A Study of

Castorseed Futures Market in India

by

Kiran Karande

A thesis

submitted to the

Indira Gandhi Institute of Development Research

in partial fulfilment of the requirements

for the award of the degree of

Doctor of Philosophy

MMVI

Indira Gandhi Institute of Development Research

Mumbai, India

Electronic copy of this paper is available at: http://ssrn.com/abstract=983342

Electronic copy of this paper is available at: http://ssrn.com/abstract=983342
Contents

List of Tables viii

List of Figures ix

Abstract x

1 Introduction 1

1.1 Evolution of commodity futures markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Thesis design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.4 Basis risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.5 Price discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.6 Spot price volatility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.6.1 Destabilising effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.6.2 Stabilising effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.6.3 Volatility and information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Institutional Features 11

2.1 The castorseed economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.2 Harvest of castorseed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.3 Castorseed production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.4 Futures trading volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.5 Basis behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.6 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.7 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3 Basis Risk 19

3.1 Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

v
 vi CONTENTS

3.2 Basis risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.3.1 Basis models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.4 Performance evaluation of alternative models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.4.1 Root mean squared error (RMSE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.4.2 Ashley F test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.4.3 Henrikson Merton (H-M) timing test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.5 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.6.1 Root mean squared error (RMSE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.6.2 Ashley F test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.6.3 Henrikson Merton (H-M) timing test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4 Price Discovery 28

4.1 Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4.2 Price discovery across markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.3 Price discovery over time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.4 Price discovery and cointegration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.5 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

4.6 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.7 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.7.1 Unit root test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.7.2 Cointegration test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.7.3 Price discovery within market at Mumbai . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.7.4 Price discovery within market at Ahmedabad . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.7.5 Price discovery across markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.7.6 Price discovery over time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.8 Inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5 Spot Price Volatility 42

5.1 Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

5.2 Spot price volatility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

CONTENTS vii

5.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5.3.1 Traditional regression technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5.3.2 Generalised AutoRegressive Conditional Heteroscedasticity (GARCH) . . . . . . . . . . . . . . 47

5.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

5.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.5.1 Traditional regression technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.5.2 Generalised AutoRegressive Conditional Heteroscedasticity (GARCH) . . . . . . . . . . . . . . 50

5.5.3 Likelihood ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

5.5.4 Futures trading volume and spot price volatility . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

6 Conclusion 53

6.1 Institutional features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

6.2 Basis risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

6.3 Price discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

6.4 Spot price volatility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.5 Policy recommendation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Bibliography 58

Index 67
List of Tables

2.1 Average annual production of castorseed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.2 Average annual futures trading volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.3 Average annual castor oil export . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.4 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.5 Descriptive statistics – spot returns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.1 Time series GARCH model of the basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.2 Root mean squared error (RMSE) results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.3 Ashley F test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.4 Henrikson-Merton (H-M) timing test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4.1 Unit root test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.2 Cointegration test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.3 Error correction model results: Mumbai . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.4 Error correction model results: Ahmedabad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.5 Error correction model results: Across markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.6 Cointegration and price discovery inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

5.1 Traditional regression technique results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.2 GARCH(1,1) model for castorseed spot returns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5.3 Likelihood ratio test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

viii
List of Figures

2.1 Basis behaviour for different contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

ix
Abstract

Historically, the Government has intervened at every stage of the marketing of major agricultural commodities. The
Government policy has been to protect and promote the agriculture sector through procurement and administered price
mechanism. However, in view of reduced direct support to agriculture under the Agreement on Agriculture with the World
Trade Organisation (WTO), there is a policy shift towards market oriented approach. Government intervention declined
after initiation of economic liberalisation and reforms in 1991. To help manage price risk, the Government has been
encouraging commodity derivatives. A key aspect of the process of strengthening agricultural markets is the question of
obtaining efficient derivative market for commodities. If derivative markets function adequately, some of the policy goals
regarding price volatility of agricultural commodities can be addressed in a market oriented manner.

In this thesis, we examine commodity futures markets in India with reference to the Government policy of encouraging
derivative markets (commodity, equity, interest rate and currency) to manage price risk. We study three aspects of futures
market viz. basis risk, price discovery and spot price volatility. Specifically, we examine the castorseed futures market at
Mumbai and Ahmedabad with respect to the above three aspects. We study the castorseed futures market since inception
(May 1985) up to August 1999. We study all four contracts (March, June, September and December) that are traded each
year. This will provide insight into the role of commodity futures markets as a means of managing price risk. The thesis
will also serve as an independent evaluation of the rationale underlying the shift in Government policy from intervention
to a market oriented approach. If the castorseed futures market performs its functions well, there is a strong case for
further development of derivative markets in India.

There are a few institutional features of the castorseed market that are instrumental in explaining the results obtained
in this thesis. The most important factor is the harvest of castorseed from November to March. Of the four contracts
traded each year, the March contract is termed as the harvest or new contract since it coincides with the harvest season. It
is the only contract in which futures trading is driven mainly by supply side information. Another institutional feature of
the castorseed market is that Gujarat is the dominant producer of castorseed in India. Internationally, India is the largest
producer of castorseed. Domestically, Gujarat is the largest producer of castorseed. A third important institutional feature
of castorseed futures market is the relatively higher futures trading volume at Ahmedabad. Compared to Mumbai, futures
trading volume is higher at Ahmedabad. Since 1994, futures trading volume has risen at Ahmedabad while it has fallen
at Mumbai.

The key to successful commodity production and marketing decisions is to understand the basis and factors which
affect its behaviour. Unanticipated basis movement reduces the ability of futures market to transfer risk from hedgers to
speculators and results in lower income to hedgers. Commercial firms and traders who buy in either the spot or futures
market and sell in the other market do so in anticipation of a specific change in the basis. Unexpected change in basis
creates additional risk for the hedger associated with this market position and makes it less desirable to hold. Empirical
assessment of the magnitude and volatility of the basis and analysis of the factors influencing basis risk may permit the
development of select management practices designed to minimise basis risk impact on market participants’ decisions.

We investigate the nature of basis risk for different contracts for the castorseed futures market at Mumbai and Ahmed-

x
ABSTRACT xi

abad since inception (May 1985) up to August 1999. We use root mean squared error (RMSE) to measure basis risk. The
test developed by Ashley, Granger and Schmalensee is used to compare RMSE. We also use the Henrikson-Merton (H-M)
timing test to gauge the qualitative accuracy of forecast.

The results show that the time series model performs better (lower RMSE and superior timing ability) than the bench-
mark model. Any effort made towards developing models to forecast the basis is likely to reap results. RMSE is low for
June contract and high for December contract. For the June contract all information regarding supply of castorseed is
known before trading begins. Forecasting the futures price is not difficult resulting in lower basis risk and hence RMSE
is low. Exactly opposite is the case for December contract and hence RMSE is high. The forecast for the Mumbai market
has higher RMSE but better timing ability as compared to forecast for Ahmedabad market. The link between liquidity and
basis risk is unclear. If the lower RMSE for Ahmedabad market is due to higher futures trading volume; the H-M timing
test does not indicate superior timing ability.

Price discovery is an important function performed by futures market. It is the revealing of information about future
spot market price through the futures market and refers to the use of futures price for pricing spot market transactions.
The essence of price discovery is to establish a competitive reference (futures) price from which the spot price can be
derived and hinges on whether information is reflected first in changed futures price or in changed spot price. The futures
price serves as the market’s expectation of subsequent spot price. The significance of price discovery depends upon a
close relationship between futures and spot price. The extent to which futures market performs this function well can
be measured from the temporal relation between futures and spot price. If information is reflected first in futures price
and subsequently in spot price, futures price should lead spot price, indicating that the futures market performs the price
discovery function.

The price linkage between castorseed futures and spot market is investigated using cointegration analysis which offers
several advantages. First, cointegration analysis measures the extent to which two markets have achieved a long run
equilibrium. Another distinct advantage of the cointegration technique is that it explicitly allows for divergence from
equilibrium in the short run. Cointegration implies that each series can be represented by an error correction model which
includes last period’s equilibrium error as well as lagged values of the first difference of each variable. Hence, temporal
causality can be assessed by examining the statistical significance and relative magnitude of the error correction coefficient
and the coefficient on lagged variables.

Usually, price discovery is studied in a futures market and its corresponding spot market. We term this as price
discovery ‘within’ market. However, in India two futures markets exist for castorseed at Mumbai and Ahmedabad. This
gives us an opportunity to study price discovery ‘across’ markets by designating one of the futures market, say at Mumbai,
as the futures market and the other futures market, at Ahmedabad, as the spot market. Such a study would reveal whether
amongst futures markets for the same commodity, which are in geographical proximity, one futures market dominates the
other in terms of price discovery. Examining futures trading volume at Mumbai and Ahmedabad we find that there has
been a fall from 1994 onwards. Hence, we also examine whether the price discovery role of futures market has changed
over time by studying it separately for the two sub periods 1985–93 and 1994–99.

The results show that the futures and spot market at Mumbai and Ahmedabad as well as the futures market at Mumbai
and Ahmedabad are mostly cointegrated. There is causality and information flow from futures to spot market at both
Mumbai and Ahmedabad. There is reverse information flow as well for some contracts and periods. Results are different
for various markets, contracts and time periods indicating the role played by geographical location, harvest and futures
trading volume in influencing price discovery.

There is widespread interest in the effect of futures trading on price in the underlying spot market. It has often been
claimed that futures trading destabilises the associated spot market by increasing spot price volatility. Others have argued
to the contrary stating that futures trading stabilises price and thus decreases spot price volatility. As the debate cannot
be resolved at a theoretical level, empirical investigation is necessary. Empirical research so far has not produced any
 xii ABSTRACT

conclusive evidence regarding impact of derivative trading on spot market volatility.

Theoretically, it is not clear why derivatives should influence spot market volatility. Derivatives increase market
liquidity by bringing more investors to the spot market. This results in a less volatile spot market unless derivatives attract
mainly uninformed speculators who destabilise the market. The introduction of futures market improves risk sharing but
lowers the informativeness of price resulting in destabilisation and welfare reduction. The advent of new speculators into
the market increases liquidity but makes the spot price more noisy and reduces net social welfare if the new speculators
are less informed than traders existing in the market. Futures trading reduces the cost of entry of small traders into the
financial market. Futures trading increases spot price volatility if increased liquidity causes spot price to reflect new
information more quickly. In this case, rise in spot price volatility increases net social welfare.

Spot market volatility decreases due to the liquidity provided by speculators. This additional liquidity allows spot
traders to hedge their position and thus curb volatility attributable to order imbalance. The availability of risk transference
afforded by futures market reduces spot price volatility because it eliminates the need to incorporate risk premium in spot
market transaction to compensate for the risk of price fluctuation. Futures trading attracts more traders to the spot market
making it more liquid and therefore less volatile.

The debate about speculators and the impact of futures on spot price volatility suggests that increased volatility is
undesirable. This is misleading as it fails to recognise the link between information and volatility. Price depends on infor-
mation currently available in the market. Futures trading can alter the information available for two reasons. First, futures
market attracts additional traders to a market. Second, as transaction cost is lower in futures market, new information is
transmitted to the spot market more quickly. In an arbitrage free economy, volatility of price is directly related to the flow
of information. If futures market increases the flow of information, volatility in the underlying spot market will rise. The
variance of price change is equal to the rate or variance of information flow. The implication is that volatility of the asset
price will rise as the rate of information flow increases. It follows therefore that if futures market increases the flow of
information, volatility of the spot price must change.

We study the impact of castorseed futures market on spot price volatility for castorseed in India. We study the market
at both Mumbai and Ahmedabad which allows us to compare results across markets and gives us additional insight into
the influence of futures trading on the underlying spot market. We use both the traditional regression technique as well as
GARCH analysis.

The results show that the introduction of castorseed futures market at Mumbai and Ahmedabad has had a beneficial
effect on the castorseed spot market in the futures early period. This effect has remained stable in the futures later period,
a possible reason being the rise in futures trading volume post 1994 at Ahmedabad.

We conclude that the castorseed futures market at Mumbai and Ahmedabad performs the function of price discovery.
Also, the introduction of castorseed futures market has had a beneficial effect on castorseed spot price volatility. Thus,
there is a strong case for promoting derivative markets in India.

Keywords: Basis risk, Castorseed, Cointegration, Commodity futures, Derivatives, Forecasting, GARCH, India, Price
discovery, Volatility

JEL classification: G13

Chapter 1

Introduction

Agriculture is a key sector occupying an important position in the Indian economy. The agriculture sector constitutes
almost a quarter (23%) of India’s Gross Domestic Product (GDP). The effective contribution of agriculture to the national
economy is far greater on account of its backward and forward linkage with other sectors. The Government policy has
been to protect and promote the agriculture sector through procurement and administered price mechanism. However,
in view of reduced direct support to agriculture under the Agreement on Agriculture with the World Trade Organisation
(WTO), there is a policy shift towards market oriented approach. In recent years, a major theme of economic liberalisation
and reforms in agriculture sector is improvement in the functioning of product markets. It is increasingly felt that efficient
product markets serve to further the interests of agriculture sector [65].

Historically, the Government has intervened at every stage of the marketing of major agricultural commodities. Mini-
mum support prices were announced for twenty one commodities and every activity of marketing such as transportation,
storage, credit supply and international trading of these commodities was regulated. Government intervention declined
after initiation of economic liberalisation and reforms in 1991. The reduction in Government intervention accelerated
with the implementation of the Agreement on Agriculture under the World Trade Organisation (WTO). Prices of agri-
cultural commodities are determined increasingly by market forces; hence fluctuating demand and supply of agricultural
commodities is expected to result in high price risk for agri-business.

In a competitive market, price behaviour of a commodity is determined mainly by demand and supply. A disquieting
movement in demand and supply results in price volatility. In the case of agricultural commodities – such as castorseed
– supply variability is high as production in a geographical area is affected by natural factors such as weather, pests and
diseases. To help manage price risk, the Government has been encouraging commodity derivatives [63, 64, 65, 116].

A key aspect of the process of strengthening agricultural markets is the question of obtaining efficient derivative
market for commodities. If derivative markets function adequately, some of the policy goals regarding price volatility
of agricultural commodities can be addressed in a market oriented manner. The National Agriculture Policy (2000) has
articulated this reasoning well and the ban on futures trading was lifted for all commodities in April 2003. Domestic
entities facing price risk abroad are permitted to utilise foreign derivative markets to address their price risk management
need [64, 65].

A peculiar institutional feature of commodity markets in India is that there is no large nation wide market for any
commodity. Instead, there are multiple, isolated, independent, regional markets for all commodities. An apt description
of commodity markets in India is that they are highly fragmented. As a result, commodity futures markets in India too are
dispersed with separate trading communities in different regions. For example, there are three separate commodity futures
markets for castorseed at Mumbai, Ahmedabad and Rajkot [154]. The recently set up national level multi-commodity Ex-
changes such as the National Commodity and Derivatives Exchange (NCDEX) and Multi-Commodity Exchange (MCX),

1
 2 INTRODUCTION

both based at Mumbai, also trade in castorseed futures and are doing well.

1.1 Evolution of commodity futures markets

Commodity futures markets in India have a long history. The first organised futures market was established for cotton at
Mumbai in 1875. This was followed by futures markets for oilseeds (Mumbai, 1900), jute (Calcutta, 1912), wheat (Hapur,
1913) and bullion (Mumbai, 1920). Most commodity futures markets were established at the beginning of the twentieth
century. After a few years of lacklustre trading, the markets underwent rapid growth between the two World Wars. As a
result, before the outbreak of the Second World War, a large number of commodity exchanges trading futures contracts
in several commodities such as cotton, jute, oilseeds, groundnut, wheat, rice, sugar, silver and gold flourished at various
locations across the country. The Defence of India Act, 1943 was invoked to prohibit futures trading in some commodities
during the Second World War. After independence, commodity futures markets again picked up especially during the late
1950s and early 1960s [45, 100, 116, 132].

Commodity futures markets were in existence for a long time before the legal and regulatory framework evolved grad-
ually. In the early stage, the markets functioned under the rules and procedures laid down by individual trade associations.
There was wide difference in regulations followed by various associations. This resulted in a variety of malpractices
leading to frequent disputes among traders. The need for regulation was fulfilled by enactment of the Bombay Forward
Contract Control Act, 1947. The Indian Constitution (1950) placed the subject of futures market in the Union List, hence
the responsibility of regulating the market devolved on the Central Government. The Government drafted a Bill, modelled
on the Bombay Act and set up a Committee under the chairmanship of Shri A. D. Shroff to frame model Rules for Asso-
ciations. The Committee submitted its report in August 1950. The Forward Contracts (Regulation) Bill, 1950 was revised
in light of the A. D. Shroff Committee and referred to a Select Committee of Parliament. The Committee submitted its
report in August 1951, but the Bill lapsed with the dissolution of the Parliament in 1952. A new Bill was drafted and after
scrutiny by another Select Committee, the Forward Contracts (Regulation) Act was passed by Parliament in December
1952. The Forward Markets Commission (FMC) was established in 1953 to regulate and develop commodity futures
market in India [45, 100].

Futures markets prospered in India during the early 1960s. In the mid 1960s, due to the war in 1965 and natural
calamities, there was a shortage in commodities. As a result, in order to have control on price movement of many
agricultural and essential commodities, futures trading was banned in 1966 in most commodities except pepper and
turmeric. Futures trading in castorseed was suspended in 1977. On the recommendation of the A. M. Khusro Committee
(1980) futures trading in some commodities like gur (Muzaffarnagar and Hapur, 1982), potatoes (Hapur, 1985) and
castorseed (Mumbai and Ahmedabad, 1985) was permitted. With liberalisation of the Indian economy in early 1990s,
there was renewed emphasis on development of commodity futures market in India. The Government in 1993 set up a
Committee under the chairmanship of Shri Kamal Nayan Kabra to examine the feasibility and role of commodity futures
market in India. The Committee submitted its report [63] in September 1994 [45, 100].

The main recommendations of Kabra Committee that have been implemented are introduction of futures trading for
several commodities such as coffee (Bangalore, 1998), cotton (Mumbai, 1999), soya oil (Indore, 1999), sugar (2001),
tea (2002) and bullion (2003) and introduction of international futures contract for pepper (Cochin, 1997) and castor
oil (Mumbai, 1999). Acting on another recommendation of Kabra Committee to strengthen the FMC, the Government
upgraded the post of FMC Chairman to Additional Secretary. The Government in 1995 set up a separate Department of
Consumer Affairs to focus on commodities. The idea of a National multi-Commodity Exchange (NCE) was proposed in
1999. A Core Group was set up to work out modalities for constituting the NCE and a decision to set up the NCE was taken
in July 2000. Four new nation wide multi-commodity Exchanges have been approved. Of these four, two Exchanges,
On-line Commodity Exchange, Ahmedabad and National Board of Trade, Indore are operational. The prohibition on
INTRODUCTION 3

futures trading was removed for 27 commodities in August 2002 and 54 more commodities in February 2003 bringing the
total number of commodities for which futures trading is permitted to 94 [45, 100, 116, 132]. The Government lifted the
ban on futures trading for all commodities in April 2003.

1.2 Thesis design

The Government of India has established organised futures market for various commodities with the following objectives
[63]:

1. To enable various trading interests to shift the risk arising out of adverse price fluctuation through hedging. To bring
about an element of stability in seasonal price fluctuation.

2. Besides providing hedging facilities, the futures market performs the function of price discovery and price reference.
The existence of such futures quotation enables exporters to quote price for different shipments ahead

3. A futures market provides to the farmer at the time of sowing an indication of the expected price for different
commodities during the marketing period, thereby enabling him to undertake proper crop planning. The continuous
publication of quotation in the futures market all over the country ensures a close integration of price between
different centres and even between allied commodities.

In this thesis we examine the role of castorseed futures market at Mumbai and Ahmedabad in fulfilling the above objec-
tives. This thesis will thus be helpful in evaluating the performance of commodity futures markets in India. This work
assumes significance in light of the Government policy to actively encourage commodity futures market for price risk
management. We study three aspects of futures market viz. basis risk, price discovery and spot price volatility. We study
the castorseed futures market since inception (May 1985) up to August 1999. We study all four contracts (March, June,
September and December) that are traded each year. This will provide insight into the role of commodity futures market
as a means of managing price risk.

The thesis will also serve as an independent evaluation of the rationale underlying the shift in Government policy from
intervention to a market oriented approach. If the castorseed futures market performs its functions well, there is a strong
case for further development of derivative markets in India. The remaining part of this introductory chapter discusses
these aspects in detail. In Chapter 2 we discuss institutional features of the castorseed futures market. In chapters 3, 4 and
5 we study basis risk, price discovery and spot price volatility respectively. These chapters together constitute the core of
this thesis. In chapter 6 we conclude by offering policy recommendation.

In the remaining part of this introductory chapter we discuss the main theoretical features of commodity derivative
markets and the manner in which we propose to empirically validate various operationally testable hypotheses pertaining
to these theoretical features.

1.3 Outline of the thesis

As documented in the National Agriculture Policy (2000), the Central Government will continue to discharge its respon-
sibility to ensure remunerative prices for agricultural produce through announcement of Minimum Support Price (MSP)
policy for major agricultural commodities. Domestic market prices will be closely monitored to prevent distress sales by
the farmers. The Government will enlarge the coverage of futures markets to minimise the wide fluctuations in commodity
prices as also for hedging their risks. The endeavour will be to cover all important agricultural products under futures
trading in course of time [64].
 4 INTRODUCTION

In any society which has to make arrangement for a continuous supply of commodities, trading activity cannot be
confined to the immediate present needs. Forward contracts evolved in the early stage of development of mercantile
activity to cater to the needs of a manufacturer, processor or producer to ensure that he gets his supply of raw material at
the appropriate time. Such forward contracts provide for giving as well as taking actual delivery of goods at a specified
price and a specified future date. The futures contract evolved at a later stage to enable the various functionaries engaged
in merchandising a commodity, or its processed products, to protect themselves against the risk of unpredictable price
fluctuation over time in the commodity or its products.

A typical market for a commodity or a financial instrument can be divided into the spot market and the futures market.
The primary characteristic of the spot market is immediate delivery and payment. In contrast, a futures market is simply a
spot market for deferred delivery and payment. A futures market is an organised forward market dealing in standardised
contracts. Futures markets are derivative markets. They exist because a spot market exists. Price uncertainty in the spot
market contributes to the creation of a futures market.

A futures market serves three important functions viz. price risk management, price discovery and spot price volatility.
The primary social purpose of a futures market is that of price risk management. Another significant goal of a futures
market is that of price discovery. A futures market also contributes to society by way of reducing price volatility in the
underlying spot market. These are the main attributes of a futures market and the justification for their existence and
continued functioning in society. However, the above features of a futures market are only theoretical in nature. It is
necessary to empirically validate these features for every new futures market before a case can be made for its continued
existence. If a newly introduced futures market for any commodity fails to perform any of the above functions satisfacto-
rily, then there is no justification for its continued existence. Alternatively, if futures markets are indeed performing the
above mentioned social functions satisfactorily, then there is a strong case for introducing new futures markets for other
similar commodities. In this thesis, such an attempt is made to empirically validate the theoretical features of a futures
market for castorseed. In the remaining part of this section, we briefly discuss the theoretical features of a futures market.
An empirical validation of these features for castorseed futures market forms the core of this thesis.

In the context of a futures market, the difference between futures and spot price is termed as the basis. The significance
of the basis lies in its use as an information source to predict future spot price of the commodity underlying the futures
contract. Studying the basis is important because it is fundamental to understanding the most important function of a
futures market viz. hedging. Hedgers seek to control the risk exposure due to adverse spot price change. This exposure
is called price risk and exists because of uncertainty about future spot price levels. The focus of hedging is on ways to
reduce spot price risk.

Hedging reduces price risk because the gain (loss) in either (futures or spot) market is offset by an equivalent loss
(gain) in the other market. However, this is true only to the extent that futures and spot price change by exactly the same
magnitude, which may not always be the case. If futures and spot price change by unequal magnitude, i.e. the basis
changes, hedging may not be successful. Gain in one market may not be equal to loss in the other market resulting in a
net loss. The risk that futures and spot price may not change by the same amount is called basis risk and is defined as the
variance of the basis. Hedging effectiveness depends primarily upon basis risk, the lower the basis risk, the more effective
is hedging. Hence, a complete understanding of the basis is essential before undertaking any hedging activity. In this
thesis, we undertake such a study of basis risk for castorseed futures market in India.

A futures market exists for the purpose of price risk management. The risk of price variability of an asset can be
managed by the mechanism of hedging. It is defined as the use of futures contract to minimise the risk of a current or
future spot position. Hedging is the most important economic function of futures market. Price discovery is the second
most important role of futures market. It is the revealing of information about future spot market price through the futures
market. The essence of the price discovery function of futures market hinges on whether new information is reflected first
in changed futures price or in changed spot price.
INTRODUCTION 5

The essence of price discovery is that new information is first reflected in futures market and then in spot market. Thus,
futures price change leads change in spot price. It is quite possible that both futures and spot price change concurrently;
neither market leads the other and price discovery occurs equally in both markets. We take the above idea of information
transmission between futures and spot price and study lead lag between castorseed futures and spot market in an attempt
to study information transmission and price discovery. In this thesis, we undertake such a study of price discovery for
castorseed futures market in India.

