Journal of Business and Behavioral Sciences Volume 23, Number 1 issn 1946-8113 Spring 2011 inthis issue

There is a significant impact on the performance appraisal system adopted by various banks in India. The employees found that salary and promotion decisions were not based on appraisal ratings. The nationalized banks have time based promotions inspite of poor or excellent performance. This is quite de-motivating for the young employees, who join an organization with high aspirations. It was also pertinent from results that rating errors do have an impact on the appraisal system and these errors are also interrelated. Although biasness cannot be reduced to zero but can be minimized by training the rater and involving them in designing the appraisal system.

Armstrong, M., & Baron A. (1998). Performance appraisal and management,

Jaico Publishing House, Mumbai.

Beach, D. S. (1980). Personnel, Macmillan, NY. p- 290.

Bernardin , H.J., and Pence, E.C. (1980). Effects of rater training :

Creating new response sets and decreasing accuracy, Journal of

errors, Journal of Applied Psychology, 64, pp 410-421. De Nisi, A., Cafferty, T.P., & Megino, B. (1984). A cognitive view of the

performance appraisal process: A model and research propositions,

Organizational Behaviour and Human performance, Vol 3, Greenwich,

CT: JAI PRESS. Drucker, Peter F. (1994). Managing the non - profit organizations. Elsevier

Limited. Kavanagh, M.J. (1982). Evaluating performance, In K.M.Rowland and G.R.

Ferris (Eds) Personnel Management, Boston , MA:Allyn and Bacon, pp

187-226.

Landy, F.J., & Farr, J.C. (1980). Performance Ratings, Psychological Bulletin, vol. 87, pp 72-107.

164

Journal of Business and Behavioral Sciences

Murphy, K.R., & Cleveland, J.N. (1991). Performance appraisal : An organizational perspective, Boston, MA: Allyn and Bacon.

165

Journal of Business and Behavioral Sciences Vol 23, No 1; Spring 2011

Indian Institute of Technology, Madras

This paper examines whether gold, silver and crude oil futures market serve as price discovery vehicle for spot market transactions in developed and emerging commodity markets. The analysis shows that gold, silver and crude oil futures and spot prices are cointegrated and there is causality and information flow between futures and spot markets at MCX, COMEX and TOCOM for all the commodities. The Error Correction Model shows that, in general, the futures market for gold, silver and crude oil lead the spot market in price discovery. Variance decomposition analysis and Hasbrouck‘s information share show that gold and silver futures market contribute more towards price innovation in MCX and NYMEX. For crude oil, futures market contributes more towards price innovation in NYMEX while it is not so in MCX. NYMEX and MCX, are efficient in price transmission as well as information sharing with respect to futures trading while TOCOM plays a central role in spot market compared to futures trading for all the commodities. The predictive ability of futures prices is very significant for gold, silver and crude oil and the futures market serves as an information leader for NYMEX, MCX and TOCOM.

Futures price have been found to be efficient unbiased forecasts of future spot prices and support the semi-strong form of efficient market hypothesis (Leuthold 1972, Peck 1975, Giles and Goss 1981, Martin and Garcia 1981 & Fatimah and Zainal (1994)) while Leuthold and Hartman (1979), Stein (1981) and Bigman et al. (1983) have found that futures market has not at all times reflected information. Most of the empirical literature before the early 1980‘s, summarized in Kamara (1982), states that an exact functional relationship exists between cash and futures prices for storable commodities described by cost of carry model. Bidirectional causal relationship is found to exist between futures and spot for live beef, live cattle, Aluminium and crude oil markets (Oellerman and Farris, 1985; Figuerola-Feri and Gilbert, 2005; Silvapulle and Moosa, 1999). However, futures market is focused to serve as a focal point for information assimilation than the spot market for live beef cattle (Oellerman and Farris, 1985). Subsequently, studies have focused to examine the evidence of causality

Journal of Business and Behavioral Sciences

in the long run and short run. Futures and spot price series were found to be cointegrated, for corn, soybean, canola, barley, oats, wheat, cotton, pork bellies, hogs, live cattle and feeder cattle and carbon. (Quan 1992; Schwartz and Szakmary 1994; Covey and Bessler 1995; Karbuz and Jumah 1995; Forternberry and Zapata 1993; Brockman and Tse ; Yang et al. 2001, Milunovich and Joyeux 2007) while Baille and Myers 1991; Chowdhury 1991; Bessler and Covey 1991; Schroeder and Goodwin 1991; Forenbery and Zapata 1997 have found no evidence of cointegration between cash and futures markets with respect to cattle, live hogs, cheese, diammonium phosphate (DAP) and anhydrous ammonia (NH3). While testing the cointegration for different varieties of shrimp cash and futures market Maynard (2001) has found mixed results. Thus, researchers have found mixed evidence for cointegration for storable and non-storable commodities.

