An examination of the relationship between trading volume and price volatility on the CME-SIMEX link

ST

':np

FACULTY WORKING
PAPER NO. 1175

u&

TH§

UNiv

An Examination of the Relationship Between Trading
Volume and Price Volatility on the CME-SIMEX Link

Joseph E. Finnerty
Kuang-Chung Chen
Michael J. Shelley

College of Commerce and Business Administration
Bureau of Economic and Business Research
University of Illinois, Urbana-Champaign

BEBR

FACULTY WORKING PAPER NO. 1175
College Qf Commerce and Business Administration
University of Illinois at Urbana- Champaign
September, 1985

An Examination of the Relationship Between Trading
Volume and Price Volatility on the CME-SIMEX Link.

Joseph E. Finnerty, Associate Professor
Department of Finance

Kuang-Chung Chen, Assistant Professor
Department of Finance

Michael J. Shelley
Ford Motor Company

An Examination of the Relationship Between Trading Volume
and Price Volatility on the CME-SIMEX Link

ABSTRACT

For DM futures contracts traded on the Chicago Mercantile Exchange
(CME) and the Singapore International Monetary Exchange (SIMEX) a posi-
tive relationship between trading volume and price volatility was found,

A strong positive relationship was found between the volume of
trading on the CME and the subsequent trading volume on the SIMEX.
Whereas the price volatility on the CME did not appear to be strongly
related to the subsequent price volatility on the SIMEX.

Digitized by the Internet Archive

in 2011 with funding from

University of Illinois Urbana-Champaign

http://www.archive.org/details/examinationofrel1175finn

An Examination of the Relationship Between Trading
Volume and Price Volatility on the CME-SIMEX Link

The opening of the CME-SIMEX Link in September 1984 provides a
unique opportunity to study the relationship between price and volume
movements for the IMM (International Monetary Market) Deutschemark
Futures Contracts. The SIMEX (Singapore International Monetary
Exchange) , formerly the Gold Exchange of Singapore is a futures market
almost identical to the IMM of the CME (Chicago Mercantile Exchange),
except that it is considerably newer and smaller than the CME, and
that it is, of course, located in Singapore instead of Chicago. The
Deutschemark Futures Contracts traded on either exchange are identical
and thus fungible. There exists a fourteen hour difference in time
between Chicago and Singapore, so that a 24 hour trading day with
respect to Chicago time would consist of an 8:00 a.m. opening in
Chicago trading until 2:00 p.m., then a five hour non trading period
followed by a 7:00 p.m. opening on the SIMEX and trading until 1:00
a.m., followed by a seven hour non trading period.

The existence of these two non overlapping trading periods, during
which the exact same contract is traded allows for the investigation of
the relationship between price volatility and volume at divergently
different levels of trading activity. In the next section a review of
the literature on the relationship between price and volume is presented.
Section II describes the hypotheses to be tested and the data and
methodology used in the study. The results and conclusions are given
in the final section.

-2-

I. Literature Review

Much of the work in the area of price and volume relationships has
been done in the evaluation of equity securities. An early study by
Ying (1966) looked at the relationship between the S&P 500 stock index
and the New York Stock Exchange daily trading volume. He found that
large price changes appeared to be associated with a large volume of
trading, that large volumes are associated with price increases, and
that small volumes accompany price declines. Although Ying's findings
are flawed by methodological problems, he has presented an interesting
issue.

Epps (1975) develops a theoretical argument for the relationship
between volume and price change. He shows that the number of shares
exchanged on a transaction in which price rises exceeds the volume
accompanying a price decline of the same magnitude. Empirically he
shows this to be true for the secondary market for publicly traded
bonds. In a follow on study, Epps (1977) also confirms this finding
for common stocks.

Cohen, Maier, Schwartz and Whitcomb (1978) develop a theoretical
model which establishes that the variance of a security's return will
be inversely related to thinness of trading. Under a specific set of
assumptions, the thinly traded issues will be more volatile.