An important concern of regulatory authorities is what effect introducing a futures market has on an existing spot
market. Through the process of arbitrage, the futures market is linked to the underlying spot market. The effect of trading
in futures market on price in associated spot market(s) remains an important but unresolved issue. One of the recurring
arguments made against futures market is that, by encouraging or facilitating speculation, they give rise to price instability
in the underlying spot market. The introduction of a futures market for any commodity where the spot market already
exists leads to a reduction in the volatility in spot market. This is because trading and its related volatility shifts to the
futures market. The existence of a futures market stabilises the corresponding spot market. We examine this idea by
observing the change in volatility prior and after the introduction of futures market for castorseed in India.

The general approach adopted in the literature, to resolve the above issue, is to examine spot price volatility prior to
the onset of futures trading and to compare this with spot price volatility post futures. This is typically done by using
a measure of volatility (such as variance of returns) as the dependent variable and regressing it on a proxy variable (to
control for market wide volatility) and a dummy variable (to capture the impact of futures). The proxy variable usually
is another commodity with similar characteristics for which there is futures trading during the study period e.g. we have
used turmeric as a proxy variable in this thesis.

However, to be more reliable, this approach must reasonably account for all other determinants of volatility that could
have influenced the results. It is well documented that the variance of returns is temporally dependent (heteroscedastic)
and hence the above approach is incomplete if not misleading. Recent literature has overcome this objection by utilising
the GARCH family of models wherein spot market returns are modelled pre and post futures as an ARIMA process while
simultaneously modelling the conditional variance as a GARCH process. The parameters of the conditional variance
equation are tested for structural change to make inference about the influence of futures trading on spot market volatility.
In this thesis, we undertake such a study of futures and spot market volatility for castorseed futures market in India.

1.4 Basis risk

The difference between futures and spot price is termed as the basis. For the futures market to serve as a successful price
risk management vehicle, the basis must be predictable. This is not to say that the spot price must be predictable. Futures
price is not an accurate forecast of the spot price that will prevail at a future delivery date, but futures price change as
the delivery date nears must be correlated with observed change in spot price if the futures market is to provide price
risk management opportunity [50]. The basis provides information to hedgers about the tendency of future prices in both
futures and spot market and hence it operates on the margin of hedging decisions. The basis is positively correlated with
the average future change in spot price and negatively correlated with that of futures price. This interpretation of the
basis is more general in nature compared to the classical view of the basis as a return to storage especially in the case
of seasonally produced commodities [44, 52]. The basis provides information on the likely direction of futures price.
As time to maturity shortens, spot price rises because of increasing storage cost and consequently the basis declines on
average. A positive basis thus becomes associated with a rise in spot price and vice versa. The information revealed by
the basis includes the cost of storage but is not limited to it [98].

A futures market is used to hedge the risk associated with price fluctuation in the underlying spot market [70]. If the
expected convergence between spot and futures price does not occur at settlement date (due to high arbitrage transaction
 6 INTRODUCTION

cost) this will introduce a risk for the hedger which will negatively affect participation in futures market [123]. The basis
reflects macro-economic risk common to all asset markets [7]. The basis between a futures contract and its underlying
commodity is an important measure of the cost of using the futures contract to hedge. In a cross hedge, the relative size of
the basis of alternative hedging vehicles plays a decisive role in selection of the optimal hedging vehicle [24]. Basis risk
can be attributed to location, timing and quality discrepancies between commodities traded in the spot market and those
deliverable on futures [120]. The variability in the basis also contributes to basis risk [20, 47].

Unanticipated change in basis affects the hedger’s income level and price risk management position, so understanding
basis risk and developing appropriate market strategies designed to minimise basis risk is important. Knowledge of this
basis risk based on day-to-day variation within contracts is important because the producer has the option of making spot
sales over a period of several days. Abrupt change in basis influences the final value of hedged commodities purchased
and sold [56].

Basis risk which is specific to futures market and does not exist in spot market introduces an element of speculation
because hedgers are still exposed to this risk while hedging their physical commodity. The difficulty in forecasting the
basis (unpredictable basis) presents hedgers with a risk that cannot be hedged [123]. Basis risk not only affects the futures
position but also the spot market position in all hedging situations [117]. For agricultural commodities, forecasting the
basis is an important aspect of a successful marketing strategy. An unforeseen change in basis can adversely affect the
net price received, increase risk and alter producer behaviour [121]. Forecasting the basis permits producers to assess
alternative forward pricing mechanisms such as futures hedging, spot forward contract and basis contract. The success of
a futures contract also hinges on a predictable basis. Increased basis risk relative to price risk can reduce the attractiveness
of futures market as a price risk management vehicle. If basis risk increases, the ability of futures to transfer risk is
reduced [57].

Basis risk exists because the basis does not move as expected during the delivery period [146]. The basis is expected
to change over the life of the hedge. An unexpected change in the basis affects the riskiness of the hedge. Anticipated
change in the basis is not considered basis risk. The fact that basis risk reduces the futures market position relative to
spot position has important implication for contract design and policy makers. By designing contracts that minimise basis
risk, a futures exchange can encourage use of the futures contract directly (because the hedge ratio rises with a reduction
in basis risk) and indirectly (because the spot position and hence the amount to be hedged also rises). To the extent that
policy makers wish to encourage spot market activity (agricultural storage to smooth incomes and reduce agricultural
subsidies) or to encourage price risk management (encourage banks to hedge their risk through use of financial futures),
the Government also should be interested in futures contract designed to minimise basis risk [117].

Thus we see that basis risk is detrimental to hedgers in using commodity derivative markets for price risk manage-
ment. In this thesis we model the basis in the time series frame-work and show that these class of models are effective
in minimising basis risk thereby improving the utility of commodity derivative markets for price risk management. Con-
sequently, the burden on the Government to mitigate price risk for farmers is reduced and the resources released can be
better utilised where they are more required. Chapter 3 deals with basis risk in more detail.

1.5 Price discovery

Futures markets exist for the purpose of price risk management. The risk of price variability of an asset can be managed
by the mechanism of hedging. It is defined as the use of futures to minimise the risk of a current or future spot position.
Hedging is the primary economic function of futures market. Price discovery is the second most important role of futures
market. It is the revealing of information about future spot market price through the futures market and refers to the use
of futures price for pricing spot market transactions. The essence of price discovery is to establish a competitive reference
(futures) price from which the spot price can be derived and hinges on whether information is reflected first in changed
INTRODUCTION 7

futures price or in changed spot price. The futures price serves as the market’s expectation of subsequent spot price
[55, 58, 81].

The significance of price discovery depends upon a close relationship between futures and spot price. The extent to
which futures market performs this function can be measured from the temporal relation between futures and spot price. If
information is reflected first in futures price and subsequently in spot price, futures price should lead spot price, indicating
that the futures market performs the price discovery function. In an efficient market where all available information is fully
and instantaneously utilised to determine market price, futures price should move concurrently with its corresponding spot
price without any lead or lag in price movement from one market to another. In the absence of market friction, price in the
futures market and its corresponding spot market should move contemporaneously in response to arrival of information.
Since futures and spot market represent the same commodity, their price should exhibit a mutual (similar) response to a
given information event, a process facilitated by arbitrage [51].

However, if one market processes information faster than the other, a lead lag relation will exist. There are many
reasons why one market will react more rapidly to the arrival of new information. Possible explanatory factors include
ease of short sale, lower transaction cost, institutional arrangement and market microstructure effect. The lead lag charac-
teristics of futures and spot market illustrate how rapidly one market incorporates information relative to the other. These
characteristics also indicate the efficiency of their functioning as well as the degree of integration between the two markets
[51, 139]. Traders act faster and at lower cost in the futures market as compared to spot market resulting in a lead lag
relation between futures and spot price [71].

Futures trading facilitates the allocation of production and consumption over time by providing market guidance in
the holding of inventories [81]. If the futures price for distant delivery is above that for early delivery, postponement of
consumption becomes attractive. Thus, a change in futures price results in subsequent change in spot price. Speculators
prefer to hold a futures contract because they are not interested in the physical commodity per se and a futures position
can be offset easily. Further, hedgers who are interested in the physical commodity and have storage constraint will hedge
themselves by buying a futures contract. Therefore, both hedgers and speculators will react to information by transacting
in futures rather than spot market. Consequently, futures price tends to lead spot price.

Chan [27] contends that the adverse selection cost faced by discretionary liquidity traders who can choose the timing
of their transactions strategically (such as large institutional traders) may be minimised by trading in futures rather than
spot market. Again, this implies that the transmission of information is from futures to spot market. It has been suggested
by Grunbichler, Longstaff and Schwartz [72] that difference in liquidity between the two markets also creates a lead
lag relationship. They point out that if the average time between trades in spot market is longer than in futures market,
information will be incorporated more rapidly in futures rather than spot price. Fleming, Ostdiek and Whaley [48] find
that there are differential trading costs between the equity, index futures and index options markets and assert that the
market with the lowest trading cost will react more quickly to the arrival of information.

In practice, futures market is commonly observed to update price more frequently than spot market [93, 143]. Investors
with strong beliefs about the direction of market, trade in futures rather than spot market because transaction cost is lower
and the degree of leverage attainable higher. Such trading moves futures price first and then pulls spot price by means of
arbitrage creating a lead lag relation. Considering the relation between futures and spot price, Kawaller, Koch and Koch
[93] put forward the general principle that spot price is affected by past spot price, current and past futures price and other
market information. Similarly, futures price is affected by past futures price, current and past spot price and other market
information. Thus, causality is likely to be bi-directional. They further argue that the lead lag pattern between futures
and spot price changes as new information arrives. Each may lead the other as market participants sift information for
clues that are relevant to their position which may be spot or futures. The conclusion that can be derived from the above
discussion is that there is some rationale for the hypothesis that futures price leads spot price and also for the hypothesis
that spot price leads futures price. However, the case for the first hypothesis is stronger and more compelling.
 8 INTRODUCTION

Thus we see that the utility of commodity derivative markets in price risk management is based primarily on their
ability to perform price discovery. In this thesis we examine the role of commodity derivative markets in performing
the function of price discovery and show that they do perform the role of price discovery effectively thereby improving
their utility for price risk management. Consequently, the burden on the Government to mitigate price risk for farmers
is reduced and the resources released can be better utilised where they are more required. Chapter 4 deals with price
discovery in more detail.

1.6 Spot price volatility

An important concern of any regulatory authority is the effect a futures market has on an existing spot market. A futures
market is linked to the underlying spot market by arbitrage. It has often been claimed that futures trading destabilises
the associated spot market by increasing spot price volatility. Others have argued to the contrary stating that futures
trading stabilises price and thus decreases spot price volatility. As the debate cannot be resolved at a theoretical level,
empirical investigation is necessary. Empirical research so far has not produced any conclusive evidence regarding impact
of derivative trading on spot market volatility [124]. Witherspoon [150] shows that the effect of futures market on spot
market volatility depends on whether the futures or spot market is dominant in terms of price discovery. If price discovery
by futures market exceeds a critical threshold level, then spot market autocorrelation and long term volatility increases.
On the other hand, if price discovery by futures is not excessive, it leads to lower long term volatility, enhanced liquidity
and increased efficiency.

Theoretically, it is not clear why derivatives should influence spot market volatility. Derivatives increase market
liquidity by bringing more investors to the spot market. This results in a less volatile spot market unless derivatives attract
mainly uninformed speculators who destabilise the market. The introduction of futures market improves risk sharing but
lowers the informativeness of price resulting in destabilisation and welfare reduction. The advent of new speculators into
the market increases liquidity but makes the spot price more noisy and reduces net social welfare if the new speculators
are less informed than traders existing in the market [142]. Futures trading reduces the cost of entry of small traders into
the financial market. Futures trading increases spot price volatility if increased liquidity causes spot price to reflect new
information more quickly. In this case, a rise in spot price volatility increases net social welfare [92]. Futures trading
enhances both the incentive and means for speculation. The speculative trade associated with futures trading has been
accused of destabilising the underlying spot market inducing price volatility. Theoretical discussion regarding influence
of speculative activity in futures on primary market does not find a consensus of opinion. Harris [75] suggests that an
increase in well informed speculative trade brought to the market by futures trading has two opposite effects on volatility
– destabilising and stabilising. These are discussed next.

1.6.1 Destabilising effect

The spot market may experience an increase in price volatility as futures trading enhances the price discovery mechanism
and so information concerning fundamentals is more rapidly assimilated into price. A major economic function of futures
market is its price discovery role. Since a futures market faces less friction than spot market particularly with respect to
lower transaction cost, futures price responds more quickly to new information. Through a process of arbitrage the price
adjustment is transmitted to the underlying spot market [35, 75]. Chassard and Haliwell [25] comment that speculation
artificially distorts price movement so as to exaggerate the normal response to fundamentals. Kaldor [90] argues that
sophisticated speculators exacerbate price change by selling to less well informed agents at a price above that a competitive
equilibrium would dictate. Baumol [10] proposes that speculators amplify a price trend by buying or selling only after
price has changed, thus increasing volatility. Edwards [41] comments on the notion of investors searching for bandwagon
profits as another reaon for increased volatility.
INTRODUCTION 9

Destabilising speculation in futures market is transmitted to the underlying spot market through arbitrage and results in
increased price volatility. Speculation produces a net loss with some speculators gaining (and others losing), in the process
destabilising the market [91]. An increase in uninformed speculative trade increases price volatility by interjecting noise
into a market with limited liquidity. The inflow and existence of speculators in futures market produces destabilising
forces which create undesirable bubbles [41, 75]. Futures market activity increases spot price variability when futures
price is distorted by technical factors or manipulation. Sometimes, futures market induces a significant amount of hedge
trading without attracting enough speculation to permit effective risk transfer. The hedging pressure in futures then spills
over to the spot market where traders end up bearing risk transferred through both the futures and spot market [46]. At
times, the deliverable supply of commodity becomes low (a short squeeze) so that price distortion in the spot market is
precipitated by attempts of futures traders to cover their position in the spot market [10, 90].

Futures trading increases spot price volatility if traders in futures market do not have as good information as partic-
ipants in the spot market. Even if the futures price accurately reflects information available to traders in that market,
their collective action pushes spot market price away from its most appropriate value. This situation presents profitable
opportunity to the better informed spot market participants whose trading acts to stabilise futures price while allowing
greater volatility in spot price [46]. Fraudulent activity in futures market results in price distortion in spot market if it is
used to corner or squeeze the spot market [141]. Stein [142] argues that futures trading introduces new speculators to the
spot market. He demonstrates that if information is asymmetrically distributed, the information content of spot price is
altered, spot price variability increases and welfare is reduced.

1.6.2 Stabilising effect

Spot market volatility decreases due to the liquidity provided by speculators. This additional liquidity allows spot traders
to hedge their position and thus curb volatility attributable to order imbalance. There are several ways a futures market
increases efficiency and smoothens price variation in the underlying spot market. Futures market provides a mechanism
for those who buy and sell the actual commodity to hedge themselves against unfavourable price movement. Through
the futures market, risk can be spread across a large number of investors and transferred away from those hedging spot
position to professional speculators who are more willing and able to bear it. The availability of risk transference afforded
by futures market reduces spot price volatility because it eliminates the need to incorporate risk premium in spot market
transaction to compensate for the risk of price fluctuation [46]. Futures trading attracts more traders to the spot market
making it more liquid and therefore less volatile [41].

Speculation in futures market also leads to stabilisation of spot price. Since futures market is characterised by a high
degree of informational efficiency, the effect of such stabilisation permeates to the spot market. Profitable speculation
stabilises spot price because well informed speculators tend to buy when price is low pushing it up and sell when price
is high causing it to fall. These opposing forces constantly check price swings and guide price towards its mean level.
Uninformed speculators are not successful and are eliminated from the market. Thus, profitable speculation in futures
market leads to a decrease in spot market volatility [53].

The introduction of futures market leads to more complete markets enhancing the information flow. Futures market
allows for new positions and expanded investment sets or enables existing positions to be taken at a lower cost. Futures
trading brings more information to the market and allows for quicker dissemination of information. The transfer of
speculative activity from spot to futures market dampens spot market volatility [112].

1.6.3 Volatility and information

The debate about speculators and the impact of futures on spot price volatility suggests that increased volatility is undesir-
able. This is misleading as it fails to recognise the link between information and volatility. Price depends on information
 10 INTRODUCTION

currently available in the market [4]. Futures trading can alter the information available for two reasons. First, futures
market attracts additional traders to a market. Second, as transaction cost is lower in futures market, new information is
transmitted to the spot market more quickly. A theoretical model developed by Cox [35] establishes that in the presence
of an active futures market, a spot market is more efficient by way of faster price adjustment. The issue then is how the
rate of information flow relates to spot price volatility. This issue is addressed at the theoretical level by Ross [131]. He
argues that in an arbitrage free economy, volatility of price is directly related to the flow of information. If futures market
increases the flow of information, volatility in the underlying spot market will rise [4, 131]. Ross shows that the variance
of price change is equal to the rate or variance of information flow. The implication is that volatility of the asset price will
rise as the rate of information flow increases. It follows therefore that if futures market increases the flow of information,
volatility of the spot price must change [131]. Return volatility is a manifestation of information in the market [107].

Factors which make futures attractive to speculators also improve information flow and the rate at which information
is assimilated into price. Increased spot price volatility during futures trading reflects such improvement. Increased spot
price volatility results from improvement in operation of the market rather than adverse destabilisation. If futures trading
attracts additional informed traders and leads to increased information flow, the benefit from assimilating information in
spot market continues even after futures trading ceases. From a policy perspective, it is important not only to determine
whether spot price volatility has changed but also why it has changed. Increased volatility may be the result of destabilising
speculation or it may be due to improved information flow leading to more rapid assimilating of information into price.
Determining the reason for change in volatility is therefore important for futures market regulation [80].

Thus we see that the effect on underlying spot market volatility is a key feature of commodity derivative markets and
this aspect has been studied in more detail in Chapter 5 of this thesis.
Chapter 2

Institutional Features

In this thesis we study the futures market for castorseed at Mumbai and Ahmedabad since inception (May 1985) up to
August 1999. Four contracts maturing in March, June, September and December are traded each year at both Mumbai
and Ahmedabad. A total of 114 contracts (57 each for Mumbai and Ahmedabad) are studied. Each contract runs for five
months. There are 100 observations per contract, making a total of more than 11,000 observations. Daily closing futures
and spot price is obtained for both markets for each contract from the Forward Markets Commission (FMC), which is the
regulatory authority for commodity futures markets in India.

2.1 The castorseed economy

Castorseed is a non-edible oilseed grown for its oil which is used mainly in the pharmaceutical, paints and lubricants
industry. India is the largest producer of castorseed in the world while Gujarat is the largest castorseed producing state in
India (Table 2.1). Castorseed is an annual crop grown mainly in tropical, semi-arid region. It grows best in region where
both heat and humidity are in abundance. In India, the crop is sown from May to July and harvested from November to
March. The crop duration is 4–5 months. Castorseed contains ricinoleic acid which has medicinal properties and hence
makes it economically useful and commercially viable.

Castorseed is cultivated for the commercial importance of its oil. The Indian variety of castorseed has 48% oil content
of which 42% can be extracted. India’s castorseed production fluctuates between 6 to 9 lakh tonne per annum. Gujarat
accounts for more than 80% of castorseed production followed by Andhra Pradesh, Karnataka and Rajasthan. Castorseed
is grown mainly in Mehsana, Banaskantha and Saurashtra-Kutch region of Gujarat and Nalgonda and Mahboobnagar
districts of Andhra Pradesh. India exports around 3 lakh tonne of commercial castor oil mainly to USA, Europe and
Japan. Though the production of castorseed is concentrated in Gujarat and Andhra Pradesh, its consumption is spread
across the country. The domestic spot market is well developed in Gujarat and Andhra Pradesh. The major mandi (spot
market) price near the producing centres, upcountry price and export price from ports is available in the public domain on
a daily basis. The major castorseed markets in Gujarat are Rajkot, Ahmedabad, Gondal, Gadwal, Bhabar, Disa and Kadi
while in Andhra Pradesh they are Jedcheria and Yemignoor. The world castorseed production fluctuates between 12 to 18
lakh tonne. India is the largest producer with more than 60% share followed by China and Brazil.

Castorseed is used mainly in the following industries (application): agriculture (organic manure), paper (water proof-
ing additive), cosmetic (emulsifier and deodarant) paint, ink and adhesive (wetting and dispersing additive), food (viscos-
ity reducing additive), plastic and rubber (coupling agent), electronics and telecommunication (capacitor fluid), lubricant
(corrosion inhibitor), textile (pigment wetting agent) and pharmaceutical (castor oil). The major factors influencing cas-

11
 12 INSTITUTIONAL FEATURES

torseed futures and spot market are

• Variation in castorseed domestic acreage based on yield and price realisation

• Crop development based on progress of monsoon in key growing areas

• Chinese and Brazilian crop size

• Comparative price of other vegetable oils in domestic market

• Both castorseed and castor oil can be stored without spoilage for a long time

2.2 Harvest of castorseed

There are a few institutional features of the castorseed market that are instrumental in explaining the results obtained
in this thesis. The most important factor is the harvest of castorseed from November to March. Of the four contracts
traded each year, the March contract is termed as the harvest or new contract since it coincides with the harvest season.
The March contract is open for trading from November to March which is also the harvest season. This makes the
March contract different from other contracts. It is the only contract in which supply side information about the new crop
becomes available to the market on a daily basis. It is only in the March contract that futures trading is driven mainly by
supply side information. For the remaining three contracts there is no new supply side information that can be used for
trading in futures market. Futures trading in these contracts is driven only by demand side information. Hence we expect
price discovery results to be different for March contract. Specifically, we expect a bi-directional (futures to spot and
vice-versa) flow of information. We expect the spot market to influence the price discovery process significantly (Sections
4.7.3 and 4.7.4).

2.3 Castorseed production

Another institutional feature of the castorseed market is that Gujarat is the dominant producer of castorseed in India. The
international and domestic production of castorseed is tabulated below.

Table 2.1: Average annual production of castorseed

Country/State 1985–93 1994–99

World production (’000 tonne)

India 468.6 826.3
China 287.2 215.0
Brazil 164.6 46.0
Source: FAO

Domestic production (’000 tonne)

Gujarat 314.2 654.5
Andhra Pradesh 71.9 64.5
Karnataka 21.5 21.1
Rajasthan 10.2 37.4
Orissa 17.7 12.1
INSTITUTIONAL FEATURES 13

Country/State 1985–93 1994–99

Tamil Nadu 7.4 10.3
Source: Government of India

From the table above we note that internationally, India is the largest producer of castorseed. We also observe that
domestically, Gujarat is the largest producer of castorseed. The average annual production of castorseed in Gujarat is
more than all other states in India combined. We expect the dominance of Gujarat in castorseed production to have three
effects on the price discovery process. First, we expect the Ahmedabad spot market to play an influential role in the
price discovery process. Second, we expect the dominance of spot market to be more pronounced for the March contract
(Section 4.7.4). Third, between futures markets, for the March contract we expect the Ahmedabad futures market to be
dominant in terms of price discovery (Section 4.7.5).

2.4 Futures trading volume

A third important institutional feature of the castorseed market is the relatively higher futures trading volume at Ahmed-
abad. The average annual futures trading volume in ’000 tonne for all contracts at Mumbai and Ahmedabad for the period
May 1985 to August 1999 is tabulated below.

Table 2.2: Average annual futures trading volume

Period March June September December

Mumbai (’000 tonne)

1985–93 171.70 144.18 125.17 146.60
1994–99 46.24 30.00 20.26 30.62

Ahmedabad (’000 tonne)

1985–93 437.27 443.57 420.08 474.86
1994–99 1042.23 1228.74 1063.35 891.39
Source: Forward Markets Commission

From the table above we observe that compared to Mumbai, futures trading volume is higher at Ahmedabad. We also
note that since 1994 futures trading volume has risen at Ahmedabad while it has fallen at Mumbai. Due to the higher
futures trading volume at Ahmedabad, we expect basis risk to be lower at Ahmedabad as compared to Mumbai (Section
3.6.1). Between futures markets, we expect the Ahmedabad futures market to dominate Mumbai futures market in terms
of price discovery (Section 4.7.5). Also, we expect the Ahmedabad futures market to have a more stabilising influence on
the underlying spot market as compared to Mumbai (Section 5.5.2).

The reason for the shift in futures trading volume from Mumbai to Ahmedabad is the opening of a sea port and setting
up of a special economic zone (SEZ) at Kandla, Gujarat in 1994. Before 1994, castorseed was crushed into castor oil
at Mumbai for export to USA and Europe via Mumbai sea port. Gujarat did not have a viable sea port. However, after
opening of Kandla sea port and setting up of the SEZ, castorseed began to be crushed into castor oil in Gujarat and
exported via Kandla sea port by utilising export benefit of the SEZ. This led to a fall in export of castor oil from Mumbai
sea port with a corresponding rise at Kandla sea port. The average annual castor oil export from Mumbai and Kandla sea
ports is tabulated below.
 14 INSTITUTIONAL FEATURES

Table 2.3: Average annual castor oil export

Port 1985–93
1994–99
’000 tonne
Mumbai 37.30 53.20
Kandla 20.25 140.97
Source: Government of India

From the table above we note the shift in castor oil export from Mumbai to Kandla sea port in 1994. We expect the
price discovery process as well as the effect of futures trading on spot price volatility to be different for the two sub periods
1985–93 and 1994–99 (Sections 4.7.6 and 5.5.2). Specifically, we expect the futures and spot market at Ahmedabad to be
more strongly cointegrated post 1994 (Sections 4.7.2 and 4.7.4). We expect the futures market at Mumbai and Ahmedabad
also to be more strongly cointegrated post 1994 (Sections 4.7.2 and 4.7.5). We expect the futures market at Ahmedabad
to have a more stabilising influence on the underlying spot market post 1994 (Section 5.5.2).

