Meta-Analysis of Efficiency of Indian Spot and Commodity Futures Markets

Table 1 – Researches on Price discovery and Volatility spillover futures market Information Efficiency reduces spot price volatility on spot market before and after its introduction Fut

Meta-Analysis of Efficiency of Indian Spot and Commodity Futures

Markets

Pankaj Kumar Gupta

Jamia Millia Islamia University, India

Abstract

Examination of market efficiency has conventionally being an area of interest to the market participants, analysts, investors and regulators We find a large number of researches in commodity markets on a global basis covering various aspects of efficiency like macro-economic impact of trading of spot and futures, discovery mechanism of prices, volatility spillover effects In many developing countries including India, the issue of market efficiency has been mind boggling for researchers since research outcomes are confusing making it extremely difficult to derive reliable implications for the market participants and regulators The researchers have used a variety of techniques in statistics and econometrics These include e.g descriptives, F-ratios and various parametric and non-parametric tests The econometric estimations include examination of Casualty, Error correction, Integration, Auto regressions, GARCH models etc

It is seen that generalization of results is a complex and difficult task since the results derived show varying behavior and complexity In Indian markets, our examination of researches reveals that inferences derived therefrom on the issue of efficiency and interrelationships connote a contrasting view It seems difficult to decide whether the operation of commodity derivatives affects the volatility and efficiency of spot markets and also the trading response In addition, the macro-impact of commodity derivative trading is an unresolved issue

We analyze the methodologies used in these researches in Indian and other countries to find out the major causes of contrast in the inferences on selected parameters like selection of exchange or asset (commodity), time frames, statistical or econometric technique used We also analyze the policy responses of the regulators and attempt to suggest a workable solution to solve these problems We derive that Indian commodity markets are inefficient for majority of commodity sets and the policy response is weak, sub-optimal and haphazard

Keywords: Market Efficiency, Price Discovery, Volatility Spillover, Error Correction, Integration

JEL Classification: Q02, G13, E31, G12

1 Introduction

Forecast of Futures prices of traded assets essentially involves the notion of market efficiency In commodity markets, using the time series data, many researchers have examined varying perspectives of market efficiency Fama (1970) has established that growth of commodity futures market is dependent upon the level of market efficiency Efficiency examination essentially involves the analysis of the co-integrating relationships, causality effects and ability of the markets to conform to the forecasts of the selected models Three principle forms of market efficiency are classified by Fama (1991) - weak, semi strong form and strong form The question of efficiency is still unsolved in the large number of studies throughout the world For

example Gupta and Mayer (1981) tested semi-strong form by comparing futures market and ARIMA models forecasts and found them to be outperforming Similar result has been obtained by Rausser and Carter (1983) with a caution that “unless the forecast information from the models is sufficient to provide profitable trades, their superior forecasting performance in a statistical sense has no economic significance” Elam and Dixon (1988) reject the validity of results using conventional F tests in a non-stationary price series Maberly (1985) reveal that “the inference that the market is inefficient for the more distant futures contracts is a direct result

of the bias inherent in using OLS to estimate the parameters in models with censored data” Engle and Granger (1987) provide a co-integration based method for efficiency test give the non-stationery nature of time series Allen and Som (1987) research on London rubber market support the hypothesis of weak form efficiency Various researches have been carried out conducted globally to examine commodity futures market efficiency of particularly futures like Canarella and Pollard (1987), Baura (1987), Dickinson and Muragu (1994),

Chan et al (1997), Min et al (1999) for Korea and Taiwan market which establish the weak form of efficiency

Similarly, Groenewold and Kang (1993) establish that Australian market is efficient in semi strong form Yalawar (1988) have used correlation and run test to test the stock market efficiency and found to be efficient Brorsen, Oellermann and Farris (1989) argue that futures trading introduction improves the efficiency of spot market efficiency though increasing the price risk Kaminsky and Kumar (1990) find excess return to be positive for some commodities time horizon being greater than three months Chowdhury (1991) suggest the use of co-integration over conventional methods of testing efficiency of futures markets Studies conducted by