Morse (1980) examined the effect of asymmetrical information on
securities prices through an analysis of the relationship between
trading volume and prices. He found that during periods of high trading
volume, serial correlations of returns residuals are high. And that

-3-

trading on the day before large return residuals is significantly dif-
ferent from trading on the day before small return residuals, thereby
indicating a linkage between trading volume and returns.

James and Edmister (1983) hypothesize that if differences in
trading activity are the cause of the return differences associated
with firm size due to the existence of a liquidity premium, then an
inverse relationship between mean daily returns and trading activity
should be observed. This is similar to the relationship expressed by
Cohen, Maier, Schwartz, and Whitcomb (1978). James and Edmister find
that there is no significant difference between the mean returns of the
highest and lowest trading activity portfolio, nor is there any evidence
of an inverse relationship between trading activity and mean daily
returns.

The literature on the volume and price relationship for common
stock trading is in a state of flux, with no definitive theory or evi-
dence predominating. The literature on the volume and price relation-
ship for futures trading is rather sparse. Rutledge (1984) and
Stansell (1983) investigate the direction of causality in the volume
and price relationship for 15 commodities futures and Treasury Bill
Futures respectively. Rutledge finds that in 31 of 136 contracts
studied, that price variability "causes" trading volume. In only 2
cases does he find that volume "causes" price variability. And in the
remaining 103 cases he can't identify a relationship or no significant
relationship exists. Stansell investigates nine different T Bill
contracts using four different methodologies. He finds four cases in

-4-

which causality goes from volume to prices; five cases in which causa-
lity goes from prices to volume; eight cases of bidirectional causality;
and 19 cases where there was no statistical relationship.

Cornell (1984) studied the relationship between volume of trading
and price variability for futures contracts of 18 commodities. He found
a significant, positive, contemporaneous correlation between changes in
average daily volume and changes in the standard deviation of daily log
price relatives for 14 of the 18' commodities studied. For the remaining
4, the relationship was found to be positive although not significant.

As it is in the equity markets, the relationship between volume and
price volatility has not been clearly identified nor empirically tested
to a satisfactory degree. This paper serves to fill this gap and pro-
vide a basis for inquiry into the interesting question. A useful exten-
sion of the findings of this paper can be made in linking the volume of
trading and the systematic risk of securities. Dimson (1979) provides
a methodology for estimating the Beta of infrequently traded (low
volume) securities. Building on Dimson' s methodology, if it can be
shown that volume of trading is systematically related to price volati-
lity a better measure of systematic risk may be developed. In the next
section of this paper, a linkage between volume and price volatility
will be hypothesized and the data and methodology used to test the
hypothesis is described.

II. Data, Hypotheses and Methodology

This study examines the relationship between price volatility and
trading volume for the Deutschmark Mark Futures Contract, which has
traded on the CME and the SIMEX since the 7th of September 1984. The

-5-

period of study is from the initiation of trading (7 September 1984) to
21 June 1985. This period encompassed the trading of three contracts,
December, March, and June. The measure of price volatility was the
daily range, or the high price of the day minus the daily low price.
This measure of volatility was used rather than the more conventional
variance of the rate of return on investment, because the initial
investment in a futures position is zero. Additionally using the rate
of return on initial margin would not be of any use, because not all
investors face the same margin requirements and T-Bills can be used to
meet the margin requirements, thereby making the opportunity cost of
the margin equal to zero. The volume is measured by the number of con-
tracts traded on a given day on each of the exchanges.

Price range information and trading volume were gathered for the
near DM contract on both exchanges. Missing data and/or holidays on
one or both of the exchanges resulted in a different number of trading
days for each contract. Table I presents descriptive statistics for
the data set used in this study.

The relationship between volume and price volatility are investi-
gated by four hypotheses:

H The average price volatility on the CME is greater than the
average price volatility on the SIMEX for the DM contract.
The average volume of trading in the CME is greater than the
average volume of trading on the SIMEX for the DM contract.

