Você está na página 1de 8

Entropy and Futures Contracts

Isaac Carruthers
August 6, 2015

Introduction Dierent futures products show markedly dierent distributions of


activity across contract months. For some products, trading is concentrated almost
exclusively in the front-month contract; in other products many dierent contracts
are active on any given day. Futures contracts fall into three robust categories:
A. Traded volume is concentrated in the front month, as for the 10-year Treasury futures contract (Figure 3). Products in category A have all their traded
volume concentrated in the single front-month maturity, except for brief wellstructured roll periods in which volume is briey split between the front month
and the deferred month. Government debt, equity index, and foreign exchange
products belong to this category.
B. Traded volume is divided among a handful of maturities, often with seasonal
eects such as a concentration in the December contract. Rolls are complex
events, as trade activity shifts among a collection of dierent maturities. Examples are crude oil (Figure 4) and corn (Figure 5). Energy futures and agricultural
futures both belong to category B.
C. Traded volume is spread across many maturities . Products in category C are
generally short-term interest rate (STIR) futures such as Eurodollars (Figure 6).
Our goal is to systematize and quantify this categorization. For example, are
these three categories distinct, or are contracts placed along a continuous range
from one active maturity to many? Do the categories as described above align with
product class? We nd that there is a surprisingly neat separation of products
under these categories, and that with the possible exception of precious metals and
the Canadian Bankers Acceptance (BAX) futures, these categories align well with
product class.
We use intraday and end-of-day data obtained from CME, Eurex, ICE Europe (formerly LIFFE), and Montral exchanges, looking at all products in which our rm has
consistently traded signicant volumes. The data shown in Figure 1 and Figure 2
were taken from across one year, from July 2014 through June 2015. Table 1 shows
summary statistics over the same period. The products are in six distinct classes:
equity index, energy, foreign exchange (FX), metal, agricultural, and interest rate
(which we subdivide into STIR and government debt).
1

Isaac Carruthers

August 6, 2015

Class

Product

Electronic
Symbol

Bloomberg
Symbol

ADV
(Thousands)

STIR

Eurodollar
Three-Month Sterling
Three-Month Euro EURIBOR
Canadian Bankers Acceptance

GE
L
I
BAX

ED
L
ER
BA

2,488
490
337
94

Government Debt

US Ten-Year Note
US Five-Year Note
Euro-Bund
Euro-Bobl
US Two-Year Note
US Thirty-Year Bond
Euro-Schatz
Euro-BTP
US Ultra Bond
Euro-OAT
Canadian Ten-Year Note
Euro-Buxl
Short-Term Euro-BTP

ZN
ZF
FGBL
FGBM
ZT
ZB
FGBS
FBTP
UB
FOAT
CGB
FGBX
FBTS

TY
FV
RX
OE
TU
US
DU
IK
WN
OAT
CN
UB
BTS

1,140
713
678
410
293
292
228
84
76
72
66
31
18

Equity Index

E-mini S&P 500


Euro STOXX 50
E-mini NASDAQ 100
DAX Index
FTSE 100 Index
S&P Toronto 80
E-mini S&P Midcap 400

ES
FESX
NQ
FDAX
Z
SXF
EMD

ES
VG
NQ
GX
Z
PT
FA

1,467
1,038
253
121
112
19
17

Energy

WTI Crude Oil


Natural Gas
RBOB Gasoline
Heating Oil

CL
NG
RB
HO

CL
NG
XB
HO

631
267
121
116

Foreign Exchange

Euro
Japanese Yen
British Pound
Australian Dollar
Canadian Dollar
Mexican Peso
Swiss Franc
New Zealand Dollar

6E
6J
6B
6A
6C
6M
6S
6N

EC
JY
BP
AD
CD
PE
SF
NV

241
151
98
96
63
42
32
19

Metal

Gold
Copper
Silver

GC
HG
SI

GC
HG
SI

151
56
37

Agricultural

Corn
Soybeans
Wheat
Soybean Oil
Soybean Meal
KC Wheat

ZC
ZS
ZW
ZL
ZM
KE

C
S
W
BO
SM
KW

263
184
105
91
77
24

Table 1: Futures products traded on CME, Eurex, ICE Europe (LIFFE), and Montral exchanges. Average daily volume is measured across one year from July 2014 through
June 2015. Colors are those used in Figures 1 and 2. We have divided interest rate
futures into short term interest rate (STIR) and government debt.

