuffing as “entering large numbers of

12 Panther Energy Trading (a HFT firm) and its owner Michael J. Coscia, were charged under the Dodd-Frank Act for engaging in spoofing by “utilizing a computer algorithm that was designed to illegally place and quickly cancel bids and offers in futures contracts” (CFTC, July 22, 2013, Press release). From August 8, 2011 until August 18, 2011, the firm accumulates a staggering profit of US$1.4 million through its spoofing algorithms.13 Cooper et al. (2016, BEQ) states that an acceptable trading strategy (1) should be prudent, (2) should not block price discovery, and (3) should not circumvent transparent price discovery.

23

orders and/or cancellations/updates to orders so as to create uncertainty for other participants, slowing

down their process and to camouflage the manipulator’s own strategy”. The high rate of orders

entering and canceling involves in quote stuffing is viewed as a way to manipulate markets, and luring

other traders into making mistakes (Narang, 2013).

Unlike spoofing and layering that use limit order near the best bid and ask price, quote

stuffing involves placing large amounts of nonexecutable orders ‒ i.e. limit orders that are far from

the best quote, aimed to congest the market and slow down other competitors (Lhabitant & Gregoriou,

2015). An exchange’s network bandwidth might be congested from receiving unusually large

numbers of trade messages (e.g. rapid orders and cancellations), thus impairing other traders’ access

to the market (Angel & McCabe, 2013). The impairment leaves the slower traders with an unclear

picture of the actual market situation and affected their ability to execute trades. The faster traders on

the other hand, able to get a better understanding of what is happening in the market, allowing them to

profit at the expense of slower traders (Biais & Woolley, 2011). Since quote stuffing strategy seeks to

make a profit by preventing others from adding their private information into the security, it lacks the

criteria of an acceptable HFT strategy, and thus, should be prohibited (Cooper et al., 2016).

2.7 The effect of HFT on market quality

A large and growing body of literature has investigated the effect of HFT on market quality. The term

"market quality" in itself is broadly defined, but it is commonly associated with price discovery and

efficiency, liquidity, and volatility (e.g.: Harris, 2002; The U.K. Government Office for Science,

2012). Based on HFT's characteristics (see section 2.2 ‒ Defining HFT), it can be thought as a new

breed of intermediary, which may improve or harm the market. The issue of whether HFT is

beneficial or detrimental to the market is a still a hot topic, debated among market participants,

regulators, media, as well as academics (Menkveld, 2016). The many perspectives on HFT may have

stemmed from the lack of consensus on the mechanics of HFT, which may act as market-makers,

arbitrageurs, predators, or some combination (Carrion, 2013).

In his seminal paper, Fama (1970) states “a market in which prices always fully reflect

available information is called efficient” (p. 383). The EMH postulates that in an informationally

efficient market, security prices will adjust rapidly to the arrival of new information, and thus, the

prevailing prices reflect all existing information about the security. There are three assumptions

underlying the hypothesis: (1) an efficient market requires that a large number of profit-maximizing

participants analyse and value securities, each independently of the others; (2) new information

regarding securities comes to the market in a random fashion, and the timing of one announcement is

generally independent of others; and (3) the competition between the many profit-maximizing

24

investors to profit from the new information causes the security prices to adjust rapidly, and thus, the

impact of new information is reflected in the security prices. Thus, the price changes are hypothesized

to be independent and random and require a certain minimum amount of trading by the numerous

competing investors in making the market more efficient.

EMH asserts that the existing securities prices in an efficient market should unbiased, and able

to reflect all currently available information. Thus, should EMH holds, once an information is

publicly disclosed, it is quickly reflected in prices (Fox, Glosten, & Rauterberg, 2017), and any

mispricing and associated arbitrage opportunities should be rapidly eliminated (Goodhart & O’Hara,

1995). Furthermore, in the era of HFT, the term “immediately”, “rapidly”, or “current” need to be

refined, as their (HFT) definition and perception on these terms are very much different than ordinary

human traders. Comparatively, it takes 400–500 milliseconds for a human being to blink an eye, while

HFT might have traded hundreds or thousands of times during a similar period (O’Hara, 2015).

There are two main functions of a financial market, i.e. to provide liquidity and to promote

price discovery; in which both are vital for asset pricing. The process of incorporating new

information into asset prices is known as price discovery, and together with liquidity, they play an

important role for an efficient capital allocation in the economy. An efficient market allows

individuals to reallocate their asset holdings, resulting in risk sharing among investors (O’Hara,

2003). The market is deemed as efficient when the price of a security fully reflects all currently

available information about its economic value, both current and historical information. Since

financial market is not naturally efficient, the market will move towards efficiency through price

discovery (Cooper et al., 2016), and trading activities by informed traders, either through market or

limit orders, will incorporate their private information about a security on its price (Cao, Hansch, &

Wang, 2009). Therefore, the maximization of price discovery is seen as an important objective by

regulators and academics alike (Cespa & Foucault, 2014).

In addition, as noted by Aldridge (2013), the process of impounding information from news to

price is hardly instantaneous. The price will first swing due to the implicit “negotiation” among the

many buyers and sellers which can be seen in the order flow before eventually finds its optimal post-

announcement price range. This process is commonly referred to as tâtonnement – a French word for

“trial and error”. The price fluctuation gives HFT an opportunity to profit from the arbitraging

surrounding news release and bring the market one step closer to its efficient state – as per EMH.

Using directional event-based strategies, HFT will place its trades based on forecasted market reaction

towards an event.

A vast majority of the empirical study on HFT and automated trading find a positive influence

on the market quality, in the sense that it reduces the bid-ask spreads, improves market liquidity, and

25

makes stock prices more efficient (Jones, 2013). Hasbrouck and Saar (2013) study the effect of low-

latency activities on market quality using the NASDAQ HFT dataset and find that an increase in HFT

activities reduce quoted spreads, reduce price impact, increase depth, and lowers short-term volatility.

They also test the relationship between normal and heightened uncertainty periods in the U.S. and find

evidence that higher low-latency activities improve market quality in both periods. This is also

consistent with Conrad, Wahal, and Xiang (2015) that uses the full cross-section of securities in the

U.S. equity markets and three hundred largest stocks on the Tokyo Stock Exchange (TSE).14,15 They

find that high-frequency quotation activity not only has no detrimental effect on market quality but in

fact, the presence of high-frequency quotes improves the efficiency of the price discovery process and

reduce the trading costs. These findings are further supported by the evidence from Boehmer, Fong,

and Wu (2015) that find co-location services facilitate HFT, which causally improves market

quality.16

Market-making HFT provides liquidity by matching buyer and seller orders, or by buying and

selling securities from their own inventories should they failed to immediately match buyers and

sellers (Shorter & Miller, 2014). HFT that engage as market-maker use their speed advantage to

quickly update quotes, and they profit from the difference between the price buyers are willing to pay

and the ask prices sellers are willing to accept for a security. Since this activity requires HFT to

maintain limit orders on both sides of the trades, it provides liquidity to the market (Chung & Lee,

2016). Hagstromer and Norden (2013) studied the event of changes in minimum tick size to examine

the effect of HFT activities on market quality using 30 Swedish large-cap stocks traded on the

NASDAQ-OMX Stockholm Exchange. Their findings suggest that HFT market-making activities

reduce short-term volatility, which is healthy for the overall market quality. Similarly, Riordan and

Storkenmaier (2012) study the effect of decreasing in latency on market quality following the release

of Xetra 8.0 by Deutsche Boerse in 2007 find significant improvement in the market quality post

upgrade, determined by narrower spread measures and higher relative quotes contribution to price

discovery.17

Even though the increasing competition of market-making in general benefits the market, the

fact that HFT does not have affirmative obligation to make market unlike the traditional market-

maker or specialists, raised concern that they might cause disruptions by fleeing the market at their

will (Carrion, 2013), e.g. when it is no longer profitable to do so (Anand & Venkataraman, 2013). The

14 Conrad et al. (2015) sample is from 2009 ‒ 2011 for the U.S. markets, and from 2010 and 2011 for the TSE.15 The three hundred largest stocks are from the First Section of the TSE by beginning-of-month market capitalization.16 Boehmer et al. (2015) use data from 42 markets to study the effect co-location on AT and HFT. The first implementation date of co-location in each country is used to capture the effect on latency prompted by the co-location service.17 The new trading platform was introduced with a sole reason to reduce the system latency, with no other meaningful microstructure change. Following the introduction, system latency is reduced from 50 millisecond to 10 millisecond.

26

absence of the constraining obligations also gives HFT more flexibility to formulate market-making

strategies beyond the traditional means (Brogaard et al., 2014). To gain more volume, certain trading

venues offer liquidity rebates to market-making HFT, which benefits both the HFT and the exchanges

themselves, as the HFT has the motivation to route the orders to their exchanges (Harris, 2015). The

aim for such rebates is to encourage and reward the liquidity supply provided by the market-makers

(The U.K. Government Office for Science, 2012). This is justified by the finding of Hendershott and

Riordan (2013) which suggest that HFT market-makers would lose money in the absence of rebate.

Similarly, Brogaard, Hendershott, and Riordan (2014) find that HFT liquidity supplying revenues are

negative without the fee rebates, especially during transactions with tighter spreads.

Theoretically, HFT could have both positive and negative effects on liquidity. The light-speed

trading activity by HFT is claimed to promote liquidity through rapid price adjustments, allowing for

narrower bid-ask spreads within a market, strengthening the inter-market linkage and activity

(Goldstein, Kumar, & Graves, 2014), and lowering the cost of intermediation (Jones, 2013). However,

the higher level of trading activity by HFT cannot simply be the indicator of better liquidity in the

market, as the HFT could be in either side of the trades. A dominance in the supply side would lead to

higher liquidity and narrower spread, while a greater number of trading activity in the demand side

would take liquidity away from the market and widened spreads (Goldstein et al., 2014). For instance,

CFTC-SEC (2010b) report suggest that even though HFT usually provide liquidity, during the Flash

Crash, they turned to consume liquidity. Easley et al. (2011) suggest that the action produces toxic

order flow and has exacerbated the ongoing liquidity crisis. This behavior of HFT has called for

regulatory discussion and debate on whether to impose HFT with quotation obligation and/or prevent

them from doing high-speed quotation entering/deleting (Gomber et al., 2011).

Brogaard et al. (2014) find that HFT, in general, has a positive role in the price discovery

process, especially contributing to the speed of price adjustment to new information, and smaller

pricing errors. However, they also contest that even though the price informativeness is commonly

viewed as something positive for the economy, the information that HFT used is short-lived, lasting

for only 3-4 seconds. Should the information eventually become public without HFT’s intermediation,

the adverse selection costs that slower traders have to bear might cause the potential welfare gains

from the faster price discovery becomes trivial or even negative.

