Investor demand, market efficiency and spot-futures relation: Further evidence from crude palm oil

Resources Policy 53 (2017) 135–146

Contents lists available at ScienceDirect

Resources Policy journal homepage: www.elsevier.com/locate/resourpol

Investor demand, market efficiency and spot-futures relation: Further evidence from crude palm oil You-How Goa, Wee-Yeap Laub, a b

MARK

⁎

Faculty of Business and Finance, Universiti Tunku Abdul Rahman (UTAR), Kampar 31900, Perak, Malaysia Faculty of Economics and Administration, University of Malaya, Kuala Lumpur 50603, Malaysia

A R T I C L E I N F O

A BS T RAC T

JEL classification: G13 G14 Q10

This study examines the hypothesis of Tilton et al. (2011) that assert investor demand affects commodity prices when spot and futures prices are closely correlated during strong contango in a hard commodity like copper. However, using daily data of crude palm oil (CPO) spot and futures prices from January 2000 to July 2016, after taking into account the variance of the increments in a random walk as measured by variance ratio, our study finds that spot and futures prices are highly correlated during backwardation period. It is further observed that: First, investor demand on the futures market is highly correlated with spot and futures prices during backwardation, but lesser during weak contango, and the least correlated during strong contango. Second, the efficiency of the futures market is related to the degree of correlation between spot and futures price changes. High efficient information transmission in the futures market is linked to a high correlation between spot and futures markets and vice versa. Therefore, we extend the hypothesis that the preference of holding a long position in the futures market is due to the anticipation of insufficient supply of CPO which happens during the backwardation period.

Keywords: Contango and backwardation Weak-form market efficiency Investor demand Spot-futures relation Crude palm oil

1. Introduction Explaining the relationship between commodity spot and futures prices has been a long-standing agenda in financial economics. Such price relationship either price level in the long run or price changes in the short run is frequently determined by investor demand, in part because not all market participants are involved in producing or consuming a commodity, but also due to their expectation to make a profit by holding physical stock of a commodity from subsequent hike in the price. For those who have such expectation, they tend to intervene in the futures markets by selling futures contracts with higher prices to those who wish to acquire stocks or inventories. As a consequence, producers or stock owners are required to pay a high premium in the form of a difference between spot and futures prices at maturity of the contract. To protect income, producers make decisions by pushing a commodity price until the futures price is sufficiently higher than the spot price. To obtain riskless profit, rational arbitrageurs who recognize this inefficient market tend to buy a commodity in the spot market and sell it in the futures market simultaneously to cover net carrying costs. Their participation in the futures markets

⁎

Corresponding author. E-mail addresses: [email protected] (Y.-H. Go), [email protected] (W.-Y. Lau).

http://dx.doi.org/10.1016/j.resourpol.2017.06.009 Received 11 May 2016; Received in revised form 29 April 2017; Accepted 12 June 2017 Available online 23 June 2017 0301-4207/ © 2017 Elsevier Ltd. All rights reserved.

as a counterparty of hedging strategy theoretically enhances market liquidity and improves the prediction of future spot prices based on futures prices (Sanders et al., 2010; Sanders and Irwin, 2010, 2011a, 2011b). In contrast, a change of spot price due to information flow from the futures market gives opportunities for investors to implement non-standard transactions through over-the-counter markets. Therefore, it is noted that the futures market can facilitate the entry of speculators, thereby affecting the spot price of a commodity. According to Loayza et al. (2007), speculation will result in price variation which is disproportionate to the underlying changes in supply and demand that reduce investment and economic growth. Such price variation due to speculation will potentially disrupt the pattern of spot-futures convergence in the long run. Therefore, a pressing question is to ascertain how the difference of market transition contributes to the investor demand and causes the commodity price to oscillate beyond its normal range? This question has been of utmost importance to participants in the spot and futures markets to offset their position in strengthening their portfolio investments. There is no clear-cut conclusion on the spot-futures relation, depending on types of commodities, market features and perceived