While examining the short run relationship, most of the studies have found that information tends to be discovered in futures market and then transferred to spot market. Futures price play a dominant role in price discovery and transmission of price information with respect to beef, steer, canola, barley, oats, wheat, corn, soybean, hogs, live cattle, feeder cattle, sugar, crude oil, cheese, gasoline, heating oil, gasoil, DAP and NH3 (Oellerman et al 1989; Hunson and Wayne 1983; Brockman and Tse 1995; Forterbery and Zapata 1996, 1997; Yang et al. 2001; Zapata et al. 2005; Serletis and Banack 1990; Hanmoudu et al. 2003; Frank 2002). However, Oellerman et al. (1989), Quan (1992) and Moosa (1996) have found evidence of spot prices influencing futures prices for crude oil and live stock. Oellerman et al. (1989) have found that lagged cash prices had little effect on futures prices. However, Maynard et al. (2001) have found that shrimp futures prices failed as a price discovery mechanism. Thus, there is mixed evidence on price discovery process in commodity markets and there are very few studies that have examined the price discovery in gold and silver markets. Though Chan and Mountain (1988) explored the pricing relationship and tested the causality between gold and silver prices, no attempt has been made to examine the futures price discovery mechanism in gold and silver market independently. There is lack of evidence of transmission of price information in emerging commodity markets vis-à-vis developed markets and there are very limited studies that have focused on determining a markets contribution to price discovery in the form of information share in commodity markets.

Hence, this study differs from previous studies in the following ways: Firstly, this study examines the price discovery mechanism for gold, silver and crude oil which are the top three commodities being traded globally. Secondly, the information share of futures market on spot and vice-versa using variance decomposition analysis and Hasbrouck‘s (1995) information share of market are examined. Thirdly, since commodity derivatives have been introduced in many emerging economies recently, it is pertinent to know the efficiency of the

167

Thenmozhi and Priya

In India, organized commodity derivatives trading was introduced in November 2003 through the setting up of Multi Commodity Exchange (MCX). This was followed by setting up of National Commodities and Derivatives Exchange (NCDEX) in December 2003. MCX accounts for 70% of the market share of the total commodity futures volume in India. Moreover, MCX is one among the world‘s top ten commodity derivatives exchange and ranks among the top three in Bullion, energy and copper bourses globally in terms of contracts traded. Hence, the price discovery mechanism has been examined for the commodities traded in MCX and the results are compared with the most popular and developed economies, namely U.S and Japan. We find that there is long run relationship between futures and spot market for gold, silver and crude oil and the futures market leads the spot market in price discovery in all the exchanges. Gold futures and spot markets in COMEX are more efficient in price discovery process compared to other two exchanges. Silver futures and spot markets in MCX are more efficient compared to the other two exchanges. Light sweet crude oil futures and spot markets in NYMEX are efficient in price discovery process while Brent Crude oil trading in MCX and TOCOM has only a unidirectional relationship from futures to spot and it shows that crude oil has a strong long run relationship than in the short run in these two markets.

The price discovery mechanism has been examined for the top three commodities namely, gold, silver and crude oil for three exchanges, namely, Multi Commodity Exchange, India (MCX), New York Mercantile Exchange, U.S. (NYMEX) and Tokyo Commodity Exchange, Japan (TOCOM). These exchanges are topmost commodity exchanges according to the statistics on Gold, Silver and Crude Oil futures volume trading. World‘s leading bullion exchanges are NYMEX, TOCOM and MCX in terms of volume. New York is the home to the largest gold futures trading through its commodity exchange "COMEX" which is a division of NYMEX. TOCOM, Tokyo Commodities Market in Japan, is the second largest futures trading centre for gold. MCX is now third largest bullion exchange after New York Metal Exchange and Tokyo Commodity Exchange. The total average daily bullion trading in the Indian commodity exchange is between Rs 5,000 (Rs.50 billion) and Rs 5,500 (Rs.55 billion) crore. MCX has 70% market share of the total commodity futures trading volume in the country and being among the world‘s top ten commodity derivatives exchanges, ranks among the top three bullion, energy and copper bourses globally in terms of contracts traded. The average daily turnover of MCX is