Biais and Woolley (2011) argue that while the development of sophisticated and rapid trading

algorithms may benefit the markets and investors through better price discovery and liquidity, they

might be detrimental to the slower traders due to adverse selection problem. In a similar note, Biais,

Foucault, and Moinas (2015) claim that even though the investment in fast trading does help to deal

with the issue of market fragmentation, it also comes with the risk of adverse selection to the slow-

traders, which lowers the social welfare. Scholtus, van Dijk, and Frijns (2014) find evidence of

27

deterioration of market quality around the U.S. macroeconomic announcements. Using 60 seconds

event window from the release of the news [0, 60], they find that higher algorithmic trading activity

leads to lower depth, and higher quoted spreads, adverse selection costs, and volatility measures.18

Froot, Scharfstein, and Stein (1992) show that in theory, short-term traders may bank on short-

term information too much, and less concern on fundamentals value of a firm, which in turn,

dampened market efficiency. Vives (1995) suggests that short-horizon traders reduce price

informativeness with the concentrated arrival of information, which is likely to be the case around

earnings news events. Zhang (2010) estimated the volume of HFT in the U.S. capital market for the

year 2009 and find that HFT accounts for 78% of the total trading volume, which is very close to

Tabb Group’s estimate at 73%. He finds that HFT is positively correlated with price volatility even

after controlling for stock’s fundamentals and explanatory variables for volatility. The result is

stronger especially in 3,000 largest stocks by market capitalization, in stocks with high institutional

holdings, and during high market uncertainty periods.

Froot et al. (1992) and Zhang (2010) demonstrate that on account of their relative emphasis on

the short-term horizon, the HFT firms hamper the price discovery process in the market. In fact, the

HFT activities may cause the markets to be “too efficient” (overshooting fundamental values) and

therefore, need to be restrained. Zhang (2010) shows that in the short run, HFT activity causes stock

prices to move excessively in the direction of the news about fundamentals making it detrimental to

the price discovery process. For instance, after positive fundamental news about a stock is released,

HFT firms will rapidly enter a long position in the stock, and consequently, raising its price. At a later

time, fundamental investors make their moves to buy the stock too, causing the stock price to rise

more than the news about the fundamentals warranted, and thus, leads to “overshooting”. Another

reason for this phenomenon could be that HFT firms try to front run fundamental investors by

anticipating the general direction of the subsequent trades. These firms will buy/sell the stock before

the fundamental investors can do so, and when they (fundamental investors) eventually execute their

trades it causes the price to move excessively.

In fact, some argue that it is the sheer speed of HFT that cause other slower investors bearing

the cost of adverse selection (Jones, 2013). In a theoretical paper, Budish et al. (2015) develop a

model in which market-makers or traders that invest in speed will be the first to react and make a

profit from the newly arrived public information. In the event of the traders receive and react to a

news before the market-makers do, they (the fast traders) will trade with the stale quotes, which

impose adverse selection cost on the market-makers. This situation will discourage liquidity

provision, and consequently, the market-makers include the cost of them being adversely selected in

18 To measure market quality, Scholtus et al. (2014) examine liquidity and volatility in the market. Liquidity is measured using depth, volume, and spread. Volatility is measured using two realized measures calculated overintervals of one (s = 60) and five (s = 300) minutes.

28

their quote, resulting in the wider spread and higher cost for other slower investors. This could be

made possible due to HFT’s speed and resources, which allow them to quickly process and take

appropriate action whenever a new publicly available information arise. Slower traders on the hand,

take a longer time to revise their orders, allowing HFT adversely select other participants’ orders

(Brogaard et al., 2014).

It is also possible for the algorithm to be fed with false information – either intentionally or

accidentally. For instance, the United Airlines (UAL) stock price suddenly crash from US$12 to US$3

on September 8, 2008, in a mere 12 minutes, in which the shareholders lost (in value) of

approximately US$1 billion (New York Times, 2008). An investigation later revealed that the rapid

drop was mainly due to the interplay between algorithms that reacted to a six-year-old headline that

mistakenly hit the news feed since human traders might not be deceived by the headline blunder

(Donefer, 2010). Similarly, two weeks prior to the UAL’s unfortunate event, on August 27, 2008,

Bloomberg News accidentally published an obituary for Steve Jobs – the CEO of Apple (APPL)

(Fortune.com, 2008). Luckily, the blunder happened during off-trading hours and was quickly

retracted. Should it be otherwise, then Apple’s stock price might suffer the same fate that befell

United Airlines’ stocks two weeks later (Donefer, 2010). This leads to another question – does

immediacy of information dissemination always a good thing?

In a nutshell, scholars' understanding of the impact of HFT on market quality is still lacking

due to its young literature and the lack of high-quality data (Carrion, 2013; Boehmer et al., 2015).

However, in general, there is mixed empirical evidence on the impact of HFT on market quality.

Despite the majority of empirical studies find positive effects of HFT's participation in the market,

they cannot rule out the possibility that HFT, in theory, may harm the market through their predatory

trading strategy (Manahov et al., 2014). Regulatory bodies around the world are either implemented

or mulling over rules to contain and mitigate any HFT activity that may potentially detriment market

quality (Benos & Sagade, 2016). It is agreed that any abusive or predatory trading activity which goes

against market integrity should be eradicated. Nonetheless, regulators must be extra careful in

formulating their arrangement to avoid any excessive regulations and constraints that may be

counterproductive and have unanticipated effects on market quality. For instance, the newly

formulated regulation should not prevent beneficial HFT strategies that have positive effects on

liquidity (e.g.: market-making strategies) or price discovery and market efficiency (e.g.: arbitrage

strategies) from taking place (Gomber, Arndt, Lutat, & Uhle, 2011).

29

2.8 Controversies on HFT

This section is aimed to highlight the negative sentiment and controversies surrounding the HFT. The

identified controversies are (1) the flash crash of May 6, 2010; (2) the economic welfare of the arms

race; and (3) HFT’s market-making obligation.

2.8.1 The Flash Crash of May 6, 2010

On May 6, 2010, the US financial markets were shocked with a short-lived, yet severe drop in prices,

all happened within minutes. The sudden market crash of May 6, 2010, is later dubbed as the "flash

crash", given the brief moment of the event. The U.S. Commodity Futures Trading Commission

(CFTC) and U.S. Securities & Exchange Commission (SEC) released joint preliminary findings with

regards to the event on May 18, 2010 (CFTC-SEC, 2010a), and full findings were released later on

September 30, 2010 (CFTC-SEC, 2010b).

The US market opened on May 6 with unsettling political and economic issues surrounding

the European debt crisis. The concern over the future direction of the European market has heightened

the level of uncertainties in the US market, evidenced by high volatility, a flight to quality, and rise in

premiums for buying protection against default by the Greek government on their sovereign debt.

Consequently, the Euro experienced a sharp decline against the U.S. Dollar and Japanese Yen around

midday. In the U.S., the financial market was shrouded by negative market sentiment, causing the

S&P500 volatility index (VIX) to rise by 22.5 percent at around 2.30 p.m. (Central Time, CT) from its

opening level. This has triggered investors to engage in flight to quality, created a selling pressure

which has pushed down the Dow Jones Industrial Average (DJIA) by 2.5%.

At 2.32 p.m., Waddell & Reed (a large fundamental trader) initiated a sell order algorithm

(Sell Algorithm) to sell 75,000 E-mini (S&P500 futures) contracts to hedge its existing equity

position (Reuters, 2010; CFTC-SEC, 2010a). The Sell Algorithm is programmed to target an

execution rate set to 9% of the trading volume calculated over the previous minute, without regard to

price or time. Normally orders at such scale (valued at approximately US$4.1 billion) are fed in

multiple stages to avoid shocks to the market, but apparently, this time it was not. Initially, the selling

pressure was absorbed by HFT and other intermediaries in the futures market, followed by the

fundamental buyer and cross-market arbitrageurs, in which the latter transferred the selling pressure to

the equities market. Within 13 minutes of execution (between 2:32 p.m. and 2:45 p.m.), 35,000 E-

mini contracts (valued at approximately US$1.9 billion) out of the intended 75,000 were sold.

30

From the Sell Algorithm order, HFT has accumulated a net long position of about 3,300

contracts.19 Between 2:41 p.m. to 2:44 p.m., HFT aggressively sold about 2,200 E-mini contracts they

held to reduce their inventories. Nearly 140,000 E-mini contracts (over 33% of total trading volume)

were traded by HFT.20 The dramatic increased in trading volume increased volatility in the market,

which in turn shied long-term traders away from the market. The lack of demand in the market caused

HFT to buy and sell from one another, generating a "hot-potato" volume effect. Enormous selling

pressure from the combination of the Sell Algorithm, HFT, and other traders drove the price of the E-

mini down by 3% in this 4 minutes period. At the same time, cross-market arbitrageurs who bought

the E-mini simultaneously sold equivalent amounts in the equities markets, driving the price of S&P

500 SPDR (SPY) also down by approximately 3%.21

The combined selling pressure was so tremendous it almost wiped clean the entire buy-side

orders of the E-mini, creating an order imbalance in the market. At that moment, there were less than

1,050 buy-side orders unmatched, and still, more than 50% of the Sell Algorithm's orders yet to be

matched. This severe liquidity absence pushed the E-mini prices down by another 1.7% in a mere 15

seconds, reaching its intraday low of 1,056 points. At 2:45:28 p.m., the Chicago Mercantile

Exchange's (CME) Stop Logic Functionality was triggered due to the rapid prices decline of the E-

mini, causing all trading on the E-mini to be halted for five seconds. After the trading resumed at

2:45:33 p.m., the E-mini prices stabilized and starting to recover, thanks to opportunistic and longer-

term traders who re-entered the market and rapidly accumulated long positions (Kirilenko et al.,

2017). Subsequently, SPY also recovered.

Despite the E-mini recovering, the prices of other affected securities continued to decline. The

sell orders placed on some individual securities and ETFs experienced reduced buying interest,

mainly due to a high level of uncertainty among market participants in the market. Accordingly, some

market-makers and other liquidity providers either widened their spreads and/or reduced offered

liquidity, while others simply withdrew their position off the market. HFT in the equity markets

traded proportionally more as volume increased, and overall were net sellers in the fast-declining

market. Some of the HFT continued their trading and tap on the opportunities arose from the severe

price dislocations in individual securities as the market started to recover, while some others just

stopped trading completely.