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

the sample period of 1980–1990, the author's result provides an evidence to reject this stabilization hypothesis for all four markets. Kaufmann and Ullman (2009) consider the roles of hedgers and speculators in the West Texas Intermediate (WTI), Brent-Blend, Maya, Bonny Light and Dubai-Fateh crude oil markets. Their result supports that speculation activities are likely to exacerbate the market fundamentals when high fluctuations in futures prices happen as opposed to spot prices. Among twenty-eight commodities traded in the United States, Emekter et al. (2012) use a duration dependence test on the stochastic interest-adjusted basis and find evidence of rational speculative bubbles in eleven commodity markets. Bos and Van der Molen (2012) develop and use their own non-parametric test and an empirical model to examine the impact of futures speculation on prices of coffee in the Arabica for the sample period of 1989–2008. They find that factors such as harvest sizes, inventories, futures market microstructure and price elasticity of demand enhance the impact of futures speculation on rising coffee price. They also suggest that speculation is one of the contributors of input for the other commodities. With the application of unit root and cointegration tests, Lucey and O'Connor (2013) detect the possibility of a rational speculative bubble in the gold price during the sample period of 1989–2013. They use Markov switching augmented Dickey-Fuller tests to account the periodically collapsing bubbles. Their adequate result based on the imposition of a constant variance in two regimes provides evidence of existing bubble for 2, 3 and 12-month lease rates. Similarly, Zhang (2013) uses linear and non-linear approaches to detect such bubble in the case of WTI market for the sample period of 2007–2010. As a result, he finds that an instantaneous feedback of speculators’ position change on crude oil return appears significantly in the linear form than the non-linear form. This linear feedback appears symmetrically across the sample period with different crude oil volatilities, but the non-linear feedback that takes asymmetric feature does not exist. Following the study by Mahalik et al. (2014) in the case of Indian commodity markets during 2005–2008, they find that the effect of past innovation in the futures market on spot volatility happens frequently in the agricultural future index, energy future index and aggregate commodity index. Guilleminot et al. (2014) find a high correlation between speculative and index positions during periods of liquidity stress, providing that a strong impact of index flows on prices for twelve traded agricultural commodities in the United States. Taking multiple periods of price changes for these agricultural commodities during 1970–2011, Etienne et al. (2014) find that 1.5–2% of them belong to episodes of price bubble. They further find 80–90% of bubbles are short-lived that usually last fewer than 10 days, accounting for more than one-third of the explosive episodes. Using a momentum threshold autoregressive approach in testing speculative bubbles in the United States during the sample period of 1993–2012, Adämmer and Bohl (2015) find that speculation has a reinforcing influence on wheat prices. However, their empirical results provide inconclusive finding for corn and soybean. Additional, Brooks et al. (2015) estimate fundamental values based on convenience yields and macroeconomic factors. Along with these estimated values, they use a switching regression approach. Their result provides a reliable evidence of pure speculation to be unsustainable in causing the extreme price movement for crude oil and feeder cattle over a 40-year period since the late 1960s. Furthermore, Huchet and Fam (2015) report that coffee, sugar, corn and wheat returns during the sample period of 1998–2013 are systematically modified by speculative transactions in futures markets. The speculative pressure from futures markets seems to

