Trading and information in futures markets
| Author | Guillermo Llorente,Jiang Wang |
| Published date | 01 August 2020 |
| DOI | http://doi.org/10.1002/fut.22079 |
| Date | 01 August 2020 |
J Futures Markets. 2020;40:1231–1263. wileyonlinelibrary.com/journal/fut © 2019 Wiley Periodicals LLC
|
1231
Received: 20 November 2019
|
Accepted: 20 November 2019
DOI: 10.1002/fut.22079
RESEARCH ARTICLE
Trading and information in futures markets
Guillermo Llorente
1
|
Jiang Wang
2
1
Facultad de C. Económicas, Universidad
Autónoma de Madrid, Madrid, Spain
2
Sloan School of Management, CAFR and
NBER, Massachusetts Institute of
Technology, Cambridge, Massachusetts
Correspondence
Guillermo Llorente, Facultad de C.
Económicas, Universidad Autónoma de
Madrid, 28049 Madrid, Spain.
Email: guiller@uam.es
Funding information
Ministerio de Economía y Competitividad,
Grant/Award Numbers: ECO02012‐
32554, ECO2017‐85356‐P;
Fundación BBVA
Abstract
This paper studies the trading behavior of different types of traders (customer
type indicators [CTI’s]) in corn futures. Nonmembers (CTI4) consume most of
the intraday liquidity while local traders (CTI1) as market makers are its main
provider. Both groups combine most of the intraday trading volume. Interday
trading comes mainly from proprietary accounts (CTI2) and other local traders’
trades (CTI3), reflecting their longer‐term needs for hedging and speculation.
Changes in the overnight positions of the general public (CTI4) and clearing
members (CTI2) contribute mostly to daily price discovery, while the positions
of CTI3 group reflect possible information advantage about future price
movements.
KEYWORDS
CTI, futures, liquidity, price discovery
JEL CLASSIFICATION
G10; G12; G13; G14; G18
1
|
INTRODUCTION
A financial market allows different market participants to meet their trading needs. In doing so, it serves the purpose of
liquidity provision and price discovery. Despite a body of theoretical analysis, our understanding of how the market
performs these functions remains limited, especially empirically.
1
This is mainly due to the lack of comprehensive and
detailed information on who trade in the market, why they trade, and how they trade. This type of information is often
proprietary, even at relatively aggregated levels. In this paper, we utilize a unique data from the futures market, which
categorizes trades according to trader types, to analyze the trading behavior of each type. In particular, we are able to
characterize their trading behavior, their gains and losses, their role in liquidity provision and price discovery, and their
potential informational advantage. These results allow us to gain more insight into how the futures market functions
and serves its participants as a financial market, different from the usual “commodity”perspective.
The Liquidity Data Bank (LDB) compiled by the Chicago Board of Trade (CBOT)/Chicago Mercantile Exchange
(CME) categorize each trade by its Customer Type Indicator (CTI).
2
It identifies each trade according to four customer
types: for a member’s own account (CTI1), a commercial clearing member’s proprietary accounts (CTI2), another
member’s own account (CTI3), or a customer (CTI4). Roughly speaking, CTI1 represents market makers, CTI2
1
Theoretical analysis often relies on somewhat abstract and simplified formulation of the heterogeneity in investors’trading needs and information. Grossman (1976), Grossman and Stiglitz (1980),
Kyle (1985) provide the basic analytical framework in analyzing the price discovery process in financial markets. Grossman and Miller (1988), among others, consider the underlying mechanism for
liquidity consumption and provision. Wang (1994) describes a dynamic model incorporating both the needs to trade for risk‐sharing/liquidity and for information motivated speculation. See also
Vayanos and Wang (2012) for a survey on the related theoretical literature.
2
During the period of our sample, the data was compiled and provided by CBOT, which merged with CME to form the CME Group in July 2007.
represents the proprietary accounts of exchange clearing members, CTI3 represents other member traders, and CTI4
represents the general public (i.e., nonmembers).
3
We only consider corn futures, for the reason that private
information may be more important for commodity futures, especially agricultural. From LDB, we can construct the
transactions by each CTI group during a day and by cumulating these transactions we can construct their closing
positions for each trading day.
We find that for corn futures, the four types of traders exhibit very different trading behavior and play different roles
in the market. First, CTI group 1 and 4 constitute most of the intraday trading and maintain little overnight positions. In
particular, group CTI1 contributes around 50–60% of the intraday volume and group CTI4 contributes around 30–40%.