19 16 out of over 15,000 trading accounts are classified as HFT, and traded over 1,455,000 contracts on May 6, equivalent to almost one-third of the total daily trading volume (CFTC-SEC, 2010b).20 This is consistent with the HFT’ typical practice of trading a very large number of contracts, but not accumulating an aggregate inventory beyond three to four thousand contracts in either direction.21 The E-mini and SPY are the two most active stock index instruments traded in the electronic futures and equity markets. Both are derivative products designed to track stocks in the S&P 500 Index, which in turn represents approximately 75% of the market capitalization of U.S.-listed equities. Since the E-mini and SPY both track the same set of S&P 500 stocks, cross-market arbitrage between these two products kept their prices closely aligned during their rapid declines.

31

There were approximately 2 billion shares with a total volume of more than US$56 billion

traded between 2:40 p.m. and 3:00 p.m. on that day. During this 20 minutes window, more than 98%

of all shares were traded within 10% of their value at 2:40 p.m. Due to the unusually high level of

uncertainty in the market, orders sent to the market found no immediate interest, caused trades being

executed at irrational prices. For instance, Accenture plc (ACN) rapidly declined in 7 seconds from

about US$30 at 2:47:47 p.m., to US$0.01 by 2:47:54 p.m., and recovered within a matter of seconds.

An ETF, iShares Russell 1000 Growth Index Fund's (IWF) share price plummeted from about US$45

just before 2:46 p.m. to the lowest price of US$0.0001 at 2:47:28 p.m., and slowly recovered to its

prior level by 2:56 p.m. On the contrary, Sotheby's (BID) was traded at an extremely high price of

US$99,999.9999 at 2:57:08 p.m., from around US$30 only minutes before that (CFTC-SEC, 2010a).

These extreme cases were caused by orders executed against stub quotes, which was triggered due to

the sudden loss of liquidity during the flash crash (Gomber, Arndt, Lutat, & Uhle, 2011).22

Overall, over 20,000 trades (amounting to 5.5 million shares) across 300 separate securities

and ETFs have executed at prices 60% or more away from their 2:40 p.m. prices. By 3:00 p.m., prices

for most securities had reverted back to trading at their rational values. After the market closed, the

SEC and the Financial Industry Regulatory Authority (FINRA) have met and agreed to adopt the

“clearly erroneous" trade rules, and thus all trades classified as "clearly erroneous" were canceled

(broken).23,24 Almost two-thirds of shares in the broken trades were executed at prices of less than

US$1.00, and approximately five percent were executed at prices of greater than US$100 (CFTC-

SEC, 2010b). From the joint report, it is evident that HFT did not trigger the Flash Crash. However,

the repeated buying and selling of contracts executed by the automated systems created the hot-potato

effect as HFT competed for liquidity. Thus, their trading behavior during the unusually large selling

pressure on May 6 is perceived to have exacerbated the price decline and market volatility (Kirilenko

22 Stub quotes are quotes generated by market-makers at levels far away from the current market in order to comply with its obligation to maintain a continuous two-sided quoting obligations. However, the stub quotes are not intended to be executed (CFTC-SEC, 2010a).23 Under the "clearly erroneous" trade rules, the regulatory body may declare a trade to be null and void, should the trade in question was considered to be "clearly erroneous" (CFTC-SEC, 2010a). On September 10, 2010, the SEC approved new rules submitted by the national exchanges and FINRA that clarify the process for breaking erroneous trades (https://www.sec.gov/rules/sro/bats/2010/34-62886.pdf).24 Following the wide-scale disruption of May 6, 2010, the exchanges and FINRA settled on the relatively high 60% standard for breaking trades (CFTC-SEC, 2010a, 2010b):• For stocks priced US$25 or less, trades will be broken if the trades are at least 10% away from the circuit breaker trigger price.• For stocks priced more than US$25 to US$50, trades will be broken if they are 5% away from the circuit breaker trigger price.• For stocks priced more than US$50, the trades will be broken if they are 3% away from the circuit breaker trigger price.Where circuit breakers are not applicable, the exchanges and FINRA will break trades at specified levels for events involving multiple stocks depending on how many stocks are involved:• For events involving between five and 20 stocks, trades will be broken that are at least 10% away from the "reference price," typically the last sale before pricing was disrupted.• For events involving more than 20 stocks, trades will be broken that are at least 30% away from the reference price.

32

et al., 2017). Due to this event, HFT has received considerable critical attention from both the CFTC

and SEC for creating "excessive" short-term volatility (CFTC-SEC, 2010b, 36-37).

2.8.2 HFT arms race and welfare issues

HFT contribution in the process of price discovery is indeed beneficial, as more informative stock

prices might lead to better resource allocation in the economy. Nonetheless, Brogaard et al. (2014)

find that the information used by HFT are short-lived, lasted for less than 3 to 4 seconds. Should the

information will eventually become public without HFT’ intermediation, then the potential welfare

contribution by HFT might be minuscule, or even negative in the situation where longer-term

investors are significantly affected by the adverse selection costs from trading with HFT. In a similar

note, Menkveld (2014) agree that the presence of market-making HFT in electronic markets does

improve welfare by reducing informational frictions from non-simultaneous orders arrival in the

market. However, the net welfare from HFT is questionable – the positive contribution from market-

making activity might be destroyed when HFT pick off investors’ quotes at lightning speed on

information that will surely arrive at the slower investors at a lower frequency.

HFT acknowledge the importance of investing in hardware, software and network capabilities

to reduce latency in an automated trading process, motivated by the nature of the game where winner-

takes-all. The upgrades allow them to continuously refine their trading algorithms, and emerge

victorious in the arms race (Kauffman, Liu, & Ma, 2017). Regardless, the technology arms race to

shave-off several seconds raised concerns about the excessive spending of money without meaningful

progress in market quality (Chung & Lee, 2016). The race among institutions to be the fastest is

deemed as unproductive, and the unwarranted investments in technological infrastructure to reduce

trading latency creates doubts of whether HFT adds value overall (Chordia et al., 2013; Jones, 2013).

In addition, Menkveld (2014) asserts that the technology investment itself may as well be the source

of negative externality through the relative speed disadvantage it creates for others.

From another point of view, Budish et al. (2015) claim that the arms race is indeed socially

wasteful, but their existence is actually a symptom, stemmed from a flaw in the architecture of

modern financial exchanges that use continuous-time trading, which also creates adverse selection

rents that attract HFT. Budish et al. (2015) suggest that the problem can be addressed using a frequent

batch auction, which will create a discrete-time market to replace the current market design that is

based on the continuous limit order book. This will make the tiny speed advantage less valuable,

which intuitively put an end to the arms race. In a similar notion, Yao and Ye ( forthcoming) find

evidence that even with discrete timing, HFT might continue to race each other – this time to compete

for rents from the queuing channel, originated from yet another microstructure design – tick size.

33

Either way, both types of rents are lucrative by-products of market’s imperfections and can be

dominated by being the fastest, which leads to an arms race in speed.

Regardless, even without the issue of arms race, HFT still pose a threat to many as they may

use high-speed predatory trading strategies (see section 2.6 ‒ Detrimental HFT strategies), such as

introducing "microstructure noise" that generates an unnecessary extra layer of intermediation

between buyers and sellers, leading to increased price volatility and worsened market quality (Cartea

& Penalva, 2012).

2.8.3 Market-making obligations

Anand and Venkataraman (2013) study the trades of two types of market-makers, the Designated

Market-makers (DMMs) and Endogenous Liquidity Providers (ELPs). The main difference between

DMMs and ELPs lies in their obligation to make a market. DMM or Specialists are bounded by

specific obligations imposed by the exchange, i.e. to maintain a market presence by continuously

posting quotes with reasonable depth. ELP on the other hand, employs market-making strategies

because of its profitability, with no affirmative obligations to maintain markets. Anand and

Venkataraman (2013) states the HFT are the most active market-makers in financial markets today, in

which some position themselves as ELP – meaning that they are likely to supply liquidity whenever it

is profitable for them to do so and cease from providing liquidity when facing large adverse selection

risks (Chung & Chuwonganant, 2018), or whenever the market conditions are unfavourable for them

to make profits, which is more likely to happen in times of high market uncertainty (Zhang, 2010).

The lack of commitment to make market especially in times of market stress and in thinly

traded securities raised concern among practitioners and regulators. HFT’s optional market-making

may exacerbate execution uncertainty, and thus, the liquidity supplied by HFT are deemed unreliable,

which might reduce investors’ confidence and participation. Liquidity withdrawal by HFT might thin

out the order book, which may induce extreme market movements (Gomber et al., 2011). This might

also be the underlying reason for the heightened sensitivity of liquidity and returns to market volatility

in the era of HFT. Furthermore, the non-HFT are playing at an uneven playing field due to their

technological inferiority to HFT, and they might find that the market is unfair, and consequently, stop

participating altogether (Anand & Venkataraman, 2013). In response to this potential problem,

regulators consider imposing quotation obligations on HFT, and/or preventing them from engaging in

high-speed order entering and cancellation (Gomber et al., 2011).

34

CHAPTER THREE:

ESSAY TWO

HIGH-FREQUENCY TRADING AND RELATIVE TICK SIZE

3.1 Chapter overview

This chapter shows the effect of effective tick size and relative tick size on HFT’s participation.

Section 3.2 presents the introduction of the essay, which covers the background of the study, problem

statement, and research motivation, followed by a statement of research objectives. The chapter

continues with hypotheses, expected contributions, and methodology.

3.2 Introduction

The introduction of computer-based trading takes the financial world by storm, leading to the

birth of a new breed of investors in the market known as algorithmic traders (ATs), which use

sophisticated algorithms to trade in the financial market. On top of the standard ATs’ modus operandi,

a sub-group of ATs known as high-frequency traders (HFT) use state-of-the-art technology and high-

speed connections such as colocation, for lightning speed information processing and order

submissions, quoted in milliseconds and even in nanoseconds. While the entrance of HFT in the

market coincides with the increased level of limit orders submissions and cancellations and extreme

intraday price volatility (Hagstromer & Norden, 2013), it is also associated with the lower bid-ask

spreads, which induced trading rate of both informed and uninformed investors, leading to higher

price efficiency and quickens price discovery (The Government Office for Science, 2012).

On March 31, 2010, the Australian Government announced the support for the approval of a

market license to Chi-X Australia Pty Limited, corresponding with Recommendation 4.5 of the

Johnson Report to increase competition on exchange-traded markets.25 This action is seen as a part of

the effort to position Australia as a leading financial center. Chi-X Australia offers a low-latency and

HFT-friendly trading platform as an alternative to ASX TradeMatch, which is the primary trading and

listing venue for equities trading in Australia. Following the approval, on November 9, 2011, Chi-X

25 The original name of the report is “Australia as a Financial Centre – Building on our Strengths”, produced by Australian Financial Centre Forum in 2009. The Forum was helmed by Mr Mark Johnson, former Deputy Chairman of Macquarie Bank.