H A positive relationship exists between the relative volume
the CME-SIMEX and the relative price volatility on the CME-
SIMEX for the DM contract.

on

-6-

Table I. Statistics of the Data Set

Contract &

# of

Standard

Exchange

Variable

Observation

Mean

Deviation

Minimum

Maximum

A. December

CME

Range

66

0.0033

0.0017

0.0012

0.0128

Volume

66

22030

9325

2579

47498

SIMEX

Range

66

0.0013

0.0009

0.0001

0.0050

Volume

66

579

352

8

1635

B. March

CME

Range

48

0.0025

0.0014

0.0010

0.0085

Volume

48

17115

7427

2441

37861

SIMEX

Range

48

0.0009

0.0008

0.0002

0.0033

Volume

48

422

481

7

3132

C . June

CME

Range

61

0.0038

0.0018

0.0014

0.0122

Volume

61

24494

10116

2546

50125

SIMEX

Range

61

0.0021

0.0017

0.0000

0.0111

Volume

61

865

550

2

2417

-7-

H~ The intercept and slope of the regression equation for the

volatility and volume relationship on the CME is equal to the
intercept and slope for the volatility and volume relationship
on the SIMEX.
H, The volume of trading on the CME is positively related to the

subsequent volume of trading on the SIMEX. The price volatility
on the CME is positively related to the subsequent price vola-
tility on the SIMEX.
The first hypothesis was tested by use of a standard t-test for the
equality of means. It tested whether the price volatility and volume
were significantly greater for the CME than for the SIMEX. Affirmation
of this hypothesis would indicate on average that for the DM futures
contract high volume was associated with high price volatility and low
volume was associated with low price volatility.

The second hypothesis examines the relationship between the relative
volume on the CME-SIMEX and the relative price volatility on the CME-
SIMEX. The correlation coefficient, P, was calculated for the natural
log of the volume on the CME divided by the natural log of the volume
on the SIMEX, with the price range of the CME divided by the price range
of the SIMEX. In notational form the second hypothesis is shown in
equation 1:

in Volr Rr

p(i^r- r> > ° (1)

A high correlation would indicate that a large volume of trading
was associated with a large amount of price fluctuation.

-8-

The third hypothesis was investigated by the use of a regression
equation with qualitative and quantitative explanatory variables. The
dummy variable regression is:

R - e + 3D + 82£n(Vol) + 33D*in(Vol) + e (2)

where

R. is the daily price volatility
Vol is the daily volume of contracts

D is tj

0 for CME

for SIMEX

3 , 3 , 3 , 3^ are the regression parameters

The third hypothesis in notational form is:

Bj = 33 = 0 (3)

which would indicate that there is no difference between the relation-
ship of price and volume on the CME and price and volume on the SIMEX.

The final hypothesis dealt with the relationship between trading on
the CME and the subsequent trading on the SIMEX. Two regressions were
run. The first regression evaluated the relationship between the volume
of trading on the SIMEX and the previous days volume of trading on the
CME. Likewise the price volatility on the SIMEX was regressed against
the previous days price volatility on the CME. The regressions are
shown in equation 4.

Vols,t ■ ao + V0lc,t-i + e . (4a)

Rs,t - % + 6iRc,t-i + e (4b)

-9-

We felt that the trading behavior on the CME and SIMEX were similar
enough to cause the residuals of equations 4a and 4b to be correlated.
Hence we used the Seemingly Unrelated Regression (SUR) approach on both
of the equations. We hypothesized that the a, and 3 would be signifi-
cantly positive. High volume of trading on the CME are followed by
high volume of trading on the SIMEX. And large price fluctuations on
the CME are followed by large price fluctuations on the SIMEX.

The results of the four parts of the study are presented in the
following section.

III. Results and Conclusion

Table II presents the results of the t-tests for the equality of
the mean trading volume and mean price fluctuations on the CME and
SIMEX.

In all cases the volume of trading and price volatility on the CME
significantly exceeded the volume of trading and price volatility on
the SIMEX. We can conclude that a low volume of trading is associated
with a small amount of price volatility and a large amount of trading
is associated with a large amount of price volatility. This empirical
finding supports the finding of Cornell (1984) of a positive relation-
ship between volume and price volatility.