In comparison, group CTI2 and CTI3 contribute about 5% each. To the contrary, CTI group 2 and 3 contribute the most
to interday trading, carrying most of the overnight positions, for about 30–40% each. The contribution to the interday
trading volume is minuscule for group CTI1 and less than 10% for group CTI4. We also find that the relative shares of
the intraday trading is quite stable for groups CTI1 and CTI4, the two dominant groups, but the relative shares of
interday trading is highly variable for the two main contributing groups, CTI2 and CTI3. This pattern of trading
behavior suggests that group CTI4 conducts mostly intraday trading, likely to speculate on very short‐term price trends.
Groups CTI2 and CTI3 are mostly trading for longer‐term motivations, such as hedging, market making or speculation.
And these trading needs can change substantially over time.
The trading patterns of different CTI groups suggest that group CTI4 is the main consumer of intraday liquidity
while group CTI1 is the main provider. As a result, we find that group CTI1 consistently earns profits from its intraday
trading while group CTI4 consistently loses in intraday trading. In addition, both the profits of groups CTI1 and CTI4
exhibit strong time consistency, with low variability. The Sharpe ratio for CTI1 group’s intraday trading exceeds 0.79,
while it is −0.58 for CTI4 traders. Group CTI2 generally breaks even in intraday trading while group CTI3 also loses
money on intraday trading, but at a lower level than group CTI4 as it trades significantly less. We further show that the
profits and losses of CTI1 and CTI4 groups are positively related to unexpected changes in daily turnover and price
volatility.
Using the day to day changes in each CTI group’s end of day positions, we also examine its relative importance in
price discovery. We start by looking at the contemporaneous correlation between changes in market prices and market‐
wide variables. We find that predicted change in turnover and imbalance between buy and sell volume exhibit a
significant correlation with daily price changes over the same day. After controlling for these variables, we further show
that only changes in group CTI4’s daily closing positions exhibit additional explanatory power for contemporaneous
price changes.
Given the different trading needs of the four CTI groups, we further examine their potential information advantage.
We find that changes in the overnight positions of group CTI3 can forecast the price change of the following day. In
contrast, changes in positions of other CTI groups have no predictive power for future price movements. Using
nonparametric analysis, we further show that changes in CTI3’s overnight positions can also predict higher moments of
the price change in the following day. In particular, an increase in their overnight position predicts a positive skewness
and higher kurtosis, while a decrease predicts negative skewness and lower kurtosis. In addition, while other CTI
groups’overnight positions tend to have mixed correlations across contracts with different maturities, group CTI3’s
positions are significantly positively correlated across maturities. These results suggest that group CTI3 collectively
possess private information that is not fully reflected in market prices.
The results above paint an informative picture about the overall trading needs of different CTI groups, how they
trade in the market, and their role in liquidity provision/consumption and in price discovery. Clearly, group CTI1 serves
as market makers, mainly providing intraday liquidity to other traders, mostly to the general public. Group CTI2 trades
for longer‐term (more than a day), with time‐varying needs. Their trades have limited information content, indicating
that they trade mostly for risk‐sharing or hedging reasons. Group CTI3 also trades for the longer term, but its trades
clearly contain information beyond what is reflected in market prices. Group CTI4, the general public, consists of
mostly short term traders. It consumes most of the short‐term liquidity, incurring nontrivial loses. However, it is also
through group CTI4’s trades (at least some of them), market prices move to reflect more fundamental information.
Compared with many other financial markets, more disaggregated data is available for the futures market, partially
because of the exchanges’desire to promote transparency and the unique regulatory environment it faces. Many
researchers have obtained various types of disaggregated data on trading and analyzed the behavior of different market
3
The actual description of the CTI is given later in the paper, with more details in Appendix A.2. As of July 2, 2015, the CME closed most of its trading floors, including the corn pit. Nevertheless, the
CTI classification continues to apply for electronic trades. See CME Rule Book.
1232
|
LLORENTE AND WANG
participants. Earlier work primarily focus on intraday trading and liquidity provision. Kuserk and Locke (1993) are
among the earliest to use CTI related data to analyze their trading behavior. They have a short sample period (only 3
months), but more detailed information on the trading records of a set of traders within each CTI group for a set of
futures.
4
They mainly focus on the short‐term trading profit of “scalpers,”those in group CTI1 identified as market
makers, and find similar results as we do for the whole CTI1 group.