Aulton, Ennew and Rayner (1997) and Kellard et al (1999) respectively show the partial integration and

short-run efficiency of markets respectively But, Andrew and Mathew (2002) using GQARCH-in-mean processes find that markets are unbiased in long run Phukubje and Moholwa (2010) on South African Futures commodity markets and Paschali (2007) on Bulgarian agricultural Commodity Markets during transition show that weak policy interventions results in poor integration between local and international markets with

a functional lacuna

Researches on Indian commodity futures markets show contrasting results Thiripalraju et al (1997) shows

that commodity markets in India are efficient on various parameters like price discovery and inter market feedback Singh (2005) argue that opinions on market efficiency are mixed and vary across commodities Sahi and Raizada (2006) conclude poor price discovery and volume trading impact of futures fueling inflation (in spot market) Bose (2008) find two way price discovery function of spot and futures commodity market, reducing the volatility thus helping the hedging function Contrary results have been indicated by Eswarana and Ramasundaram (2008) Singh, (2009) have realized the “genuine shortcomings accruing due to lack of detailed investigation of seasonality, overlapping data and unspaced observations in examining the efficiency and spill overs”

Goyari and Jena (2010) find the Indian commodity markets to be inefficient in weak form and sufficient opportunities exits to make profits through a structured trade Singh (2010) have shown the persistence of a long-run equilibrium relationship between futures and spot price series of selected commodities

Kaur and Rao (2010), Chakrabarty and Sarkar (2010) find support for market integration that is suitable for hedging However, Dinkar and Nagpal (2011), Ali and Gupta (2011) find the market to be partially efficient Agnihotri and Sharma (2012) argue that quality of market need not to be compromised with separation of contracts It is, therefore, inferred that the issue of co-integration and market efficiency is still unresolved

We therefore find interesting to explore the reasons for these varying results of these researches and possible explanation of the methodological implications For this purpose, we split the issue of market efficiency into three components – (a) Price discovery and Volatility Spill over, (b) Interrelationship between futures and spot markets and (c) Macro economic impact of futures commodity markets We use the published researches on commodity markets with a focus on Indian markets1

1 Research Papers/Articles in various table indicated in bold represent researches on Indian commodity markets

2 Price Discovery and Volatility Spillover

An important function of commodity futures market is discovery of price that aids the producers to plan their various operational activities starting from production to distribution of commodities Analysis of volatility spillover coupled with price discovery function of market continuously attracts the interest of analysts and investors In spite of similar availability of information to both spot and futures markets the reactions are not identical causing a spillover effect Various researches in India and abroad have been conducted on the examining the issues of price discovery and volatility spillovers Table 1 shows a gist of these researches focusing the research technique used, issues examined and the implications

Table 1 – Researches on Price discovery and Volatility spillover

futures market

Information Efficiency reduces spot price volatility

on spot market before and after its introduction

Futures trading introduction either reduced or did not increase volatility of spot prices

Garbade and Silber

(1983)

Granger Casualty, Co-integration

Impact of Futures market

on spot market

Flow of information from spot to futures market is reverse and important role of liquidity and market size in price discovery

Oellermann et al

(1985)

Granger causality Lead lag relationship

between spot and futures prices

Futures market discover price

model based on prices interactions of informed speculators and hedgers

Opening Futures trading improves risk sharing and reduces price volatility

Oellermann and

Farris (1989) Garbade Silber framework Impact of Futures trading on spot markets Futures market discovers price Bessembinder and

Seguin (1993)

market volatility

Volatility of spot prices has positive relation with unexpected volume and negative relation with the expected open interest Yang and Leatham

(1999)

futures markets

In a search for equilibrium price, futures market implicitly gather more information than available

in the cash markets alone

variable model Impact of futures trading on seasonal price

fluctuations

Futures trading reduced price volatility

countries Strength of the volatility spillovers increased after major

events

discovery function of futures markets

Cash and Futures market co-integration not affected by storage

of traded assets, yet creating a bias in price estimates

Thomas and

Karande (2001) Price dynamics model suggested by Garbade

and Silber (1983)