An extension of this research would be to determine the nature of
this positive relationship between volume and volatility. Various
forms such as a step function, linear relationship, or non linear rela-
tionship can be envisioned suggesting various types of phenomena occur-
ring as the volume of trading increases.

-10-

Table II. T-test Results

Contract &

# of

Exchange

Variable

Observation

Mean

T-statistic

A. December

CME

Range

66

0.0033

8.21**

SIMEX

Range

66

0.0013

CME

Volume

66

22030

27.30**

SIMEX

Volume

66

579

B. March

CME

Range

48

0.0025

7.20**

SIMEX

Range

48

0.0009

CME

Volume

48

17115

20.46**

SIMEX

Volume

48

422

C . June

CME

Range

61

0.0038

5.67**

SIMEX

Range

61

0.0021

CME

Volume

61

24494

18.55**

SIMEX

Volume

61

865

** = Significant at 1% level.

-11-

Additional evidence supporting a strong positive relationship
between volume and volatility was the large highly significant correla-
tion coefficients of relative volume with relative price volatility.
These results are presented in Table III.

The third hypothesis which dealt with the relationship between
volume and volatility in both markets was supported by the results for
the December and March contracts. However for the June contract the
results indicate that there is a significant difference in the volume
and volatility relationships between both markets. Table IV presents
the results of the dummy variable regression.

The mixed results of the B and 3_ coefficients being not signifi-
cantly different from zero for the December and March contracts and
being very significant for the June contract prevents us from reaching
a conclusion about the similarity or difference of the volume and vola-
tility relationship of the CME and SIMEX. As more contract cycles are
completed and more data become available, this issue will be able to be
resolved.

The final area of interest was the relationship between volume on
the CME and subsequent volume on the SIMEX and price volatility on the
CME and subsequent price volatility on the SIMEX. This issue was
approached from two different ways. Independent regressions for each
relationship and seemingly unrelated regressions (SUR) for both rela-
tionships simultaneously. The results of these two approaches are
presented in Tables V and VI respectively.

Under either method, a strong positive relationship was found
between trading volume on the CME and next day trading volume on the

-12-

Table III. Correlation Coefficients for Relative Volume and Volatility

in Volp Rp
PHn Vols' Rg;

Contract
December
March
June

Correlation Coefficient

.82**

,80**
,85**

** =

Significant at 1% level

-13-

Table IV. Dummy Variable Regression Results

R = 0 + 3D + 3 In(VOL) + 33D*£n(VOL) + e

where R ■ range ;

1 if SIMEX

D = i

0 if CME

VOL = Volume

Contract

0

1

r

December -0.00666 0.00504 0.00010 -0.00053 33.63** 0.441
(-2.55)* (1.79) (3.82)** (-1.68)

March

-0.00319 0.00342 0.00059 -0.00047 19.75** 0.392
(-1.19) (1.24) (2.13)* (-1.55)

June

-0.01184 0.01035 0.00157 -0.00101 28.90** 0.424

(-3.86)** (3.24)** (5.12)** (-3.00)**

** = Significant at 1% level.
* = Significant at 5% level.

-14-

Table V. Independent Regression Results

£n(V0LSt) = 3Q + 81£n(V0LCt_1) + e

where S = SIMEX and C = CME

Contract

6o

-5.035
(-3.75)**

1.127
(8.29)**

F
68.66**

R2

December

0.518

March

-6.906
(-2.62)*

1.288
(4.71)**

22.18**

0.425

June

-12.056
(-7.11)**

1.841
(10.84)**

117.50**

0.666

Rst " 8o + 6iRct-i + £

Contract

0.0012

6i

0.0471

F
0.53

R2

December

0.001

(4.76)**

(0.72)

March

0.0002
(1.01)

0.270
(3.61)**

13.04**

0.22:

June

0.0014
(2.80)**

0.168
(1.40)

1.96

0.03:

** = Significant at 1% level.
* = Significant at 5% level.

-15-

Table VI. Seemingly Unrelated Regression Results

*n(VOLSt) = eQ + 31in(V0LCt_1) + e

Rst = ao + aiRct-i + e

Contract

0

R

December

-9.247
(-3.88)**

1.531
(6.20)**

0.470

March

-11.015
(-6.84)**

1.736
(10.76)**

0.499

June

-3.836
(-2.96)**

1.006
(7.66)**

0.419

Contract

December

March

June

a

a.