5
Using a similar data set, Manaster and Mann (1996) further examine the market‐making behavior of a sample of
scalpers in their intraday trading. In particular, they look at the relationship between market makers’inventory,
customer spread, market depth and price variability over short (1‐,5‐or 15‐min) intervals. They also find that at 1‐min
intervals, the net flow of customers (CTI4) can predict price changes over the next minute. They did not examine the
behavior of CTI2 and CTI3 groups.
6
Wiley and Daigler (1998) use data similar to ours on several financial futures to
analyze the daily trading volume for all four CTI groups, in particular, their dynamics and cross‐correlation. Daigler and
Wiley (1999) further examine the relationship between price volatility and unexpected CTI group’s trading volume.
They do not separate intraday versus interday trading and their connection with price discovery and information
asymmetry.
Brandt, Kavajecz, and Underwood (2007) use the CTI data on Treasury futures to examine the role of different CTI
groups in the price discovery process for both the futures and cash market at a daily frequency. Their methodology is
similar to ours, as in Evans and Lyons (2002), among others. They find that changes in group CTI4’s overnight positions
are significantly positively correlated with contemporaneous price changes for Treasury futures, which we also find for
the corn futures. Since they bundle together groups CTI2 and CTI3, they find a significant negative correlation between
changes in their total overnight positions and price changes. We consider the two groups separately and show that such
a relationship is only weakly significant for group CTI2. Our results on the predicative power of group CTI3’s position
changes for future price changes indicate its distinctive information advantage.
More recently, several authors have conducted fruitful analysis based on new data sets with detailed
informationonpositionrecords.Dewally,Ederington, and Fernando (2013) utilize the data from Commodity
Futures Trading Commission’s(CFTC’s) Large Trader Reporting System (LTRS) to study the interday trading
profits of large traders in the futures on energy‐related products. The LTRS contains information of the daily
closing positions of all large traders whose open positions exceed a certain threshold. The advantage of this data is
it contains more information on trade types for these large traders.
7
This information allows a finer grouping of
traders according to their “physical”characteristics, such as refiners versus hedge funds and market makers.
8
In
fact, Dewally et al. (2013) formed 11 groups from the set of large traders and analyze their interday trading
patterns, profits, and possible determinants. Grouping by “physical”characteristics, however, may not best reflect
the “financial”characteristics of these traders such as liquidity demanders/providers versus information motivated
speculators. The market makers using LTRS characteristics contain mostly CTI1, but also CTI3. The other groups
refine but also mix CTI2 and CTI4 groups in the LDB data set. Instead of using the physical characteristics, Fishe
and Smith (2012) rely on the performance of different traders to help identifying informed traders and then
examine the nature of their potential information, its relationship with trader characteristics and trading profits.
This empirical approach can potentially reveal additional information about a trader, but it is also subject to the
limitations of statistical inference, including the accuracy of the underlying hypotheses.
9
Based on the complete
transaction record on S&P 500 futures, Locke and Onayev (2007) examine the relationship between customer order
flow, price change, and dealer predictability at 5‐min intervals. They find significant connections among these
Several papers, such as Sanders, Irwin, and Merrin (2009), Brunetti, Buyuksahin, and Harris (2016), Buyuksahin and
Harris (2011), utilize CFTC’s reported daily closing positions from the Commitments of Traders (COT) data to test the
Granger causality and led–lag relationships between changes in futures prices and trades. The evidence mostly finds
4
The data Kuserk and Locke (1993) used is based on the audit transaction trail data from the exchanges, which records all the trades occurred on the exchange floor with the trader identifications. It
covers a set of futures, including eight financials and four agricultural. Trader ID can be further grouped into CTI categories. The data also gives the time stamp of each trade up to a minute, with some
errors. This data used by researchers typically consists of a random sample from the whole population.
5
See also Silber (1984) for an earlier study on the trading records of two scalpers in stock index futures, showing their role in liquidity provision at a profit.
6
Based on the LDB data on 30‐year Treasury futures, Menkveld, Sarkar, and van der Wel (2012) show thatnet customer volume (CTI4) contributes to price discovery over 15‐min intervals. They do not
consider the order flow of other CTI groups.
7
See Haigh, Hranaiova, and Overdahl (2005) for a more detailed discussion of the LTRS data set.
8
The CFTC trader classification is commonly referred to as done by the “business line of activities.”Ederington and Lee (2002) discuss the accuracy of this classification.
9
For example, to make inferences about the intraday performance and its source, Fishe and Smith (2012) have to rely on several hypotheses in measuring performance and its determinants.
LLORENTE AND WANG
|
1233
To continue reading
Request your trial