35

Australia commenced its full operation to trade all S&P/ASX 200 constituent stocks and ASX-listed

Exchange Traded Funds (ETFs). To remain competitive, ASX reduced its headline trade execution fee

from 0.28 basis points to 0.15 basis points in June 2010 and launched its very own low-latency

alternative trading venue, ASX PureMatch in November 2011. The new trading platform, which

aimed to “meet the growing needs of the trading community for order books that offer the most liquid

stocks across the fastest available platform” (ASX, 2011), directly competes with Chi-X Australia to

attract HFT.

Comerton-Forde (2012) states seven (7) factors that may attract HFT’s participation in a

market, namely, small tick sizes, fragmented markets, low-latency trading platforms, low explicit

trading fees, high liquidity, and trade through protection. In Australia., the tick size, which is also the

main highlights of this study, has remain unchanged since December 4, 1995, despite the many crises,

issues, and changes that happened in the financial markets all over the world since then, such as the

Asian Financial Crisis in 1997, Global Financial Crisis in 2008, European Debt Crisis from 2009 –

2012, Flash Crash of May 2010, the rise of algorithmic traders, and the emergence of alternative

trading venues. During a similar period, the tick size in the U.S. was changed twice, i.e. from one-

eight (US$0.125) to one-sixteenth (US$0.0625) of a dollar in 1997, and from the one-sixteenth to the

decimalization in 2001.

Tick size, or minimum price variation, is defined as “the minimum amount by which share

prices are allowed to vary. It determines the prices at which orders may be entered. Orders may only

be entered at prices that are evenly divisible by the minimum tick size” (ASIC, 2010, p. 84). The

importance of small tick sizes on HFT’ strategy is pretty clear, which obviously leads to narrower bid-

ask spreads, and thus, directly affecting the minimum trading costs (Comerton-Forde, 2012; Harris,

1994).26 Regardless, this does not mean that smaller tick sizes are always favourable – a tick size that

is too small may negatively influenced the interaction between different type of investors in the

market, and reduced the willingness of investors to expose their orders, which may hinder HFT’s

participation in the market, leading to reduction in depths (Chordia, Roll, & Subrahmanyam, 2011;

O’Hara, Saar, & Zhong, 2016; Yao & Ye, forthcoming).

Angel (1997, 2012) emphasizes that tick size plays an important role in governing the market,

and he outlined the reason for why the optimal tick size is not zero; (1) a non-zero tick simplifies

trader’s information sets, (2) an economically significant (i.e. non-trivial) tick size preserve the price

and time priority in an order book, which incentivises traders to supply liquidity by posting limit

orders, and, (3) tick size creates a floor for the quoted bid-ask spread, which works as an incentive for

dealers to make markets. On the other hand, this also means that the bid-ask spread, and tick size as

its foundation, increases the minimum transaction costs for investors. In addition, tick size limits the 26 Other factors that may attract HFT are fragmented markets, the use of low-latency trading system, low explicit trading fees, high liquidity, and trade through protection (Comerton-Forde, 2012).

36

potential price points and numbering that can be quoted, which makes it easier for human traders to

comprehend the current view of the market. This issue, however, is not applicable to algorithm-based

traders, as they should have no problem to understand complex numbering.

The tick size structure currently being practiced in ASX and the U.S. market is different. In

the U.S., stocks below US$1.00 has a tick size of US$0.0001, and stocks greater than or equal to

US$1.00 have a tick size of US$0.01. This creates a leap on relative tick size for stocks crossing the

price border, i.e. from ≈0.01% (US$0.9999) to 1.00% (US$1.00), which is an increase by 100 folds.

Conversely, there are two tick size borders practiced in the Australian market, which are A$0.10 and

A$2.00. These borders create a jump in relative tick size for stocks crossing the borders from ≈1.01%

(A$0.099) to 5.00% (A$0.100), and from ≈0.25% (A$1.995) to 0.50% (A$2.000).

The relative tick size represents the profit potential that HFT able to gain from a transaction,

from one tick of price movement. O’Hara, Saar, and Zhong (2016) for instance, find that “a larger

relative tick size benefits HFT firms that make markets on the NYSE: they leave orders in the book

longer, trade more aggressively, and have higher profit margins”, which signifies that a large relative

tick size may incentivise HFT to participate. The differences in tick size and price steps mandated in

these two markets might influence HFT’ potential profitability and strategy. Therefore, the results

obtained using Australian dataset might be different from studies which are based on U.S. dataset.

Moreover, unlike in the U.S., there is no dedicated market specialist or market-maker in Australian

equity market. This absence can be seen as an opportunity for HFT, as there is less competition to

make the market. In addition, the maker-taker pricing model that is offered by Chi-X Australia should

give further encouragement for HFT to actively participate.

Table 3.1 compares the tick size structures applied in both markets, while Figure 3.1 illustrates

the changes in relative tick size in Australian stock prices ranging from A$0.010 until A$5.000. A

clearer detail on the changes surrounding the price borders of A$0.10 and A$2.00 is illustrated in

Figure 3.2 and Figure 3.3 respectively.

37

Table 3.1

Comparison of tick size structure between Australian and U.S. equity market

Australia (A$)

Price Range $0.001 - $0.099 $0.100 - $1.995 $2.00 - $99,999,990

Tick Size $0.001 $0.005 $0.01

Highest relative tick size 100.00% 5.00% 0.50%

Lowest relative tick size ≈1.01% 0.25% ≈0.00%

United States (US$)

Price Range p < $1.000 p ≥ $1.000

Tick Size $0.0001 $0.01

Highest relative tick size 100.00% 1.00%

Lowest relative tick size ≈0.01% ≈0.00%

Furthermore, the minimum trading unit applied in Australia is also different from the U.S., in

which the latter requires all stocks to be traded at a minimum of 100 units per board lot. Any odd lot

trading will have a greater trading cost associated with them. On the contrary, the mandated minimum

trading unit in Australia is one unit, which gives all traders, including HFT, the opportunity to trade at

any unique combination with no additional cost.27 In this sense, trading in Australia gives HFT greater

flexibility in formulating their strategy, without being constrained by the minimum trading unit.

Therefore, large traders in Australia might have a better opportunity to hide their large orders amongst

retail orders in the market by “packaging” their orders in various smaller pack sizes, compared to

those in the U.S. market, thanks to the smaller minimum trading unit. This characteristic might also

affect HFT’ ability to sniff the incoming large order and makes it harder for them to “ride the wind”.

Most empirical studies on tick size are based on the U.S. financial markets (see for example

Angel, 1997; Bessembinder, 2003; Gibson, Singh, & Yerramilli, 2003; Goldstein & Kavajecz, 2000;

Jones & Lipson, 2001; Lipson & Mortal, 2006; O’Hara et al., 2016; Schultz, 2000; Yao & Ye,

forthcoming). However, given the differences in the tick size structure and minimum trading unit

practiced in Australia and the U.S., HFT might have different preference and pursue different trading

strategy in these markets, as the two aforementioned factors may have a direct effect on typical HFT’

strategy that holds to the principle of “little and often fills the purse”, and thus, calls for a further

investigation.

27 This is compared to the higher fees charged for odd-lot trading in markets with minimum trading unit of greater than one.

38

0.000 0.500 1.000 1.500 2.000 2.500 3.000 3.500 4.000 4.500 5.0000.000%

1.000%

2.000%

3.000%

4.000%

5.000%

6.000%

7.000%

8.000%

9.000%

10.000%

Figure 3.1: Relative tick size for stock prices ranging from A$0.01 to A$5.00

39

0.000 0.050 0.100 0.150 0.200 0.250 0.300 0.350 0.400 0.450 0.5000.000%

2.000%

4.000%

6.000%

8.000%

10.000%

12.000%

Figure 3.2: Relative tick size for stock prices surrounding the A$0.10 border

0.000 0.500 1.000 1.500 2.000 2.500 3.000 3.500 4.000 4.500 5.0000.000%

0.200%

0.400%

0.600%

0.800%

1.000%

1.200%

Figure 3.3: Relative tick size for stock prices surrounding the A$2.00 border

40

3.3 Research objectives

To the extent of the researcher’s knowledge, there are only two empirical papers on tick size using

data from Australian equity markets. Aitken and Comerton-Forde (2005) use the event of Australia’s

tick size changes in 1995 to study the impact of tick size reduction on liquidity, while Frino, Mollica,

and Zhang (2015) use ASX listed companies’ split events to investigate the effect of relative tick size

changes (due to stock split) on HFT activity. This study, on the other hand, intends to:

RO1: To determine the effect of tick size borders on HFT’s activity, using stocks priced within the

proximity of the borders.

RO2: To determine the effect of crossing tick size borders on HFT’s activity.

RO3: To determine the effect of relative tick size on HFT’s activity, using stocks within a similar

tick size structure.

3.4 Hypotheses development

Relative tick size is measured by dividing a stock’s tick size with its price – i.e. it will change as the

price move and it is not uniform across stocks, making it the more relevant economic measure than

the tick size. O’Hara et al. (2016) find that larger relative tick size promotes HFT’s market-making

activity on the NYSE. Yao and Ye (forthcoming) find that tick size is a driver for HFT in NASDAQ –

it acts as a queuing channel in which price competition is constrained by the existence of tick size, and

creates rents for liquidity provision especially for lower-priced securities.

This study intends to study the effect of relative tick size on HFT activity in Australian equity

market. There are three reasons for this selection: (1) the tick structure applied in Australia is different

than the one practiced in the U.S., and this difference might affect HFT’s competition; (2) the tick size

structure in Australia was last revised in 1995, and has never been revised since then, indicating a

more stable environment with lesser regulatory intervention compared to the U.S. market; and (3) the

mandated minimum trading in the U.S. is 100 unit per board lot while it is only one unit in Australia –

giving HFT greater flexibility in formulating its strategy in the latter market. These differences might

influence HFT’s strategy, activity, and behavior, and therefore, findings from Australian equity

market might give a further understanding on HFT. To answer the research objectives, this study will

three different scenarios that directly affect relative tick size.

41

1. The effect of tick size borders on HFT activity

A tick size border refers to the price set by regulators which determine a stock’s default tick size

whenever its price fell into a particular tick size band. In Australia, there are two tick size borders

(A$0.10 and A$2.00), which creates three price bands (A$0.001 – A$0.099; A$0.100 – A$1.995; and

A$2.00 - A$99,999,990.00). Stocks priced lower than the border will have smaller relative tick size

compared to those priced higher than the border. The lowest stock price within a band is equivalent to

the price set to be the tick size border, which also represents the highest relative tick size in the band.