Effectiveness of Price discovery function of futures markets

Reaction of markets to information altogether varies in price discovery process

Assoe (2001) Co-integration Linkage between various

markets

Small and negative mean spillover effects

(1983) model

Effectiveness of Price discovery function of futures markets

Futures role in Price Discovery is not substantial

management Significant Co-integration relationship (a strong long run

relation Raju and Karande

(2003) Granger causality, Co-integration, GARCH Volatility spill over and Price discovery function of

futures markets

Futures markets impact cash market and brings volatility

stabilization of spot prices, inventory and production decisions

“Unnecessary hoarding increase the carrying cost leading to lower responsiveness of inventory to future prices”

Co-integration

Information efficiency and Price discovery function of futures markets in India

Futures market unable to fully incorporate information Bhar and Hamori

(2005)

relationship between volatility and trading volume

Higher order lagged returns affect trading volume

Chopra and Blesser

(2005) Co-integration, error correction models Price discovery function in cash markets Cash contracts and first distant futures contracts behavior shows

inefficiency of market

markets

Futures trading ineffective in reducing seasonal price volatilities

Yang et al (2005) Generalized forecast

error variance decomposition Granger Causality

Lead-lag relationships between futures trading and spot prices

Unidirectional impact on spot prices, “weak causal feedback between open interest and cash price volatility”

Slade and Thille

(2006) Descriptive measures Levels and volatilities of the spot prices Relationship between trading volume and price instability is

positive

markets

No change in volatility Qing-fu and

Jin-qing(2006)

Johansen Co-integration, VECM, Bivariate EGARCH

Price discovery function of futures

Significant bi-directional information flows Praveen and

Sudhakar (2006)

Commodity futures market and stock market and

Unidirectional impact

Bryant et al (2006) Co-integration

framework

Generalized theory of normal backwardation

Rejected the hypothesis Kiran and

Mukhopadhyay

(2007)

GARCH on intra-day Volatility spillover from the

US to India

Established, Basic ARMA-GARCH specification outperforms MGARCH Piyamas and

Pavabutr (2007)

futures contracts

Partial performance Nath and

Lingareddy (2008)

Hodrick-Prescott filter (1997)

Trading impact of futures markets (India)

Futures do not increase cash price volatility

Biswas (2009) Cointegration and

Error correction dynamics

Price discovery function of futures contracts

Co-integrated and long-run relationship established

Forecast error variance decompositions, Impulse response

Linkages between futures trading and volatility of spot market

Lagged unexpected volatility causes spot price volatility for all commodities

futures contracts

Efficient in price discovery Maitra and

Narayanan (2010)

Linear regression forecasting

Forecast performance of volatility

Found some evidence

response

Futures and Spot price convergence

Causality is bi-directional and spot market for selected commodity flexible

BEKK-GARCH and CCC-GARCH

Volatility and hedging behavior of selected commodity futures indices

Models effective in spot price discovery

Joseph and Sajan

(2010)

Mean, Deviations, Correlations

Futures trading effectiveness

Markets offers profitable investment opportunities Dash and Andrews

(2010)

futures prices

Price discovery mechanism is quite effective in general Chakravarty and

Ghosh (2010)

Bivariate EGARCH, Parkinson(1980) range based estimator

Volatility using intra day interval data

Evidence of intra day periodicity, seasonality, and the net inverse relation between time to maturity and realized volatility

Debasish and

Kushankur (2011) GARCH, EGARCH CGARCH, MGARCH,

Diagonal VECH, BEKK

Volatility and its spill over effects Bi-directional spill over captured under GARCH and unidirectional

spill over found under EGARCH

Co-integration Price discovery function Futures market dominates price discovery

model Price discovery mechanism in commodities and assess

the long-term trends in their prices

Model robust to capture trends

Chauhan, Singh and

Arora (2013)

discovery

Information flow from futures market to spot market

Masood and

Chary(2016)