2.5 Basis behaviour

In the context of futures market, the basis is defined as the difference between futures and spot price as a percent of spot
price for a given commodity at a specific location and a particular time. The interpretation of the basis is important,
particularly for agricultural commodities. The basis is used as an information source to predict future spot price of the
commodity that underlies the futures contract. Since price is measured in natural log, the basis at any time t is
ln(Ft,T ) − ln(St )
Bt = (2.1)
ln(St )
where Bt is the basis, Ft,T the futures price and St the spot price at end of day t. The above specification of the basis is in
percentage terms (basis as a per cent of spot price); the basis is in effect deflated by the spot price and thus adjusted for
inflation [109]. An important feature of the basis is that as the contract approaches maturity, the basis tends to zero. When
a contract begins trading, there are five months to go before the contract matures. The basis is large at the beginning of
the contract. As time to maturity for the contract decreases to zero, the cost of carry reduces to zero and futures and spot
price converge. Consequently, the difference between futures and spot price (the basis) also reduces to zero. The basis in
a contract maturing post harvest is more volatile since information about the new crop is not known at the beginning of the
contract but becomes available as the contract approaches maturity. The additional information flow relating to harvest
increases volatility of the basis. For an agricultural commodity with annual harvest, such as castorseed, the basis exhibits
different behaviour for contracts maturing at different times of the year.

Castorseed is produced seasonally but consumed uniformly throughout the year. The fact that new supply of cas-
torseed becomes available periodically and that the harvest is large relative to existing stock has a significant effect on
the pattern the spot price of castorseed will follow. Because the harvest occurs at a certain time, the spot price exhibits a
seasonal trend, with important implication for futures price. For agricultural commodities with seasonal production, such
as castorseed, the single most important factor determining spot price is the erratic availability of the commodity. The
inventory of castorseed is high immediately after harvest. Over the crop year, inventory falls due to consumption, till the
next harvest. At harvest, spot price is low because of abundant supply. As the crop year progresses, nothing is added to
inventory because the harvest is over. However, continuing consumption causes inventory to drop and spot price to go up
until the next harvest, when the commodity is plentiful again.

For a contract that matures just before harvest, the expected spot price at maturity is high so the futures price is high
as well. Similarly, for a contract that matures just after harvest, the expected spot price at maturity is low so the futures
price is low as well. Futures price matches the spot price that will prevail when the contract matures. Futures price is an
INSTITUTIONAL FEATURES 15

unbiased estimate of the expected spot price at maturity. Castorseed is harvested annually from November to March. The
spot price for castorseed rises continuously from June to August, remains high up to November and then falls steadily up
to March. The harvest of castorseed begins in November increasing its supply in the spot market. Increased supply leads
to a fall in price from November to March. From April to June spot price of castorseed fluctuates as uncertainty about
the exact quantity of castorseed produced during the year is resolved. From June to October spot price of castorseed rises
steadily since increased demand is not met by a corresponding increase in supply.

For a contract maturing pre harvest, such as June and September, futures price is higher than spot price and hence the
basis is positive. Similarly, for a contract maturing post harvest, such as December and March, futures price is lower than
spot price and hence the basis is negative. The basis is close to zero for March contract. The basis is large in magnitude
for September and December contract as compared to March and June contract. The behaviour of basis is the same for the
castorseed futures market at Mumbai and Ahmedabad, indicating that both markets function similarly. Also, the behaviour
of basis has remained the same each year from 1985 up to 1999. This result is expected because basis behaviour depends
mainly on the harvest season of castorseed which has remained the same over the study period. The harvest determines
futures price and hence the basis for all contracts. Thus, we see that the basis behaves differently for contracts maturing
in different months. This is also evident from the graph of the basis shown below.

0.012
September
. ....
0.009 . ... ... ... . .
... .. . .... ..... ..... . .
..... .. . ... . . .... ...
...... ....... ..... ...... ..... ..... ..... .
........
.... .. ... ..... ... . .
. .. . .
. . . . .
. .. ... . .. . .... .... .... .
.....
.. ... ..............
. .... .... .... ..... .... ............ .....
. .......
. .. .... .. ... . .
. .
.
.. ..... .
... .. ... .... ... ...
. ... .. ... ... ... ...... ..... .... ...... ............ ... .. . ..
.. .... ... .. ...... ... ... .. ... ... .. ....... .. .... . ... ... ...... .... . ..... ...
....... .. . .....
. . ... ...
... ..
......
.. .
..... .
..... ......
.
............ ... ... .. .... . ... .
. .... ... . .... ... . . .
. ..
.. .. .. .. .. .. .. ... ... ....... . .. .. .. .. .. ... . ... ... .. .
.. .... .... .... .... ... ... ... .. ....
0.006 ...... .... ... .. .
..... .......... .. ... ....
. .
. .
.... ....... .... ....
. ... ..
..... .
... ...
... ..
... .. .
. ....
.
.
.
... ................ .. .
. . ....
.
.
..... ..... .....
... .... .
. . ... .
.... ..
.
.
..
.
.
.
..... .
. ..
.
.
..
. .. . . . . . . ... . .... .... . ..
....... .... ...... .. ................. .......... .. .... ....
.
.....
.
......
. .. ..
. . . June. ...
. .. . . .
. .. ...... . ..
. . .
.
...
. .. .... ...
..... .... .. ... ....... . ..... .. . ... .......... ... . . ..... . .. .. .. ......... ... . .
....... ... ... ..... ..... ..... ... .. .... ... . . . .. . .... ... ... ..... ... .. . ...... .. ... ..
.
....... .. . ... . .. . ... . . . . ..
.. . .. ..
. .. .. ..
.
. .. . ....... . .. . .. ..
... . .... .. ...... ............
....... .. ... ...... ...... .
.
...
.. .... .... . .. ...
...
.
....
... ..
. . .. ......
... ..... ... ..... .... ....
0.003 ... ..... ... ..
... ......... ..... . . . ...
..
. ....
. .
..
.
... .
.
. .... ..
...
....... ........
...... .... ...... ...
.....
.
.
.
...... ................................................... ... ..... ......
.
.
. .... ... .... ... .. ............. .. .... ... ..........
.. . . . . .... . . . . ...
.... .. .... .... ....... .. ... . . .
.. ...... .................. ....... ................
.. ........... .. ..... .... .. .....
B ....... ......... ........ ... ... ....
...
.. .
.. ...
... .... .
......
.... .... .. .
.... ......
........ ..
. . . . .. .. ... .... .... .. ...... .....
.. .... ........... ... . ................ ........... ...... .... . .................... ... ............
... ...... ..... . . . . . . . .
a ... .. ... ...... ...
... ... ... ... ... ......
.. . .
.... .... . ...
. ... . .
. ..
.. .
.. . ..
.
..
March .
... ..
... ...
. ..
..........
.. ... .. ....
. .....
.....
. . . .
. .. . . ... .
... .. ... .... ......... ... ....... ...................
.. .. .. ........ .. . .. .
. . ... .
... . .. ...... .....
s ... ... ...... ......... ........ ..... .. .... ........ . ... ......... . .. . . . ... .... ....... .... ........ .....
. ... ........ .. ... ......... ... .... . ............ .... .........
.
... ..
.
... ... ..... ........ . ... ... ....... . . .. .. . . . . . . . . .. . . . ... .. . .
... .. ... ..... ...... .. .... .. .. .. .. ...... . . . .
.. .. .. .. .. ... ...... .. .... .... .... ... .... ... ........ ..... ... . ........ . .. ........
.. ....
... .. ..... .. .. ....
. . . .... .......
. ....... .... ..... .... ....
. .. ....
. ... .... ...... .... .... .....
.
i ... ..
... ..
.
.
. .
... .. ...
... ... .
.
.
... ... .. ..... . ...
.
.... ..... .
.
.
..... ...
.
....
.
.
.
.
....
.
....
. .
.
.
.
....
...
.
. .........
.
.
. ......... .. ... ...... ...
. . .
. ... ... .. ... ... ... ..
.
.
. .
.
.
. ... ...
. .
. ... ..
. .
..... .....
...
.
..... .
.
.
.
.
.
.... .
.
.
...
.. . .
.
. .
. .. ..... ...... .. . .. ....... ..
. . ..... ... .
.
. .
. . .
.
.
.. . .
...
.
..
.
.. .. .. .. .
.
. .
.
.
.
... ... . . .
s ... ..
... ...
... ...
... ...
... ..
. .
........ .. .. .... .... . . .
...... ...... ... ... .... ...... .... .... .............. .... ...... ....
.
. .. .
. .. .
. . . .
... .... .... ....
. ..
... .. ........ .... ....
.
. . .
... .
... ..
... ...
. ... .
... ...
.
...... ..
.
.. ... .
.... . . . ....... ...
.
. .. .. . .. .. ...........
..... ... . . ......
. . .
... .
.
..
.. .
..
. . . ... .. .... ... .. ... .. .. ... .. ... .. .. . .. .. . .. .. .. ...
... .. ... ... ... .. .. ... .. ... ... ... .. . . .. .. ...... ...... .. ... .... .. .. .. .. . .... ... .... .. ...
-0.003 ... ... ... ...
......
..... .... ...
.
... ...
... ...
..... ...
. ... ... ... ..... ........ ..... ...
... ... . ... ... ... ... ...
... ..
. .
.
.. .. .... .
. ...
.
. .
. ... .. .. . . . ....
......
..... ..... . .. ..... ..... ... ... ...... ... ... .. ... ... . . ... .... . . ... . .. ......
...... ...... ... ... ... ...
.. ... .. ... ... .. .. ... .. ...... .. . .. .... ...
...... ...... . ..... ........ . ... ... .. ... ... . ..... ...... ...... .. ..... ... .... ........ ... ...
..... .... ........ ..... .... ....... .......... ...... . .
..... .. .. . . .
.
. December .
. .
...... .. . .. . ... . ... . ... . . .... .... .. .. ........ ...
.. .. .. ... ...... . .... .. . .. . ...
...... .
.. .... ..... ... .... .. . .... .. .
-0.006 .....
...... .. ..
.
. ...
.
... . .. ..
.. ... .. ..
. ... ....... ..
. . .
.
... .... ..
... . ...
. . .
.
... ..
. .
.. .. .
.
... ... ... .. .. .. ...
.
. . . . ... .
...... . .
. ... . . . . .
.
. . .
...... .... .. . . ... .. ... .. ... ..
..... ..... . ... . . .. .. .
... .
.. . ... . . .
... ..
.. ..
....
...... .. ..
.. .....
.. .. .. ... .. .. .... ... . .. ...
. . ... .
. .. .. ..... . ... ... ... .. ... ..
....... ...... ..
. ....
. ..... ... . .
. .
.
.....
.
...
.. .. .
.. .
.
... . . .. ... .. ... .. .
-0.009 .... .. ... ....
. .. .. . ...... ... ..
.. .. ....
... .. . .... ... .... ...
...
.
..... ... .. . .
125 .......
.
.
....
.
.
.....
.. .
.
100 .. ..
. 75 ......
....
50 25
. . ..
.
. ...
...... .....
.. ....
.....
... Time to maturity in days
.
-0.012 ...

Figure 2.1: Basis behaviour for different contracts

As observed from the graph above and descriptive statistics for the basis (Table 2.4), the mean is negative for December
contract. This is because the futures price is lower than the spot price as it is known that the harvest would lead to a fall
in spot price by end of December. The mean for June and September contract is positive. The mean for March contract
though negative, is close to zero. The range and standard deviation is highest for December contract and lowest for June
contract. This is because the quantity of castorseed (and hence the spot price) is known completely during the life of a
June contract but not a December contract. There is little uncertainty regarding spot price of castorseed for June contract
but a high degree of uncertainty regarding spot price of castorseed for December contract. The reason is the harvest from
November to March. The December contract trades from August to December. The harvest is to begin and nothing is
known about the quantity of castorseed expected in the market. As a result, the futures price fluctuates widely resulting
in a wide range and high variability for December basis. For a June contract, since the harvest is over, the quantity of
castorseed in the market is known with certainty, hence the range of June basis is narrow and variability is low.
 16 INSTITUTIONAL FEATURES

The standard deviation of the basis is different for various contracts but similar for Mumbai and Ahmedabad (Table
2.4). The standard deviation of the basis is lowest for June contract and highest for December contract. The standard
deviation of basis is low for June contract because the harvest is over and there is no additional information available
about the expected spot price at maturity. The futures price does not vary much over the life of a June contract and hence
the variability of the basis for June contract is low. The standard deviation of basis is high for September and December
contract because of the higher uncertainty prevalent during these contracts. For September contract, the inventory is low,
spot price high and information about new harvest is not complete. Hence futures price varies considerably and basis
variability is high. For December contract, the harvest has begun, but is not yet over. Spot price falls rapidly but there is
uncertainty about the extent of the fall. Futures price fluctuates widely and consequently basis variability is high.

2.6 Data

All data employed in this thesis has been obtained from the Forward Markets Commission (FMC). This data is publicly
available via the FMC Bulletin and Annual Report as also through their web-site (fmc.gov.in).

2.7 Descriptive statistics

The descriptive statistics for castorseed basis, futures returns and spot returns are tabulated below.

Table 2.4: Descriptive statistics

Statistic March June September December

Break-up of contracts
Contracts 14 14 15 14
Observations 1375 1426 1560 1420

Castorseed basis, Mumbai

Minimum -0.0277 -0.0131 -0.0265 -0.0414
Maximum 0.0155 0.0186 0.0279 0.0192
Mean -0.0009 0.0055 0.0040 -0.0047
SD 0.0080 0.0057 0.0086 0.0117
Skewness -0.5361 -0.1698 -0.0858 -0.7677
Kurtosis 0.2898 -0.2765 0.7990 0.7178

Castorseed basis, Ahmedabad

Minimum -0.0193 -0.0094 -0.0309 -0.0376
Maximum 0.0189 0.0257 0.0274 0.0155
Mean -0.0015 0.0050 0.0049 -0.0046
SD 0.0080 0.0048 0.0072 0.0105
Skewness 0.3258 0.4532 -0.0612 -0.8434
Kurtosis -0.3655 1.3934 2.2605 0.5058

Castorseed futures returns, Mumbai

Minimum -0.0430 -0.0322 -0.0389 -0.0432
INSTITUTIONAL FEATURES 17

Statistic March June September December

Maximum 0.0445 0.0380 0.0363 0.0542
Mean 0.0002 0.0001 0.0005 0.0002
SD 0.0088 0.0075 0.0083 0.0090
Skewness 0.3863 0.0351 -0.0526 0.2043
Kurtosis 3.3408 2.3027 1.8648 3.0427

Castorseed spot returns, Mumbai

Minimum -0.0437 -0.0336 -0.0387 -0.0444
Maximum 0.0384 0.0397 0.0440 0.0449
Mean 0.0001 0.0002 0.0013 -0.0001
SD 0.0090 0.0078 0.0095 0.0103
Skewness -0.2227 0.2188 0.3843 -0.2113
Kurtosis 2.8121 2.4870 2.5403 3.3757

Castorseed futures returns, Ahmedabad

Minimum -0.0531 -0.0407 -0.0404 -0.0474
Maximum 0.0415 0.0549 0.0484 0.0529
Mean 0.0002 0.0001 0.0004 0.0003
SD 0.0094 0.0094 0.0099 0.0101
Skewness 0.0186 0.1557 0.1411 0.1174
Kurtosis 2.7542 2.9515 2.3417 2.8034

Castorseed spot returns, Ahmedabad

Minimum -0.0556 -0.0556 -0.0432 -0.0530
Maximum 0.0608 0.0608 0.0633 0.0500
Mean 0.0001 0.0003 0.0012 -0.0001
SD 0.0106 0.0093 0.0109 0.0118
Skewness 0.0249 0.1766 0.2708 -0.1663
Kurtosis 3.9610 4.3476 3.0048 2.7324
(S D is standard deviation)

From the table above, we observe that for both futures and spot returns the mean is close to zero and standard deviation
is close to one per cent. Skewness is close to zero and kurtosis is close to three. Hence, we conclude that futures and
spot returns follow a standard normal distribution. The Augmented Dickey Fuller (ADF) and the Phillips Perron tests
[38, 39, 125] indicate that futures and spot returns are stationary. The descriptive statistics for spot returns are tabulated
below.

Table 2.5: Descriptive statistics – spot returns

Statistic Pre-futures Futures period

period Early Later

Castorseed spot returns, Mumbai

Observations 803 815 785
Minimum -0.0529 -0.0437 -0.0269
Maximum 0.0556 0.0440 0.0282
Mean 0.0002 0.0010 0.0001
 18 INSTITUTIONAL FEATURES

Statistic Pre-futures Futures period

period Early Later
SD 0.0123 0.0116 0.0070
Skewness 0.0291 0.0399 0.1419
Kurtosis 4.3153 1.9309 1.5768

Castorseed spot returns, Ahmedabad

Minimum -0.0529 -0.0430 -0.0311
Maximum 0.0556 0.0483 0.0282
Mean 0.0002 0.0012 0.0007
SD 0.0123 0.0123 0.0074
Skewness 0.0291 0.1443 -0.1669
Kurtosis 4.3153 1.4170 1.4516

Turmeric spot returns, Sangli

Minimum -0.1088 -0.0777 -0.1146
Maximum 0.0862 0.1660 0.1466
Mean 0.0008 -0.0004 0.0007
SD 0.0147 0.0172 0.0159
Skewness -0.0554 1.3541 0.8277
Kurtosis 10.2794 13.8836 15.7988
(S D is standard deviation)

From the table above, we observe that the standard deviation of spot returns for castorseed has decreased in the futures
later sub period, but not in the futures early sub period. This indicates a decrease in volatility in the futures later sub
period. However, as the literature cautions, inferring from static value of variance may not be completely correct.
Chapter 3

Basis Risk

The key to successful commodity production and marketing decisions is to understand the basis and factors which affect its
behaviour. Unanticipated basis movement reduces the ability of futures market to transfer risk from hedgers to speculators
and results in lower income to hedgers. Commercial firms and traders who buy in either the spot or futures market and sell
in the other market do so in anticipation of a specific change in the basis. Unexpected change in basis creates additional
risk for the hedger associated with this market position and makes it less desirable to hold. Empirical assessment of the
magnitude and volatility of the basis and analysis of the factors influencing basis risk may permit the development of
select management practices designed to minimise basis risk impact on market participants’ decisions [56].

Consider the simple case of a producer who takes a long position in the spot market and a short position in the futures
market at time t1 and reverses the position at time t2 . The producer buys the spot commodity at c1 and sells at c2 . He
also sells the futures contract at f1 and buys it at f2 . The profits are (c2 − c1 ) + ( f1 − f2 ). In terms of the basis (futures -
spot), profits are b1 − b2 . Thus, the change in basis over the life of the hedge determines the producer’s profit. The risk
or variance of the producer’s profit is a function of the variance of basis (basis risk). The importance of basis and basis
risk arises because few contracts are offset through delivery. If delivery occurs, in the case of storage profits are given
by f1 − c1 . The case is similar to a forward contract, the agent contracts today for delivery tomorrow. In this case the
existence of a futures market allows the agent to shift risk [117].

Basis risk has a significant impact on inventory as well as the hedge ratio. Basis risk reduces inventory because it
limits the effectiveness of futures market as a price risk management tool. As producers face more basis risk, they reduce
their exposure by reducing the level of inventory. For a storable commodity, the basis will change by the reduction in
storage cost. The theory of storage suggests that the basis is equal to the cost of storage between today and expiry of
the contract. Thus, if a hedge is initiated two months before contract expiration date and lifted one month later, the basis
narrows by a month’s storage cost. This change in basis is anticipated; hence it is not considered basis risk [117].

The basis also changes due to the expected relative movement in spot price (vis-a-vis futures price). Hence it is a
function of expected shift in demand and supply. The impact of new information on the magnitude of basis is unclear.
The basis either contracts or expands depending on the nature of information. The arrival of new information influencing
price occurs randomly. Its frequency and impact on current and future demand and supply is uncertain. In early period
of the contract, new information about expected demand and supply influences the futures price more than the spot price.
Also, in this period forecast of subsequent demand and supply based on new information is subject to error. Thus, change
in local demand and supply conditions results in different levels of basis risk in various markets [56].

Basis risk differs across markets because the type and quality of commodity grown locally is not the same as what is
specified in the contract. Announcement of crop expectation or change in weather affects the expected basis and creates
basis risk if information in the announcement is unexpected. Basis risk varies across time and location because random

19
 20 BASIS RISK

shocks in each period affect the local market (impacting spot price) and the national market (impacting futures price)
differently [117].

3.1 Literature review

Garcia and Sanders [57] study basis risk in live hogs futures market and find it to be stable. They find that the basis is
neither more variable nor less predictable than historical standard.

3.2 Basis risk

We investigate the nature of basis risk for different contracts for the castorseed futures market at Mumbai and Ahmedabad
since inception (May 1985) up to August 1999. Following Peck [121] and Garcia and Sanders [57], we use root mean
squared error (RMSE) to measure basis risk. The test developed by Ashley, Granger and Schmalensee [6, 99] is used
to compare RMSE. We also use the Henrikson-Merton (H-M) timing test [57, 109] to gauge the qualitative accuracy of
forecast. Futures market for castorseed exist in India since May 1985 at two geographically close locations, Mumbai
and Ahmedabad. Four contracts maturing in March, June, September and December are traded each year at both places.
Mumbai is the commercial and marketing centre whereas Ahmedabad is the production centre (Section 2.3). The average
annual futures trading volume at Ahmedabad is far higher as compared to Mumbai (Section 2.4).

Castorseed is harvested each year from November up to March the next year (Section 2.2). This fact imparts a distinct
pattern to basis risk for various contracts. The December contract matures during harvest; hence there is uncertainty
regarding the quantum of new production. Spot price variability is high for December contract resulting in greater basis
variability. We hypothesise that compared to other contracts, basis risk will be higher for December contract. On the other
hand, exactly opposite is the case for June contract. Since the harvest is over, spot price variability and consequently basis
variability is low for June contract. We hypothesise that compared to other contracts, basis risk will be lower for June
contract. Further, we expect basis risk for March and September contract to be higher than June contract but lower than
December contract (Sections 2.5 and 2.7).

As compared to Mumbai, the futures market at Ahmedabad is relatively more liquid (Section 2.4). In a market with
large futures trading volume, futures and spot price is likely to be linked together resulting in lower basis risk. As
compared to Mumbai, spot price variability is lower at Ahmedabad due to its proximity to the production centre (Section
2.3). Lower spot price variability reduces basis variability resulting in reduced basis risk. Hence, we hypothesise that
compared to Mumbai, basis risk will be lower at Ahmedabad (Sections 2.5 and 2.7).

To examine the above hypotheses, we proceed as follows. We divide our data period into in-sample period (May 1985
to September 1995) and out-of-sample period (October 1995 to August 1999). The in-sample period is three times out-
of-sample period [109]. We utilise data from the in-sample period to develop a time series model of the basis (Equation
3.1) and use the model to forecast the basis for the out-of-sample period. We also develop a benchmark model (Equation
3.2) in which the basis is modelled as last year’s basis on the same date. The benchmark model is used to evaluate the
performance of the time series model. We measure basis risk and evaluate performance of the models [57, 121] using the
mean squared error of forecast (Section 3.6.1). The test developed by Ashley, Granger and Schmalensee [6, 99] is used
to compare RMSE (Section 3.6.2). We also evaluate the models on their timing ability – whether the models are able to
predict sign of the basis correctly – using the Henrikson Merton (H-M) timing test [57, 109] in its non-parametric form
(Section 3.6.3).
BASIS RISK 21

3.3 Methodology

The basis is defined earlier in Equation 2.1. Previous research and analysis [109] has suggested that a simple time series
model is as good at forecasting the basis as an economic model based on structural representation of the basis.

3.3.1 Basis models

Liu et al. [109] stress the importance of short term dynamics (lagged basis) in modelling the basis and find that they are
at least as important as fundamental variables in predicting the basis. As a distinct alternative to an economic model,
the basis is modelled here with the use of a set of variables to account for seasonality and time to maturity with the
residuals from the regression being used to identify an ARMA process following standard Box-Jenkins procedure. The
trend, seasonality and ARMA components are then jointly re-estimated.