0.0001
(0.66)

0.304
(4.50)**

0.0010
(2.11)*

0.270
(2.37)*

0.0010
(4.29)**

0.091
(1.44)

R

0.370

0.499

0.319

** = Significant at 1% level.
* = Significant at 5% level.

-16-

SIMEX. For all three contracts the slope coefficient was significantly

2
positive and the R were relatively high. This indicates that a high

volume day on the CME will usually precede a high volume day in the

SIMEX.

For the pricing relationship between volatility on the CME and next

day volatility on the SIMEX the results are not as clear cut. From

Table V we see that only the March contract has a significantly positive

slope coefficient indicating support for the hypothesis. However the

December and June contracts have a slope not significantly different

2
than zero and very low R indicating very little relationship between

CME and SIMEX price volatility. The SUR results for the price volatility

are somewhat more encouraging. The December contract has a significantly

positive slope at the 1% level and the March contract has a significantly

2
positive slope at the 5% level and the R s are much higher.

Overall we can conclude that there is a very strong relationship
between the volume of trading on the CME and the volume of trading on
the SIMEX. The relationship between the markets for price volatility
is not as strong.

The research studied the relationship between the volume of trading
and price variability for DM futures contracts being traded on the CME
and SIMEX. A significant positive relationship between volume of
trading and price volatility was found. The correlations between rela-
tive price and relative volume movements on the CME-SIMEX link was
found to be quite high.

In assessing the relationship between volume and trading on the CME
and volume of trading on the SIMEX, a very strong positive relationship

-17-

was found. Whereas these results on the price volatility relationship
between the two models is mixed.

The results of this research are encouraging given that less than
one years worth of data on only one type of contract was available.
Further research when more data becomes available in terras of time and
contracts is warranted.

-18-

Ref erences

K. Cohen, Maier, S. , Schwartz, R. & D. Whitcomb, "The Returns Generation
Process, Returns Variances and the Effect of Thinness in Securities
Markets," Journal of Finance, 33, March 1978, pp. 149-168.

R. Cornell, "The Relationship Between Volume and Price Volatility in
Futures Markets," in Selected Writings in Futures Markets, Ed.
A. Peck, Chicago Board of Trade, 1984, pp. 253-266.

E. Dimson, "Risk Measurement When Shares are Subject to Infrequent
Tradings," Journal of Financial Economics, 7, June 1979, pp.
197-226.

T. Epps , "Security Price Changes and Transaction Volume: Theory and
Evidence," The American Economic Review, 65, September 1975, pp.
586-597.

, "Security Price Changes and Transaction Volume: Some Addi-

tional evidence," Journal of Financial and Quantitative Analysis,
12, March 1977, pp. 586-597.

C. James and R. Edmister, "The Relation Between Common Stock Returns,

Trading Activity, and Market Value," Journal of Finance, 38,
September 1983, pp. 1075-1086.

D. Morse, "Asymmetrical Information in Securities Markets and Trading

Volumes," Journal of Financial and Quantitative Analysis, 15,
December 1980, pp. 1129-1143.

D. Rutledge, "Trading Volume and Price Variability: New Evidence on
the Price Effects of Speculation," in Selected Writings in Futures
Markets, Ed. A. Peck, Chicago Board of Trade, 1984, pp. 237-251.

S. Stansell, "A Study of the Causal Relationships Between Treasury
Bill Futures Prices and the Volume of Futures Traded," Quarterly
Journal of Business and Economics , 22, December 1983, pp. 3-24.

C. Ying, "Stock Market Prices and Volumes of Sales," Econometrica,
July 1966, 34, pp. 676-685.

D/265A