Therefore, the highest stock price in a band will have the lowest relative tick size. This situation

creates a drastic relative tick size change in a continuous price structure, in which the tick size is

binding. For instance, a stock priced at one-tick smaller than the A$2.00 border (i.e. the highest point

in its price band) is priced at A$1.995, has a tick size of A$0.005 and a relative tick size of 0.251%.

On one hand, a stock priced exactly at the border (i.e. the lowest point in its price band) will have a

price of A$2.00, a tick size of A$0.01, and a relative tick size of 0.500% – almost double than its

previous price point. This situation creates a sudden “jump” in the relative tick size (see Figure 3.1,

3.2, and 3.3) Based on this information, this study will determine whether the discontinuity of relative

tick size due to the “treatment” of tick size affect HFT activity.

H1: There is a significant difference in HFT activity between stocks priced lower than the tick size

borders and stocks priced higher, due to the binding tick size.

2. The effect of stocks crossing a tick size border on HFT activity

Stock prices constantly change, especially in actively traded stocks, driven by demand and supply in

the market. In addition, corporate events such as stock split or stock dividends (bonus issues) may as

well impact the stock price directly, causing the price to change by a certain degree in accordance

with the pre-determined split factor or dividend percentage. Thus, it is possible for the stock price to

cross the tick size borders, either naturally (market force) or intentionally (corporate actions). Using

stock split events, Frino et al. (2015) find lower order-to-trade ratio and longer order resting time in

the post-split period. Their finding suggests that a sudden increase in relative tick size is associated

with lower HFT activity. Complementing Frino’s et al. (2015) research, this study intends to

determine the effect of relative tick size change due to a natural event (i.e. purely driven by demand

and supply) on HFT activity. In particular, this study will observe HFT activity in stocks before and

after crossing the tick size borders and compare them with matching firms that did not cross the

borders.

42

H2: There is a significant difference in HFT activity in the period before and after crossing a tick

size border, even after controlling for matching firms.

3. The effect of relative tick size on HFT activity

Tick size represents the smallest price variation. All stocks within a similar price band will have a

uniform tick size. Even though the tick size is fixed, the stock price is not. The variation in stock price

results in variation of relative tick size across stocks within a similar price band, which are bounded

by the same tick size. For instance, the price band of A$0.100 – A$1.995 will have a relative size

ranging from 5.000% to 0.251%, while it ranges from 0.500% to ≈0.000% in the next price band (see

Table 3.1), showing that relative tick size can be considerably different depending on stock price

levels. Based on this relationship, O’Hara et al. (2016) find that market-making HFT trade more

aggressively in stocks with larger relative tick size – they increase their undercutting of resting limit

orders in the book and they leave limit orders in the order book longer, compared to stocks with

smaller relative tick size. This study intends to supplement O’Hara’s et al. (2016) work by comparing

HFT activity in stocks with substantially different relative tick size within a similar tick size structure

using a dataset from a pure order-driven market.

H3: There is a significant difference in HFT activity in stocks with different relative tick size

within a similar tick size structure.

3.5 Expected contribution of the Study

Even though there is many research studied the impact of changes in relative tick size on HFT, there

is only one study to date, that uses Australian dataset to investigate the issue. Thus, the study is

expected to enrich the existing literature on the impact of market microstructure design on HFT

activity in Australia. The study is intended to reflect the actual HFT activity by using the widely-used

HFT proxy, i.e. the order-to-trade ratio. Evidence from the study is expected to be able to measure

how HFT will react to the changes in relative tick size. The tick size structure, coupled with the

minimum trading of one unit mandated in Australia, is different than what being practiced in the U.S.

for instance. This unique market design is expected to significantly affect HFT strategy, and thus, the

results obtained from this study might give a better understanding on HFT, while complementing the

existing literature on the similar issue.

43

3.6 Methodology

This section discusses the methodology that will be used in this study. The methodology is designed

to address the research objectives stated in Section 3.3, and thus, the methods presented in this section

is also arranged based on the objectives, preceded by the general description on the data used in this

study.

3.6.1 Data and Sample

The study will employ order book data from AusEquities database provided by Securities Industry

Research Centre of Asia-Pacific (SIRCA). The information provided by the database allows for

reconstruction of the order book at any point in time. The database provides the following

information28, which is vital for this study, timestamped precise at milliseconds:

Record type:

ENTER: Entry of a new order into the order book.

DELETE: Deletion of an order from the order book.

AMEND: Modification of existing order.

TRADE: A trade between two orders.

CANCEL_TRADE: A trade cancellation.

OFFTR: Off-market trade.

Price: The price at which an order was traded or entered. For an AMEND message, this is

the new order price

Volume: The total volume of the order or trade

Value: Price*Volume

The study will use the order book data made during ASX normal trading hours only, which is from

10:10:00 until 16:00:00 (Sydney time), Monday to Friday, and close during weekends (Saturday and

Sunday) and public holidays. The study will exclude all messages sent during the opening phase29,

and all messages sent outside the normal trading hours, i.e. any messages recorded from 16:00:00.001

until 10:09:15.000 in the next trading day. Therefore, the study will observe all order book messages

that are recorded into AusEquities database during intraday trading, thus, excluding all records

28 Retrieved from http://help.sirca.org.au/display/AUSEQ/4.+Choosing+Data+Fields.29 Retrieved from ASX’s official website at https://www.asx.com.au/about/trading-hours.htm. “Opening takes place at 10:00 am Sydney time and lasts for about 10 minutes. ASX Trade calculates opening prices during this phase. Securities open in five groups, according to the starting letter of their ASX code... Opening takes place at 10:00 am Sydney time and lasts for about 10 minutes. ASX Trade calculates opening prices during this phase. Securities open in five groups, according to the starting letter of their ASX code”

44

marked as OFFTR, as it refers to the orders made outside the normal trading hours. The study will

also exclude messages recorded as CANCEL_TRADE, which refers to a trade that is voided by the

system. Even though the excluded messages might have some information, they cannot be used to

observe how rapid the market, especially HFT, post their messages.

Chi-X Australia and ASX PureMatch platform were launched in Australia at the end of 2011,

aimed to attract low-latency traders. To guarantee the results obtained able to represent HFT activity,

this study will only use the data from the year 2012 onwards. The study will also limit the sample to

S&P/ASX 200 constituent stocks only, because HFT prefers highly liquid stocks, as well as to ensure

the sample do not suffer from the thin trading problem. In addition, since the estimation of HFT

activity requires a great deal of high-frequency data analysis, the study will randomly select 20

percent of all trading days in a year (approximately 50 trading days), unless stated otherwise.

3.6.2 Measures of HFT activity

The study will use several measures of HFT activity following ASIC (2013, 2015) methods in

identifying HFT in Australian equity market, which are the order-to-trade ratio, total turnover per day,

and average resting time.

1. Order-to-trade ratio

The order-to-trade ratio (OTR) is widely used to proxy for HFT activity (e.g. Aquilina & Ysusi, 2016;

ASIC, 2013, 2015; Brogaard et al., 2015; Friederich & Payne, 2015; Frino et al., 2015; Hagstromer &

Norden, 2013). HFT typically places a large number of orders across various price levels, and they

will revise their orders with the arrival of new information in the market, resulting in large OTR in

stocks with high HFT activity. The OTR is defined as the sum of order-book transaction message on a

given day, divided by the sum of trade executed on a given day. The formula excludes any messages

recorded as CANCEL_TRADE and OFFTR. A ratio of 1:1 means every order submitted results in a

trade, a large (small) OTR indicates a greater (lesser) proportion of HFT activity in stock i on day t.

Equation 3.1 shows the formula used to calculate OTR.

OTR i ,t=∑ Message i ,t

∑TRADEi , t

(Equation 3.1)

Where:

45

∑ Message i ,k = Sum of orders recorded as ENTER, AMEND, and DELETE in the order

book for stock i on day t

∑Tradei , k = Sum of executed trades in the order book for stock i on day t

2. Total turnover per day

HFT typically follows a low-margin strategy, in which they execute many transactions and make

marginal profits, amplified by the high trading volume in a day, suggesting that HFT needs to stay

active during the day to be profitable (AFM, 2010; ASIC, 2013, 2015; Zhang, 2010). Therefore,

stocks with high HFT activity tend to have high total turnover per day, which is defined as total dollar

value bought plus the total dollar value sold. Equation 3.2 shows the formula used to calculate total

turnover per day.

Total turnover i , t=Total Buy i ,t+Total Sell i ,t

(Equation 3.2)

Where:

Total Buy i ,t = The total amount (in dollar) of stock i bought on day t

Total Selli , t = The total amount (in dollar) of stock i sold on day t

3. Average resting time

HFT mainly engage in market-making activities, which involves frequent updates of quotes with the

arrival of new information in the market. HFT’s algorithm will delete or update its existing quote

should they face the risk of adverse selection, or they are trying to capture profitable trading

opportunities. This will cause an order to rest shortly before being updated/deleted. Therefore, the

average resting time will be shorter in stocks with high HFT activity. The average resting time is

calculated by determining the duration that an order sits in an order book before being revised (i.e.

amended or deleted). Equation 3.3 shows the formula used to calculate the average order resting time.

ORT j , i ,t=OrderID j , d ,i , t−OrderID j ,d−1 , i ,t

46

ORT i ,t=∑k=1

n ORT j ,i , t

n

(Equation 3.3)

Where:

OrderID j , d ,i , t = Order ID j timestamped at d of stock i on day t

OrderID j , d−1 ,i , t = Order ID j timestamped at d-1 of stock i on day t

ORT j , i ,t = Order ID j of stock i on day t

ORT i ,t = Average duration of all Order ID of stock i on day t

3.6.3 Method to address RO1

RO1: To determine the effect of tick size borders on HFT’s activity, using stocks priced within the

proximity of the borders.

In the Australian equity market, there are two tick size borders (A$0.10 and A$2.00), and three price

bands (A$0.001 – A$0.099; A$0.100 – A$1.995; and A$2.00 - A$99,999,990.00) applied. Based on

the two tick size borders, the study will use all constituent stocks of S&P/ASX 200 that suffice the

following criteria:

(1) Priced between A$0.08 and A$0.20, i.e. twenty (20) tick size movement in both directions

from the A$0.10 tick size border, or

(2) Priced between A$1.90 and A$2.20, i.e. twenty (20) tick size movement in both directions

from the A$2.00 tick size border, and

(3) Stocks that crossed the tick size borders during the selected day will be removed.

The existence of the tick size borders creates a discontinuity in the relative tick size, as shown in

Figure 3.1, Figure 3.2, and Figure 3.3 earlier. The tick size border can be perceived as the “treatment”

that cause the sudden jump in relative tick size. This situation suffices the condition of sharp

regression discontinuity design (Sharp RD) mentioned in Angrist and Pischke (2008), which states:

Sharp RD is used when treatment status is a deterministic and discontinuous function of a

covariate, x i. Suppose, for example, that:

47

Di={1 if xi ≥ x0

0if x i<x0

Where x0 is a known threshold or cutoff. This assignment mechanism is a deterministic

function of x i because once we know x i we know Di. It’s a discontinuous function because

no matter how close x i gets to x0, treatment is unchanged until x i=x0.”