risk among market participants toward the nature of a commodity. For crude palm oil (CPO), there is a growing demand for biofuels and foods in the emerging countries. The government has implemented some national policies on energy and food and influenced the consumption pattern on CPO. For example, being the world's second largest producer of CPO, Malaysia has implemented the National Biofuel Policy on March 21, 2006 to promote the use of biodiesel derived from palm oil as environmentally friendly and sustainable energy source in order to reduce dependency on fossil fuels. It also aims to stabilize and boost palm oil prices through export, research and development activities (Gain Report, 2014). On the other hand, the National Agro-Food Policy has been launched on September 28, 2011 to solve the issue of inequality of income distribution and poverty. It aims to ensure steady and resilient food related industries through the development of agricultural sector. This, in turn, would increase farmers’ revenue and directly curb inflation to maintain sufficient amount of food supplies for consumption in the country (Ministry of Agricultural and Agro-Based Industry Malaysia, 2014). Tilton et al. (2011) and Östensson (2011) state that shifting investor demand for the supply of commodity spot varies when the market transits from contango to backwardation or vice versa. However, examining the relationship between spot and futures prices without taking into account the efficiency as measured by the variance of the increments of a random walk for futures pricing could result in finding a spurious relationship between investor demand and commodity price. In this regard, correlation coefficients between spot and futures price changes during strong contango, weak contango and backwardation periods by given convenience yields of 0%, 1%, 2.5%, 5% and 10% are tested. The degree of efficient futures pricing is measured using variance ratio. The empirical finding of this study provides implications for market participants to adjust their response to their making decisions under market transition. For investors, if the futures market is found to be efficient, they can adjust their decisions in executing inter-temporal arbitrage strategies between spot and futures markets by trading liquid and physical stocks of the commodity. Furthermore, both stocks and futures are treated as precautionary instruments. They can relate the efficient futures market to their precautionary behavior towards output and price risk under a certain market condition. The efficiency of futures price allows them to adjust their decisions of holding stocks in obtaining convenience yields in the future using futures contracts as hedging instruments. Section two reviews the literature. The subsequent section provides the explanation of data and methodology, followed by findings and empirical results. The last section concludes the discussion and suggests the implication. 2. Literature review In the following section, past findings on speculative bubbles in commodity prices are discussed. Then, the arguments on linkages between market transition and investor demand/supply by Tilton et al. (2011) and Östensson (2011) are briefly explained. Finally, a review of market efficiency for the respective commodity spot and futures markets is outlined. 2.1. Financial speculation in commodity markets Numerous studies look at the aspect of whether investors in the futures markets act as a major force that distorts and drives up commodity prices in a variety of situations. There are some studies that support the existence of such speculative pressure. For instance, Kocagil (1997) tests the hypothesis of speculation that stabilizes spot prices for copper, gold, silver and aluminum. From 136

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

There are some surveys done by researchers to challenge that the view of financial speculation is an important determinant of rising commodity prices. For instance, Irwin and Sanders (2011) conclude that variation in fundamental supply and demand factors play important role in leading the sharp increases in agricultural prices as compared to index investors’ participation. In terms of the role of financial speculation particularly index funds during the 2008 commodity price boom, Östensson (2012) finds that most of past studies with statistical and anecdotal evidence provide a limited credible explanation on how speculators are responsible for the rising commodity prices in 2008. He concludes that the influence of futures activities on the commodity price spikes in 2008 being subject to backwardation. Through a brief review of speculation hypothesis, Fattouh et al. (2013) discuss to what extent different methodologies from past studies published during 2008–2012 in detecting the role of speculation in oil futures markets. They conclude that the comovement of spot and futures prices reflects common economic fundamentals instead of the financialization. However, along with such comovement, they find that speculative activities after 2003 have no impact on oil spot prices. Will et al. (2015) further review thirtyfive empirical studies which are published during 2010–2012. They synthesize and conclude that most existing findings provide less evidence to support the increase of financial speculation will raise the price level or price volatility in agricultural markets. The reasoning is that fundamental factors such as demand and supply play as a major driver of spikes in commodity prices. For the latest survey, Haase et al. (2016) review the major findings from one hundred empirical studies are published during 2003–2016. They state that the number of studies is about the same in supporting and contradicting the criticized effects of speculation on commodity futures markets. They find that these reviewed studies demonstrate speculation has reinforcing effects on price levels, while it provides weakening effects on returns and risk premiums. These effects of speculation on spreads, volatilities and spillover effects are found can be either as reinforcing or weakening effects. Furthermore, their reviewed studies majority show that two ways in measuring effects of speculation are markedly different. For example, using the indirect way in measuring such effects through proxy variables supports that speculation provides reinforcing effects. For the direct way, it supports that speculation provides weakening effects. Most of the past studies mildly support that investor demand is put forward as a factor of increasing commodity prices, especially for agricultural commodities. However, these findings are considered as insufficient to demonstrate that how investors act towards to fundamental factors such as weather shock, declining inventory and consumption growth in demanding agricultural commodities. Their action is subject to whether they are willing to pay a premium in order to have the commodity at some point in the future (strong contango) or pay the costs of storage and the carry costs of buying such commodity today (weak contango or backwardation). To quantify their impact on the increase in commodity prices, this study investigates whether CPO prices are exposed to investor demand during strong contango, weak contango and backwardation periods, respectively.