Isaac Carruthers

August 6, 2015

Diversity measures Suppose that 1 , , are a partition across bins, with


each 0 1 and 1 + + = 1. In our application, the are the fractions
of traded volume in each of futures maturities. We want to calculate a diversity
index or eective number describing how spread-out these values are.
Diversity measures have been widely studied in ecology and in economics. A
useful family can be constructed as follows, tracing the analysis of Hill [1973] and
Jost [2006]. For = (1 , , ) with restrictions on as above, and for 0, we
write the generalized mean as
1/

() =

1/

where = () denes the standard mean. In particular, 1 () = ,


() = max , and 0 () = exp log (in the limit 0). Then the effective number of order is
() =

1
1 ()

If has the value 1/ for dierent indices and is zero elsewhere, so that the
weight is spread evenly across dierent maturities, then () = for > 0.
This motivates the use of the term eective number.
It is easy to see that () is increasing in and hence is decreasing in .
Smaller values of assign a more uniform weight to elements of the partition, giving
relatively more weight to less common elements. In the limit 0, converges
to the total number of nonzero .
Two values of in particular are in widespread use:
1 () = exp log is the exponential of the Shannon entropy, originally
developed for use in signal processing (Shannon [1948]).
2 () = 1/ 2 is directly related to the Simpson index in ecology and
to the Herndahl-Hirschman index in economics, after Simpson [1949] and
Hirschman [1945] respectively. The latter is used by the US Justice Department to evaluate the monopoly risk of proposed mergers.
For our study we prefer 1 , because of its theoretical elegance and because of its
focus on the broad distribution.
We therefore x our denition of eective number
e () = exp log .

This denition systematizes our intuitive ideas about the number of actively traded
contracts of a futures product, extending our ability to rank products by the diversity of their trades.
In this study we apply this construction to daily traded volume, normalizing the
values on each day so that the values for all contracts on the same product sum to
1. For the purple lines in Figures 36, we plot e as the value on each day. The
range of observed values of e for each product can be seen in Figure 1.

Isaac Carruthers

August 6, 2015

15

10
8
6
4
3
2
1
EMD
SXF
ES
NQ
Z
FDAX
FESX
R
FGBX
FBTS
FOAT
CGB
FBTP
UB
FGBM
FGBL
ZN
ZB
FGBS
ZF
ZT
6N
6B
6J
6A
6S
6E
6C
6M
SI
GC
HG
ZW
KE
ZC
ZS
ZL
ZM
RB
CL
HO
NG
BAX
L
I
GE

Effective Number of Traded Contracts

20

Symbol

Figure 1: Eective number of daily traded contracts for each product considered.
For each product we show the median (bold midline), center 50% interval (box), and
total range (whiskers) for the daily value of e for that product.
Empirical Findings As shown in Figure 2, government debt, foreign exchange, equity index, and precious metal products all trade a very small eective number of
contracts. These category A products are all almost entirely traded on just one or
two contracts, but show signicant variation in their average daily traded volume,
as shown by the spread along the vertical axis. Further right on the plot, category
B energy and agricultural products tend to be traded on 37 contracts, indicating
a more gradual roll period. Finally, the short-term interest rate products tend to
trade a high eective number of contracts, indicating that these category C products spread their trading activity relatively evenly over the available contracts.
Detailed Examination We can improve our understanding further by examining
detailed day-by-day graphs of daily traded volume and entropy. Data is compiled
from daily and end-of-day data from CME, Eurex, ICE Europe (LIFFE), and Montral
exchanges. In all of these graphs, the daily traded volume is shown by green or
gold lines and measured in thousands of lots, and the eective number of traded
contracts is shown in purple.
Figure 3 shows the 10-year Treasury futures contract (ZN), as an example of a
typical roll structure. This pattern is typical of government debt, equity index, and

Isaac Carruthers

August 6, 2015

GE

Average Daily Volume (lots)

1M

100k

ES
ZN
FESX
ZF
FGBL
FGBM
ZT
ZB
ZC
NQ
6E
FGBS
ZS
R
6JGC
FDAX
Z
ZW ZL
6B
6A
FBTP
ZM
UB
FOAT
CGB
6C HG
6M
6SSI
FGBX
KE
6N
SXF
FBTS
EMD

CL

L
I

NG
RB
HO

BAX

10k

5
10
Effective Number of Traded Contracts

15

Figure 2: Eective number versus average daily volume. Category A products have
an eective number close to one, and cluster near the left of the graph. Other classes
of instrument can be accurately distinguished by their eective number. The vertical
dashed line marks an eective number of 1, which is the lowest value possible.
foreign exchange (FX) futures. On most dates, all the trade volume is concentrated in
the single front month contract. The entropy is near zero and the eective number
of contracts is one. Only around the roll periods is traded volume split between two
contracts. The eective number of contracts rises as high as two.
Figures 4 and 5 show the WTI Crude Oil (CL) and Corn (ZC) futures. These products show an intermediate roll structure, which is typical of energy and agricultural
contracts. In this structure, trade volume is divided among a handful of dierent
maturities, but divided in a predictable way which repeats itself from one roll period to the next. In these plots, December contract volumes are shown in gold to
highlight the seasonal variation in trading patterns.
Figure 6 shows the Eurodollar complex, which is typical of STIR products in having its traded volume distributed across a wide range of product maturities.
Conclusion By evaluating the entropy of a products traded contracts, we can quantitatively categorize the nature of that products roll-structure. This categorization
also strongly aligns with the class of the asset underlying the futures contracts. This
quantication of the roll-structure yields a useful tool for understanding the diverse
dynamics of dierent futures markets.