(Angrist & Pischke, 2008, p. 189)

3.6.4 Method to address RO2

RO2: To determine the effect of crossing tick size borders on HFT’s activity.

To address the second objective, the study will use difference-in-difference analysis, based on the

event of a stock crossing the tick size borders due to demand and supply in the market. The following

condition is set to be selected as the sample in this study:

(1) Any price changes triggered by stock splits or stock dividends events will be removed.

(2) If the crossing stocks happened to make any price-sensitive announcement on the same day as

the crossing, it will be removed to avoid confounding effect.

(3) After crossing a tick size border, the stock should stay within the same price borders for at

least 10 days (following Hasgtromer & Norden, 2013).

The difference-in-difference methodology requires an identification of a control stock for each of the

selected sample. To determine a control stock, the following conditions are set to ensure the situation

where “everything else equal” can be sufficed, allowing for an accurate estimation on the effect of

crossing the tick size border on HFT activity:

(1) Control stocks did not cross any tick size border at any time within 10 days period

surrounding the event day.

(2) They are in the same industry (based on ASX industries classification).

(3) Has the closest market capitalization to the selected stocks, based on the value recorded at the

end of the previous calendar year.

3.6.5 Method to address RO3

RO3: To determine the effect of relative tick size on HFT’s activity, using stocks within a similar

tick size structure.

48

To answer the third and final objective of this essay, the study will use all S&P/ASX 200 constituent

stocks priced within the largest price band, i.e. from A$2.00 to A$99,999,990.00, resulting in a

relative tick size ranging from 0.500% (highest) to ≈0.000% (lowest) during the randomly selected

day. This study will adapt O’Hara et al. (2016) method for determining the effect of relative tick size

on HFT activity.

Stocks are segregated into two groups based on its relative tick size, creating a group of high

(HIGH) and low (LOW) relative tick size. From HIGH, the stocks are then sorted based on its market

capitalization, and 50 stocks will be chosen using a stratified sampling method to fairly represent the

entire range of market capitalization within the group. The selected stocks from HIGH are then

matched to a control stock selected from LOW. The matching criteria are (1) they are in the same

industry (based on ASX industries classification); and (2) has the closest market capitalization to the

selected stocks, based on the value recorded at the end of the previous calendar year. The reason for

controlling based on industry and market capitalization is because stocks from different industry may

be of interest to a different clientele, and stocks of larger size are likely to get more news coverage

and have more investors holding their shares (O’Hara et al., 2016).

The matching process is important to make sure that any results obtained are free from other

contaminating factors that may influence HFT activity in the selected sample. Therefore, after the

determination of sample and control stocks, the following step is to determine the level of HFT

activity in each group, and then test whether there is a significant difference in HFT activity between

the group.

49

CHAPTER FOUR:

ESSAY THREE

HIGH-FREQUENCY TRADING AND EXPECTED VOLATILITY

4.1 Chapter overview

This chapter examines the behavior of high-frequency traders during low and high expected volatility

in the ASX. Section 4.2 introduced the essay by explaining the problem and motivations of the

research. The statement of research objectives, hypotheses development, and expected contribution of

this study is presented in Section 4.3 4.4, and 4.5 respectively. Section 4.6 describes the data and

methodology used in this study.

4.2 Introduction

The VIX index (RIC: VIX), also known as the “investor fear gauge”, represents the expected stock

market volatility over the next 30 calendar days. The main difference between the VIX and other

stock market indices is the former measures volatility, while the latter tracks price. The VIX, which

was developed by Whaley (1993) for Chicago Board Options Exchange (CBOE), provides a

benchmark for short-term market volatility and serves as a volatility “standard” upon which futures

and options contracts on volatility could be written (Whaley, 1993, 2009). The VIX is constructed

based on a highly liquid underlying security market – e.g. S&P 500 (RIC: SPX) in the U.S., EURO

STOXX 50 (RIC: STOXX50) in Europe, and S&P/ASX 200 (RIC: AXJO) in Australia, and the index

level is implied by the current prices of options of its underlying market.

Whaley (2009) emphasizes the VIX is a forward-looking index, which means it measures the

volatility investors expect to see. Technically, volatility represents an unexpected market movement

which can either be upwards or downwards. Regardless, investors are more concern about the

potential of the latter happening, and thus, they predominantly use the VIX as insurance for the value

of their stock portfolios. Whaley (2009) documents that the investors’ fear for a bearish market is

greater than their excitement (or greed) in a bullish market, evident by the asymmetric rates of change

in the VIX and the SPX. Therefore, high levels of the VIX indicates greater levels of anxiety

regarding a potential drop in the stock market, while low levels imply stronger investors’ confidence.

50

This study is interested to investigate the effect of different levels of expected volatility on

HFT activity. HFT do not have emotions, and naturally, they do not suffer from “anxiety” or “fear”.

Nevertheless, they would still be affected by the ripples created by the aggregate market conditions,

such as other investors’ reactions towards the sudden change of expected volatility. HFT execute their

course of action according to the algorithm they are programmed for and use the information available

in the market to make their moves. By examining HFT activity during periods with different level of

expected volatility, this study will be able to understand how HFT perceived the expected volatility,

or more precisely, how their algorithms are designed to react in such situation. Ideally, the algorithms

applied by HFT should incorporate this factor, which in turn, will determine their movement and

activity in accordance with the prevailing level of expected volatility in a period.

During the period of high expected volatility, typical human investors might choose to hedge

their position in the market to ensure their investment value is not affected, or they might simply

liquidate their positions and stay away from the market, at least until the condition is favorable again

for them to return. Either way, they will choose to adjust their portfolios accordingly. But what about

the HFT? Similar to human traders, logically HFT should have a dynamic approach in its trading

strategies, and thus, the strategy that they apply in the period of low and high expected volatility

should be comparatively different. This difference might differently affect the market and other

participants. In extreme cases such as the Flash Crash, where the expected volatility is excessively

high, HFT are found to abandon the market given the extreme level of future uncertainty, as they have

no obligation or commitment to stay within (Anand & Venkataraman, 2013; Chung & Chuwonganant,

2018; Zhang, 2010).

Studies find that the HFT are attracted to stocks that meet certain criteria, such as those with

high relative tick size and high liquidity (ASIC, 2010; Brogaard et al., 2014; O’Hara et al., 2016; Yao,

& Ye, forthcoming). However, will they change their preference and based on the level of expected

volatility? This study will investigate whether there is a difference in the effect of HFT activity on

liquidity during the period of low and high expected volatility. Studies also suggested that HFT

activity has a significant effect on transaction costs, but they are also sensitive to transaction cost

(Carrion, 2013; Viljoen, Westerholm, & Zheng, 2014). Therefore, this study will determine whether

there is a causal relationship between HFT activity and transaction costs and whether the relationship

holds in the period with different level of expected volatility.

51

4.3 Research objectives

There are several research objectives that this study proposes to achieve as listed below:

RO1: To examine the effect of different level of expected volatility on HFT activity and preference.

RO2: To determine the effect of HFT activity on liquidity during the period of low and high level of

expected volatility.

RO3: To investigate whether there is a causal relationship between HFT activity and transaction

costs in a different level of expected volatility

4.4 Hypotheses development

Unlike human traders, HFT are not driven by emotion or sentiment. However, the aggregate action of

non-HFT traders in the market will influence HFT activity in the market due to the ripple created by

their collective action. For instance, during a period of high expected volatility, most non-HFT traders

might choose to liquidate their positions, creating a herding effect in the process. HFT will

incorporate this information and as a response, they might change their appetite and trade in stocks

with negative beta or proceed with any other course of action that will maximize their profit in such

conditions. Furthermore, since HFT are not obliged to maintain their presence in the market at all

time, if the situation is unfavorable for them, they might choose to abandon the market. Should the

HFT have been supplying most of the liquidity in the market during other periods, then the absence of

HFT at this particular time would likely cause a liquidity crisis in the market, which will exacerbate

the worsening market condition. In addition, in times of heighten uncertainty in the market,

transaction cost will increase due to a wider spread in the market which is harmful to HFT strategies

that mainly make marginal profits from every trade. Therefore, it is logical to assume that HFT

activity is negatively affected by an increase in transaction cost. However, it is worth noting that the

higher transaction cost might be due to the fleeting liquidity as HFT stop making the market.

Therefore, it can also be deduced that HFT activity is the one that negatively influence transaction

cost, at least in times of high expected uncertainty. Based on these potential scenarios, the following

hypotheses are proposed:

H1: There is a significant difference in HFT activity and preference in the period of low and high

expected volatility.

H2: There is a significant effect of HFT activity on liquidity during the period of low and high

level of expected volatility

52

H3: There is a causal relationship between HFT activity and transaction costs, and the different

level of expected volatility significantly affect this relationship.

4.5 Expected contribution of the Study

Even though there is many research studied the impact of different level of expected volatility on HFT

activity, there is no study to date, that uses Australian dataset to investigate the issue. Thus, the study

is expected to enrich the existing literature on the impact of market anxiety on HFT activity in

Australia. The study is intended to reflect the different level of expected volatility using the percentile

approach. The actual HFT activity is gauged using the order-to-trade ratio. Evidence from the study is

expected to be able to measure how HFT will react under a different level of expected volatility, and

subsequently, to examine the effect of HFT activity during those periods on liquidity. In addition, this

study is expected to provide some information on the HFT – transaction cost puzzle, by determining

the causal relationship between these variables, and investigate whether the relationship holds in a

different level of expected volatility. Overall, the results obtained from this study might give a better

understanding on HFT, while complementing the existing literature on the similar issue.

4.6 Methodology

This section discusses the methodology that will be used in this study. The data, sample, and HFT

measures in this essay is similar to the one employed in Essay Two, and details of the study period to

be used in this essay is described in Section 4.6.2. This essay ends with a brief description of the

proposed methods used to answer the research objectives.

4.6.1 Data, Sample, and HFT measures

Similar to Essay Two, the study will use the database provided by SIRCA to get the order-book data

needed to achieve the research objectives. The study will also use the constituent stocks of S&P/ASX

200 as the sample, from the year 2012 until 2017. As for HFT measures, this study will employ

similar proxy for HFT activity as described in Essay Two, which are the order-to-trade ratio, total

turnover per day, and average resting time.