have a weak effect or no effect on rice, cocoa and soybean prices due to their different market features. For example, the rice returns are found to be insufficiently correlated with the size of futures markets and relative share of long positions taken by speculators among the sum of open positions. Next, cocoa returns are not excessively speculated even fundamental factors can explain a continuous rise in its price. In addition, soybean returns are found as not sensitive to positions taken by non-hedgers. Consequently, its returns do not depend on speculation, even though its market seems to be highly efficient and liquid. There are some studies shift against the criticized effect of financial speculation on commodity prices. For instance, Irwin et al. (2009) provide a discussion on the controversy about the influence of speculative bubble in driving commodity prices up during the period of 2007–2008. They provide four points: First, the increasing energy prices during such period are due to economic fundamentals (strong demand from China, India, and other developing nations). Second, a substantial bubble in commodity prices happens inconsistently under different situations. Third, their statistical evidence indicates that a long-only index fund does not affect changes in commodity futures prices. Fourth, there is a historical pattern of attacks upon speculation during high volatile periods. Using the empirical analysis based on the monthly data of 2006–2009, Gilbert (2010) demonstrates that investors’ view on the demand of raw material from China drives index-based futures investment and generates the food price spike of 2007/08. In addition, Sanders et al. (2010) use Working's index of speculation after decomposing open interest in commodity futures markets into commercial and non-commercial positions. They find that there is no change in speculative activities during 2006–2008. For agricultural and energy futures markets, Sanders and Irwin (2010, 2011a, 2011b) find that there is no significant relationship between returns and long positions held by index funds. These studies further argue that speculation is a necessary counterpart to hedging activities because it contributes to efficiency in transmitting information and increases the market liquidity. Furthermore, Capelle-Blancard and Coulibaly (2011) use panel Granger causality tests to take the possible contemporaneous dependence across agricultural and livestock futures markets. Their empirical findings based on the weekly data of 2006–2010 suggest that index-based trading and futures price do not exhibit the causal effect. Irwin and Sanders (2012a) subsequently use cross-sectional regression and time series tests. Their empirical results based on the quarterly data of 2007–2011 indicate that a long-only index investment is not a major driver in agricultural commodity and energy futures markets. By looking at the aspect of three structural changes over 2000–2011, Irwin and Sanders (2012b) report that the minimal impact of passive index investment can cause a massive bubble in futures markets for corn, wheat, soybeans, live cattle and lean hogs. For the case of wheat futures market during 1989–2011, Liu et al. (2013) use a regime-switching model and their result contradicts the hypothesis of periodically collapsing bubbles drive prices of agricultural commodities. Bohl et al. (2013) use a stochastic volatility model and find that index investors have no impact on futures prices for twelve financialized grain, livestock and soft commodities. Miffre and Brooks (2013) use eight trading strategies to examine whether long-short investors destabilize futures markets for twenty-seven commodities during 1992–2011. Using Granger causality tests, they find little evidence that speculators for long-short portfolios have an impact on volatility or cross-market correlation with S & P500 and Barclays bond indexes. Based on the data of 1992–2012, Kim's (2015) empirical analysis shows that speculators in the United States futures markets which consist of twenty-one commodities during high volatile periods can either have no effect or dampen commodity spot prices.

2.2. Linkages between market transition and investor supply/ demand In considering the market transition from contango to backwardation or vice versa, Keynes (1930) who firstly observes that speculators who are holding long positions in the futures contract will obtain a risk premium during the backwardation period. When they take short positions in order to receive a risk premium, the contango period will be in the existence. This demonstrates that the 137