168

Journal of Business and Behavioral Sciences

Thenmozhi and Priya

Given the time-series nature of the data, stationarity of the series is tested using the Augmented Dickey-Fuller (ADF) (1979, 1981) test and Phillips-Perron (PP) test. The long-run relationship between futures and spot prices has been examined using Johansen‘s cointegration and the causality is examined using Granger Causality test. The short-run relationship between futures and spot prices has been examined using the Error Correction Model (ECM). In order to examine the relative impact that one variable has upon another variable within the ECM system, the Variance Decomposition Analysis is carried out. Many closely-linked commodities trade in different markets and the prices of commodities share a common implicit efficient price and hence, the price discovery function in all these spot and futures market is examined using the Hasbrouck information share. Hasbrouck (1995) defines price discovery as innovations in the efficient price. Because the sources of variation in this efficient price can be attributed to different markets, a market's contribution to price discovery is defined as its information share, which is a market's proportion of the efficient price innovation variance. Hasbrouck (1995) uses cointegration system to describe the common implicit efficient price. If Π is a diagonal covariance matrix of e_k, the variance of Yekis ^X¥U^X¥' and we can decompose

the ^X¥U^X¥' into proportions contributed by each series. Hasbrouck defines the information share of market i(ISi) as the proportion of the contribution of market i to the total innovation variance, i.e., the prices of the market with higher information share impound more new information and this market plays the role

of price discovery, namely: ISi =

The ADF and PP test indicate that the series at level are non-stationary but the returns series are stationary at 1% level of significance. Thus, it is found that gold futures and spot prices are integrated of order one in MCX, COMEX and TOCOM. Similarly, the silver price and returns series and crude oil price and returns series for the three exchanges also exhibit similar characteristics of gold series, and both silver and crude oil returns are also stationary at first level.

The Granger causality test shows that, we cannot reject the hypothesis that spot price does not Granger cause futures MCX gold price but we do reject the hypothesis that futures Gold does not Granger cause spot price. Therefore, it appears that Granger causality runs one-way from futures MCX gold price to spot gold price and not the other way. This shows that the Indian futures market is more efficient in gold trading. Though the findings are similar to Oellerman et al. (1989), it is contradictory to Schroeder and Goodwin (1991) may be because they have used two different time periods in their model. But for COMEX gold

170

and TOCOM gold, the result shows that spot does Granger causes futures while futures does not Granger cause spot, which shows that these markets are not actively trading gold futures compared to MCX gold futures trading. With reference to MCX silver futures and spot trading, bidirectional relationship exists and thus both futures and spot markets are more efficient. However, the results for COMEX and TOCOM silver also show that spot does Granger cause futures while futures does not Granger cause spot which means that there exists an unidirectional causal relationship between them. For Crude oil, all the exchanges are more efficient and active as the Granger Causality test shows that there exists a bidirectional relationship between futures and spot for all the exchanges.

Johansen (1988) procedure shows that there exists one cointegration equation between futures and spot gold, silver and crude oil prices for all the exchanges. This result shows that there exists long run equilibrium between futures and spot prices which means that the shocks that increase or decrease the ―normal‖ distance between the variables tend to be corrected over time. Thus the short-term changes in the commodity futures reflect not only its own past changes and those of the spot price, but also current adjustment to correct for past deviations (errors) from the stable long-term relationship. The ECM shows that MCX gold futures price is influencing spot but the reverse is not true; and this result is similar to Granger Causality test results. The long run causality is expressed by the EC term and the correction factor is significant at 5% level. The speed of adjustment coefficient AZtfor spot is significant at 5% level and

this shows that about 24% of disequilibrium corrected everyday by changes in spot price but the speed of adjustment coefficient for futures is insignificant at 5% level and this shows that only 8% of disequlibrium corrected everyday by changes in futures price. The error correction estimates show that spot price is caused by its own first lag and also influenced by the futures price upto 2 lags. On the other hand, MCX gold futures price is influenced by none of its own lags or the lags of spot price. This shows the existence of unidirectional short run relationship between futures and spot prices and futures price strongly predict spot prices. The results of TOCOM gold are similar to MCX gold and gold futures influences spot more than spot price influencing futures prices. TOCOM gold spot price is caused by its own 2 lags and is also influenced by the first lag of futures close price. However, the speed of adjustment coefficient for spot is 31.5% while for futures it is 87.3% which is significant and greater than that in MCX.