53

4.6.2 Event selection

The determination of periods with the high and low level of expected volatility serves as the main

foundation for this essay, which is interested to determine the effect of different level of expected

volatility on HFT activity, using data from Australian equity market. Following the method used by

Whaley (2009), this study characterizes the normal and abnormal range of expected volatility using

percentile approach.

In Australia, the S&P/ASX 200 VIX index (RIC: AXVI) is constructed based on the

S&P/ASX 200, an index measuring the performance of the 200 largest index-eligible stocks listed on

the ASX by float-adjusted market capitalization, and widely considered as Australia’s primary

benchmark index (S&P Dow Jones Indices, 2017). Table 4.1 presents the ranges for S&P/ASX 200

VIX daily levels, from September 2010 – February 2018, which covers the entire history of the

AXVI. There are 1,880 trading days observed during the period, with a median and average closing

level of 15.02 and 16.29 respectively. The AXVI closed between 12.83 and 18.30 (a range of 5.47

index points) 50% of the time, between 11.62 and 22.08 (a range of 10.46 index points) 80% of the

time, and between 11.10 and 26.33 (a range of 15.23 index points) 90% of the time.

Table 4.1 shows large year to year variations in the values within the 25 th and 75th percentile,

in which 2011 and 2017 are the highest and lowest respectively. In 2011, the median daily closing

level of the AXVI was 20.34, ranging from 16.83 to 28.11 (a range of 11.29 index points) 50% of the

time. In 2017 on the contrary, the median was 12.34, and 50% of the time the closing level will fall

between 11.45 and 13.05, a range of only 1.60 index points. Figure 4.1 illustrates the AXJO and

AXVI price index level during the same period. From the chart, it is pretty obvious that the AXVI is

highly volatile in 2011, more so than other years, suggesting that there is a high uncertainty about the

expected direction of the market at that time. Consequently, this negative perception is manifested by

the drop in AXJO closing level during that period. In addition, the chart also shows that an increase in

AXVI value is usually associated with the fall in AXJO value.

The presence of investors’ fear can be gauged by exploring the period in which the AXVI

level persistently remains above certain threshold levels. From the data reported in Table 4.1, the odds

to see the AXVI level of higher than 22.08 is 10.0%. Using this threshold, by re-examining the

historical value of AXVI since its inception, it is possible to determine the number of consecutive

days in which the price index level consistently stays above a level of 22.08. There are two periods of

more than 10 consecutive days (equivalent to 2 trading weeks) can be identified; August 2, 2011 –

November 28, 2011 (85 days); and August 25, 2015 – October 7, 2015 (32 days).

54

Table 4.1: Normal ranges for S&P/ASX 200 VIX daily levels, from September 2010 – February 2018

Period Obs. MeanPercentile

5.0% 10.0% 25.0% 50.0% 75.0% 90.0% 95.0%All 1880 16.29 11.10 11.62 12.83 15.02 18.30 22.08 26.33

2010 72 18.15 12.71 13.56 16.36 19.04 20.04 20.63 20.922011 252 22.87 14.77 15.15 16.83 20.34 28.11 34.57 37.752012 253 16.59 12.22 12.52 14.11 15.97 18.88 21.20 23.222013 253 14.50 11.35 12.19 13.17 14.10 15.39 17.63 19.272014 253 12.91 10.27 10.76 11.44 12.53 14.01 15.57 16.632015 254 17.93 13.64 14.11 15.20 17.00 19.41 24.32 26.312016 253 16.79 11.98 12.40 13.62 16.46 18.72 22.17 23.992017 252 12.29 10.32 10.90 11.45 12.34 13.05 13.67 13.962018 38 13.71 9.46 10.32 11.09 12.21 15.94 19.51 21.53HFT 1592 15.32 10.99 11.44 12.57 14.39 17.26 20.46 23.28

As mentioned in Essay Two, HFT can be said to officially enter Australian market with the

commencement of Chi-X Australia on November 9, 2011. Using this date as the starting point for the

post-HFT period, this study finds that there are 1,592 trading days since the inception, with a median

of 14.39 and an average closing level of 15.32. The AXVI closed between 12.57 and 17.26 (a range of

4.69 index points) 50% of the time, between 11.44 and 20.46 (a range of 9.02 index points) 80% of

the time, and between 10.99 and 23.28 (a range of 12.29 index points) 90% of the time. Using the 90 th

percentile value of 20.46 from this period as the new threshold level, this study identified five (5)

periods with at least 10 consecutive days of persistent AXVI level, which are: December 9, 2011 –

January 10, 2012 (20 days); May 16, 2012 – June 15, 2012 (22 days); August 21, 2015 – October 13,

2015 (38 days); January 7, 2016 – January 29, 2016 (16 days); and February 2, 2016 – March 1, 2016

(21 days). Similarly, the period of low expected volatility can also be determined by changing the

threshold level to a lower percentile (e.g. 10.0%).

55

21-Sep-10

12-Dec-1

0

4-Mar-1

1

25-May-1

1

15-Aug-11

5-Nov-1

1

26-Jan-12

17-Apr-1

2

8-Jul-1

2

28-Sep-12

19-Dec-1

2

11-Mar-1

3

1-Jun-13

22-Aug-13

12-Nov-1

3

2-Feb-14

25-Apr-1

4

16-Jul-1

4

6-Oct-

14

27-Dec-1

4

19-Mar-1

5

9-Jun-15

30-Aug-15

20-Nov-1

5

10-Feb-16

2-May-1

6

23-Jul-1

6

13-Oct-

16

3-Jan-17

26-Mar-1

7

16-Jun-17

6-Sep-17

27-Nov-1

7

17-Feb-183600

3800

4000

4200

4400

4600

4800

5000

5200

5400

5600

5800

6000

6200

0

5

10

15

20

25

30

35

40

45

.AXJO .AXVI

S&P/

ASX

200

(pric

e in

dex

leve

l)

S&P/

ASX

200

VIX

(pric

e in

dex

leve

l)

Figure 4.1: S&P/ASX 200 and S&P/ASX 200 VIX price index level from September 2010 – February 2018

56

4.6.3 Method to address RO1

RO1: To examine the effect of different level of expected volatility on HFT activity and preference.

To fulfill the first research objective, the study will first observe HFT activity in a market during

periods of high expected uncertainty and have it compared with HFT activity during periods of low

expected uncertainty. As for HFT preference, the study will identify the characteristics of stocks that

experience a high level of HFT activity during the two periods and determine whether HFT preference

is affected to the level of unexpected uncertainty in the market.

4.6.4 Method to address RO2

RO2: To determine the effect of HFT activity on liquidity during the period of low and high level of

expected volatility.

The study will use ordinary least square (OLS) model to answer the second research objective. The

study will regress the dependent variable (liquidity measures) against key independent variable

(proxies for HFT activity) and control for the level of expected volatility in the market.

4.6.5 Method to address RO3

RO3: To investigate whether there is a causal relationship between HFT activity and transaction

costs in a different level of expected volatility.

To address the final objective, the study will use Granger causality approach, which allows for an

estimation of the causal relationship between two variables, which are HFT activity and transaction

costs. The results of from this method will determine whether HFT activity is affected transaction

costs, or is it vice versa while controlling for the level of expected volatility in the market.

57

BIBLIOGRAPHY

Agarwal, A. (2012). High frequency trading: Evolution and the future. London, UK.

Aitken, M., & Comerton-Forde, C. (2005). Do reductions in tick sizes influence liquidity? Accounting

and Finance, 45, 171-184.

Aldridge, I. (2013). High-frequency trading: A practical guide to algorithmic. New Jersey: John

Wiley & Sons, Inc.

Anand, A., & Venkataraman, K. (2013). Should exchanges impose market maker obligations?

Working paper.

Angel, J. J. (1997). Tick Size, Share Prices, and Stock Splits. Journal of Finance, 52(2), 655-681.

Angel, J. J. (2012). Tick Size study mandated by the JOBS Act. Washington, D.C.

Angel, J. J. (2014). When finance meets physics: The impact of the speed of light on financial markets

and their regulation. The Financial Review, 49, 271–281.

Angel, J. J., & McCabe, D. (2013). Fairness in financial markets: The case of high frequency trading.

Journal of Business Ethics, 112(4), 585–595 |.

Angrist, J. D., & Pischke, J.-S. (2008). Mostly Harmless Econometrics: An Empiricist's Companion.

Princeton university press.

Aquilina, M., & Ysusi, C. (2016). Are high-frequency traders anticipating the order flow? Cross-

venue evidence from the UK market. Working paper.

Australian Securities and Investments Commission. (2010). Australian equity market structure.

Victoria.

Australian Securities and Investments Commission. (2013). Dark liquidity and high-frequency

trading. Report 331.

Australian Securities and Investments Commission. (2015). Review of high-frequency trading and

dark liquidity. Report 452.

Australian Securities Exchange. (2011). PureMatch to go-live 28 November. Sydney.

58

Avramovic, A., Lin, V., & Krishnan, M. (2017). We’re all high frequency traders now. Credit Suisse.

Benos, E., & Sagade, S. (2016). Price discovery and the cross-section. Journal of Financial Markets,

30, 54–77.

Bessembinder, H. (2003). Trade execution costs and market quality after decimalization. Journal of

Financial and Quantitative Analysis, 38(4), 747-777.

Biais, B., & Woolley, P. (2011). High Frequency Trading. Working paper.

Biais, B., Foucault, T., & Moinas, S. (2015). Equilibrium fast trading. Journal of Financial

Economics, 116, 292–313.

Boehmer, E., Fong, K., & Wu, J. (2015). International evidence of algorithmic trading. Working

paper.

Brogaard, J. (2010). High frequency trading and its impact on market quality. Northwestern

University Kellogg School of Management Working Paper, 66.

Brogaard, J., Hendershott, T., & Riordan, R. (2014). High-frequency trading and price discovery. The

Review of Financial Studies, 27(8), 2267–2306.

Budish, E., Cramton, P., & Shim, J. (2015). The high-frequency trading arms race: Frequent batch

auctions as a market design response. Quarterly Journal of Economics, 130(4), 1547–1621.

Cao, C., Hansch, O., & Wang, X. (2009). The information content of an open limit-order book.

Journal of Futures Markets, 29(1), 16-41.

Carrion, A. (2013). Very fast money: High-frequency trading on the NASDAQ. Journal of Financial

Markets, 16, 680–711.

Cartea, A., & Penalva, J. (2012). Where is the Value in High Frequency Trading? Quarterly Journal

of Finance, 2(3), 1-46.

Cespa, G., & Foucault, T. (2014). Sale of Price Information by Exchanges: Does It Promote Price

Discovery? Management Science, 60(1), 148-165.

Chordia, T., Goyal, A., Lehmann, B. N., & Saar, G. (2013). High-frequency trading. Journal of

Financial Markets, 16, 637-645.