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

occurrence of backwardation or contango depends on whether speculators are “net long” or “net short”. However, Working (1953) opposes this theory because speculators require a risk premium based on their different opinions on future price changes of a commodity, where the futures price is regarded as being equal to the expected spot price. The futures price is also interrelated with current spot price based on storage theory, where the futures price should not be greater than the current spot price plus various carry charges such as storage cost and convenience yield. In strong contango, the futures price exceeds the expected future spot price, as well as the discounted futures price, is greater than the current spot price. The rising futures price relative to spot price provides the situation for a commodity to be available for sale to prompt market at a discount rate with the same commodity for delivery at future dates. The weak backwardation occurs when the current spot price is lesser than futures price and greater than the discounted future spot price, while zero backwardation occurs if the current spot price equals to the discounted future spot price. According to Pindyck (2001: p.17), weak backwardation and zero backwardation are said to be in contango, but this study refers both market conditions as weak contango. When the futures market is in backwardation, both contemporaneous futures and discounted futures price are lesser than the expected future spot price. In the other word, the futures price of a commodity is either below the spot price or insufficiently above the spot price to cover storage cost, which allows participants to buy a commodity in the futures market and sell the same commodity in the spot market. In this situation of scarcity, future stocks will be not physically available for sale today because a greater storage and inventories of a commodity are needed to reallocate for the shortrun production by reducing production costs. Consequently, demand for storage and convenience yield will be quite high because market participants anticipate that near-term supplies are inadequate. To provide a buffer against a high fluctuation in production due to the unpredictable shift in demand-supply during that period, high short-run production costs are required instead of the long-run production costs. Without using the available data for copper prices, Tilton et al. (2011) develop curves for producer supply, consumer demand, investor demand and total demand. They illustrate that speculator or investor demand in the futures market can comparably influence spot prices when a market in the contango (exceeds the cost of storage and interest). They conclude that the investor demand, which is associated with rising futures prices in excess of supply for future production will depress spot prices. As the result, the investor demand most likely occurs during the strong contango period for copper due to investors’ decision in buying stocks drives commodity prices up even their stocks are declining.1 Overall, their hypothesis of investor demand in the case of copper, stating that futures market dominates the role of investor demand during the strong contango period because future stocks are physically available for sale in the futures market. As a result, spot and futures prices would be closely correlated in strong contango. They further provide two possible explanations to argue that the investor demand in the copper futures market may also play its role in backwardation or weak contango. First, investors anticipate that inadequacy of short-term supply of actual physical copper before the maturity date of the futures contracts. Second, investors are willing to pay a premium to hold physical copper. However, investor demand on the futures market is determined by the short-term consideration which contributes to a weak effect of

Fig. 1. Investor demand for and supply of spot material. Source: Adopted from Östensson (2011: p. 373).

futures prices on spot prices. Furthermore, higher spot prices than futures prices cannot allow investors to buy physical stocks from the futures markets and sell them immediately in the spot market. This unfeasible of inter-temporal arbitrage makes the correlation between spot and futures prices is turned to be weaker during the backwardation period. They depict an investor demand curve for spot material that is a function of spot prices. In this sense, Östensson (2011) concurs with their basis of conceptual and theoretical arguments on spot and futures prices during the periods of strong contango and backwardation by considering investor demand for and supply of spot material should be a function of the difference between futures and spot prices. As shown in Fig. 1, when this difference is larger than the cost of holding stocks, the futures market is in strong contango and investors demand spot material. When this difference is lower than the cost of holding stocks but larger than zero, the futures market is in weak contango and investors supply spot material to the market. Finally, if such difference is less than zero, the futures market is in backwardation and again, investors supply spot material (Östensson, 2011: p. 373). With empirical evidence for copper, Gulley and Tilton (2014) further find that spot and futures price changes are closely correlated during the period of strong contango. Meanwhile, they find that this correlation should be considered as high during the periods of backwardation and weak contango due to market participants’ concern on their near-term shortages. Moreover, Fernandez (2015) extends the scope of examining this hypothesis of investor demand by considering aluminum, copper, lead, nickel, tin, and zinc from the London Metal Exchange during 1992–2014. The author uses various robustness tests by controlling the conditional heteroscedasticity in returns, detecting unconditional meanreturn breakpoints, and detecting and removing outlying observations. These tests indicate that a linkage between spot and futures markets for six industrial metals traded is weak during the contango period. Under the situation of no-arbitrage, the relationship between spot and futures prices of a commodity is explained by the cost-ofcarry model. With the application of the model which is given by Eq. (1), the expected (future) spot price or futures price is obtained for delivery for T months forward (Ft , T ).

Ft , T = St (1 + rt + Ct − ψt )t , T

(1)

where,

St = current spot price at time t; Ct = annualized storage cost in percent, which includes the costs of handling, spoilage, shrinkage, shipping and others at time t; ψt = convenience yield at time t ; rt = the deposit rates at time t ; and T = the number of months divided by 12.

1 Refer to Tilton et al. (2011), they address that the investor demand likely occurs during the period of strong contango (p.191, para 7) and it can push up the price of a commodity even when investors’ stocks are declining (p.193, para 2).