For any futures contract, the basis declines over time reaching a level close to zero as the contract approaches maturity.
Hence, the time remaining to maturity is one of the explanatory variables affecting the basis. In a time series framework,
lagged basis serves as an explanatory variable. We first regress the basis on a constant and time to maturity and plot the
auto correlation function (ACF) and partial auto correlation function (PACF) of the residuals. We find that the ACF
decays exponentially and the PACF has two spikes at first and second lag. Hence, we model the basis as a Generalised
AutoRegressive Conditional Heteroscedasticity GARCH(1,1) Auto Regressive AR(2) process with two lags and trend
(TTM, time to maturity)

Bt = a + b1 ∗ T T M + b2 ∗ Bt−1 + b3 ∗ Bt−2 + ut ut ∼ N(0, ht ) (3.1)

ht = α0 + α1 ∗ ut−1
2
+ β1 ∗ ht−1

The naive or benchmark model that is used in the literature [108] as well as by traders, is one where the basis is modelled
as last year’s basis on the same date (Bl )
Bt = Bl + et (3.2)

3.4 Performance evaluation of alternative models

We measure, evaluate and compare basis risk across contracts and markets using the twin criteria of a) Root mean squared
error (RMSE) of the basis forecast [57, 121] and b) the Henrikson Merton (H-M) timing test in its non-parametric form
[57, 109].

3.4.1 Root mean squared error (RMSE)

The definition of risk employed is Root Mean Squared Error (RMSE) used in the literature. Peck [121] identifies the
mean squared forecast error (MSE) as a measure of basis risk. In a hedging context, the risk in returns from a completely
hedged position is directly proportional to the basis forecast error. We define a forecast of xt as ft , so that the forecast
error is εt = ft − xt . The root mean squared forecast error (RMSE) over n forecasts can be expressed as
s
1 n 2
RMSE = ∑ εt
n t=1
(3.3)

We employ the following steps to calculate RMSE [57]. First, we develop a time series model for the basis (Table 3.1)
using data from the in-sample period (May 1985 to September 1995). Next, we forecast the basis for the out-of-sample
 22 BASIS RISK

period (October 1995 to August 1999) for each contract for both markets separately using the benchmark model as well
as the time series model. We then calculate RMSE (Table 3.2) for the entire out-of-sample period for each contract, for
both markets (Mumbai and Ahmedabad) and for both models (benchmark and time series).

3.4.2 Ashley F test

Differences in mean squared error (RMSE) can be tested across models by employing the procedure set forth by Ashley,
Granger and Schmalensee [6] and illustrated by Kolb and Stekler [99]. For some out-of-sample observation t, let ut and
et be the forecast error made by time series and benchmark model respectively. The difference between the mean squared
error (MSE) is

∆t = et − ut (3.4)
Σt = et + ut (3.5)
MSE(et ) − MSE(ut ) = (ē − ū ) + (cov(∆, Σ))
2 2
(3.6)

We conclude that the time series model outperforms the benchmark model if we can reject the joint null hypothesis that
the mean and covariance of difference in MSE is zero (∆¯ = 0 and cov(∆, Σ) = 0) in favour of the alternative hypothesis
that both quantities are non negative and at least one is positive [6, 40, 99]. Now consider the regression equation

∆t = α + β(Σt − Σ̄t ) + ηt (3.7)

The test outlined above is equivalent to testing the null hypothesis α = β = 0 against the alternative that both are non
negative and at least one is positive. If either of the two least squares estimates α̂ or β̂ is significantly negative, the time
series model is not superior to the benchmark model. If both estimates are positive, the usual F test for joint significance
can be performed. The test for α = 0 is equivalent to testing mean of differences is zero while the test for β = 0 is
equivalent to testing covariance is zero [6, 40, 99]. For each market and contract we test whether the time series model
is significantly better than the benchmark model by estimating the above regression and examining whether α and β are
jointly non negative by employing the usual F test (Section 3.6.2).

3.4.3 Henrikson Merton (H-M) timing test

The RMSE provides quantitative information concerning the goodness of fit of forecast. However, detecting directional
change or timing is more crucial to the practitioner. When considering the basis and its role in futures pricing, a relevant
question is whether the basis will be positive or negative. That is, will the effective marketing price be above or below the
futures price at which the hedge is placed [57, 109]. The ability to correctly forecast the sign of the basis can be evaluated
using the Henrikson Merton (H-M) timing test [78]. A non-parametric test statistic is used to measure a model’s market
direction prediction ability. In its non-parametric form, we perform the H-M test as follows.

We define p1 (t) as the conditional probability of a correct forecast given that the basis has decreased or remained
constant. Similarly, we define p2 (t) as the conditional probability of a correct forecast given that the basis has increased.
As shown by Merton [113], a necessary and sufficient condition for a forecaster’s prediction to have zero information value
is that p1 (t) + p2 (t) ≤ 1. The prediction has a positive value only if p1 (t) + p2 (t) > 1. Therefore a test of forecaster’s
market timing ability is to determine whether or not p1 (t)+ p2 (t) > 1. To determine this probability we proceed as follows
[78, 109]. By definition
n1 n2
E( ) = p1 E( ) = 1 − p2 (3.8)
N1 N2

N1 is the number of observations with downward movement or no change in the basis

BASIS RISK 23

N2 is the number of observations with upward movement in the basis

N is the sum of N1 and N2 , total number of observations
n1 is the number of successful predictions of downward movement or no change in the basis, given N1
n2 is the number of unsuccessful predictions of downward movement or no change in the basis, given N2
n is the sum of n1 and n2 , number of times forecast of a downward or no change in the basis

Both n1 and n2 are sums of independently and identically distributed random variables with binomial distribution. Hence,
under the null hypothesis of zero information value in the forecast, the probability distribution for n1 – the number of
correct forecasts, given a downward movement or no change in the basis – has the form of a hypergeometric distribution
and is independent of both p1 and p2 [78]. For large samples (N > 50), the hypergeometric distribution can be approx-
imated by the normal distribution. The parameters used for this normal approximation are the mean and variance of the
hypergeometric distribution

nN1 n1 N1 (N − N1 )(N − n)
E(n1 ) = σ2 (n1 ) = (3.9)
N N 2 (N − 1)

A zero information value (null hypothesis) for a forecast means that the probability of obtaining a correct forecast is no
more than one half. Similarly, a positive information value for a forecast (alternative hypothesis) means that the probability
of the forecast being correct is more than one half. Henrikson and Merton [78] have tabulated the normal approximation
critical value at the 1% significance level for N = 50, 100 and 200 as 1.4, 1.3 and 1.2 respectively. The required value of
p1 (t) + p2 (t) decreases as total sample size increases. In our case, with a sample size of more than 300 observations, using
equation 3.9 we obtain a critical value of 1.03. We reject the null hypothesis of zero information value in the forecast at
the 5% significance level if calculated value is greater than 1.03 (Section 3.6.3).

3.5 Data

The standard deviation of basis is different for various contracts as well as for Mumbai and Ahmedabad. The standard
deviation of basis is lowest for June contract and highest for December contract. Also, the standard deviation of basis is
equal for March and September contract. Excepting March contract, standard deviation for the remaining three contracts
is lower for Ahmedabad as compared to Mumbai (Section 2.7). Performing an F test for equality of variance, we find that
basis variance is statistically not different for March contract at Mumbai and Ahmedabad. For the remaining contracts
at Mumbai and Ahmedabad the F test rejects equality of basis variance between contracts and across markets at the 1%
significance level. Though there is no appreciable difference in trading volume between contracts in the same market,
trading volume is far higher at Ahmedabad for all contracts (Section 2.4).

3.6 Results

The results of the basis modelled as a GARCH(1,1) auto regressive AR(2) process with two lags and trend for Mumbai
and Ahmedabad are tabulated below.

Table 3.1: Time series GARCH model of the basis

Coefficient March June September December

Mumbai
a 0.0002 -0.0002 0.0004 0.0001
 24 BASIS RISK

Coefficient March June September December

(0.0001) (0.0001) (0.0002) (0.0001)
b1 -0.0008 0.0010 -0.0014 -0.0004
(0.0003) (0.0003) (0.0007) (0.0004)
b2 0.7224 0.7870 0.8908 0.8930
(0.0263) (0.0279) (0.0267) (0.0296)
b3 0.2543 0.1903 0.0774 0.0922
(0.0266) (0.0278) (0.0273) (0.0297)
α0 0.0002 0.0002 0.0002 0.0002
(0.0001) (0.0001) (0.0001) (0.0001)
α1 0.2066 0.0545 NA 0.2604
(0.0502) (0.0160) (0.0678)
β1 NA 0.9346 0.9995 NA
(0.0193) (0.0016)
R̄2 0.97 0.95 0.94 0.98
Ljung Box 0.94 0.40 0.03 0.01

Ahmedabad
a 0.0002 -0.0002 0.0004 0.0001
(0.0001) (0.0001) (0.0002) (0.0001)
b1 -0.0007 0.0013 -0.0010 -0.0005
(0.0003) (0.0003) (0.0007) (0.0002)
b2 0.7274 0.7060 0.8428 0.8130
(0.0290) (0.0273) (0.0197) (0.0267)
b3 0.2534 0.2606 0.1175 0.1711
(0.0294) (0.0271) (0.0209) (0.0266)
α0 0.0002 0.0002 0.0002 0.0002
(0.0001) (0.0001) (0.0001) (0.0001)
α1 0.3601 0.0799 NA 0.2796
(0.0765) (0.0240) (0.0706)
β1 NA 0.8958 0.9988 NA
(0.0337) (0.0019)
R̄2 0.97 0.94 0.92 0.98
Ljung Box 0.86 2.30 0.02 0.14
(Standard error in parenthesis)

From the table above we observe that for all four contracts at both Mumbai and Ahmedabad the AR(2) model with
trend is a good fit (R̄2 > 0.9). The Durbin Watson statistic is close to two in all cases indicating an absence of first
order serial auto-correlation. As an additional check, the Ljung Box Q statistic for first order serial correlation at the 5%
level of significance is well below its critical value of 3.84. The trend is significant at the 5% level of significance for all
contracts except December contract at Mumbai and September contract at Ahmedabad. The coefficient for both lags is
statistically significant at the 5% confidence level for all contracts.

We have utilised data from May 1985 up to September 1995 (in-sample period) to build the model and employed
the model to forecast the basis from October 1995 up to August 1999 (out-of-sample period). To safeguard against the
possibility of a structural break between the in-sample and out-of-sample period, we perform a Chow test . We estimate
the model separately for the in-sample and out-of-sample period and test for a significant change in parameters. We find
BASIS RISK 25

that we cannot reject the null hypothesis of no change at the 5% level of significance for all contracts at both Mumbai and
Ahmedabad.

3.6.1 Root mean squared error (RMSE)

The RMSE for the benchmark and time series model for all contracts at both Mumbai and Ahmedabad is tabulated below.

Table 3.2: Root mean squared error (RMSE) results

Market Model March June September December

In-sample period (1985–95)

Mumbai Benchmark 1.00 0.63 1.31 1.54
Time series 0.46 0.45 0.88 0.85
Ahmedabad Benchmark 1.13 0.64 1.10 1.39
Time series 0.53 0.32 0.78 0.66

Out-of-sample period (1996–99)

Mumbai Benchmark 1.19 0.62 0.63 1.26
Time series 0.56 0.35 0.64 0.79
Ahmedabad Benchmark 1.00 0.52 0.48 1.04
Time series 0.35 0.23 0.45 0.64

From the table above we observe that for the out-of-sample period, RMSE for the benchmark model is higher than for
the time series model for all contracts (except September) at both Mumbai and Ahmedabad indicating the superiority of
the time series model in forecasting the basis. RMSE is low for June contract and high for December contract for both
models at Mumbai as well as Ahmedabad. RMSE is lower at Ahmedabad as compared to Mumbai for all contracts for
both models.

3.6.2 Ashley F test

The result of the Ashley F test for difference between RMSE is tabulated below. If both α̂ and β̂ are individually signifi-
cantly non negative, we examine the F statistic for joint significance. If either estimate is significantly negative, the F test
is not applicable (N A).

Table 3.3: Ashley F test results

Market March June September December

In-sample period (1985–95)

Mumbai N A 349.86 NA 964.88
Ahmedabad N A 783.36 NA 1809.67

Out-of-sample period (1996–99)

Mumbai 482.93 NA NA 60.55
Ahmedabad 642.04 NA NA 126.95
 26 BASIS RISK

Market March June September December

Employing the procedure set forth by Ashley, Granger and Schmalensee [6] and illustrated by Kolb and Stekler [99],
we find that RMSE for the September contract for time series model is not statistically different from benchmark model
at the 5% level of significance at Mumbai as well as Ahmedabad. RMSE for all remaining contracts for both markets is
statistically different at the 5% level of significance.

3.6.3 Henrikson Merton (H-M) timing test

The correct forecast (H-M test) for the benchmark and time series model for all contracts at both Mumbai and Ahmedabad
is tabulated below.

Table 3.4: Henrikson-Merton (H-M) timing test results

Market Model March June September December

In-sample period (1985–95)

Mumbai Benchmark 1.06 1.04 1.02 0.98
Time series 1.00 1.02 1.02 1.09
Ahmedabad Benchmark 1.02 0.98 1.06 1.00
Time series 0.99 1.02 1.05 1.06

Out-of-sample period (1996–99)

Mumbai Benchmark 1.04 1.08 1.05 0.96
Time series 1.10 1.07 1.04 1.10
Ahmedabad Benchmark 1.00 0.94 0.99 0.97
Time series 1.07 1.02 1.02 1.02
(Critical value: 1.03)

Employing the procedure outlined by Henrikson and Merton [78] we find that the critical value for rejecting the null
hypothesis of zero information value in the forecast at the 5% level of significance is 1.03. A zero information value for a
forecast means that the probability of obtaining a correct forecast is no more than one half. A positive information value
for a forecast means that the probability of the forecast being correct is more than one half. From the table above we
observe that for the Mumbai market, all contracts have positive information value except the December contract for the
benchmark model. In contrast, we observe that for the Ahmedabad market, no contract has a positive information value
except the March contract for the time series model.

3.7 Conclusion

The time series model performs better (lower RMSE and superior timing ability) than the benchmark model. Any effort
made towards developing models to forecast the basis is likely to reap results. RMSE is low for June contract and high for
December contract. For the June contract all information regarding supply of castorseed is known before trading begins.
Forecasting the futures price is not difficult resulting in lower basis risk and hence RMSE is low. Exactly opposite is the
case for December contract and hence RMSE is high.
BASIS RISK 27

The forecast for the Mumbai market has higher RMSE but better timing ability as compared to Ahmedabad market.
The link between liquidity and basis risk is unclear. If the lower RMSE for Ahmedabad market is due to higher futures
trading volume, the H-M timing test does not indicate superior timing ability.
Chapter 4

Price Discovery

Futures markets have evolved and grown rapidly because they contribute to the organisation of economic activity by
performing at least two important functions – risk transfer and price discovery. The essence of price discovery hinges on
whether new information is reflected first in changed futures price or in changed spot price. There are many reasons why
one market will react more rapidly to the arrival of new information. Possible explanatory factors include ease and lower
cost of trading, institutional arrangement and market micro structure effect. In practice, a futures market is commonly
observed to update price more frequently than spot market [93, 143]. Measurement of lead lag relation and identification
of the more influential market (futures or spot) in the pricing process is important because it identifies arbitrage opportunity
and strategy to reduce risk [101].

The introduction of new information results in price difference for short intervals of time between futures and spot
market due to positive information and communication cost. Due to increased availability and lower cost of information,
a futures market assimilates information faster than a spot market [101]. The price linkage between castorseed futures
and spot market is investigated using cointegration analysis which offers several advantages. First, cointegration analysis
measures the extent to which two markets have achieved a long run equilibrium. Another distinct advantage of the
cointegration technique is that it explicitly allows for divergence from equilibrium in the short run [50].

4.1 Literature review

Garbade and Silber [55] applied their framework to seven commodities (wheat, corn, oats, orange juice, copper, gold and
silver). They found that while futures market dominated spot market, the spot market also played a role in price discovery.
There was reverse information flow from spot to futures market as well. They also found that market size and liquidity
played a positive role in the price discovery function. Oellermann and Farris [118] investigated lead lag relation between
change in futures and spot price for live beef cattle between 1966 and 1982. The futures price led spot price during nearly
every sub period analysed. Based on Granger causality test for various sub samples of their data, they conclude that
change in live cattle futures price led change in live cattle spot price. They also found that the spot market responded to
change in futures price within one trading day. The authors conclude that futures market was the centre of price discovery
for live cattle. They suggest that a likely explanation for the results is that the futures market serves as a focal point for
information assimilation. They conclude that the cattle futures market contributes towards a more efficient price discovery
process in the underlying spot market for live beef cattle.

Herbst, McCormack and West [79] studied S&P index futures and Value Line Index (VLI) futures from February 1982
to September 1991 and found that futures price led the spot price. They also found some evidence of feedback. They

28
PRICE DISCOVERY 29

conclude that the spot index adjusted quickly so that knowledge of the lead lag relation could not be used for profitable
trading opportunity. Kawaller, Koch and Koch [93] examined the intraday price relation between S&P 500 futures and
the S&P index using minute by minute data for the period 1984–85. Their results suggest that futures price movement
led index movement by 20 to 45 minutes while movement in the index rarely affected futures beyond one minute. They
attribute the stronger futures leading spot relation to infrequent trading in the stock market.

Brorsen, Oellermann and Farris [22] investigated the direct impact of futures trading on the live cattle spot market.
They conclude that the introduction of futures trading in live cattle improved spot market efficiency, but also increased
short run spot price risk. Oellermann, Brorsen and Farris [119] investigated dominance for feeder cattle using daily price
data over an eight year period divided into two sub periods of four year each to account for structural change. The authors
found that feeder cattle futures price led spot price in incorporating new pricing information indicating that the futures
market serves as the centre of price discovery for feeder cattle. Using the Garbade Silber framework, the authors found
that futures market exhibited strong dominance over spot market in the first sub period and weak dominance in the second
sub period. Cheung and Ng [30] studied S&P index futures using data over 15 minute interval from April 1982 to June
1987. They found that futures led spot by 15 minutes with weak evidence of spot leading futures.

Koontz, Garcia and Hudson [101] employed Granger causality to examine dominance in live cattle market using
weekly data. Their study divided the sample period into three sub periods of four year each to examine the dynamic
nature of dominance. Their results showed changing dominance pattern due to structural change in the industry. Their
main finding was that the price discovery process is dynamic in nature and dependent on the structure of futures and
spot market. The authors conclude that the spot market was becoming less reliant on futures market for price discovery.
Stoll and Whaley [143] investigated the time series properties of five minute intra-day returns of stock index futures and
stock index. They found that the S&P 500 and Major Market Index (MMI) futures returns led stock market returns by
15 to 20 minutes, even after purging effect of infrequent trading. However, the lead lag relation was not completely uni-
directional, with lagged stock index returns having a mild positive predictive impact on futures returns in the inception
period of futures trading.

Schwarz and Laatsch [135] applied the Garbade Silber framework to the Major Market Index (MMI). Their results
based on both daily and weekly data showed that spot market dominated during the early period of futures trading, with
futures market exhibiting increasing dominance in later years. Their investigation highlighted the temporal nature of
dominance through the use of different subperiods and suggested that neither market maintains dominance at all times.
Bessler and Covey [13] applied cointegration analysis to the live cattle market and found slight evidence of cointegration
between nearby futures and spot price but no evidence of cointegration when more distant futures contract was considered.
Schroeder and Goodwin [134] examined short and long run price relation between Omaha spot and Chicago Mercantile
Exchange (CME) futures daily price for live hogs. They found that the futures and spot series operated somewhat inde-
pendently and the long term basis was generally non stationary. They found similarly that futures market dominated spot
market for live hogs.

Chan, Chan and Karolyi [26] studied S&P index futures and Major Market Index (MMI) futures using data over five
minute interval from August 1984 to December 1989. They examined the intra-day pattern of lead lag relation using
a bi-variate GARCH framework and found strong intra-market dependence. They found much stronger bi-directional
dependence between stock index and stock index futures price change when the volatility of price change was also con-
sidered. Khoury and Martel [98] study the informational content of basis using barley, oats and canola futures during
1980–88. They found that futures price led spot price on a day-to-day basis. They also found that the feedback from
spot to futures was weak. Chan [27] investigated the stock index market on an intra-day basis using lead lag regression.
He examined the effect of good and bad news, trading intensity and market wide price movement on dominance. The
results showed weak dominance by futures market and no effect due to good or bad news. While the study found no com-
pelling evidence that trading intensity affected lead lag relation, Chan suggests that response to market wide movement
was supportive of dominance by futures market.
 30 PRICE DISCOVERY

Quan [128] used monthly crude oil price in a two step procedure to study price discovery. The steps involved first
establishing a long term relation between futures and spot price using cointegration and then testing lead lag relation with
Granger causality test. Quan found that spot price led futures price. This finding is however questionable due to the
frequency of data used [136]. Wahab and Lashgari [147] used cointegration analysis to examine the temporal causal link
between stock index futures and stock index for both S&P 500 and FTSE 100 and found that futures and spot market is
cointegrated. They found that futures price exhibited a stronger response to disequilibrium. Although feedback existed
between futures and spot market for both S&P 500 and FTSE 100, the spot to futures lead was more pronounced.

Fortenbery and Zapata [49] applied cointegration analysis to futures and spot market for corn and soyabean. Evidence
of cointegration was detected for all spot and futures market pairs considered. Their results suggest that where the cost
of maintaining inventory is high, a more appropriate specification of the relation between spot and futures price is to
include interest rate also as an explanatory variable. Zapata and Fortenbery [156] introduced interest rate as an argument
in the cointegration model and found that it was important in describing the price discovery relation between futures
and spot market for storable commodities. Schwarz and Szakmary [136] replicated Quan’s [128] analysis with daily
observations (instead of monthly) and found that futures market dominated spot market, a result opposite that of Quan
but in conformity with other studies. They attribute Quan’s failure to find support for the price discovery function of
futures market to inappropriate choice of data and time interval. They found the petroleum futures and spot market to be
cointegrated. Their study suggests that futures dominate in price discovery in all three petroleum product markets.

Grunbichler, Longstaff and Schwartz [72] studied the German index futures market using minute by minute data from
November 1990 to September 1991 and found that futures price led spot price by 15 to 20 minutes. As stocks are floor
traded while futures are screen traded they attribute the significant lead lag relation to the hypothesis that screen trading
accelerates the price discovery process. Harris et al. [76] study price discovery for IBM stock on New York, Midwest and
Pacific Exchanges. They found that all three markets react to independent information reflected in each Exchange’s price.
They found that though price discovery occurs primarily on NYSE, the two smaller Exchanges also play a significant role
in the price discovery process. Hasbrouck [77] studied price discovery for the 30 Dow stocks traded on the New York
Stock Exchange (NYSE) as well as on other regional Exchanges. The results suggest that price discovery is concentrated
at the NYSE. Interestingly, the study found that for most stocks price discovery occurs predominantly on the NYSE even
though trading volume for these stocks is lower on NYSE as compared to other Exchanges. Significantly, the author also
found that price discovery takes place primarily in the medium sized (2,500–10,000) trades. It is these trades that convey
more information.

Foster [51] studied the behaviour of crude oil futures and spot market in the USA and UK during the 1990–91 Gulf
conflict. The objective was to examine price discovery with emphasis on its time varying nature. This was achieved using
the technique of Kalman filter to obtain time varying estimates. The results indicate that such relationship is strongly
temporal and market conditions affect the price discovery process. The author found that dominance is time varying
implying that the use of time invariant point estimation cannot adequately explain dominance relationship. Fleming,
Ostdiek and Whaley [48] using minute by minute data from January 1988 to March 1991 found that S&P 500 futures
led the S&P 500 index by 5 minutes with no evidence of feedback. They also found that the contemporaneous relation
between the S&P 500 futures and S&P 500 index has grown stronger over time and feedback from futures to spot market
has almost disappeared. The authors suggest that this is a sign of increased integration between index futures and the
stock market.

Shyy, Vijayraghavan and Scott-Quinn [138] find reverse causality from spot to futures for the CAC index (France).
They suggest that previous results indicating futures market leading the spot market are primarily due to market asyn-
chronous trading and difference in trading mechanisms used in futures or spot market. Pizzi, Economopolous and O’Neill
[126] examined price discovery in the S&P 500 index and its three month and six month index futures using intra-day
minute by minute data. Using cointegration analysis several error correction models were developed. Using the Engle
Granger two step procedure, the authors found that both the spot index and three month futures and the spot index and six
PRICE DISCOVERY 31

month futures are cointegrated, indicating market efficiency and stability. Their results indicate that the futures market
led spot market by 20 minutes while the spot market led futures market by 3 minutes. They found that while the futures
market had a stronger lead effect, there was causation from spot to futures as well.

Fortenbery and Zapata [50] examined whether futures and spot market for cheese have established a long run equilib-
rium relation in terms of pricing behaviour and whether one market dominated as the centre for price discovery. Based
on cointegration test they found no evidence of a stable long run relation between futures and spot market for cheese.
Abhyankar [2] studied price discovery in the FTSE 100 index futures market using intra-day minute by minute data.
Employing the technique of linear and non linear Granger causality the author found that FTSE 100 index futures led the
spot index by 5 to 15 minutes. The author also found that, in contrast to the linear test, if non linear effect is accounted
for, neither market leads or lags the other. Silvapulle and Moosa [139] examine the lead lag relation between futures and
spot crude oil price using both linear and non linear causality tests. Their study concludes that though the futures market
plays a bigger role in the price discovery process, the spot market also plays a role in this regard. Min and Najand [114]
investigate the lead lag relationship in returns between the newly established futures and spot market in Korea. They find
that the futures market leads spot market by 30 minutes.