Chordia, T., Roll, R., & Subrahmanyam, A. (2011). Recent trends in trading activity and market

quality. Journal of Financial Economics, 110, 243-263.

59

Chung, K. H., & Chuwonganant, C. (2018). Market volatility and stock returns: The role of liquidity

providers. Journal of Financial Markets, 37, 17-34.

Chung, K. H., & Lee, A. J. (2016). High-frequency trading: Review of the literature and regulatory

initiatives around the world. Asia-Pacific Journal of Financial Studies, 45, 7–33.

Comerton-Forde, C. (2012). Is Australia HFT-friendly? JASSA The Finsia Journal of Applied

Finance, 3, 12-14.

Committee of European Securities Regulators. (2010). Call for evidence on micro-structural issues of

the European equity markets. Paris.

Conrad, J., Wahal, S., & Xiang, J. (2015). High-frequency quoting, trading, and the efficiency of

prices. Journal of Financial Economics, 116, 271-291.

Cooper, R., Davis, M., & Vliet, B. V. (2016). The mysterious ethics of high-frequency trading.

Business Ethics Quarterly, 26, 1-22.

Donefer, B. S. (2010). Algos gone wild: Risk in the world of automated trading strategies. Journal of

Trading, 5(2), 31-34.

Durbin, M. (2010). All about high-frequency trading: The easy way to get started. New York:

McGraw-Hill Companies, Inc.

Easley, D., Prado, M. M., & O’Hara, M. (2011). The microstructure of the “Flash Crash”: Flow

toxicity, liquidity crashes, and the probability of informed trading. Jounal of Portfolio

Management, 37(2), 119-128.

Easley, D., Prado, M. M., & O'Hara, M. (2012). The volume clock: Insights into the high-frequency

paradigm. Journal of portfolio management, 19-29.

European Securities and Markets Authority. (2011). Guidelines on systems and controls in a highly

automated trading environment for trading platforms, investment firms and competent

authorites. Paris.

Financial Industry Regulatory Authority. (2012). FINRA Joins Exchanges and the SEC in fining Hold

Brothers more than $5.9 million for manipulative trading, anti-money laundering, and other

violations. Washington, D.C.

Foucault, T., Hombert, J., & Rosu, I. (2016). News trading and speed. Journal of Finance, 71(1),

335–382.

60

Frino, A., Mollica, V., & Zhang, S. (2015). The impact of tick size on high frequency trading: The

case for split.

Froot, K. A., Scharfstein, D. S., & Stein, J. C. (1992). Herd on the street: Informational inefficiencies

in a market with short‐term speculation. Journal of Finance, 47(4), 1461-1484.

Garvey, R., & Wu, F. (2010). Speed, distance,and electronic trading:New evidence on why location

matters. Journal of Financial Markets, 13, 367–396.

Gibson, S., Singh, R., & Yerramilli, V. (2003). The effect of decimalization on the components of the

bid-ask spread. Journal of Financial Intermediation, 12, 121–148.

Goldstein, M. A., & Kavajecz, K. A. (2000). Eighths, sixteenths, and market depth: changes in tick

size and liquidity provision on the NYSE. Journal of Financial Economics, 56, 125-149.

Goldstein, M. A., Kumar, P., & Graves, F. C. (2014). Computerized and high-frequency trading.

Financial Review, 49, 177–202.

Golub, A., Dupuis, A., & Olsen, R. B. (2013). High-frequency trading in FX markets. In D. Easley,

M. L. Prado, & M. O’Hara, High-Frequency Trading New Realities for Traders, Markets and

Regulators (pp. 65-90). London: Risk Books.

Gomber, P., Arndt, B., Lutat, M., & Uhle, T. (2011). High-frequency trading.

Hagströmern, B., & Nordén, L. (2013). The diversity of high-frequency traders. Journal of Financial

Markets, 16, 741–770.

Harris, L. (2003). Trading and exchanges: Market microstructure for practitioners. New York:

Oxford University Press.

Harris, L. (2013). What to do about high-frequency. Financial Analysts Journal, 69(2), 6-9.

Harris, L. (2015). Trading and Electronic Markets: What Investment Professionals Need to Know.

Research Foundation of CFA Institute.

Harris, L. E. (1994). Minimum price variations, discrete bid-ask spreads, and quotation sizes. Review

of Financial Studies, 7(1), 150-178.

Hasbrouck, J., & Saar, G. (2013). Low-latency trading. Journal of Financial Markets, 16(4), 646-679.

Hendershott, T., & Riordan, R. (2013). Algorithmic trading and the market for liquidity. Journal of

Financial and Quantitative Analysis, 48(4), 1001–1024.

61

Jarnecic, E., & Snape, M. (2014). The provision of liquidity by high-frequency participants. The

Financial Review, 49, 371-394.

Jones, C. M. (2013, March 20). What do we know about high-frequency trading? Columbia Business

School Research Paper No. 13-11.

Jones, C. M., & Lipson, M. L. (2001). Sixteenths: direct evidence on institutional execution costs.

Journal of Financial Economics, 59, 253-278.

Kauffman, R. J., Hu, Y., & Ma, D. (2015). Will high-frequency trading practices transform the

financial markets in the Asia Pacific Region? Financial Innovation, 1, 1-27.

Kauffman, R. J., Liu, J., & Ma, D. (2015). Innovations in financial IS and technology ecosystems:

High-frequency trading in the equity market. Technological Forecasting and Social Change,

99, 339–354.

Kaya, O. (2016). High-frequency trading reaching the limits. Deutsche Bank Research.

Kirilenko, A., Kyle, A. S., Samadi, M., & Tuzun, T. (2017). The flash crash: High-frequency trading.

The Journal of Finance, 72(3), 967–998.

Kosinski, R. J. (2013). A literature review on reaction time. Working paper.

Kozhan, R., & Tham, W. W. (2012). Execution risk in high-frequency arbitrage. Management

Science, 58(11), 2131-2149.

Lhabitant, F.-S., & Gregoriou, G. N. (2015). High-frequency trading: Past, Present and future. In G.

N. Gregoriou, Handbook of high frequency trading (pp. 155-165). Academic Press.

Lipson, M. L., & Mortal, S. (2006). The effect of stock splits on clientele: Is tick size relevant?

Journal of Corporate Finance, 12, 878– 896.

Malinova, K., & Park, A. (2015). Subsidizing liquidity: The impact of make/take fees on market

quality. Journal of Finance, 70(2), 509–536.

Manahov, V., Hudson, R., & Viktor, B. G. (2014). Does high frequency trading affect technical

analysis and market efficiency? And if so, how? Journal of International Financial Market,

Institutions and Money, 28, 131– 157.

Markham, J. W. (2002). A financial history of the United States. 1. From Christopher Columbus to

the Robber Barons (1492 -1900). New York: M. E. Sharpe.

62

Menkveld, A. J. (2014). High-frequency traders and market structure. The Financial Review, 49, 333-

344.

Moosa, I., & Ramiah, V. (2015). The profitability of high-frequency trading: Is it for real? In The

Handbook of high-frequency trading (pp. 25-45). Academic Press.

Narang, R. K. (2013). Inside the black box: A Simple Guide to Quantitative (Second Edition ed.).

New Jersey: John Wiley & Sons, Inc.

Netherlands Authority for the Financial Markets. (2010). High frequency trading: The application of

advanced trading technology in the European marketplace. Amsterdam.

O’Hara, M. (2015). High frequency market microstructure. Journal of Financial Economics, 116,

257–270.

O’Hara, M., Saar, G., & Zhong, Z. (2016). Relative tick size and the trading environment.

Unpublished working paper.

O'Hara, M. (2003). Presidential Address: Liquidity and price discovery. Journal of Finance, 58(4),

1335-1354.

Riordan, R., & Storkenmaier, A. (2012). Latency, liquidity and price discovery. Journal of Financial

Markets, 15, 416-437.

Scholtus, M. n., Dijk, D. v., & Frijns, B. (2014). Speed, algorithmic trading, and market quality

around macroeconomic news announcements. Journal of Banking and Finance, 38, 89-105.

Schultz, P. (2000). Stock Splits, Tick Size, and Sponsorship. Journal of Finance, 55, 429-450.

Securities and Exchange Commission. (2010). Concept Release on Equity Market Structure.

Washington, D.C.

Securities and Exchange Commission. (2014). Equity market structure literature review Part II: High

frequency trading. Washington, D.C.

Serbera, J.-P., & Paumard, P. (2016). The fall of high-frequency trading: A survey of competition and

profits. Research in International Business, 36, 271–287.

Shorter, G., & Miller, R. S. (2014). High-frequency trading: Background, concern and regulatory

developement. Congressional Research Service.

The Government Office for Science. (2012). Foresight: The future of computer trading in financial

markets - Final Project Report. London.

63

U.S. Commodity Futures Trading Commission and U.S. Securities & Exchange Commission.

(2010a). Preliminary findings regarding the market events of May 6, 2010. Washington, D.C.

U.S. Commodity Futures Trading Commission and U.S. Securities & Exchange Commission.

(2010b). Findings regarding the market events of May 6, 2010. Washington, D.C.

Viljoen, T., Westerholm, P. J., & Zheng, H. (2014). Algorithmic trading, liquidity, and price

discovery: An intraday analysis of the SPI 200 Futures. Financial Review, 245-270.

Vives, X. (1995). Short-term investment and the information efficiency of the market. Review of

Financial Studies, 8(1), 125-160.

Whaley, R. E. (1993). Derivatives on market volatility: Hedging tools long overdue. Journal of

Derivatives, 1(1), 71-84.

Whaley, R. E. (2000). The investor fear gauge. Journal of Portfolio Management, 26(3), 12-17.

Whaley, R. E. (2009). Understanding the VIX. Journal of Portfolio Management, 35(3), 98-105.

Wissner-Gross, A. D., & Freer, C. E. (2010). Relativistic statistical arbitrage. Physical Review, 82(5),

1-7.

Yao, C., & Ye, M. (Forthcoming). Why trading speed matters: A tale of queue rationing under price

controls. Review of Financial Studies.

Zhang, S., & Riordan, R. (2011). Technology and market quality: The case of high frequency trading.

ECIS 2011 Proceedings. European Conference on Information Systems.

Zhang, X. F. (2010). High-frequency trading, stock volatility, and price discovery. Working paper.

64

PROPOSED TIMELINE FOR THE COMPLETION OF DISSERTATION

Schedule Task

March 2018 Ph.D. confirmation

April 2018 – June 2018 Completing Essay 1

July 2018 – December 2018 Essay 2

January 2019 – May 2019 Essay 3

June 2019 Linking the three essays into the dissertation

July 2019 Submitting a draft copy of the dissertation

September 2019 Submitting a bound copy of the dissertation

65