Variance Ratios, Non-parametric tests

Growth of Commodity Futures Market

Volume and Value relationship is linear

We derive from the literature is that futures and spot markets absorb the same set of information in an identical period leading to discovery of price and spillover effects However, the conflict is the point of reaction i.e flowing from spot to futures or futures to spot We have further investigated these researches, particularity

on Indian commodity markets Our examination indicates that the research inferences differ mainly because

of the data used for analysis The data available on commodity exchanges – MCX, NCDEX, MCX, ICX differ

significantly with respect to the time frequency, price-volume, and high low statistics On some exchanges the data is not continuous which raises questions on the applicability of an appropriate times series method We

also raise questions on use of econometric techniques A vast number of studies have used the GARCH model

during the periods when the market conditions were not relatively stable, even though the model captures

time varying variances We find instances of fat tails in the price distribution of some commodities that limit the

application of GARCH Use of Johansen Co-integration test is appropriate compared to Engle-Granger (EG)

method since EG captures only a single set of co-integrating relations However, Johansen Co-integration test requires a stationery filter, which many are violated for a significant number of commodities in India There

is an ongoing debate on the use of VAR vs VECM VECM is able to explain the price discovery built in

short-run dynamics framework The inherent requirements of a particular model when applied to a given price series of

commodity in different time periods may produce differing results Therefore, it becomes difficult to generalize the

results of researches We also find that the commodities as per their nature be classified as – agro type (immediate consumption/intermediate production), asset type (gold/silver), industrial type (metals) However, there exists a relationship for the same commodity classified in two sets For example, silver is for industrial as well as household consumption Business cycles, festivals and hedging needs of economic participants would altogether affect the price series Consequentially, the results of price discovery and spillovers would be different in different circumstances These may possible explain the confusion between the researchers

3 Spot and Future Markets Integration

We further explore the relationship between futures and spot markets used to test the notion of efficiency under a distinct framework The selected list of researches on these aspects is given in Table 2

Table 2 – Researches on Spot and Future Markets Integration

relationships

Long-run relationships found, cash markets lead futures markets

GARCH

market

Futures market found to be efficient

Bessler and Covey

(1991)

relationships for slaughter

Found evidence

lag regression

Cash-futures price relationships

Trading intensity affected lead lag relation-no convincing evidence Fortenbery and

Zapata (1993) Co-integration Interest rate in describing price discovery Interest rate as an explanatory variable useful in explaining

relationship Baharumshah and

Habibullah (1994)

among prices in six Malaysian markets

Markets in Malaysia were highly integrated

Karbuz and Jumah

(1995)

between spot and futures

Prices of commodities tend to move together in the long run Herrero and John

(1997)

Co-integration, Granger Causality

Stability of long-run co-movement between commodity prices of world and UK retail

No casualty found

Fortenberry and

Zapata (1997) Co-integration Lead-lag relationship between spot and futures in

US

Partial evidence of integration

Fung and Patterson

volatility and open interest

Open interest and volume are not endogenously determined

Silvapulle and

Moosa (2000)

Non-linear causality test Bivariate VAR

Price volume relationship

in futures market

Linear causality from volume to price

Maynard et al (2001) Short-run analysis

based on Two-sided distributed lag model

Performance of Minneapolis Grain Exchange futures contracts with thin volumes

Co-integration found in only one variety(commodity class)

McMillan and

Speight (2002)

Sequential Information arrival Hypothesis and Distribution

Hypothesis , VAR

Relationship between volume and absolute return

From volumes to absolute returns,

no apparent causality, but some cases of causality flowing from return to volume seen

Asche and

Guttormsen (2002)

between spot and futures

Futures lead cash spot prices and distant futures lead near futures

market depth, trading volumes and absolute returns

Lack of efficient and modern infrastructure facilities, existence

of the gray market and lack of participation in the futures markets

modeling, OLS, two stage least squares regression

Lead-lag relationship between futures and spot prices of stock indices

Futures react faster than the spot market in disseminating

information

causality, VAR Relationship between Tokyo commodity futures

price changes and volumes

Volume contains no information

to forecast return

Chen and Firth

(2005)