4.2 Price discovery across markets

Usually, price discovery is studied in a futures market and its corresponding spot market. We term this as price discovery
‘within’ market. However, in India two futures markets exist for castorseed at Mumbai and Ahmedabad. This gives us
an opportunity to study price discovery ‘across’ markets by designating one of the futures market, say at Mumbai, as
the futures market and the other futures market, at Ahmedabad, as the spot market. Such a study would reveal whether
amongst multiple futures markets for the same commodity, which are in geographical proximity, one futures market
dominates the other in terms of price discovery (Section 4.7.5).

4.3 Price discovery over time

Examining futures trading volume at Mumbai and Ahmedabad (Section 2.4) we find that there has been a fall from 1994
onwards. Hence, we also examine whether the price discovery role of futures market has changed over time by studying
it separately for the two sub periods 1985–93 and 1994–99 (Section 4.7.6).

4.4 Price discovery and cointegration

To study price discovery, different econometric techniques, such as regression analysis and spectral analysis have been
developed over time. The use of cointegration analysis and error correction models takes into account non stationarity
and enables one to distinguish between short run deviation from equilibrium indicative of price discovery and long run
deviation that accounts for efficiency and stability [147]. If two series (such as futures and spot price) are non stationary
but a linear combination of the two variables (such as the basis) is stationary so that both are cointegrated then a bi-variate
dynamic model that uses only first difference with lags is mis-specified because it ignores interim short run adjustment
to long run equilibrium [42]. A percent change specification is also mis-specified because it fails to consider short run
dynamics by leaving out lagged terms.

Cointegration methodology considers non stationarity and allows for both short term and long run adjustment [147].
Economic forces such as arbitrage prevent futures and spot price from drifting too far apart over time. Cointegration acts
as evidence of this long run equilibrium in which deviation is due to short run shocks [33]. The link between cointegration
 32 PRICE DISCOVERY

and causality stems from the fact that if futures and spot price are cointegrated; causality must exist in at least one direction
[67]. The possibility that one variable in a system of cointegrated series is exogenous (independent) within the error
correction process motivates the use of error correction models in evaluating price discovery. The cointegrating vectors
define the long run equilibrium while error correction dynamics characterise the price discovery process, whereby markets
attempt to find equilibrium [133]. An appropriately specified error correction model allows the long run component of
variables to obey equilibrium constraints, while the short term component has a flexible dynamic specification which may
diverge from the long run trend. The model indicates not only the proportion of disequilibrium from one period that is
corrected in the next, but also the relative magnitude of adjustment in both markets towards equilibrium. This also reveals
the nature of causal relationship [147].

Cointegration analysis represents an attempt to determine whether two markets (such as futures and spot) are pricing
information similarly by investigating the price difference between the two markets. A major advantage of cointegration
analysis is that it allows for the possibility that price for identical commodities in two separate markets may respond
differently to information in the short run, but would return to a long run equilibrium if both are efficient. There are
several reasons for an asymmetric response from different markets in the short run. One is that the markets have different
access times to the information being delivered. Another is that the information is interpreted differently. However,
because the markets are for the same commodity, arbitrage opportunity between the markets eventually results in a multi
market consensus concerning the value of information.

4.5 Methodology

Evidence of price change in one market (futures or spot) generating price change in the other market (spot or futures) so
as to bring about a long run equilibrium relation is

Ft − α − βSt = ût (4.1)

where Ft and St are futures and spot price at time t. Ordinary least squares (OLS) is inappropriate if Ft or St is non
stationary because the standard errors are not consistent. If Ft and St are non stationary but the deviation ût is stationary,
Ft and St are said to be cointegrated (Section 4.7.1). If each series (Ft and St ) is non stationary in the level but the
first difference (∆Ft and ∆St ) and deviation (ût ) are stationary, the series are said to be cointegrated of order (1,1) with
β as the cointegrating coefficient (Section 4.7.2). An error correction model (ECM) exists for each series which is
not subject to spurious results [126]. Cointegration implies that each series can be represented by an error correction
model which includes last period’s equilibrium error as well as lagged values of first difference of each variable1 . Hence,
temporal causality can be assessed by examining the statistical significance and relative magnitude of the error correction
coefficient and coefficient on lagged variables. The error correction model [126, 147] is

∆Ft = δ f + α f ût−1 + β f ∆St−1 + γ f ∆Ft−1 + e f ,t (4.2)

∆St = δs + αs ût−1 + βs ∆Ft−1 + γs ∆St−1 + es,t (4.3)

Each of the two equations above is interpreted as having two parts. The first part (ût−1 ) is the equilibrium error. This
measures how the left hand side variable adjusts to the previous period’s deviation from long run equilibrium. The
remaining portion of the equation is lagged first difference which represents the short run effect of previous period’s
change in price on current period’s deviation. The coefficient of the equilibrium error, α f and αs , are the speed of
adjustment coefficient and have important implication in an error correction model. At least one speed of adjustment
coefficient must be non zero for the model to be an ECM. The coefficient serves the role of identifying the direction of
causal relation and shows the speed at which departure from equilibrium is corrected. If α f is statistically insignificant,
1 Clive Granger received the Nobel prize in Economics (2003) for this insight.
PRICE DISCOVERY 33

the current period’s change in futures price does not respond to last period’s deviation from long run equilibrium. If both
α f and β f are statistically insignificant; the spot price does not Granger cause futures price [126, 147].

The primary purpose of estimating the ECM is to implement price leadership test between futures and spot price.
Tests of causality between cointegrated variables are conducted in an error correction framework because standard tests of
causality overlook the reversion to equilibrium channel of causality represented by ût−1 (Equation 4.1). Causality test in
the ECM framework involves testing exclusion restriction on the coefficients α f , β f (Equation 4.2) and αs , βs (Equation
4.3). If the coefficients are jointly insignificant, there is no Granger causality. If futures and spot price are cointegrated,
there is causality in at least one direction [42, 68, 136]. The dynamics of price discovery are studied using the maximum
likelihood approach of Johansen and Juselius [87, 88]. An important feature of this approach is that it jointly incorporates
the long run equilibrium between spot and futures market resulting in median unbiased long run coefficients [49].

An issue that needs to be resolved before estimation is regarding pooling of data. Preliminary analysis using dummies
to account for different maturities as well as different time periods indicated that there was no significant difference
amongst the various contracts over time (the dummies turned out to be insignificant). However, a significant difference
was found between contracts of different maturities (Section 2.5). Hence we pool data for all March, June, September and
December contracts and estimate separately for each contract at Mumbai and Ahmedabad.

4.6 Data

From Section 2.7 we observe that for both futures and spot returns the mean is close to zero and the standard deviation
is close to one percent. Skewness is close to zero and kurtosis is close to three. Hence, we infer that futures and spot
returns at both Mumbai and Ahmedabad follow a standard normal distribution. The Augmented Dickey Fuller (ADF) and
Phillips Perron tests [38, 39, 125] indicate that futures and spot returns at both Mumbai and Ahmedabad are stationary.
From Section 2.4 we note the higher futures trading volume at Ahmedabad as compared to Mumbai.

4.7 Results

In this section the results of cointegration, causality and price discovery for the futures market at Mumbai, Ahmedabad
and across markets are discussed.

4.7.1 Unit root test

To determine the order of integration of futures and spot price we perform the Augmented Dickey Fuller (ADF) unit root
test [38, 39] on the level of futures and spot price. A constant and a time trend are included in the equation. In all cases,
a lag length of two is found to be sufficient in removing autocorrelation. The critical value at the 5% level of significance
is -3.41. The results are tabulated below.

Table 4.1: Unit root test results

Period Series March June September December

Mumbai
1985–93 Futures -2.49 -2.00 -1.89 -2.16
Spot -2.19 -2.04 -1.51 -2.83
1994–99 Futures -3.05 -2.18 -1.97 -2.14
 34 PRICE DISCOVERY

Period Series March June September December

Spot -2.42 -1.71 -1.44 -1.96

Ahmedabad
1985–93 Futures -2.44 -2.03 -1.62 -2.15
Spot -2.19 -2.02 -1.35 -1.96
1994–99 Futures -2.69 -2.11 -1.94 -2.06
Spot -2.44 -1.67 -1.61 -1.89
(Critical value: -3.41)

From the table above, we observe that the null of no unit root cannot be rejected and hence we conclude that both
futures and spot price series are I(1). The first difference of both futures and spot price is found to be stationary. This is
true for all four contracts at Mumbai as well as Ahmedabad. Results are qualitatively unchanged using the Phillips Perron
unit root test [125] .

4.7.2 Cointegration test

The results relating to cointegration test using the Johansen technique are tabulated below. In the table below r denotes
the number of cointegrating vectors. In our case since there are only two series, the number of cointegrating vectors can
be at most one. The critical value at the 10% level of significance for r = 0 and r ≤ 1 is 17.96 and 7.56 respectively. For
two series to be cointegrated, the test statistic for r = 0 should be siginificant jointly with an insignificant test statistic for
r ≤ 1 [49].

Table 4.2: Cointegration test results

Period Null March June September December

Mumbai
1985–93 r=0 18.36 19.41 22.82 16.76
r≤1 5.09 4.08 2.37 5.59
1994–99 r=0 16.17 14.11 20.90 17.76
r≤1 4.66 2.77 2.15 3.76
1985–99 r=0 25.92 29.21 34.58 24.93
r≤1 7.27 5.47 2.18 7.04

Ahmedabad
1985–93 r = 0 16.77 19.00 22.85 18.75
r≤1 4.82 4.91 1.98 6.34
1994–99 r = 0 15.55 18.90 29.66 18.70
r≤1 5.81 2.79 2.72 3.26
1985–99 r = 0 23.77 30.51 42.59 23.56
r≤1 7.09 6.77 2.92 7.10

Across markets
1985–93 r = 0 23.21 16.27 15.18 17.58
r≤1 5.97 3.96 2.31 4.56
1994–99 r = 0 36.32 29.52 29.53 22.35
PRICE DISCOVERY 35

Period Null March June September December

r≤1 6.97 4.52 3.90 6.62
1985–99 r=0 34.47 25.85 22.33 23.19
r≤1 6.95 6.42 3.70 6.26
(Critical value for r = 0: 17.96)
(Critical value for r ≤ 1: 7.56)

From the table above we observe that futures and spot market are mostly cointegrated. If we consider results for the
entire period, futures and spot market at Mumbai and Ahmedabad as well as futures market at Mumbai and Ahmedabad
are cointegrated and this is true for all four contracts. At Mumbai, futures and spot market are not cointegrated (except
September) for the period 1994–99. A possible reason could be the fall in futures trading volume post 1994 at Mumbai
(Section 2.4). At Ahmedabad, futures and spot market are not cointegrated for the March contract. Ahmedabad is the
production centre for castorseed (Section 2.3) and March is the harvest contract (Sections 2.2 and 2.5). An absence of
long term price linkage for March contract at Ahmedabad is surprising especially during the 1994–99 period when futures
trading volume had risen considerably (Section 2.4). Across markets, futures market at Mumbai and Ahmedabad were
not cointegrated in the early period (1985–93) but are cointegrated in the later period (1994–99). This indicates that long
term price linkage between the futures market at Mumbai and Ahmedabad has strengthened over time possibly due to the
increase in futures trading volume at Ahmedabad post 1994 (Section 2.4).

4.7.3 Price discovery within market at Mumbai

In this section the results for cointegration, causality and price discovery for Mumbai are discussed. Results for all four
contracts (March, June, September and December) are shown in the following table. We have used only one lag as it
was found to be sufficient in removing autocorrelation. The Durbin Watson test statistic is close to two in all cases
indicating an absence of first order autocorrelation. As an additional check, the Ljung Box Q statistic for first order serial
correlation at the 5% level of significance is well below its critical value of 3.84.

Table 4.3: Error correction model results: Mumbai

Coefficient March June September December

Period: 1985–93
αf -0.0129 0.0022 0.0003 0.0016
(0.0061) (0.0076) (0.0047) (0.0035)
βf 0.1312 0.1161 0.0935 0.1239
(0.0426) (0.0433) (0.0325) (0.0381)
γf -0.1313 -0.0763 -0.0527 -0.0044
(0.0500) (0.0510) (0.0461) (0.0428)
R̄2 0.02 0.01 0.01 0.02
Ljung Box 0.01 0.01 0.01 0.03

αs 0.0186 0.0234 0.0122 0.0143

(0.0091) (0.0086) (0.0054) (0.0142)
βs 0.2864 0.2928 0.3117 0.2381
(0.0389) (0.0499) (0.0414) (0.0408)
γs 0.0463 0.0271 -0.0347 0.1223
(0.0415) (0.0445) (0.0446) (0.0451)
 36 PRICE DISCOVERY

Coefficient March June September December

R̄2 0.09 0.08 0.06 0.09
Ljung Box 0.10 0.35 0.15 0.90

Period: 1994–99
αf -0.0180 0.0059 -0.1040 -0.0040
(0.0167) (0.0120) (0.0097) (0.0081)
βf 0.1279 0.0652 0.0860 -0.0391
(0.0506) (0.0523) (0.0512) (0.0516)
γf 0.0172 -0.0613 0.1272 0.1956
(0.0566) (0.0546) (0.0616) (0.0618)
R̄2 0.03 0.01 0.03 0.03
Ljung Box 0.01 0.02 0.01 0.14

αs -0.0052 0.0144 -0.0067 0.0157

(0.0068) (0.0097) (0.0105) (0.0097)
βs 0.4204 0.2912 0.3725 0.3551
(0.0545) (0.0520) (0.0481) (0.0666)
γs 0.0584 -0.0266 0.0157 0.1364
(0.0748) (0.0697) (0.0594) (0.1043)
R̄2 0.20 0.12 0.13 0.15
Ljung Box 0.10 0.01 0.15 0.20
(Standard error in parenthesis)

The results for the error correction model are consistent with and support the results for cointegration. At least one
error correction coefficient (α f or αs ) is significant in all cases where the Johansen technique indicates the presence
of a cointegrating vector. The coefficient βs is always significant indicating that causality exists from futures to spot for
all contracts and periods. In other words price discovery occurs in futures market for all contracts and periods. Price
discovery occurs additionally in the spot market also for some contracts and periods as indicated by a significant β f . The
magnitude of βs is at least twice that of β f indicating a stronger feedback from futures to spot market for all contracts and
periods [126, 147].

At Mumbai, causality is mostly bi-directional except for the period 1994–99. Price discovery occurs in both futures
and spot market; there is to and fro information flow. It is surprising that in spite of a fall in futures trading volume post
1994 (Section 2.4), price discovery occurs in futures market and there is uni-directional causality from futures to spot
market during 1994–99. Overall, there is evidence of cointegration and bi-directional causality. However, low futures
trading volume post 1994 (Section 2.4) has resulted in reduced price linkage and reduced information flow between
futures and spot market during 1994–99.

The results indicate that for March contract price discovery occurs in both futures and spot market. There is a bi-
directional flow of information between futures and spot market for the March contract. This is true for the entire period
as well as both sub periods. A possible explanation for price discovery occurring in both futures and spot market for
March contract is the harvest of castorseed from November to March (Section 2.5). The March contract is active from
November to March which coincides with the harvest season (Section 2.2). Information that flows into the market is
processed by both futures and spot market simultaneously. Hence, for March contract, price discovery occurs in both
futures and spot market.
PRICE DISCOVERY 37

4.7.4 Price discovery within market at Ahmedabad

In this section the results for cointegration, causality and price discovery for Ahmedabad are discussed. Results for all
four contracts (March, June, September and December) are shown in the following table. We have used only one lag as
it was found to be sufficient in removing autocorrelation. The Durbin Watson test statistic is close to two in all cases
indicating an absence of first order autocorrelation. As an additional check, the Ljung Box Q statistic for first order serial
correlation at the 5% level of significance is well below its critical value of 3.84.

Table 4.4: Error correction model results: Ahmedabad

Coefficient March June September December

Period: 1985–93
αf -0.0024 -0.0239 -0.0026 0.0053
(0.0067) (0.0126) (0.0071) (0.0062)
βf 0.0815 0.0974 0.1228 0.0354
(0.0449) (0.0741) (0.0436) (0.0355)
γf -0.0274 -0.0672 -0.0798 -0.0126
(0.0538) (0.0966) (0.0492) (0.0504)
R̄2 0.01 0.01 0.01 0.01
Ljung Box 0.01 0.01 0.02 0.01

αs 0.0214 -0.0025 0.0181 0.0200

(0.0181) (0.0142) (0.0101) (0.0055)
βs 0.3475 0.2947 0.2681 0.2907
(0.0636) (0.0965) (0.0567) (0.0518)
γs -0.1340 -0.0790 -0.0706 -0.0293
(0.0616) (0.0944) (0.0558) (0.0492)
R̄2 0.06 0.05 0.04 0.07
Ljung Box 0.02 0.04 0.15 0.41

Period: 1994–99
αf -0.0105 0.0055 -0.0269 -0.0087
(0.0087) (0.0194) (0.0117) (0.0100)
βf -0.0135 0.0285 -0.0199 0.0627
(0.0642) (0.0655) (0.0599) (0.0512)
γf 0.1138 -0.0185 0.1084 0.0777
(0.0564) (0.0634) (0.0682) (0.0606)
R̄2 0.03 0.01 0.01 0.01
Ljung Box 0.01 0.01 0.02 0.03

αs -0.0054 0.0369 0.0135 0.0231

(0.0074) (0.0122) (0.0127) (0.0111)
βs 0.5176 0.3664 0.4935 0.3055
(0.0527) (0.0554) (0.0647) (0.0776)
γs -0.0615 -0.0202 -0.1170 0.1089
(0.0645) (0.0715) (0.0628) (0.0792)
R̄2 0.24 0.17 0.16 0.10
 38 PRICE DISCOVERY

Coefficient March June September December

Ljung Box 0.54 0.02 1.43 0.01

(Standard error in parenthesis)

The results for the error correction model are consistent with and support the results for cointegration. At least one
error correction coefficient (α f or αs ) is significant in all cases where the Johansen technique indicates the presence
of a cointegrating vector. The coefficient βs is always significant indicating that causality exists from futures to spot for
all contracts and periods. In other words price discovery occurs in futures market for all contracts and periods. Price
discovery occurs additionally in the spot market also for some contracts and periods as indicated by a significant β f . The
magnitude of βs is at least twice that of β f indicating a stronger feedback from futures to spot market for all contracts and
periods [126, 147].

At Ahmedabad, causality is mostly uni-directional. Price discovery occurs mostly in futures market and information
flows from futures to spot market but not the other way round, especially during 1994–99. This is true for all four contracts.
A possible reason could be the rise in futures trading volume post 1994 (Section 2.4) resulting in futures market leading
the spot market in terms of price discovery. Overall, there is evidence of cointegration and uni-directional causality.

The results indicate that for March contract there is lack of cointegration between futures and spot market during both
sub periods and price discovery occurs only in futures market. There is uni-directional flow of information between futures
and spot market for the March contract. This is true for the entire period as well as later sub period. This is surprising
since Ahmedabad is the production centre for castorsed (Section 2.3) and March is the harvest contract (Sections 2.2 and
2.5). A possible explanation for price discovery occurring only in futures market for March contract coluld be the increase
in futures trading volume post 1994 (Section 2.4).

4.7.5 Price discovery across markets

In this section the results for cointegration, causality and price discovery across futures markets are discussed. Results
for all four contracts (March, June, September and December) are shown in the following table. The futures market at
Ahmedabad is designated as the spot market. We have used only one lag as it was found to be sufficient in removing
autocorrelation. The Durbin Watson test statistic is close to two in all cases indicating an absence of first order autocor-
relation. As an additional check, the Ljung Box Q statistic for first order serial correlation at the 5% level of significance
is well below its critical value of 3.84.

Table 4.5: Error correction model results: Across markets

Coefficient March June September December

Period: 1985–93
αf -0.0207 0.0035 -0.0067 -0.0053
(0.0082) (0.0067) (0.0077) (0.0088)
βf 0.1828 0.1463 0.0953 0.0470
(0.0507) (0.0552) (0.0399) (0.0457)
γf -0.2262 -0.1563 -0.0868 0.0070
(0.0550) (0.0674) (0.0553) (0.0531)
R̄2 0.03 0.01 0.01 0.01
Ljung Box 0.07 0.01 0.01 0.01
PRICE DISCOVERY 39

Coefficient March June September December

αs -0.0113 0.0083 0.0037 0.0148
(0.0086) (0.0084) (0.0096) (0.0095)
βs 0.0001 0.2123 0.2740 0.2169
(0.0588) (0.0943) (0.0646) (0.0597)
γs 0.0225 -0.1131 -0.1414 -0.1003
(0.0665) (0.0917) (0.0480) (0.0536)
R̄2 0.01 0.01 0.02 0.03
Ljung Box 0.01 0.01 0.01 0.09

Period: 1994–99
αf -0.0301 0.0137 -0.0348 -0.0145
(0.0161) (0.0200) (0.0155) (0.0158)
βf 0.1872 0.1191 0.0572 0.0046
(0.0702) (0.0524) (0.0680) (0.0682)
γf -0.0695 -0.1347 0.1205 0.1774
(0.0809) (0.0662) (0.0802) (0.0823)
R̄2 0.03 0.01 0.03 0.03
Ljung Box 0.01 0.02 0.01 0.13

αs 0.0066 0.0848 0.0090 0.0264

(0.0220) (0.0230) (0.0188) (0.0151)
βs 0.0333 -0.0233 0.1636 0.2678
(0.0834) (0.0800) (0.0753) (0.0766)
γs 0.0902 0.0229 -0.0076 -0.0486
(0.0782) (0.0714) (0.0739) (0.0725)
R̄2 0.01 0.02 0.02 0.06
Ljung Box 0.01 0.01 0.13 0.71

(Standard error in parenthesis)

The results for the error correction model are consistent with and support the results for cointegration. At least one
error correction coefficient (α f or αs ) is significant in all cases where the Johansen technique indicates the presence of
a cointegrating vector. The results indicate that the futures market at Mumbai and Ahmedabad are cointegrated overall
as well as for later sub period implying that the price linkage between futures market at Mumbai and Ahmedabad has
strengthened over time. A contributory factor for the stronger price linkage could be the increase in futures trading
volume at Ahmedabad post 1994 (Section 2.4).

The coefficient βs is significant for December contract indicating uni-directional causality and information flow from
Mumbai to Ahmedabad futures market. Similarly, the coefficient β f is significant for March contract indicating uni-
directional causality and information flow from Ahmedabad to Mumbai futures market. This is true overall and for both
sub periods. The harvest occurs from November to March (Sections 2.2 and 2.5) and Ahmedabad is the production centre
for castorseed (Section 2.3). Hence for the March contract, price discovery occurs in the Ahmedabad futures market and
information flows from Ahmedabad to Mumbai futures market. Information is processed faster in the Ahmedabad futures
market. For the December contract it is not clear why information flows uni-directionally from Mumbai to Ahmedabad
futures market.
 40 PRICE DISCOVERY

4.7.6 Price discovery over time

From Section 4.7.3 we observe that at Mumbai there is no cointegration between futures and spot market during 1994–99
possibly due to a steep fall in futures trading volume during that period (Section 2.4). We also observe that there is no
causality and information flow from spot to futures market (except March contract) during this period. From Section 4.7.4
we observe that at Ahmedabad futures and spot market are cointegrated (except March contract) for both sub periods. The
lack of cointegration for March contract is surprising since Ahmedabad is the production centre for castorseed (Section
2.3) and March is the harvest contract (Sections 2.2 and 2.5). From Section 4.7.5 we observe that across markets, coin-
tegration between futures market at Mumbai and Ahmedabad has strengthened over time possibly due to the increase in
futures trading volume at Ahmedabad post 1994 (Section 2.4).

4.8 Inference

The inference that can be drawn from the results is summarised below. In the following table, Yes (No) indicates the
presence (absence) of cointegration, F indicates futures market, S indicates spot market, M indicates futures market at
Mumbai and A indicates futures market at Ahmedabad. The symbol → indicates uni-directional causality while the
symbol ⇔ indicates bi-directional causality.

Table 4.6: Cointegration and price discovery inference

Period March June September December

Mumbai
1985–93 Yes Yes Yes No
F⇔S F⇔S F⇔S F⇔S
1994–99 No No Yes No
F⇔S F→S F→S F→S

Ahmedabad
1985–93 No Yes Yes Yes
F⇔S F→S F⇔S F→S
1994–99 No Yes Yes Yes
F→S F→S F→S F→S

Across markets
1985–93 Yes No No No
A→M M⇔A M⇔A M→A
1994–99 Yes Yes Yes Yes
A→M A→M M→A M→A

From the table above we note that at Mumbai futures and spot market are mostly cointegrated and there is bi-directional
information flow (except during 1994–99). At Ahmedabad futures and spot market are mostly cointegrated (except March
contract) and there is uni-directional information flow from futures to spot market. Across markets, there is cointegration
between Mumbai and Ahmedabad futures market (except during 1985–93). There is uni-directional flow of information
from Ahmedabad to Mumbai futures market for March contract and the opposite for December contract. The reasons for
PRICE DISCOVERY 41

the above result are that Ahmedabad is the production centre for castorseed (Section 2.3), March is the harvest contract
(Sections 2.2 and 2.5) and fall in futures trading volume at Mumbai post 1994 (Section 2.4).