In contrast, the COMEX gold futures and spot markets are efficient and there is a bidirectional relationship between them and all the lags exhibit statistical significance. Thus, the speed of adjustment coefficient for spot is 60.3% while for futures it is 157% which is significant and greater than MCX and TOCOM. Price discovery in gold trading is more efficient in COMEX followed by MCX and TOCOM since causality, information transmission and predictability between spot and gold futures in COMEX is more significant than

171

the MCX and TOCOM. In sum, all these three exchanges are efficient in Gold futures trading and futures leads the spot in price discovery and the findings are similar with Working (1948), Wiese (1978) and Lake (1981).

Unlike Gold, Silver futures trading in India is more efficient than spot trading and all the three lags of futures are statistically significant. The speed of adjustment coefficient is significant at 5% level and there is a long run and short run relationship between futures and spot markets. This result is similar to Granger causality test results. Moreover, the speed of adjustment coefficient for spot and futures show that about 26% and 24% of disequilibrium is corrected everyday by changes in spot and futures price respectively and are significant at 5% level. However, COMEX and TOCOM silver spot trading is more efficient than futures trading as lagged prices influence futures price more than futures prices influence spot. The error correction estimates of COMEX shows that spot price is caused by its own 2^nd lag and also it is influenced by the futures 2^nd lag. Silver trading in TOCOM results show that spot price is caused by its own first lag and 3rd lag of futures price. Since the speed of adjustment coefficients are not significant at 5% level for COMEX futures and spot markets and TOCOM spot market. Hence, there is bidirectional causality between silver spot and futures markets but price discovery in silver trading is more efficient in MCX followed by COMEX and TOCOM.

Brent Crude oil in MCX and TOCOM futures price serve as a predictor for spot since spot price is influenced by first lag of futures price. The influence of spot price on futures is not significant for both MCX and TOCOM. Light sweet crude oil in NYMEX shows bidirectional relationship between futures and spot price and the causal relationship could be identified by the significant second lag of futures and spot prices respectively. The error correction term is significant and the speed of adjustment is 55.02% for spot and 15.95% for futures market. Hence, Price discovery in Crude oil trading is efficient in NYMEX followed by TOCOM and MCX.

The variance decomposition analysis shows that for MCX gold, much of the variance in futures is from futures market and not from spot market whereas in the case of variance in spot market, major contribution is from futures as well as spot. For COMEX gold, much of variance in futures is from futures market but for spot market major contribution is from futures. For TOCOM gold, variance in futures is from futures market and variance in spot is from spot market. In sum, all these exchanges are rely more on futures market rather than spot in some extent. We can identify similar results in Silver trading in futures and spot in all these exchanges. But Brent Crude Oil in MCX behaves in a different manner, as most of the variance in futures is from futures and spot market and much of the variance in spot is from spot market. In NYMEX Light sweet Crude Oil, variance in futures is from futures market and variance in spot is also from futures market and to some extent from spot market. In contrast, for TOCOM Brent Crude oil,

172

variance in futures is from futures and spot market and variance in spot is also from both futures and spot market. But the variance explained by futures on spot is much higher than variance explained by spot on futures.

In case of MCX gold futures, the average information share of futures market is 94.66% (i.e.(98.63528% + 90.68155%)/2), while that of spot is only 5.34%. In case of COMEX Gold futures, the average information share of futures market is 83.80%, while that of spot is only 16.19%. In the case of TOCOM gold futures, the average information share of spot market is 85.48%, while that of futures is only 15.520435%. Similarly, for MCX Silver futures contributes 90.54% whereas spot contributes 9.46% and for COMEX Silver futures contributes 93.10% and spot contributes 6.90%. For TOCOM silver, futures contributes 8.01% whereas, spot information share is 91.99%. For MCX brent crude oil, futures average information share is 19.52% and spot market contributes 80.48% then NYMEX light sweet crude oil information share of futures is 98.39% and spot contribution is 1.61% whereas TOCOM brent crude oil futures contributes 65.79% and spot contributes 34.21%. This structure of information share means reveals that most of the price changes of spot and futures are because of futures innovation, and more information flows from futures to spot. The information share and variance decomposition confirms the dominant role of commodity futures in price discovery.