Are Odd‐Lot Orders Informed?

• DOI: http://doi.org/10.1111/fire.12127
• Date: 01 February 2017
• Published date: 01 February 2017
• Author: Hardy Johnson,James Upson

The Financial Review 52 (2017) 37–67

Are Odd-Lot Orders Informed?

James Upson

University of Texasat El Paso

Hardy Johnson∗

Kansas State University

Abstract

Using a version of the ITCH data set time stamped to the millisecond, O’Hara, Yao and

Ye find that odd-lot trades are highly informed. However, NASDAQ reports trades based on

the size of the resting limit order, creating a bias in the count of odd-lot trades. Using ITCH

data from 2013, time stamped to the nanosecond, we find that roughly 50% of odd-lot trades

are created by the resting limit order and are part of larger marketable orders. We show that

odd-lot marketable orders are not more informed than round/mixed lot marketable orders.

Keywords: TAQ data, odd-lots, price discovery,transparency, order imbalance, retail trading

JEL Classifications: G10, G14

1. Introduction

The information content of orders in equity markets is an area of interest for

academics, practitioners, and regulators. The size of a marketable order submitted

∗Corresponding author: College of Business Administration, Kansas State University, 2097 Business

Building, 1301 Lovers Lane, Manhattan, KS 66506; Phone (785) 532-6992; Fax: (785) 532-6822; E-mail:

bhjohnson@ksu.edu.

The authors would like to thank Thomas McInish for many helpful comments on this research. We also

thank Robert VanNess and Christine Jiang for helpful comments. In addition, we thank Mao Ye for helpful

comments. Any mistakes in this paper are our own.

C2017 The Eastern Finance Association 37

38 J. Upson and H. Johnson/The Financial Review 52 (2017) 37–67

for execution is a strategic choice of the order initiator (Brennan and Subrahmanyam,

1998). With the advent of algorithmic trading, the use of small trades to fill large

equity positions is economically feasible, including the use of odd-lot orders for less

than 100 shares. Prior to December 10, 2013, the consolidated tape did not include

odd-lot trades. A first major finding of O’Hara, Yao and Ye (2014), who use the

NASDAQ high-frequency trading data, is that 24% of trade executions are odd-lots

and therefore excluded from the consolidated tape. In a second major finding, O’Hara,

Yao and Ye show that the fact that odd-lot trades are not reported to the consolidated

tape leads to a bias in trade and volume estimations. We strongly believe that these

two findings are a significant contribution to the literature.

Our study focuses on two additional findings of O’Hara, Yao and Ye (2014),

namely, that (1) odd-lot trades are highly informed, constituting 35% of the price

discovery, and (2) odd-lot trades are used by informed traders to hide from the

consolidated tape. Using NASDAQ ITCH data, time stamped to the nanosecond,

we are able to identify the size of the marketable order submitted for execution,

which may be reported as a series of trades in the ITCH data. The NASDAQ high-

frequency trading data used by O’Hara, Yao and Ye are time stamped only to the

milliseconds, so exact marketable order identification is not possible.1We show that

odd-lot orders are not more informed than round lot or mixed lot orders (100+share

orders hereafter). Our analysis indicates that 51% of odd-lot trades printed in the

ITCH database come from the structure of the limit order book (LOB), rather than

the size of the incoming marketable order, meaning that the majority of odd-lot trades

are part of larger marketable orders.

Table 1 contains a selectiveportion of the NASDAQ ITCH data which we use to

illustrate the difference between trade prints and order size. In an electronic LOB, a

trade can be reported based on the size of the incoming marketable order or based on

the size of the resting order in the limit order queue. NASDAQ reports trades based on

the size of the resting limit order. We show (Table 1, row 23) a 300 share marketable

order printed in the ITCH data as trades of size 7, 100, 80, 6, 12, and 95 shares, all of

which execute in the same nanosecond. These six executions are printed separately

because of the orders resting in the LOB when the incoming marketable order of 300

shares arrives. O’Hara, Yao and Ye (2014) analyze odd-lot trades, round lot trades

(trades of increments of 100 shares), and mixed lot trades (trades greater than 100

shares but not increments of 100 shares) separately. In this example, we believe that

the one 300 share order may be included in the their analysis as five informed odd-lot

trades and one less informed 100+share trade. Our analysis focuses on the size of

the marketable order, not the size of the trade reports in the ITCH data.

1O’Hara, Yao and Ye (2014) attempt to consolidate trades in table 8 of their analysis, but their results

remain unchanged. Webelieve that the coarseness of their time stamp does not allow the accurate estimation

of the size of the marketable order. Upson, Johnson and McInish (2015) showthat consolidation of trades

by summing all trades at the same price and same millisecond leads to a statistically and economically

biased estimation of order size and distribution.