4.9 Conclusion

The futures and spot market at Mumbai and Ahmedabad as well as the futures market at Mumbai and Ahmedabad are
mostly cointegrated. There is causality and information flow from futures to spot market at both Mumbai and Ahmedabad.
There is reverse information flow as well for some contracts and periods. Results are different for various markets,
contracts and time periods indicating the role played by geographical location, harvest and futures trading volume in
influencing price discovery.
Chapter 5

Spot Price Volatility

The effect of futures trading on price in the associated spot market remains an important but unresolved issue. One of
the recurring arguments made against futures markets is that they give rise to greater price volatility in the underlying
spot market [3, 4]. The debate about speculators and the impact of futures market on spot price variability suggests that
increased volatility is undesirable. However, this fails to recognise the link between information and volatility [4]. The
variance of price change is equal to the rate or variance of information flow. The implication is that volatility of the asset
price will increase as rate of information flow increases [131]. There is little theory regarding the impact of futures trading
on spot market volatility. It is not clear whether futures trading will affect volatility in the underlying spot market in a
positive or negative manner. It is an empirical question as to the impact of futures market on the associated spot market
[3].

The general approach adopted in the literature to resolve the above issue is to examine spot price volatility before and
after futures trading. This is typically done using a measure of volatility (such as variance of returns) as the dependent
variable and regressing it on a proxy variable (to control for macro-economic factors) and a dummy variable (to capture
the impact of futures). The proxy variable usually is another commodity with similar characteristics (turmeric, in our
case) for which there is futures trading during the study period [4]. However, this approach must reasonably account for
all other determinants of volatility that may have influenced results [107].

5.1 Literature review

In this section we provide a brief review of the empirical work regarding the influence of futures trading on spot price
volatility. The review highlights the main techniques used in empirical investigation and assesses the weight of evidence
on both sides of the debate. Past studies have examined the effect of futures trading on spot price volatility in diverse
markets. Investigation has been carried out for various physical commodities as well as financial instruments. The
empirical techniques employed range from simple regression, uni-variate Box-Jenkins analysis and Granger causality
tests to multi-variate time series analysis, Box-Tiao intervention analysis and GARCH models. The general approach is to
consider spot price variability before and after futures trading and test for significant change. Less common is the use of
cross-sectional analysis where spot volatility is compared for similar markets with and without futures trading. Harris [75]
uses both forms of analysis to test for change in stock index volatility since the onset of index futures trading. Inspite of a
great deal of research, the topic continues to be of interest mainly due to the inconclusive nature of empirical evidence.

Early research on the effect of futures trading in commodities has generally concluded that the existence of a futures
market tends to stabilise price in the spot market. Two papers on the onion market by Gray [69] and Working [151]

42
SPOT PRICE VOLATILITY 43

found that futures trading reduced the range between high and low spot price over a crop year. A study of live cattle
futures by Taylor [144] compared the variance of price between a period with and without futures trading and found
that the spot price was more stable when futures market was in existence. Another study of cattle and pork bellies by
Powers [127] using variate difference method to isolate the noise component of the price series produced evidence of the
stabilising influence of futures trading. Cox [35] investigated the effect of organised futures trading on information in the
spot market. Empirical evidence showed that futures trading increased traders’ information about underlying supply and
demand conditions. Cox concluded that the spot market operated more efficiently in the presence of futures trading.

In one of the earliest studies on financial futures, Froewiss [54] concluded that futures trading did not adversely affect
the variability of spot price for the US Government National Mortgage Association (GNMA) pass-through certificates.
Froewiss investigated the GNMA market using uni-variate Box-Jenkins analysis for weekly percent change in spot price.
Pre and post futures sub samples were modelled using Box-Jenkins technique and tests conducted for significant parameter
change between the two periods. Results indicated no significant parameter change, suggesting that spot price volatility
had not been influenced by futures trading. On the contrary, Figlewski [46] examined the effect of futures trading on
the GNMA spot market and concluded that futures trading in GNMA securities led to increased price volatility in the
GNMA spot market. The author contends that GNMA futures price is determined largely by new traders who have less
information than experienced GNMA spot market dealers. Figlewski suggests that spot price volatility increases when the
additional noise in futures price is transmitted to the GNMA spot market.

In an attempt to shed further light on this issue, Simpson and Ireland [140] investigated the effect of futures trading
on GNMA certificates. The study used first difference of both daily and weekly average price. As a preliminary search
into the effect of derivative trading the authors specified a regression model with a proxy variable to remove extraneous
influence, a dummy for the onset of futures trading and an interaction term. The regression model was constructed for both
pre futures and post futures sub samples for daily and weekly volatility measure. Tests were carried out for significant
change in model parameters for the two periods. As no significant change was recorded the authors concluded that there
was no increase in spot price volatility since the advent of derivative trading.

Continuing with the study of futures trading on spot market for GNMA securities, Corgel and Gay [34] investigated
the effect of GNMA derivative trading on spot price using Box-Tiao intervention analysis which is designed to model
the impact of events on time series data. The results confirm the general finding that derivative trading did not have a
destabilising effect on spot price; instead a long run stabilising effect was found. Moriarty and Tosini [115] replicated
Figlewski’s [46] study and found that GNMA futures trading did not cause any increase in GNMA spot market volatility.
They attribute Figlewski’s [46] results to his choice of study period and conclude that the study period considered may
alter results. Bhattacharya et al. [14] also replicated Figlewski’s [46] study using his measure to calculate weekly volatility
series for spot and futures price for GNMAs. Using Granger causality tests their results suggest no change in spot price
volatility since the introduction of futures trading in GNMA securities. In a study of Treasury bills, Simpson and Ireland
[141] found that futures trading led to a decrease in volatility initially, but the effect disappeared when futures trading
volume became large resulting in increased volatility in the spot market.

The effect of stock index futures on spot price volatility is well described by Edwards [41] who compares spot price
volatility before and after the introduction of stock index futures and finds that the long term volatility of the S&P 500
index was greater before futures trading. His analysis of the 1972–87 period did not indicate that futures trading was
associated with increased stock market volatility. He finds that volatility decreased post futures for the S&P 500 but
finds no significant difference for the Value Line Index. Edwards argues that causality tests cannot infer whether futures
trading has stabilised or destabilised the spot market. The perception of futures volatility leading spot volatility could be
explained by futures market reacting more quickly to information which will eventually reach the spot market where it
will have a similar effect on volatility. Using variance ratio tests of daily returns from 1973 to 1987, Edwards concludes
that the introduction of futures trading in interest rate and stock index has not destabilised the underlying spot market.

Harris [75] compares daily return volatility pre futures (1975–82) and post futures (1982–87) between the S&P 500
 44 SPOT PRICE VOLATILITY

stocks and a non S&P group of stocks. Controlling for difference in firm attributes (beta, price level, size and trading
frequency), Harris finds that in the post futures period individual stock returns in the S&P group were more volatile
than the non S&P stock returns. He contends that the marginal increase could be due to factors other than derivatives
trading because the increased volatility was more noticeable in daily returns than in returns measured over a longer
interval. Harris believes that other index related phenomena such as foreign ownership of American equity and growth
in index funds could account for the change. However, Harris’ approach of comparing S&P 500 stocks to other stocks
underestimates the effect of futures market since price of non S&P 500 stocks is influenced by movement in the S&P 500
index.

Brorsen, Oellermann and Farris [22] present both theoretical and empirical models which suggest that a successful
futures market for cattle improves spot market efficiency; however short run spot price risk also increases. The fact that
volatility has increased (though only in the short run) could be an important factor in understanding why cattle producers
perceive the futures market as adversely affecting spot price. Weaver and Banerjee [148] investigate whether futures
trading of cattle and other related commodities destabilises spot price. They argue that futures trading does not lead to
dynamic instability of spot price. Despite the finding of an apparent role for external information in determining cattle
price, their results support the conclusion that futures trading for cattle and other related commodities did not lead to
dynamic instability of cattle price during the study period.

Recent literature has taken a different approach by utilising the GARCH family of models wherein spot market returns
are modelled pre and post futures as an ARIMA process while simultaneously modelling the conditional variance as a
GARCH process. The parameters of the conditional variance equation are tested for structural change to make inferences
about the influence of futures trading on spot price volatility [3, 4, 5, 8, 107].

Brorsen [23] tests for homogeneity of variance in the S&P 500 index for time periods before and after futures trading
and finds that while the variance of daily price change increased significantly, the variance of five day and twenty day
price change remained the same. Brorsen suggests that short run volatility can be decreased by any measure that increases
friction such as raising futures margins or increasing transaction cost for arbitrageurs. Baldauf and Santoni [8] used
ARCH analysis to test for increased volatility in the S&P 500 index since the introduction of futures trading and growth
of programme trading. Spot price was modelled for periods before and after futures trading and the model specification
tested for significant change. The study found no evidence of a shift in the model parameters suggesting no effect of
derivative trading on volatility. Lee and Ohk [107] find that following the introduction of index futures, the volatility
of stock returns in Japan, UK and USA rose significantly but not in Australia and Hong Kong. Their results show that
although stock index futures exert a volatility increasing influence on daily stock price behaviour, it makes the stock
market relatively more efficient as volatility shocks are disseminated more quickly.

Antoniou and Foster [3] study the effect of futures trading on spot price volatility for Brent crude oil in UK using
GARCH analysis. They find no evidence to suggest that there has been a spillover of volatility from futures to spot
market. Bessembinder and Seguin [11, 12] examine whether greater futures trading activity (volume and open interest)
is associated with greater equity volatility. They partition each trading activity series into expected and unexpected com-
ponents. The authors find that while equity volatility is associated positively with unexpected futures trading volume, it
is negatively related to forecastable futures trading activity. Their results support the view that an active futures market
enhances liquidity and depth of the underlying equity market by way of arbitrage. Kamara, Miller and Siegel [92] inves-
tigate the effect of futures trading in S&P 500 on stability of the underlying index. Both parametric and non parametric
tests suggest that the volatility of daily returns in the post futures period (1982–89) was higher than in pre futures period
(1976–82), but the volatility of monthly returns remained unchanged. They find that the distribution of daily returns is
frequently (and non event induced) changing. Consequently, the authors conclude that the significant increase in volatility
of daily returns after the inception of futures trading is not futures induced.

Antoniou and Holmes [4] find that stock return volatility in the UK increased following the introduction of index
futures. They examine the impact of trading in FTSE 100 stock index futures on volatility in the underlying spot market.
SPOT PRICE VOLATILITY 45

To examine the relation between information and volatility they use the GARCH family of techniques. Their results
suggest that futures trading has led to increased volatility, but the nature of volatility has not changed post futures. Their
finding of price change being integrated pre futures but being stationary post futures, implies that the introduction of
futures trading has improved the speed and quality of information flowing into the spot market. Their results suggest
that there was an increase in spot price volatility on a daily basis, but this was due to increased information flow in the
market and not speculators having an adverse destabilising effect. The increased volatility is the result of futures trading
expanding the routes over which information can be conveyed to the spot market.

Darrat and Rahman [37] examine whether futures trading activity has contributed to jump volatility1 in the stock
market. Their results suggest that futures trading activity is not a cause for episodes of jump volatility. Holmes [80]
examines the impact of futures trading on price volatility in the underlying spot market for FTSE Eurotrack in which
trading was thin and consequently suspended. Evidence presented demonstrates that despite low futures trading volume,
the existence of futures market improved the rate at which information was assimilated into the spot price and reduced
persistence. The results suggest that futures trading brings beneficial effect to the associated spot market even if trading
is thin.

Pericli and Koutmos [124] examined the impact of S&P 500 index futures and options on volatility in the spot market.
They found that for weekly returns the unconditional variance decreased in the post futures period, though the reduction
was not significant. They found clear evidence that the introduction of index futures and index options produced no
structural change in either the conditional or unconditional variance. Antoniou, Holmes and Priestley [5] examined the
impact of futures trading on stock market volatility for six countries (Germany, Japan, Spain, Switzerland, UK and USA).
Their results suggest that although futures trading had a limited impact on the level of stock market volatility over a three
year period, it had a major effect on the dynamics of the stock market. They found that post futures, volatility asymmetry
was significantly lower for all countries except Spain.

Chang, Cheng and Pinegar [28] propose new tests to examine the influence of index futures on Nikkei stock price
volatility. The tests decompose spot portfolio volatility into the expected cross-sectional dispersion and the average
volatility of returns on the portfolio’s constituent securities. Their results indicate that futures trading increased spot
portfolio volatility but the volatility impact did not spill over to stocks against which futures were not traded. They find
that most of the variation in spot market volatility is attributable to disturbance in broad market factors. The increase in
volatility induced by futures trading can only be detected when tests properly control for these factors.

Gulen and Mayhew [73] examine stock market volatility before and after the introduction of equity index futures
trading in 25 countries. They find that, except for USA and Japan, introduction of futures market had a significantly
negative impact in 17 countries and no effect in the remaining 6 countries on the underlying spot market volatility. Shang–
Wu Yu [155] using a switching GARCH(1,1)–MA(1) model examined the impact of index futures on volatility in the spot
market. He found that volatility increased in USA, France, Japan and Australia but not in UK and Hong Kong. McKenzie,
Brailsford and Faff [112] examined individual share futures in Australia using a threshold ARCH model and found mixed
evidence concerning impact of futures on conditional volatility in the spot market. Their results demonstrate that different
institutional settings or different sample periods are not the reason for different findings. The authors suggest that trading
conditions associated with individual markets (such as liquidity) are the more likely reason for different findings. Their
results indicate that markets behave differently depending on the surrounding circumstances. They contend that futures
and volatility are unlikely to have a direct link. Rahman [129] examines the impact of trading in Dow Jones Industrial
Average (DJIA) index futures and options on the conditional volatility of component stocks. He finds that the introduction
of index futures and options on the DJIA produced no structural change in the conditional volatility of component stocks.

1 Jump volatility refers to the occasional extreme change in stock price.

 46 SPOT PRICE VOLATILITY

5.2 Spot price volatility

We study the impact of castorseed futures market on spot price volatility for castorseed in India. Futures markets for
castorseed exist in India since May 1985 at two geographically close locations, Mumbai and Ahmedabad. Four contracts
maturing in March, June, September and December are traded each year at both places. Mumbai is the commercial and
marketing centre whereas Ahmedabad is the production centre (Section 2.3). The futures trading volume at Ahmedabad is
higher as compared to Mumbai (Section 2.4). We study both markets which allows us to compare results across markets
and gives us additional insight into the influence of futures trading on the underlying spot market. Following the literature,
we use both the traditional regression technique as well as GARCH analysis.

In the traditional regression technique, we use two measures of volatility: standard deviation of daily spot returns
and the Figlewski measure [46]. Each measure of volatility is regressed on a dummy variable to account for change in
volatility due to introduction of futures and a proxy variable (turmeric) to account for macro-economic factors. Turmeric
is chosen as a proxy variable because for the entire study period there was futures trading in turmeric. Further, we model
the spot price returns for castorseed as a GARCH process to examine the structural change in volatility pre and post
futures. Following Simpson and Ireland [141], we divide our study period into three sub periods: (a) the pre futures sub
period (1982–85), (b) the futures early sub period (1985–88) and (c) the futures later sub period (1994–97). The futures
early sub period is the period immediately following the introduction of futures trading. The futures later sub period is
characterised by relatively low futures trading volume (Section 2.4). With two futures sub periods, we seek to reduce the
possibility of a bias in our choice of futures period. Also, since the two futures sub periods are a few years apart, we get
additional insight into the short term and long run influence of futures trading on spot price volatility.

5.3 Methodology

Futures trading for castorseed began in May 1985 at both Mumbai and Ahmedabad. We have data for the castorseed
spot market from September 1982 onwards. Hence, we choose the pre futures period from September 1982 to April
1985. The pre futures period is immediately prior to the introduction of futures trading. To mitigate the effect of other
extraneous factors, the futures early sub period is chosen to be immediately after the onset of futures trading. Hence, we
select the futures early sub period from September 1985 to April 1988. This allows us to analyse the short term effect
of futures trading on spot price volatility. We consider the futures later sub period from September 1994 to April 1997.
The futures later sub period is nine years after the onset of futures trading. This allows us to analyse the long term effect
of futures trading on spot price volatility. Also, the futures later sub period is characterised by relatively low futures
trading volume (Section 2.4). This allows us to examine the link between futures trading volume and spot price variability
[11, 12, 15, 41, 53]. Each period is 2 12 years and begins in September to avoid seasonal variation bias.

We have used spot price of turmeric as a proxy for macro-economic factors. The proxy variable should be such that
there exists futures trading for the proxy variable during the entire study period. Also, the proxy commodity should be
similar in nature to the one under study e.g. both commodities should be agricultural, with similar size of production.
Hence the choice of turmeric as a proxy, for which there was futures trading during the entire study period. The futures
market for turmeric is located at Sangli, a small town in Maharashtra. We study the influence of futures trading on spot
price for castorseed using both the traditional regression technique and the more recent GARCH analysis as discussed
below.
SPOT PRICE VOLATILITY 47

5.3.1 Traditional regression technique

We construct a measure of volatility and regress it on a proxy variable (to account for macro-economic factors) and
a dummy variable (to account for the impact of futures trading). To examine sensitivity of the test to the measure of
volatility used [3], we employ two measures of volatility viz. standard deviation of daily spot returns and the Figlewski
[46] measure s
1 n 2
Vm = ∑ rt
n t=1
(5.1)

where Vm is monthly volatility, rt is the daily spot return on day t and n is the number of observations in a month.
We construct a monthly series for the two measures of volatility. The series are constructed separately for castorseed
at Mumbai and Ahmedabad and turmeric at Sangli. Next, we run the following regression separately for Mumbai and
Ahmedabad and also separately for both measures. Volatility is modelled as an auto regressive AR(1) process with one
lag. The two dummies for the futures early sub period and futures later sub period are included along with a proxy variable
(turmeric) to account for macro-economic factors

Vt = a1 + a2 ∗ De + a3 ∗ Dl + b1 ∗Vt−1 + b2 ∗ PRXt + et (5.2)

Vt is the monthly volatility series constructed above (standard deviation of daily spot returns and Figlewski
measure),
De is a dummy variable representing futures early sub period,
Dl is a dummy variable representing futures later sub period,
Vt−1 is first lag of the volatility series constructed above and
PRXt is the proxy variable (turmeric)

The intercept term for pre futures period is a1 . The parameter a2 is the difference between a1 and the intercept in futures
early sub period. Similarly a3 is the difference between a1 and the intercept in futures later sub period. The intercept
parameters measure the level of volatility in the spot market. The intercept parameters a2 and a3 represent the change in
level of spot market volatility from pre futures period that can be attributed to the introduction of futures trading. This form
of the equation tests for change in intercept between early and later futures sub period. A significant positive (negative)
change in the intercept coefficient provides evidence of increase (decrease) in volatility relative to pre futures spot market
volatility. If the proxy variable is found to be insignificant it raises questions as to the reliability of inference made about
the impact of futures trading on spot market volatility since change could be due to other factors. An insignificant proxy
variable means we cannot account for other determinants of volatility which may also have changed (Section 5.5.1).

5.3.2 Generalised AutoRegressive Conditional Heteroscedasticity (GARCH)

Traditional regression analysis assumes homoscedasticity of residuals. Studies based on constructed volatility measures
implicitly assume that price change in spot market is serially uncorrelated and homoscedastic. However, findings of
heteroscedasticity in financial time series has been well documented in the literature [16, 106]. Inferences drawn from
studies failing to control for such dependence are therefore likely to be erroneous. While observed difference in volatility
may be due to the introduction of futures, it may also be the result of return dependence [4]. As shown by Ross [131] any
change in the rate of information flow will change price volatility of the spot asset. Therefore, unless information remains
constant (which is unlikely), volatility is likely to be time varying. A natural way to capture the time varying nature of
volatility is to model the conditional variance as a Generalised Autoregressive Conditional Heteroscedasticity (GARCH)
process2 . GARCH models account for change in the variance of a time series by modelling the conditional variance as a
time series [3, 4, 16, 106].
2 Robert Engle received the Nobel prize in Economics (2003) for this contribution.
 48 SPOT PRICE VOLATILITY

We account for heteroscedasticity by modelling castorseed spot returns as a GARCH model. We then test whether
the parameters of the model change significantly subsequent to the introduction of trading in futures [107, 112, 155].
We run the following GARCH model separately for Mumbai and Ahmedabad (Section 5.5.2) and examine for significant
change in GARCH parameters (Section 5.5.3). We use the GARCH(1,1) specification as it has been found to be an
adequate representation for most financial time series [16, 105, 106]. Following Lee and Ohk [107] we employ a switching
GARCH model that captures the structural change in the level and slope of time varying volatility using dummy variables
(to account for the onset of futures trading) in the conditional variance equation

rt = a1 + b1 ∗ rt−1 + b2 ∗ Mt + et et ∼ N(0, ht ) (5.3)

ht = δ0 + α0 ∗ et−1
2
+ β0 ∗ ht−1 + δ1 De + α1 ∗ et−1
2
De + β1 ∗ ht−1 De
+ δ2 Dl + α2 ∗ et−1
2
Dl + β2 ∗ ht−1 Dl (5.4)

rt is daily spot return for castorseed at Mumbai and Ahmedabad,

rt−1 is first lag of daily spot return,
Mt is daily spot return for the proxy variable (turmeric),
ht is conditional variance of et ,
De is a dummy variable representing futures early sub period and
Dl is a dummy variable representing futures later sub period

We estimate the model for castorseed spot market at Mumbai and Ahmedabad with and without the dummy variables
for the period September 1982 to April 1997. We test for significant change in GARCH parameters post futures (early
and later sub period) as compared to pre futures by employing the likelihood ratio test. If no structural change occurs
in the mean level and autoregressive structure of conditional volatility after the onset of trading in castorseed futures,
the coefficient of dummy variables for intercept and slope will not be significantly different from zero. The joint null
hypothesis of no change in both unconditional and conditional volatility is δ1 = α1 = β1 = 0 for the futures early sub
period, the alternate hypothesis is at least one of the coefficient is non zero. Similarly, the joint null hypothesis of no
change in both unconditional and conditional volatility is δ2 = α2 = β2 = 0 for the futures later sub period, the alternate
hypothesis again is at least one of the coefficient is non zero (Section 5.5.3).

The coefficient for lagged squared error term (α0 ) relates to the change in spot price on the previous day attributable
to market specific factors. The change in price is due to the arrival of information specific to the commodity under
consideration and is hence considered as news. An increase in α0 post futures means that information is assimilated into
the spot price more quickly (due to introduction of futures trading). Just as α0 reflects the impact of recent news, the
coefficient on lagged variance term (β0 ) measures the impact of price change related to days prior to the previous day. A
decrease in β0 post futures indicates that previous news has less impact on today’s price change [3, 4]. We interpret the
increase in GARCH parameters as an increase in informational efficiency of the spot market due to information content
of futures price. This increased efficiency results in a greater reaction to news as suggested by increase in the α0 news
coefficient. The decrease in lagged conditional variance term β0 is taken as evidence that volatility is considered less
important since the risk it implies can be hedged in the futures market (Section 5.5.2).

5.4 Data

From Section 2.7 we observe that descriptive statistics are similar for the castorseed spot market at Mumbai and Ahmed-
abad. The standard deviation of spot returns for castorseed at Mumbai and Ahmedabad has reduced in the post futures
later sub period, but not in the post futures early sub period. This may indicate a decrease in volatility in the post futures
SPOT PRICE VOLATILITY 49

later sub period. Performing an F test on the variance of castorseed spot returns at Mumbai and Ahmedabad and turmeric
spot returns at Sangli we reject the null hypothesis of equal variance in all cases except the pre futures and post futures
early sub period for Ahmedabad. The results for turmeric (which is a proxy for macro-economic factors) are exactly the
opposite. For the turmeric spot market, volatility has increased in both the futures sub periods. Thus, the volatility in the
castorseed spot market at Mumbai and Ahmedabad appears to have reduced in spite of a rise in general market volatility
(Section 2.7). But as the literature cautions, inferring from static values of variance may not be completely correct [3].

5.5 Results

The results of the traditional regression technique for both measures of volatility as well as GARCH for Mumbai and
Ahmedabad are discussed below.

5.5.1 Traditional regression technique

The results of the traditional regression technique for both measures of volatility for Mumbai and Ahmedabad are tabulated
below.

Table 5.1: Traditional regression technique results

Measure Standard deviation Figlewski

Coefficient Mumbai Ahmedabad Mumbai Ahmedabad
a1 0.0037 0.0040 0.0040 0.0043
(0.0008) (0.0008) (0.0009) (0.0009)
a2 0.0009 0.0004 0.0008 0.0004
(0.0007) (0.0007) (0.0007) (0.0007)
a3 -0.0011 -0.0017 -0.0012 -0.0018
(0.0005) (0.0004) (0.0005) (0.0005)
b1 0.5670 0.5163 0.5498 0.4996
(0.0583) (0.0686) (0.0606) (0.0687)
b2 0.0120 0.0701 0.0061 0.0633
(0.0382) (0.0404) (0.0418) (0.0437)
R̄2 0.39 0.38 0.37 0.35
Ljung Box 1.82 0.05 1.82 0.02
(Standard error in parenthesis)

From the table above we note that results are similar for Mumbai and Ahmedabad and also for both measures of
volatility. The Durbin Watson test statistic is close to two in all cases indicating an absence of first order autocorrelation.
As an additional check, the Ljung Box Q statistic for first order serial correlation at the 5% level of significance is well
below its critical value of 3.84. The proxy variable (turmeric) is not significant indicating that macro-economic factors
are unable to explain volatility in the castorseed spot market. An important observation is that the dummy variable for
the futures early sub period is insignificant whereas the dummy variable for the futures later sub period is significant
and negative. This is true for both measures of volatility at Mumbai as well as Ahmedabad. We infer that volatility has
reduced post futures only in the later sub period.
 50 SPOT PRICE VOLATILITY

5.5.2 Generalised AutoRegressive Conditional Heteroscedasticity (GARCH)

The results of the GARCH(1,1) model for castorseed spot returns for Mumbai and Ahmedabad are tabulated below.

