Periodicity of trading activity in foreign exchange markets

• Published date: 01 June 2022
• Author: Tao Chen,Kam C. Chan,Haodong Chang
• Date: 01 June 2022
• DOI: http://doi.org/10.1111/jfir.12280

Received: 7 May 2020

|

Accepted: 9 November 2021

DOI: 10.1111/jfir.12280

ORIGINAL ARTICLE

Periodicity of trading activity in foreign exchange

markets

Tao Chen

1

|Kam C. Chan

2

|Haodong Chang

3

1

Faculty of Business Administration,

University of Macau, Taipa, Macau

2

Research Center of Finance, Shanghai

Business School, Shanghai, China

3

School of Accounting, Zhongnan University

of Economics and Law, Wuhan, China

Correspondence

Kam C. Chan, Research Center of Finance,

Shanghai Business School, Shanghai, China.

Email: johnny.chan@wku.edu

Funding information

Multi‐Year Research Grant at the University

of Macau, Grant/Award Number:

MYRG2020‐00042‐FBA

Abstract

Using the high‐frequency exchange rates of 25 global

currency pairs, we document a striking clock‐time perio-

dicity in which trading activity surges at the beginning of a

minute. Additional analyses indicate that clock‐time spikes

are accompanied by a lower level of liquidity. Moreover,

we find that time‐clustering trades yield permanent price

impacts, are devoted to efficient pricing, and make a

significant cont ribution to price d iscovery. Finall y, we

investigate three informed scenarios to ascertain how

trades at spikes acquire information beforehand and

reflect them in markets. Taken together, our findings

reinforce the view in the literature that subminute

periodicity emanates from algorithmic trading.

KEYWORDS

clock‐time periodicity, Foreign exchange markets, Price discovery

JEL CLASSIFICATION

F31, G12, G14

1|INTRODUCTION

Studies suggest that the periodicity of trading activity appears ubiquitous in various asset classes. For example,

periodic patterns have been identified in bond (Bollerslev et al., 2000), commodity (Cai et al., 2001), futures

(Wu et al., 2005), stock (Heston et al., 2010), foreign exchange (Boudt et al., 2011), and cryptocurrency (Wang et al.,

2020) markets. Most studies have focused on seasonality in diverse low‐frequency horizons, such as annual,

monthly, weekly, daily, and intraday. Hasbrouck and Saar (2013) reveal high‐frequency periodicity in which intraday

trading volume tends to display a substantial increase at the start of a second. This trading regularity is consistent

J Financ Res. 2022;45:445–465. wileyonlinelibrary.com/journal/JFIR

|

445

© 2022 The Southern Finance Association and the Southwestern Finance Association.

with subsecond periodicity and likewise occurs every minute and hour (Broussard & Nikiforov, 2014). Despite the

extensive examination of this phenomenon in equity markets, currency markets, which capture a large trading

volume involving round‐the‐clock transactions from global dealers, have not commanded much attention. To fill this

literature gap, the first purpose of this study is to examine whether clock‐time periodicity of trading activity prevails

in foreign exchange markets.

Unlike traditional low‐frequency periodicity, the spike in transactions at the turn of a round‐time mark seems

independent of human beings in that none could react swiftly and punctually enough to produce such a discernible

pattern within a shorter time frame. By contrast, clock‐time periodicity is more likely to originate from algorithmic

participation (Broussard & Nikiforov, 2014; Chen, 2021). Depending on specific objectives, previous studies have

divided algorithmic activity into two categories: agency and proprietary algorithms (Hasbrouck & Saar, 2013).

Agency algorithms enable buy‐side large institutions, such as mutual funds, insurance firms, pension funds, and

brokers, to split large orders into smaller ones and then route them to different trading venues over time. This

activity aims to minimize execution costs and price impacts. These computer programs mostly adopt trading volume

within a preset period to determine how to divide up institutions’orders. For example, the volume‐weighted

average price (VWAP) algorithm places its slice of the overall order, whose size is proportional to the trading

volume, in the equivalent interval on the prior trading day. From this perspective, the manner by which execution

algorithms (i.e., VWAP) break up each trading day into multiple periods of fixed length presumably induces high‐

frequency periodicity (Hasbrouck & Saar, 2013). Because of psychologically biased human programmers, schedule‐

based algorithms are easily compiled to monitor market conditions and trade assets at the round‐number time

point, which lends further credence to the idea of agency algorithms generating clock‐time periodicity (Muravyev &

Picard, 2016).

Similarly, the observed periodicity may stem from proprietary algorithms attributable to astute traders (e.g.,

hedge funds) for several reasons. First, smart algorithms can harness the order anticipation strategy that exploits

the predictability of order flows from agency counterparts (Sağlam, 2020). Second, given the liquidity demand from

agency algorithms surrounding a round‐time mark, sophisticated traders are incentivized to compete with

designated dealers by implementing a passive market‐making strategy (Kirilenko & Lo, 2013). Third, to leverage

private information under imperfect competition, informed (algorithmic) traders act aggressively toward the end of

a period (Holden & Subrahmanyam, 1992).

A large body of literature provides evidence that algorithmic trading in equity markets enhances the

informativeness of quotes (Hendershott et al., 2011). In this way, algorithmic trading supplies liquidity when it is

expensive (Hendershott & Riordan, 2013), reduces execution shortfall for buy‐side institutions (Boehmer et al.,

2020), curtails the spread (Hasbrouck, 2018), and absorbs order imbalances surrounding extreme price movements

(Brogaard et al., 2018). Overall, such a positive role of algorithms in market quality primarily pertains to their

comparative advantage in speed, which benefits from proprietary data feeds from exchanges (Ding et al., 2014),

tiered information disclosures (Hu et al., 2017), and index futures information (Jovanovic & Menkveld, 2016). These

algorithms prompt clock‐time periodicity, and such biased trading behavior is expected to affect the price formation

of exchange rates. Therefore, the second purpose of this study is to shed light on the economic consequences of

clock‐time periodicity on market quality.

We analyze a sample of 25 major global currencies and find that trading activity intensifies sharply in the first 1

s of a minute and continues until it declines rapidly to its average for the remaining interval. Such a remarkable spiral

of transactions over time is consistent with Hasbrouck and Saar's (2013) discovery of an analogously dynamic

pattern in the stock market. Taken together, our findings substantiate the prevalence of subminute periodicity in

the foreign exchange market.

Next, we investigate whether trades at spikes matter to two crucial elements of market quality: liquidity and

price discovery. In terms of the liquidity dimension, we document that both quoted and effective spreads climb at

the beginning of a minute. Meanwhile, a large Amihud illiquidity measure is closely connected to trades at spikes.

Parallel to the asymmetric information model (Foster & Viswanathan, 1990), these results can be interpreted as

446

|

JOURNAL OF FINANCIAL RESEARCH