138

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

squares estimator of slope on futures prices and its test statistic would be biased and misleading. To produce consistent estimators, they incorporate the conditional heteroscedasticity into the model based on the weighted least squares method. Then, they use the model to test market efficiency and unbiasedness of gold, silver, platinum and oil futures prices for the sample period of 2005– 2011. Their result reveals that spot and futures prices for gold, silver, and platinum have cointegration relationship with a slope of one and otherwise for crude oil. Some studies show that different prevailing economic and political conditions cause futures prices to exhibit time-varying behavior in predicting future spot prices. For instance, Charles and Darné (2009) use new variance-ratio tests to explore the relationship between weak-from efficiency and deregulation in the crude oil spot markets for Brent and WTI using data of 1982–2008. Their result indicates that the Brent crude oil market seems to be more efficient than the WTI crude oil market. This attributed to the process of deregulation during 1994–2008, making the WTI crude oil returns to become less predictable. In the subsequent study, Inoue and Hamori (2014) find that spot and futures prices for the Indian commodities are cointegrated during the more recent sample period of 2009–2011. They suggest that increasing trading volume of the futures market since 2009 improves the efficiency of futures prices in producing an unbiased predictor of spot prices. To account the possibility of a structural break in testing the efficiency of a futures market for crude oil, Stevens and de Lamirande (2014) generalize the basis regression by testing the parameter stability for two sub-periods: 1985–2008 and 2008–2013. With a strong rejection of the null hypothesis of parameter stability, they further test the generalized null hypothesis of efficiency. Then, they find that the structural change of futures price movement in May 2008 as an evidence of inefficiency in the futures market. The market efficiency is also characterized by different types of commodity. For commodity futures markets, Kristoufek and Vosvrda (2014) propose the Efficiency Index and find that energy commodities during 2000–2013 are the most efficient, followed by soft commodities, grains, and metals. The other agricultural commodities such as live cattle and feeder cattle are the least efficient. For commodity spot markets, Charles et al. (2015) use an automatic portmanteau test. They find that gold and silver markets during 1977–2013 exhibit a downward trend in return predictability. Based on a variance-ratio test, their result of return predictability reveals that the gold market is the most efficient, followed by the silver market during the late 1970s. However, they do not find such downward trend of return predictability in the platinum market, indicating that such market is inefficient.

To identify the strong contango, weak contango and backwardation ∧

periods in the sample, Ft , T from Eq. (1) is compared with St . Specifically, the different conditions are shown as follows: ∧

∧

If Ft , T > St and contango;

Ft , T (1 + rt + Ct − ψt )t , T ∧

∧

if Ft , T > St and contango; and

> St , the futures market will be in strong

Ft , T (1 + rt + Ct − ψt )t , T

∧

if Ft , T < St and backwardation.

< St , the futures market will be in weak

∧

Ft , T (1 + rt + Ct − ψt )t , T

< St , the futures market will be in

2.3. Efficiency of commodity spot and futures markets Market efficiency relates to the spot-futures relation. Garbade and Silber (1983), Oellermann et al. (1989), and Schroeder and Goodwin (1991) suggest the futures price plays a vital role in price discovery process for the underlying spot market under the efficient market hypothesis. The futures market is efficient when futures prices equal to expected future spot prices plus or minus a constant, a time-varying risk premium. Furthermore, Silvapulle and Moosa (1999) find that the futures price more efficient than the spot price. The reason is futures prices respond to new information faster than spot prices as a result in lower transaction costs and flexibility of short-selling activities in the futures market. In the recent study, Caporale et al. (2014) incorporate an endogenous convenience yield into the cost-of-carry model to obtain the time-varying spot and futures markets’ contribution to price discovery in the case of WTI crude oil. They find that prices of the futures contract with shorter maturities dominant a role of price discovery during 1990–2008. Focusing on the effect of futures prices of four contracts maturing in one, two, three, and four months on spot prices for the WTI crude oil, Chang and Lee (2015) use a wavelet coherency analysis and find that dynamic correlations between both prices in time-frequency domain contribute to more significant dynamic causality between spot prices and futures prices of contract with shorter maturity during 1986–2014. This suggests that the short-term futures prices in the oil markets are more efficient in implementing price discovery mechanism than the long-term futures prices. To test the efficiency of commodity futures markets, some studies indicate that the condition for the futures market to achieve efficiency is futures and spot prices should be cointegrated. For instance, Tomek and Gray (1970), Kofi (1973), Leuthold (1974), and Martin and Garcia (1981) regress spot prices on lagged one of futures prices. They conclude that intercept of zero and a unit slope on futures prices in a simple regression model indicates the market is efficient, suggesting that futures prices should be unbiased predictors of future spot prices. Furthermore, McKenzie and Holt (2002) argue that the market may be efficient and unbiased in the long run, but may experience inefficiency and pricing biases in the short run. With the error-correction and generalized-quadratic ARCH models, their argument is found to be consistent in the case of live cattle, hogs, and corn futures markets for the sample period of 1959–2000. This finding is consistent with Liu's (2009) finding in the case of Malaysian CPO futures market for the sample period of 2001–2007. With the application of a vector error cointegration model, the author finds that CPO spot and futures prices have the long-run relationship for all forecasting horizons, but this model rejects the short-run efficiency. In terms of econometric modeling, Westerlund and Narayan (2013) state that both spot and futures prices are not necessarily cointegrated given a unit slope on futures prices. This contributes to endogeneity problem, causing the conventional ordinary least