Table 5.2: GARCH(1,1) model for castorseed spot returns

Coefficient Pre-futures Futures period

period Early Later

Mumbai
a1 0.0002 0.0002 0.0002
(0.0001) (0.0001) (0.0001)
b1 0.1695 0.1705 0.1704
(0.0156) (0.0153) (0.0143)
b2 0.0202 0.0221 0.0213
(0.0070) (0.0072) (0.0066)
δ0 0.0009 0.0007 0.0010
(0.0002) (0.0002) (0.0002)
α0 0.0699 0.0611 0.0618
(0.0055) (0.0076) (0.0054)
β0 0.9229 0.9320 0.9294
(0.0064) (0.0085) (0.0065)
δ1 0.0105 0.0102
(0.0044) (0.0021)
α1 0.0592 0.0583
(0.0277) (0.0201)
β1 -0.1323 -0.1294
(0.0520) (0.0526)
δ2 -0.0004
(0.0003)
α2 0.0081
(0.0086)
β2 -0.0105
(0.0095)
Loglikelihood 15163.4 15178.3 15180.7

Ahmedabad
a1 0.0002 0.0002 0.0002
(0.0001) (0.0001) (0.0001)
b1 0.1748 0.1765 0.1760
(0.0149) (0.0155) (0.0138)
b2 0.0151 0.0155 0.0148
(0.0083) (0.0084) (0.0081)
δ0 0.0024 0.0019 0.0027
(0.0004) (0.0003) (0.0005)
α0 0.0866 0.0720 0.0721
(0.0076) (0.0074) (0.0078)
β0 0.8936 0.9117 0.9066
SPOT PRICE VOLATILITY 51

Coefficient Pre-futures Futures period

period Early Later
(0.0093) (0.0087) (0.0097)
δ1 0.0052 0.0040
(0.0011) (0.0021)
α1 0.0846 0.0830
(0.0307) (0.0278)
β1 -0.1130 -0.1036
(0.0282) (0.0353)
δ2 -0.0016
(0.0034)
α2 -0.0049
(0.0087)
β2 0.0067
(0.0042)
Loglikelihood 14682.8 14689.9 14691.3
(Standard error in parenthesis)

From the table above we observe that the unconditional variance δ0 has increased in the futures early sub period
(significant and positive δ1 ) but remained unchanged in the futures later sub period (insignificant δ2 ). Similarly, we
observe that the news coefficient α0 has increased in the futures early sub period (significant and positive α1 ) but remained
unchanged in the futures later sub period (insignificant α2 ). We note that the lagged conditional variance term β0 has
decreased in the futures early sub period (significant and negative β1 ) but remained unchanged in the futures later sub
period (insignificant β2 ). We also note that the proxy variable for macro-economic factors (turmeric) is significant and
positive. The above results are true for both Mumbai and Ahmedabad.

5.5.3 Likelihood ratio

The likelihood ratios for the hypotheses to be tested are tabulated below. The likelihood ratio follows a χ2 distribution
with degrees of freedom equal to the number of restrictions. In our case, for three restrictions, the critical value at the 5%
level of significance is 7.81.

Table 5.3: Likelihood ratio test results

Null hypothesis Futures period

Early Later

Mumbai
δ1 = α1 = β1 = 0 29.8
δ2 = α2 = β2 = 0 4.8

Ahmedabad
δ1 = α1 = β1 = 0 14.2
δ2 = α2 = β2 = 0 2.8
(Critical value: 7.81)

From the table above we observe that compared to the pre futures period, the GARCH parameters have changed in the
 52 SPOT PRICE VOLATILITY

futures early sub period. We note that compared to the futures early sub period, the GARCH parameters have remained
the same during the futures later sub period.

With the introduction of futures trading there has been a structural change in volatility in the spot market for castorseed
in the futures early sub period. The change was beneficial in nature because α0 increased and β0 decreased. The increase
in α0 suggests that new information is assimilated into the spot price more quickly due to introduction of futures trading.
The decrease in β0 indicates that old information has less impact on today’s price change, because information is being
processed more rapidly. The significant and positive coefficient for the proxy variable (turmeric) indicates that macro-
economic factors are able to at least partly explain volatility in the castorseed spot market. The change in α0 or β0 for the
futures later sub period is not significant at both Mumbai and Ahmedabad. This suggests that the beneficial change in the
structure of volatility evident in the futures early sub period has remained stable over the futures later sub period.

5.5.4 Futures trading volume and spot price volatility

A possible explanation for the increase in α0 and decrease in β0 in the futures early sub period could be higher futures
trading volume (Section 2.4). Similarly, a possible explanation for the insignificant change in α0 and β0 in the futures
later sub period at Mumbai and Ahmedabad could be the rise in futures trading volume since 1994 at Ahmedabad (Section
2.4).

In this context, Bessembinder and Seguin [11, 12] find that for the S&P 500 index, the relation between expected
futures trading volume and spot price volatility is negative. Increased liquidity (higher trading volume) in futures market
allows spot traders to hedge their position and curb volatility attributable to order imbalance [41]. Well informed traders
trade in the spot market so as to guide spot price towards its mean level, in the process reducing volatility in the spot
market [53]. Board, Sandmann and Sutcliffe [15] find similarly that for the UK, futures trading activity (volume) does not
destabilise the spot market.

5.6 Conclusion

The introduction of castorseed futures market at Mumbai and Ahmedabad has had a beneficial effect on the castorseed
spot market in the futures early sub period. This effect has remained stable in the futures later sub period, a possible
reason being the rise in futures trading volume post 1994 at Ahmedabad.
Chapter 6

Conclusion

Commodity futures markets in India have a long history. The first commodity futures exchange was established in 1875
at Mumbai for cotton. A few more commodity exchanges were set up at the beginning of the previous century: oilseeds
(Mumbai, 1900), jute (Calcutta, 1912) and bullion (Mumbai, 1920). Commodity futures trading flourished in India till
late 1960’s when it was banned in order to have control over price. The ban was in place for two decades. Based on
the recommendation of the A. M. Khusro Committee (1980) it was lifted with the (re)introduction of futures trading
in gur (Muzaffarnagar and Hapur, 1982) and castorseed (Mumbai and Ahmedabad, 1985). In line with the economic
liberalisation of early 1990’s and based on recommendations of the K. N. Kabra Committee (1994), futures trading was
introduced for a few more commodities: coffee (Bangalore, 1998), cotton (Mumbai, 1999) and soya oil (Indore, 1999).
The ban on futures trading was lifted for all commodities in April 2003. Commodity futures markets in India are regulated
by the Forward Markets Commission (FMC) set up in 1953 under the Forward Contracts (Regulation) Act, 1952.

Historically, the Government has intervened at every stage of the marketing of major agricultural commodities. The
Government policy has been to protect and promote the agriculture sector through procurement and administered price
mechanism. However, in view of reduced direct support to agriculture under the Agreement on Agriculture with the World
Trade Organisation (WTO), there is a policy shift towards market oriented approach. Government intervention declined
after initiation of economic liberalisation and reforms in 1991. To help manage price risk, the Government has been
encouraging commodity derivatives. A key aspect of the process of strengthening agricultural markets is the question of
obtaining efficient derivative market for commodities. If derivative markets function adequately, some of the policy goals
regarding price volatility of agricultural commodities can be addressed in a market oriented manner.

In this thesis, we examine commodity futures markets in India with reference to the Government policy of encouraging
derivative markets (commodity, equity, interest rate and currency) to manage price risk. We study three aspects of futures
market viz. basis risk, price discovery and spot price volatility. Specifically, we examine the castorseed futures market at
Mumbai and Ahmedabad with respect to the above three aspects. We study the castorseed futures market since inception
(May 1985) up to August 1999. We study all four contracts (March, June, September and December) that are traded each
year. This will provide insight into the role of commodity futures markets as a means of managing price risk. The thesis
will also serve as an independent evaluation of the rationale underlying the shift in Government policy from intervention
to a market oriented approach. If the castorseed futures market performs its functions well, there is a strong case for
further development of derivative markets in India.

53
 54 CONCLUSION

6.1 Institutional features

There are a few institutional features of the castorseed market that are instrumental in explaining the results obtained in
this thesis. The most important factor is the harvest of castorseed from November to March. Of the four contracts traded
each year, the March contract is termed as the harvest or new contract since it coincides with the harvest season. It is
the only contract in which futures trading is driven mainly by supply side information. Hence we expect price discovery
results to be different for March contract. Specifically, we expect a bi-directional (futures to spot and vice-versa) flow of
information. We expect the spot market to influence the price discovery process significantly.

Another institutional feature of the castorseed market is that Gujarat is the dominant producer of castorseed in India.
Internationally, India is the largest producer of castorseed. Domestically, Gujarat is the largest producer of castorseed.
The average annual production of castorseed in Gujarat is more than all other states in India combined. We expect the
dominance of Gujarat in castorseed production to have three effects on the price discovery process. First, we expect the
Ahmedabad spot market to play an influential role in the price discovery process. Second, we expect the dominance of
spot market to be more pronounced for the March contract. Third, between futures markets, for the March contract we
expect the Ahmedabad futures market to be dominant in terms of price discovery.

A third important institutional feature of the castorseed market is the relatively higher futures trading volume at
Ahmedabad. Compared to Mumbai, futures trading volume is higher at Ahmedabad. Since 1994, futures trading volume
has risen at Ahmedabad while it has fallen at Mumbai. Due to the higher futures trading volume at Ahmedabad, we expect
basis risk to be lower at Ahmedabad as compared to Mumbai. Between futures markets, we expect the Ahmedabad futures
market to dominate Mumbai futures market in terms of price discovery. Also, we expect the Ahmedabad futures market
to have a more stabilising influence on the underlying spot market as compared to Mumbai.

The reason for the shift in futures trading volume from Mumbai to Ahmedabad is the opening of a sea port and setting
up of a special economic zone (SEZ) in 1994 at Kandla, Gujarat. Before 1994, castorseed was crushed into castor oil
at Mumbai for export to USA and Europe via Mumbai sea port. Gujarat did not have a viable sea port. However, after
opening of Kandla sea port and setting up of the SEZ, castorseed began to be crushed into castor oil in Gujarat and
exported via Kandla sea port by utilising export benefit of the SEZ. This led to a fall in export of castor oil from Mumbai
sea port with a corresponding rise at Kandla sea port. We expect the price discovery process as well as the effect of futures
trading on spot price volatility to be different for the two sub periods 1985–93 and 1994–99. Specifically, we expect the
futures and spot market at Ahmedabad to be more strongly cointegrated post 1994. We expect the futures market at
Mumbai and Ahmedabad also to be more strongly cointegrated post 1994. We expect the futures market at Ahmedabad
to have a more stabilising influence on the underlying spot market post 1994.

6.2 Basis risk

Basis risk which is specific to futures market and does not exist in spot market introduces an element of speculation
because hedgers are still exposed to this risk while hedging their physical commodity. The difficulty in forecasting the
basis (unpredictable basis) presents hedgers with a risk that cannot be hedged. Basis risk not only affects the futures
position but also the spot market position in all hedging situations. For agricultural commodities, forecasting the basis is
an important aspect of a successful marketing strategy. An unforeseen change in basis can adversely affect the net price
received, increase risk and alter producer behaviour. Forecasting the basis permits producers to assess alternative forward
pricing mechanisms such as futures hedging, spot forward contracts and basis contracts. The success of a futures contract
also hinges on a predictable basis. Increased basis risk relative to price risk can reduce the attractiveness of futures market
as a price risk management vehicle. If basis risk increases, the ability of futures to transfer risk is reduced.

The key to successful commodity production and marketing decisions is to understand the basis and factors which
CONCLUSION 55

affect its behaviour. Unanticipated basis movement reduces the ability of futures market to transfer risk from hedgers to
speculators and results in lower income to hedgers. Commercial firms and traders who buy in either the spot or futures
market and sell in the other market do so in anticipation of a specific change in the basis. Unexpected change in basis
creates additional risk for the hedger associated with this market position and makes it less desirable to hold. Empirical
assessment of the magnitude and volatility of the basis and analysis of the factors influencing basis risk may permit the
development of select management practices designed to minimise basis risk impact on market participants’ decisions.

We investigate the nature of basis risk for different contracts for the castorseed futures market at Mumbai and Ahmed-
abad since inception (May 1985) up to August 1999. We use root mean squared error (RMSE) to measure basis risk. The
test developed by Ashley, Granger and Schmalensee is used to compare RMSE. We also use the Henrikson-Merton (H-M)
timing test to gauge the qualitative accuracy of forecast.

The results show that the time series model performs better (lower RMSE and superior timing ability) than the bench-
mark model. Any effort made towards developing models to forecast the basis is likely to reap results. RMSE is low for
June contract and high for December contract. For the June contract all information regarding supply of castorseed is
known before trading begins. Forecasting the futures price is not difficult resulting in lower basis risk and hence RMSE
is low. Exactly opposite is the case for December contract and hence RMSE is high. The forecast for the Mumbai market
has higher RMSE but better timing ability as compared to Ahmedabad market. The link between liquidity and basis risk
is unclear. If the lower RMSE for Ahmedabad market is due to higher futures trading volume; the H-M timing test does
not indicate superior timing ability.

6.3 Price discovery

Price discovery is an important function performed by futures market. It is the revealing of information about future spot
market price through the futures market and refers to the use of futures price for pricing spot market transactions. The
essence of price discovery is to establish a competitive reference (futures) price from which the spot price can be derived
and hinges on whether information is reflected first in changed futures price or in changed spot price. The futures price
serves as the market’s expectation of subsequent spot price. The significance of price discovery depends upon a close
relationship between futures and spot price. The extent to which futures market performs this function can be measured
from the temporal relation between futures and spot price. If information is reflected first in futures price and subsequently
in spot price, futures price should lead spot price, indicating that the futures market performs the price discovery function.

The introduction of new information results in price difference for short intervals of time between futures and spot
market due to positive information and communication cost. Due to increased availability and lower cost of information,
a futures market assimilates information faster than a spot market. The price linkage between castorseed futures and spot
market is investigated using cointegration analysis which offers several advantages. First, cointegration analysis measures
the extent to which two markets have achieved a long run equilibrium. Another distinct advantage of the cointegration
technique is that it explicitly allows for divergence from equilibrium in the short run. Cointegration implies that each series
can be represented by an error correction model which includes last period’s equilibrium error as well as lagged values of
first difference of each variable. Hence, temporal causality can be assessed by examining the statistical significance and
relative magnitude of the error correction coefficient and coefficient on lagged variables.

Usually, price discovery is studied in a futures market and its corresponding spot market. We term this as price
discovery ‘within’ market. However, in India two futures markets exist for castorseed at Mumbai and Ahmedabad. This
gives us an opportunity to study price discovery ‘across’ markets by designating one of the futures market, say at Mumbai,
as the futures market and the other futures market, at Ahmedabad, as the spot market. Such a study would reveal whether
amongst futures markets for the same commodity, which are in geographical proximity, one futures market dominates the
other in terms of price discovery. Examining futures trading volume at Mumbai and Ahmedabad we find that there has
 56 CONCLUSION

been a fall from 1994 onwards. Hence, we also examine whether the price discovery role of futures market has changed
over time by studying it separately for the two sub periods 1985–93 and 1994–99.

The results show that the futures and spot market at Mumbai and Ahmedabad as well as the futures market at Mumbai
and Ahmedabad are mostly cointegrated. There is causality and information flow from futures to spot market at both
Mumbai and Ahmedabad. There is reverse information flow as well for some contracts and periods. Results are different
for various markets, contracts and time periods indicating the role played by geographical location, harvest and futures
trading volume in influencing price discovery.

6.4 Spot price volatility

There is widespread interest in the effect of futures trading on price in the underlying spot market. It has often been
claimed that futures trading destabilises the associated spot market by increasing spot price volatility. Others have argued
to the contrary stating that futures trading stabilises price and thus decreases spot price volatility. As the debate cannot
be resolved at a theoretical level, empirical investigation is necessary. Empirical research so far has not produced any
conclusive evidence regarding impact of derivative trading on spot market volatility.

Theoretically, it is not clear why derivatives should influence spot market volatility. Derivatives increase market
liquidity by bringing more investors to the spot market. This results in a less volatile spot market unless derivatives attract
mainly uninformed speculators who destabilise the market. The introduction of futures market improves risk sharing but
lowers the informativeness of price resulting in destabilisation and welfare reduction. The advent of new speculators into
the market increases liquidity but makes the spot price more noisy and reduces net social welfare if the new speculators
are less informed than traders existing in the market. Futures trading reduces the cost of entry of small traders into the
financial market. Futures trading increases spot price volatility if increased liquidity causes spot price to reflect new
information more quickly. In this case, rise in spot price volatility increases net social welfare.

Spot market volatility decreases due to the liquidity provided by speculators. This additional liquidity allows spot
traders to hedge their position and thus curb volatility attributable to order imbalance. Futures market provides a mecha-
nism for those who buy and sell the actual commodity to hedge themselves against unfavourable price movement. Through
the futures market, risk can be spread across a large number of investors and transferred away from those hedging spot
position to professional speculators who are more willing and able to bear it. The availability of risk transference afforded
by futures market reduces spot price volatility because it eliminates the need to incorporate risk premium in spot market
transaction to compensate for the risk of price fluctuation. Futures trading attracts more traders to the spot market making
it more liquid and therefore less volatile.

The debate about speculators and the impact of futures on spot price volatility suggests that increased volatility is
undesirable. This is misleading as it fails to recognise the link between information and volatility. Price depends on infor-
mation currently available in the market. Futures trading can alter the information available for two reasons. First, futures
market attracts additional traders to a market. Second, as transaction cost is lower in futures market, new information is
transmitted to the spot market more quickly. In an arbitrage free economy, volatility of price is directly related to the flow
of information. If futures market increases the flow of information, volatility in the underlying spot market will rise. The
variance of price change is equal to the rate or variance of information flow. The implication is that volatility of the asset
price will rise as the rate of information flow increases. It follows therefore that if futures market increases the flow of
information, volatility of the spot price must change.

The general approach adopted in the literature to resolve the above issue is to examine spot price volatility before and
after futures trading. This is typically done using a measure of volatility (such as variance of returns) as the dependent
variable and regressing it on a proxy variable (to control for macro-economic factors) and a dummy variable (to capture
the impact of futures). The proxy variable usually is another commodity with similar characteristics for which there is
CONCLUSION 57

futures trading during the study period. However, this approach must reasonably account for all other determinants of
volatility that may have influenced results.

We study the impact of castorseed futures market on spot price volatility for castorseed in India. We study the market
at both Mumbai and Ahmedabad which allows us to compare results across markets and gives us additional insight into
the influence of futures trading on the underlying spot market. We use both the traditional regression technique as well as
GARCH analysis.

The results show that the introduction of castorseed futures market at Mumbai and Ahmedabad has had a beneficial
effect on the castorseed spot market in the futures early period. This effect has remained stable in the futures later period,
a possible reason being the rise in futures trading volume post 1994 at Ahmedabad.

6.5 Policy recommendation

We conclude that the castorseed futures market at Mumbai and Ahmedabad performs the function of price discovery. Also,
the introduction of castorseed futures market has had a beneficial effect on castorseed spot price volatility. Thus, there is
a strong case for promoting derivative markets in India.
Bibliography

[1] Abhay H Abhyankar (June 1995) Return and volatility dynamics in the FTSE 100 stock index and stock index
futures markets, The Journal of Futures Markets 15(4) 457–488

[2] Abhay H Abhyankar (August 1998) Linear and non-linear Granger causality: evidence from the UK stock index
futures market, The Journal of Futures Markets 18(5) 519–40

[3] Antonios Antoniou and Andrew J Foster (June 1992) The effect of futures trading on spot price volatility: evidence
for Brent crude oil using GARCH, Journal of Business Finance and Accounting 19(4) 473–484

[4] Antonios Antoniou and Phil Holmes (April 1995) Futures trading, information and spot price volatility: evidence
for the FTSE 100 stock index futures contract using GARCH, Journal of Banking and Finance 19(1) 117–29

[5] Antonios Antoniou, Phil Holmes and Richard Priestley (April 1998) The effects of stock index futures trading on
stock index volatility: an analysis of the asymmetric response of volatility to news, The Journal of Futures Markets
18(2) 151–66

[6] R. Ashley, C. W. J. Granger and R. Schmalensee (July 1980) Advertising and aggregate consumption: an analysis
of causality, Econometrica 48(5) 1149–67

[7] Warren Bailey and K. C. Chan (June 1993) Macroeconomic influences and the variability of the commodity futures
basis, The Journal of Finance 48(2) 555–73

[8] Brad Baldauf and G. J. Santoni (April 1991) Stock price volatility: some evidence from an ARCH model, The
Journal of Futures Markets 11(2) 191–200

[9] Scott W. Barnhart and Andrew C. Szakmary (June 1991) Testing the unbiased forward rate hypothesis: evidence on
unit roots, co-integration and stochastic coefficients, Journal of Financial and Quantitative Analysis 26(2) 245–67

[10] William J. Baumol (August 1957) Speculation, profitability and stability, Review of Economics and Statistics 39(3)
263–71

[11] Hendrik Bessembinder and Paul J. Seguin (December 1992) Futures trading activity and stock price volatility, The
Journal of Finance 47(5) 2015–34

[12] Hendrik Bessembinder and Paul J. Seguin (March 1993) Price volatility, trading volume, and market depth: evi-
dence from futures markets, Journal of Financial and Quantitative Analysis 28(1) 21–39

[13] David A. Bessler and Ted Covey (August 1991) Cointegration: some results on U.S. cattle prices, The Journal of
Futures Markets 11(4) 461–74

[14] Anand K. Bhattacharya, Anju Ramjee and Balasubramani Ramjee (Spring 1986) The causal relationship between
futures price volatility and the cash price volatility of GNMA securities, The Journal of Futures Markets 6(1) 29–39

58
BIBLIOGRAPHY 59

[15] John Board, Gleb Sandmann and Charles Sutcliffe (September 2001) The effect of futures market volume on spot
market volatility, Journal of Business Finance & Accounting 28(7) 799–819

[16] Tim Bollerslev, Ray Y. Chou and Kenneth F. Kroner (April 1992) ARCH modelling in finance: a review of the
theory and empirical evidence, Journal of Econometrics 52(1)5–59

[17] G. Geoffrey Booth, Teppo Martikainen and Yiuman Tse (June 1997) Price and volatility spillovers in Scandinavian
stock markets, Journal of Banking and Finance 21(6) 811–23

[18] G. Geoffrey Booth, Raymond W. So and Yiuman Tse (September 1999) Price discovery in the German equity index
derivatives markets, The Journal of Futures Markets 19(6) 619–43

[19] William Breen, Lawrence R. Glosten and Ravi Jagannathan (December 1989) Economic significance of predictable
variations in stock index returns, The Journal of Finance 44(5) 1177–89

[20] Michael J. Brennan and Eduardo S. Schwartz (1990) Arbitrage in stock index futures, Journal of Busines 63 S7–31

[21] Robin J. Brenner and Kenneth F. Kroner (March 1995) Arbitrage, cointegration and testing the unbiasedness hy-
pothesis in financial markets, Journal of Financial and Quantitative Analysis 30(1) 23–42

[22] B. Wade Brorsen, Charles M. Oellermann and Paul L. Farris (August 1989) The live cattle futures market and daily
cash price movements, The Journal of Futures Markets 9(4) 273–82

[23] B. Wade Brorsen (April 1991) Futures trading, transaction costs and stock market volatility, The Journal of Futures
Markets 11(2) 153–63

[24] Mark G. Castelino (April 1992) Hedge effectiveness: basis risk and minimum variance hedging, The Journal of
Futures Markets 12(2) 187–201

[25] C. Chassard and M. Haliwell (1986) The NYMEX crude oil futures market: an analysis of its performance, Oxford
Institute for Energy Studies Working Paper No. M9

[26] K. Chan, K. C. Chan and G. A. Karolyi (1991) Intra–day volatility in the stock index and stock index futures
market, Review of Financial Studies 4(4) 657–84

[27] K. Chan (1992) A further analysis of the lead-lag relationship between the cash market and stock index futures
market, Review of Financial Studies 5(1) 123–52

[28] Eric C. Chang, Joseph W. Cheng and J. Michael Pinegar (May 1999) Does futures trading increase stock market
volatility? The case of the Nikkei stock index futures markets, Journal of Banking and Finance 23(5) 727–53

[29] Arjun Chatrath and Frank Song (April 1998) Information and volatility in futures and spot markets: the case of the
Japanese yen, The Journal of Futures Markets 18(2) 201–23

[30] Y. W. Cheung and N. Ng (1990) The dynamics of S&P 500 index and index futures intra–day price volatilities,
Review of Futures Markets 9(2) 458–86