Chinese commodity futures trading volume and returns

Existence of contemporaneous correlation between trading and volumes and absolute returns ruled out

Chinese commodity futures volume, daily volatility and returns

Relationship of volatility with volume is positive and with open interest is negative

Yang et al (2005) Generalized forecast

error variance decompositions Granger causality

Relationship between commodity futures volume and open interest

Weak causal feedback between volatility of cash prices and open interest

world cash prices for exported sugar and New York sugar futures

Causality in one direction from future price to cash but not vice versa

of spot and future prices Long-run equilibrium mechanism exists

Mattos and Garcia

(2004)

Brazilian agricultural market future and spot prices

Long-term equilibrium relationships between spot and future prices exists due to higher volume of trading

Casualty

Lead lag relationship between spot and futures trading

Imperfections in Indian commodity markets

Granger causality, VECM

Causality and speed of adjustment to deviations

in long run equilibrium

No single country is completely exogenous and many countries GSP prices are interlinked to some extent

Hadsell (2006) Threshold ARCH Relationship between

trading volume and price volatility

Traders react asymmetrically to new information

Babula et al (2006) Co-integrated VAR

based Johansen and Juselius methodology

Relationship between price volatility and trading volume and

Policy relevant elasticities important in US markets Azizan, Ahmad and

Shannon (2007)

Bivariate ARMA-EGARCH model

Futures-cash market relationship (Malaysia)

“Information transmission process at mean and volatility level between cash and futures”

Casualty

Cash-futures price relationships

Evidence of co-integration and operational efficiency, though at a slower rate

Iyer and Mehta

Roy and Kumar

(2007) Johansen Co- integration,

Garbade-Silber (1983) model

Lead-lag relationship between futures and spot prices and hedging effectiveness

Partial evidence

Kumar, Singh and

Pandey (2008) Co-integration, Casualty Volatility, Risk premium and seasonality Risk return relationship, Seasonality in risk and return

established

Nath and

Lingareddy (2008)

Granger causality, Correlations, Basic regression with dummies, GARCH

Futures-Spot Integration Co-integration observed between

China and U.S.A cotton futures

Casualty Market integration with global Spot markets in India are co- integrated with global

Long and Lei (2008) Co-integration,

Granger causality, Residue analysis, IRF, VECM

Commodity prices relationships at exchanges

Co-integrated

Bekiros and Diks

(2008)

GARCH BEKK model Causal linkages between

spot and futures of Linear and nonlinear type

Evidence found

on market integration Partial evidence

Raveendran et al

(2010) VECM Transmission of future price to spot price Unidirectional flow from futures to spot Bhardwaj and

Vasisht (2009) Co-integration, Granger Causality the Integration of futures and spot market Price discovery in each contract is unidirectional

Elumalai et al (2009) Johansen Cointegration

Granger causality Long-term equilibrium relationship between spot

and futures

Relationship found

competing domestic commodity futures markets

Partial evidence

Causality Unrelated regression

Role of commodity futures with thin trading in prices determination in physical market

Little evidence that futures price serves as reference price

Iyer and Pillai (2010) TVAR Futures markets role in the

price discovery process

Convergence rate of information

is slow specifically observed in non-expiration weeks

Vasisht and

Bhardwaj (2010) Johansen Cointegration tests,

Granger causality

Volatility of agricultural prices Co-integration evidence found but unidirectional causality from

futures to spot markets

relationship post migration

of Malaysian crude palm oil futures

Persistence of Volatility higher marginally in automated trading system compared to relative to the open outcry system Batten, Ciner, and

Lucey (2010)

for precious metals

“Precious metals are too distinct

to be considered a single asset class, or represented by a single index”

Saravanan and

Malabika (2010) VECM, GARCH model Role of futures on underlying spot market Partial evidence

Xaviour and Mathew

(2010)