Mar '13

Sep '13

Jun '12

Jun-19

Mar-20

Dec-19

Sep-19

Jun '13
Mar '14

1000 Mar '12

Jun-19

Mar-20

Dec-19

Sep-19

Effective Number of Contracts

Thousands of Lots

1500

Jun-19

Mar-19

Dec-19

Sep-19

August 6, 2015

Jun-20

2000 ZN

Mar-21

Isaac Carruthers

Jun '14 Sep '14

Mar '15
Dec '14

Dec '13

Sep '12 Dec '12

500

0
Jun '12

Dec '12

Jun '13

Dec '13

Jun '14

1
Jun '15

Dec '14

Figure 3: Daily traded volume and entropy of 10-year Treasury futures. Daily traded
volumes are shown in green, and eective number of contracts shown in purple.
Grey vertical lines mark contract expiration dates.
Mar '15

Apr '15
May '15

Thousands of Lots

350
NovDec
'14 '14
Jan Feb
'15 '15

300

12
10
8

250

Jun '13
JulAug
'13 '13
Sep Oct
'14 '14
Sep Oct
'13 '13 Dec '13
NovDec
'12 '12
May '13
Nov '13 Jan '14 Mar Apr
'14 Jun
'14
'14
May
'14
'13 '13
Jan '13 Mar Apr
200
Aug '14
Jul '14
Feb '13
Oct '12
Feb '14

150

100

50
0

Effective Number of Contracts

400

Sep-24
Oct-24
Nov-20
Dec-21
Jan-24
Feb-22
Mar-22
Apr-24
May-23
Jun-24
Jul-24
Aug-22
Sep-24
Oct-24
Nov-22
Dec-23
Jan-23
Feb-24
Mar-24
Apr-24
May-22
Jun-24
Jul-24
Aug-22
Sep-24
Oct-23
Nov-24
Dec-23
Jan-22
Feb-24
Mar-24
Apr-23
May-21
Jun-24

450 CL

0
Dec '12

Jun '13

Dec '13

Jun '14

Dec '14

Jun '15

Figure 4: Daily traded volume and entropy of crude oil futures. Daily traded volumes
are shown in green (December contracts in gold), and eective number of contracts
shown in purple. Grey vertical lines mark contract expiration dates.

Mar-13

Thousands of Lots

Mar '15

Dec '13

6
5

May '15

Effective Number of Contracts

Dec '14

Mar '14

150

May-14

Dec-12

Sep-12

Jul-14

Mar-14

Dec-13

200 ZC

August 6, 2015

May-14

Isaac Carruthers

May '14

Jul '14

100

Sep '14

3
50

1
Dec '13

Jun '14

Dec '14

Jun '15

Figure 5: Daily traded volume and entropy of corn futures. Daily traded volumes
are shown in green (December contracts in gold), and eective number of contracts
shown in purple. Grey vertical lines mark contract expiration dates.
Jan-16
Feb-13
Mar-19
Apr-16
May-14
Jun-18
Jul-16
Aug-13
Sep-17
Oct-15
Nov-19
Dec-17
Jan-14
Feb-18
Mar-18
Apr-15
May-13
Jun-17
Jul-15
Aug-19
Sep-16
Oct-14
Nov-18
Dec-16
Jan-13
Feb-17
Mar-17
Apr-14
May-19
Jun-16
Jul-14
Aug-18
Sep-15
Oct-13
Nov-17
Dec-15
Jan-19
Feb-16
Mar-16
Apr-13
May-18
Jun-15

600 GE

20

400

Effective Number of Contracts

Thousands of Lots

500

25

15
300

10
200 Mar '12
100

Jun '12

Mar '15
Mar '13 Jun '13

Sep '12
Dec '12

Mar '14
Sep '13 Dec '13

Jun '14

Sep '14 Dec '14

0
Jun '12

Dec '12

Jun '13

Dec '13

Jun '14

Dec '14

5
0
Jun '15

Figure 6: Daily traded volume and entropy of Eurodollar futures. Daily traded volumes are shown in green, and eective number of contracts shown in purple. Grey
vertical lines mark contract expiration dates.

Isaac Carruthers

August 6, 2015

References
M. O. Hill. Diversity and evenness: A unifying notation and its consequences. Ecology, 54(2):427432, March 1973.
A. O. Hirschman. National Power and the Structure of Foreign Trade. University of
California Press, expanded edition, 1945. ISBN 0-520-04082-1.
L. Jost. Entropy and diversity. Oikos, 113(2):363375, 2006.
C. E. Shannon. A mathematical theory of communication. The Bell System Technical
Journal, 27:379423,623656, July and October 1948.
E. H. Simpson. Measurement of diversity. Nature, 163:688, April 1949.
Disclaimer This document contains actual performance results achieved, but past performance is
not necessarily indicative of future results. Trading futures and options on futures involves signicant
risk and may result in unlimited losses. Futures trading is not suitable for all investors. QB oers
execution services to institutional investors exclusively.