[31] Abdur R. Chowdhury (October 1991) Futures market efficiency: evidence from cointegration tests, The Journal of
Futures Markets 11(5) 577–89

[32] Laurence S. Copeland (May 1991) Cointegration tests with daily exchange rate data, Oxford Bulletin of Economics
and Statistics 53(2) 185–98

[33] Dean Corbae and Sam Ouliaris (August 1988) Cointegration tests and purchasing power parity, The Review of
Economics and Statistics 70(3) 508–11
60 BIBLIOGRAPHY

[34] J. B. Corgel and G. D. Gay (Summer 1984) The impact of GNMA futures trading on cash market volatility,
AREUEA Journal 12 176–90

[35] Charles C. Cox (December 1976) Futures trading and market information, Journal of Political Economy 84(6)
1215–37

[36] William J. Crowder and Anas Hamed (December 1993) A cointegration test for oil futures market efficiency, The
Journal of Futures Markets 13(8) 933–41

[37] Ali F. Darrat and Shafiqur Rahman (August 1995) Has futures trading activity caused stock price volatility?, The
Journal of Futures Markets 15(5) 537–57

[38] David A. Dickey and Wayne A. Fuller (June 1979) Distribution of estimates for autoregressive time series with a
unit root, Journal of the American Statistical Association 74(366) 427–31

[39] David A. Dickey and Wayne A. Fuller (July 1981) Likelihood ratio statistics for autoregressive time series with a
unit root, Econometrica 49(4) 1057–72

[40] Francis X. Diebold and Roberto S. Mariano (July 1995) Comparing predictive accuracy, Journal of Business and
Economic Statistics 13(3) 253–63

[41] Franklin R. Edwards (August 1988) Futures trading and cash market volatility: stock index and interest rate futures,
The Journal of Futures Markets 8(4) 421–39

[42] Robert F. Engle and Clive W. J. Granger (March 1987) Cointegration and error correction: representation, estima-
tion and testing, Econometrica 55(2) 251–76

[43] Robert F. Engle and Byung Sam Yoo (Annals 1987) Forecasting and testing in cointegrated systems, Journal of
Econometrics 35 143–59

[44] Eugene F. Fama and Kenneth R. French (January 1987) Commodity futures prices: Some evidence on forecast
power, premiums and the theory of storage, Journal of Business 60(1) 55–73

[45] Ficci (October 2002) Background paper, International conference on commodity futures and derivatives trading,
Mumbai

[46] Stephen Figlewski (May 1981) Futures trading and volatility in the GNMA market, The Journal of Finance 36(2)
445–56

[47] Stephen Figlewski (July 1984) Hedging performance and basis risk in stock index futures, The Journal of Finance
39(3) 657–69

[48] Jeff Fleming, Barbara Ostdiek and Robert Whaley (June 1996) Trading costs and the relative rates of price discovery
in stock futures and options markets, The Journal of Futures Markets 16(4) 353–87

[49] T. Randall Fortenbery and Hector O. Zapata (December 1993) An examination of cointegration relations between
futures and local grain markets, The Journal of Futures Markets 13(8) 921–32

[50] T. Randall Fortenbery and Hector O. Zapata (May 1997) An evaluation of price linkages between futures and cash
markets for cheddar cheese, The Journal of Futures Markets 17(3) 279–301

[51] Andrew J. Foster (July 1996) Price discovery in oil markets: a time varying analysis of the 1990–91 Gulf conflict,
Energy Economics 18(3) 231–46

[52] Kenneth R. French (1986) Detecting spot price forecast in futures prices, Journal of Business 59 S39–S54
BIBLIOGRAPHY 61

[53] Milton Friedman (1953), Essays in Positive Economics, University of Chicago Press, Chicago

[54] Kenneth C. Froewiss (Spring 1978) GNMA futures: stabilising or destabilising? Federal Reserve Bank of San
Francisco Economic Review 20–29

[55] Kenneth D. Garbade and William L. Silber (May 1983) Price movements and price discovery in futures and cash
markets, The Review of Economics and Statistics 65(2) 289–97

[56] Philip Garcia, Raymond M. Leuthold and Mohamed E. Sarhan (November 1984) Basis risk: measurement and
analysis of basis fluctuations for selected livestock markets, American Journal of Agricultural Economics 66(4)
499–504

[57] Philip Garcia and Dwight R. Sanders (June 1996) Ex ante basis risk in the live hog futures contract: has hedgers’
risk increased?, The Journal of Futures Markets 16(4) 421–40

[58] Bruce L. Gardner (February 1976) Futures prices in supply analysis, American Journal of Agricultural Economics
58(1) 81–84

[59] Asim Ghosh (April 1993) Cointegration and error correction models: intertemporal causality between index and
futures prices, The Journal of Futures Markets 13(2) 193–98

[60] Lawrence R. Glosten, Ravi Jagannathan and David E. Runkle (December 1993) On the relation between the ex-
pected value and the volatility of the nominal excess returns on stocks, The Journal of Finance 48(5) 1779–1801

[61] Jesus Gonzalo and Clive W. J. Granger (January 1995) Estimation of common long-memory components in coin-
tegrated systems, Journal of Business & Economic Statistics 13(1) 27–35

[62] Barry K. Goodwin and Nicholas E. Piggott (May 2001) Spatial market integration in the presence of threshold
effects, American Journal of Agricultural Economics 83(2) 302–17

[63] Government of India (September 1994) Report of the Committee on Forward Markets (Kabra Committee), Ministry
of Civil Supplies, Consumer Affairs and Public Distribution, New Delhi

[64] Government of India (July 2000) National Agriculture Policy, Department of Agriculture and Cooperation, Min-
istry of Agriculture, New Delhi

[65] Government of India (September 2003) Draft report of the inter-ministerial task force on convergence of securities
and commodity derivative markets, Ministry of Consumer Affairs, Food and Public Distribution, New Delhi

[66] Clive W. J. Granger (Annals, 1981) Some properties of time series data and their use in econometric model speci-
fication, Journal of Econometrics 16 121–30

[67] Clive W. J. Granger (August 1986) Developments in the study of cointegrated economic variables, Oxford Bulletin
of Economics and Statistics 48(3) 213–28

[68] Clive W. J. Granger (September 1988) Some recent developments in the concept of causality, Journal of Econo-
metrics 39(1) 199–211

[69] Roger Gray (May 1963) Onions revisited, Journal of Farm Economics 45 273–76

[70] Sanford J. Grossman (1986) An analysis of the role of insider trading on futures markets, Journal of Business 59
S129–46

[71] Sanford J. Grossman and Merton H. Miller (July 1988) Liquidity and market structure, Journal of Finance 43(3)
617–37
62 BIBLIOGRAPHY

[72] A. Grunbichler, F. Longstaff and E. Schwartz (April 1994) Electronic screen trading and the transmission of infor-
mation: an empirical examination, Journal of Financial Intermediation 3(2) 166–87

[73] Huseyin Gulen and Stewart Mayhew (August 2000) Stock index futures trading and volatility in international equity
markets, The Journal of Futures Markets 20(7) 661–85

[74] Yasushi Hamao, Ronald W. Masulis and Victor Ng (1990) Correlations in price changes and volatility across
international stock markets, The Review of Financial Studies 3(2) 281–307

[75] Lawrence Harris (December 1989) S&P 500 cash stock price volatilities, The Journal of Finance 44(5) 1155–75

[76] Frederick H. deB. Harris, Thomas H. McInish, Gary L. Shoesmith and Robert A. Wood (December 1995) Cointe-
gration, error correction and price discovery on informationally linked security markets, Journal of Financial and
Quantitative Analysis 30(4) 563–79

[77] Joel Hasbrouck (September 1995) One security, many markets: determining the contributions to price discovery,
The Journal of Finance 50(4) 1175–99

[78] Roy D. Henrikson and Robert C. Merton (October 1981) On market timing and investment performance – II:
statistical procedure for evaluating forecasting skills, Journal of Business 54(4) 513–33

[79] Anthony F. Herbst, Joseph P. McCormack and Elizabeth N. West (August 1987) Investigation of a lead–lag rela-
tionship between spot indices and their futures contracts, The Journal of Futures Markets 7(4) 373–81

[80] Phil Holmes (January 1996) Spot price volatility, information and futures trading: evidence from a thinly traded
market, Applied Economics Letters 3(1) 63–66

[81] H. S. Houthakker (1992) Futures trading, in The New Palgrave Dictionary of Money and Finance (Vol. 2) ed. P.
Newman, M. Milgate and J. Eatwell, Macmillan, London 211–13

[82] Sharon Brown-Hruska and Gregory Kuserk (September 1995) Volatility, volume and the notion of balance in the
S&P 500 cash and futures markets, The Journal of Futures Markets 15(6) 677–89

[83] Darren Hudson and Keith Coble (September 1999) Harvest contract price volatility for cotton, The Journal of
Futures Markets 19(6) 717–33

[84] Yoshio Iihara, Kiyoshi Kato and Toshifumi Tokunaga (April 1996) Intra–day return dynamics between the cash
and the futures markets in Japan, The Journal of Futures Markets 16(2) 147–62

[85] M. Illueca and J. A. Lafuente (September 2003) The effect of spot and futures trading on stock index market
volatility: a non-parametric approach, The Journal of Futures Markets 23(9) 841–58

[86] Yasunori Ishii (September 1977) On the theory of the competitive firm under price uncertainty: note, American
Economic Review 67(4) 768–69

[87] Soren Johansen (June 1988) Statistical analysis of cointegration vectors, Journal of Economic Dynamics and Con-
trol 12(2) 231–54

[88] Soren Johansen and Katarina Juselius (May 1990) Maximum likelihood estimation and inference on cointegration
– with applications to the demand for money, Oxford Bulletin of Economics and Statistics 52(2) 169–210

[89] Soren Johansen (November 1991) Estimation and hypothesis testing of cointegration vectors in Gaussian vector
autoregressive models, Econometrica 59(6) 1551–80

[90] Nicholas Kaldor (Spring 1939) Speculation and economic stability, Review of Economic Studies 7 1–27
BIBLIOGRAPHY 63

[91] Nicholas Kaldor (1960) Essays on Economic Stability and Growth, Duckworth and Company

[92] Avraham Kamara, Thomas W. Miller Jr. and Andrew F. Siegel (December 1992) The effect of futures trading on
the stability of Standard and Poor 500 returns, The Journal of Futures Markets 12(6) 645–58

[93] Ira G. Kawaller, Paul D. Koch and Timothy W. Koch (December 1987) The temporal price relationship between
S&P 500 futures and the S&P 500 index, Journal of Finance 42(5) 1309–29

[94] Ira G. Kawaller, Paul D. Koch and Timothy W. Koch (August 1990) Intra–day relationships between volatility in
S&P 500 futures prices and volatility in the S&P 500 index, Journal of Banking and Finance 14(2) 373–97

[95] David Kenyon, Eluned Jones and Anya McGuirk (May 1993) Forecasting performance of corn and soyabean
harvest futures contracts, American Journal of Agricultural Economics 75(2) 399–407

[96] Nabil T. Khoury and Jean-Marc Martel (Winter 1985) Optimal futures hedging in the presence of asymmetric
information, The Journal of Futures Markets 5(4) 595–605

[97] Nabil T. Khoury and Jean-Marc Martel (December 1989) A supply of storage theory with asymmetric information,
The Journal of Futures Markets 9(6) 573–81

[98] Nabil T. Khoury and Pierre Yourougou (February 1991) The informational content of the basis: evidence from
Canadian barley oats and canola futures markets, The Journal of Futures Markets 11(1) 69–80

[99] R. A. Kolb and H. O. Stekler (April 1993) Are economic forecasts significantly better than naive predictions? an
appropriate test, International Journal of Forecasting 9(1) 117–20

[100] D. S. Kolamkar (2003) Regulation and policy issues for commodity derivatives in India, in Susan Thomas (ed.)
Derivatives Markets in India, Oxford University Press, India

[101] Stephen R. Koontz, Philip Garcia and Michael A. Hudson (April 1990) Dominant–satellite relationship between
live cattle cash and futures markets, The Journal of Futures Markets 10(2) 123–36

[102] Gregory Koutmos and G. Geoffrey Booth (December 1995) Asymmetric volatility transmission in international
stock markets, Journal of International Money and Finance 14(6) 747–62

[103] Gregory Koutmos and Michael Tucker (February 1996) Temporal relationships and dynamic interactions between
spot and futures stock markets, The Journal of Futures Markets 16(1) 55–69

[104] Kyriacos Kyriacou and Lucio Sarno (May 1999) The temporal relationship between derivatives trading and spot
market volatility in the U.K.:empirical analysis and monte carlo evidence, The Journal of Futures Markets 19(3)
245–70

[105] Christopher G. Lamoureux and William D. Lastrapes (March 1990) Heteroscedasticity in stock return data: versus
GARCH effects, The Journal of Finance 45(1) 221–29

[106] Christopher G. Lamoureux and William D. Lastrapes (April 1990) Persistence in variance structural change and
the GARCH model, Journal of Business and Economic Statistics 8(2) 225–34

[107] Sang Bin Lee and Ki Yool Ohk (October 1992) Stock index futures listing and structural change in time varying
volatility, The Journal of Futures Markets 12(5) 493–509

[108] Raymond M. Leuthold, Philip Garcia, Brian D. Adam and Wayne I. Park (February 1989) An examination of the
necessary and sufficient conditions for market efficiency: the case of hogs, Applied Economics 21(2) 193–204

[109] Shi Min Liu, B. Wade Brorsen, Charles M. Oellermann and Paul L. Farris (May 1994) Forecasting the nearby basis
of live cattle, The Journal of Futures Markets 14(3) 259–73
 64 BIBLIOGRAPHY

[110] Michael W. McCracken (December 2000) Robust out-of-sample inference, Journal of Econometrics 99(2) 195–223

[111] Christopher S. McIntosh and Jeffrey H. Dorfman (February 1992) Qualitative forecast evaluation: a comparison of
two performance measures, American Journal of Agricultural Economics 74(1) 209–14

[112] Michael D. McKenzie, Timothy J. Brailsford and Robert W. Faff (March 2001) New insights into the impact of the
introduction of futures trading on stock price volatility, The Journal of Futures Markets 21(3) 237–55

[113] Robert C. Merton (July 1981) On market timing and investment performance – I: an equilibrium theory of value
for market forecasts, Journal of Business 54(3) 363–406

[114] Jae H. Min and Mohammad Najand (April 1999) A further investigation of the lead-lag relationship between the
spot market and stock index futures: early evidence from Korea, The Journal of Futures Markets 19(2) 217–32

[115] Eugene J. Moriarty and Paula A. Tosini (Winter 1985) Futures trading and the price volatility of GNMA certificates
– further evidence, The Journal of Futures Markets 5(4) 633–41

[116] Gopal Naik and Sudhir Kumar Jain (July 2002) Indian agricultural commodity futures markets: a performance
survey Economic and Political Weekly 37(30) 3161–73

[117] Janet S. Netz (May 1996) An empirical test of the effect of basis risk on cash market positions, The Journal of
Futures Markets 16(3) 289–311

[118] Charles M. Oellermann and Paul L. Farris (Winter 1985) Futures or cash: which market leads live beef cattle
prices? The Journal of Futures Markets 5(4) 529–38

[119] Charles M. Oellermann, B. Wade Brorsen and Paul L. Farris (April 1989) Price discovery for feeder cattle, The
Journal of Futures Markets 9(2) 113–21

[120] Jacob Paroush and Avner Wolf (December 1989) Production and hedging decisions in the presence of basis risk,
The Journal of Futures Markets 9(6) 547–63

[121] Anne E. Peck (August 1975) Hedging and income stability: concepts, implications and an example, American
Journal of Agricultural Economics 57(3) 410–19

[122] Anne E. Peck (August 1976) Futures markets supply response and price stability, Quarterly Journal of Economics
90(3) 407–64

[123] Joost M. E. Pennings and Matthew T. G. Meulenberg (August 1997) Hedging efficiency: a futures exchange man-
agement approach, The Journal of Futures Markets 17(5) 599–615

[124] Andreas Pericli and Gregory Koutmos (December 1997) Index futures and options and stock market volatility, The
Journal of Futures Markets 17(8) 957–74

[125] Peter C. B. Phillips and Pierre Perron (June 1988) Testing for a unit root in time series regression, Biometrika 75
335–46

[126] Michael A. Pizzi, Andrew J. Economopoulos and Heather M. O’Neill (May 1998) An examination of the rela-
tionship between stock index cash and futures markets: a cointegration approach, The Journal of Futures Markets
18(3) 297–305

[127] Mark J. Powers (June 1970) Does futures trading reduce price fluctuations in the cash markets? The American
Economic Review 60(3) 460–64

[128] Jing Quan (April 1992) Two step testing procedure for price discovery role of futures prices, The Journal of Futures
Markets 12(2) 139–49
BIBLIOGRAPHY 65

[129] Shafiqur Rahman (July 2001) The introduction of derivatives on the Dow Jones industrial average and their impact
on the volatility of component stocks, The Journal of Futures Markets 21(7) 633–53

[130] Matthew Roope and Ralf Zurbruegg (March 2002) The intra-day price discovery process between the Singapore
exchange and Taiwan futures exchange, The Journal of Futures Markets 22(3) 219–40

[131] Stephen A. Ross (March 1989) Information and volatility: the no-arbitrage martingale approach to timing and
resolution irrelevancy, The Journal of Finance 44(1) 1–17

[132] K. G. Sahadevan (July 2002) Saving agricultural commodity exchanges: growth constraints and revival policy
options Economic and Political Weekly 37(30) 3153–60

[133] P. S. Schreiber and R. A. Schwartz (Summer 1986) Price discovery in securities markets, Journal of Portfolio
Management 12 43–48

[134] Ted C. Schroeder and Barry K. Goodwin (December 1991) Price discovery and cointegration for live hogs, The
Journal of Futures Markets 11(6) 685–96

[135] Thomas V. Schwarz and Francis E. Laatsch (December 1991) Dynamic efficiency and price leadership in stock
index cash and futures markets, The Journal of Futures Markets 11(6) 669–83

[136] Thomas V. Schwarz and Andrew C. Szakmary (Aril 1994) Price discovery in petroleum markets: arbitrage cointe-
gration and the time interval of analysis, The Journal of Futures Markets 14(2) 147–67

[137] David F. Seiders (May 1981) Discussion – futures trading and volatility in the GNMA markets, The Journal of
Finance 36(2) 484–86

[138] Gang Shyy, Vasumathi Vijayraghavan and Brian Scott-Quinn (June 1996) A further investigation of the lead-lag
relationship between the cash market and stock index futures market with the use of bid/ask quotes: the case of
France, The Journal of Futures Markets 16(4) 405–20

[139] Param Silvapulle and Imad A. Moosa (April 1999) The relationship between spot and futures prices: evidence from
the crude oil market, The Journal of Futures Markets 19(2) 175–93

[140] W. Gary Simpson and Timothy C. Ireland (Winter 1982) The effect of futures trading on the price volatility of
GNMA securities, The Journal of Futures Markets 2(4) 357–66

[141] W. Gary Simpson and Timothy C. Ireland (September 1985) The impact of financial futures on the cash market for
treasury bills, Journal of Financial and Quantitative Analysis 20(2) 371–79

[142] Jeremy C. Stein (December 1987) Informational externalities and welfare reducing speculation, Journal of Political
Economy 95(6) 1123–45

[143] Hans R. Stoll and Robert E. Whaley (December 1990) The dynamics of stock index and stock index futures returns,
Journal of Financial and Quantitative Analysis 25(4) 441–68

[144] G. S. Taylor and Raymond M. Leuthold (February 1974) The influence of futures trading on cash cattle price
variations, Food Research Institute Studies 13 29–35

[145] Yiuman Tse (December 1999) Price discovery and volatility spillovers in the DJIA index and futures markets, The
Journal of Futures Markets 19(8) 911–30

[146] W. Vollink and R. Raikes (1977) An analysis of delivery period basis determination for live cattle, Southern Journal
of Agricultural Economics 9 179–84
 66 BIBLIOGRAPHY

[147] Mahmoud Wahab and Malek Lashgari (October 1993) Price dynamics and error correction in stock index and stock
index futures markets: a cointegration approach, The Journal of Futures Markets 13(7) 711–42

[148] Robert D. Weaver and Aniruddha Banerjee (February 1990) Does futures trading destabilise cash prices? evidence
for U.S. live beef cattle, The Journal of Futures Markets 10(1) 41–60

[149] Kenneth D. West and Dongchul Cho (October 1995) The predictive ability of several models of exchange rate
volatility, Journal of Econometrics 69(2) 367–91

[150] James T. Witherspoon (August 1993) How price discovery by futures impacts the cash market, The Journal of
Futures Markets 13(5) 469–96

[151] Holbrook Working (February 1960) Price effects of futures trading, Food Research Institute Studies 1 3–31

[152] Holbrook Working (June 1962) New concepts concerning futures markets and prices, American Economic Review
52(3) 431–59

[153] Jian Yang, David A. Bessler and David J. Leatham (March 2001) Asset storability and price discovery in commodity
futures markets: a new look, The Journal of Futures Markets 21(3) 279–300

[154] Frida Youssef (October 2000) Integrated report on commodity exchanges and Forward Markets Commission, World
Bank project for the improvement of the commodities futures markets in India

[155] Shang–Wu Yu (March 2001) Index futures trading and spot price volatility, Applied Economics Letters 8(3) 183–86

[156] Hector O. Zapata and T. Randall Fortenbery (1996) Stochastic interest rates and price discovery in selected com-
modity markets, Review of Agricultural Economics 18 634–54
Index

ACF, see Auto correlation function Dickey D. A., 25, 47, 48

AR, see Auto regression Durbin J., 35, 50, 52, 54, 70
Ashley F test, 28, 29, 31, 36 Durbin Watson statistic, 35, 50, 52, 54, 70
Ashley R., 28, 29, 31, 36
Auto correlation ECM, see Error correction model
Durbin Watson statistic, 35, 50, 52, 54, 70 Economy, see Castorseed
Error correction model, 46, 51, 53, 55
Ljung Box statistic, 35, 50, 52, 54, 70
Auto correlation function, 30 FC(R)A, see Forward Contracts (Regulation) Act
Auto regression, 30, 34, 67 Figlewski measure, 65, 66
Figlewski S., 65, 66
Basis, 8
FMC, see Forward Markets Commission
behaviour, 20
Forward Contracts (Regulation) Act, 3
benchmark model, 30
Forward Markets Commission, 3
definition, 20
Fuller W. A., 25, 47, 48
descriptive statistics, 23
Futures
equation, 20
descriptive statistics, 23
graph, 22
evolution, 2
time series model, 30, 34
trading volume, 19, 74
Basis risk, 8, 27, 28
Ashley F test, 28, 29, 31, 36 GARCH, 67, 68, 71
H-M test, 32, 37 heteroscedasticity, 67
Henrikson Merton timing test, 32, 37 homoscedasticity, 67
RMSE, 31, 35 likelihood ratio, 68, 73
Root mean squared error, 31, 35 Granger C. W. J., 28, 29, 31, 36
Ashley F test, 28, 29, 31, 36
Box G., 35, 50, 52, 54, 70 H-M test, see Henrikson Merton timing test
Harvest, see Castorseed
Castorseed Henrikson Merton timing test, 32, 37
economy, 16 Henrikson R. D., 32, 37
harvest, 17 Heteroscedasticity, 67
production, 18 Homoscedasticity, 67
Chow G. C., 35
Chow test, 35 Johansen S., 48, 51, 53, 56
Cointegration, 44, 48 Johansen technique, 48, 51, 53, 56
inference, 56 Jump volatility, 64
Johansen technique, 48, 51, 53, 56 Kabra K. N., see Committee
Committee Kandla, 19
A. D. Shroff, 3 Khusro A. M., see Committee
A. M. Khusro, 3
K. N. Kabra, 3 Likelihood ratio, 68, 73

67
 68 INDEX

Ljung Box statistic, 35, 50, 52, 54, 70

Ljung G., 35, 50, 52, 54, 70

Merton R. C., 32, 37

PACF, see Partial auto correlation function

Partial auto correlation function, 30
Perron P., 25, 47, 48
Phillips P. C. B., 25, 47, 48
Price discovery, 9, 39
across markets, 44, 54
Ahmedabad, 52
cointegration, 44
Error correction model, 46, 51, 53, 55
Johansen technique, 48, 51, 53, 56
inference, 56
Mumbai, 50
over time, 44, 56
Production, see Castorseed

RMSE, see Root mean squared error

Root mean squared error, 31, 35
Ashley F test, 28, 29, 31, 36
definition, 31
equation, 31

Schmalensee R., 28, 29, 31, 36

SEZ, see Special Economic Zone
Shroff A. D., see Committee
Special Economic Zone, 19

Turmeric, 59, 65, 66, 69, 73

Unit root test, 48

Dickey Fuller, 25, 47, 48
Phillips Perron, 25, 47, 48

Volatility
destabilising, 12
futures trading volume, 74
GARCH, 67, 68, 71
likelihood ratio, 68, 73
information, 14
jump, 64
spot price, 12, 59, 65
stabilising, 14
traditional regression technique, 66
Figlewski measure, 65, 66, 70
standard deviation, 65, 66, 70

Watson G. S., 35, 50, 52, 54, 70