Palanichamy et al

(2010) Cointegration , Granger Causality

tests

Long-run equilibrium

Vasisht and

Bhardwaj (2010) Granger causality Bivariate relationship between spot and future

market

Evidence found

Hernandez and

Torero (2010) Linear and nonlinear (nonparametric)

causality tests

Relationship dynamics between spot and futures prices of agricultural commodities

Futures markets generally discover spot prices

Jayagurunathan et

Error Correction Models

Co-integrations, price discovery process “Spot markets marginally leads futures and spot prices tend to

discover new information more rapidly than future prices” Chakravarty and

Ghosh (2010) Standard GARCH model Effect of commodity futures trading on spot

prices

Positive effect of future price movements

Mukerjee (2011) Multiple regressions,

VAR, Granger Causality, GARCH

Impact of futures trading

on spot markets Significant discovery of price and risk management

capabilities

M-GARCH with error correction model

Performance of various hedge ratios “Dynamic M-GARCH model hedging strategy performs best

in reducing the conditional variance of the hedged portfolio” Kumar and Pandey

(2011a)

Johansen’s Cointegration Granger causality Variance decomposition GARCH (BEKK)

Cross market linkages- Nine commodities futures

in India with those outside India

World markets have a significant unidirectional impact on markets

in India

Kumar and Pandey

(2011b) GARCH, Granger Causality, variance

decomposition, IRF

Relationship between trading volume, volatility and open interest

Overnight volatility impact the trading volume, open interest not affected

Srinivasan (2012) Johansen

Cointegration, VECM, Bivariate EGARCH

Price discovery process and volatility spillovers Evidence of Flow of information from spot to futures

Haq and Rao(2014) Johansen

Co-integration and VECM

Commodity market efficiency

Only Long-run Efficiency observed

Gupta, Chaudhary

and Agarwal(2017) VECM, V-MGARCH Hedging Effectiveness Distant Futures have better hedge ratio

Going by the notion of efficiency, when the phenomenon of arbitrage and information flow distortion occurs, the market is said to be inefficient High volumes guide futures market of the commodities when large players can use muscle power to drive the Indian market by taking simultaneous position in futures and spot markets and playing an arbitrage The flow of information from futures to spot is obvious and is an auto

process, yet we find instances of backwardation and contango violated over fairly large time periods We also blame

the large Foreign Institutional Investors (FII) acting as fly by night operators both in the commodity and securities markets We generalize by saying that the Indian commodity market is inefficient in semi strong form but sometimes for selected data groups typically show efficiency in short runs

4 Macro-Economic Impacts of Commodity Markets

Researchers have extensively used a variety of integration and vector type models to investigate the impact

of commodity trading on various macro economic measures particularly inflation In an Indian context, the most common measure of inflation is the change in the whole price index (WPI) Expectation hypothesis explains that prices in futures markets are unbiased expectations of series of future spot prices Therefore, demand-supply effect on prices of commodities in futures markets is obvious The series of futures prices convey trading information to the market players Given the inefficiency of information dissemination, the price arbitrage can have a devastating impact An important concern to the policy makers is the impact of the market behavior on the macro economic variables India, in the last decade has witnessed serious demand-supply gaps as well as price distortions in markets, which have raised important concerns like rising prices Inflation has become more important after the global financial crisis that made the markets more volatile Table

3 shows the summary of the researches mainly carried out on Indian commodity markets

Table 3 - Researches on Macro-economic Impacts of Commodity Markets

and the real return to forward speculation

Conditional covariance consistent with the predictions of the consumption-beta model Becker and Finnerty

(2000)

inflation hedge

Portfolio comprising of long commodity futures contracts improves risk and return performance when combined with bonds and stocks Jensen, Mercer and

Johnson (2002)

Co-integration Benefits of diversification

derived from inclusion of commodity futures to a traditional portfolio

Commodity future sustainability enhances portfolio performance

commodities in Asian neighbors on Indian markets

Partial Evidence

Gorton and Horst (2006) Co-integration Correlation between traded

commodity futures and other assets classes

Commodity futures correlate positively to inflation and its changes