3. Data Data for the daily CPO spot and futures prices in Malaysian currency, Ringgit Malaysia (RM) from January 3, 2000, to July 29, 2016 are used in the analysis. These daily prices are obtained from the Bursa Malaysia and transformed to become daily changes in the logarithmic prices. The CPO futures contract is officially traded in two sessions from the trading floor of the Bursa Malaysia. The first trading session: Malaysian time from 10:30 a.m. to 12:30 p.m. The second trading session: Malaysian time from 3:00 p.m. to 6:00 p.m. The change of futures prices becomes more closely associated with the change of underlying commodity prices when the futures contract matures at its expiration. According to the study by Ma et al. (1992), the heterogeneity between consecutive contract and unusual market activity are often observed to happen when futures contracts draw closer to maturity. Consequently, there will be biases in the various time-series properties of the artificial price series produced by different rollover methods. To perform the 139

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

5. Empirical results

analysis of futures, it is important to concatenate various individual futures contracts by ensuring price jumps have been removed. To apply a method of rolling over contracts, the first step is to select a point in time to “roll over”. Switching at the delivery date is a common method of rolling over contracts, where futures contracts are rolled over some arbitrary time in the first of the delivery month or the last trading day of the previous month. In the case of CPO futures, its contract is deliverable in a fixed month that expires at noon on the 15th day of the delivery month. If the 15th day is a non-market day, a contract expires on the preceding business day. To ensure that all prices are measured at the same point in time, the constant maturity contract is chosen. Therefore, this study uses the 3-month futures contract as it is the most active and liquid contract traded in the futures exchange. The contract is rolled together proportionally over 15 days of trading in each 3 months. On the rollover date, the differences in the price levels between the two contracts are often observed due to hedging or carry costs. This introduces price jumps in the entire time series that generate seemingly excessive volatility and extremely large price changes. Consequently, the results may be flawed. To avoid the problem, the second step is to adjust the price levels by removing the price jumps that occur at each rollover date, leading to a more robust measure of underlying price changes on the contract. This can be done by subtracting the difference in the price levels between the new contract and the old contract at each rollover date for all new prices (Ma et al., 1992: p.205). For the same sample period, the daily data of deposit rates are obtained from the Central Bank of Malaysia and further used in the analysis.

5.1. Identifying strong contango, weak contango and backwardation periods for CPO By assuming convenience yields of 0%, 1%, 2.5%, 5% and 10%, Fig. 2 clearly indicates that shaded areas for backwardation period dominate the sample period of January 2000 - July 2016. This demonstrates that a period for the spot price to exceed the futures price is most visible due to seasonal factor in CPO production associated with both CPO spot and futures markets. In this regard, investors normally expect the CPO futures market in backwardation because a shortage of CPO that happens most of the time. As observed, there is a high fluctuation of daily spot and futures price changes from January 2005 - August 2006 during the backwardation period, in that the movement of both CPO returns corresponds to seasonal variation. Taking a closer look at Fig. 2, when the convenience yield is increased from 0% to 5%, there are some frequent and extended sub-periods for strong contango. This implies that investors are willing to hold CPO as inventories in order to gain a high convenience yield even the futures price tends to exceed the spot price. However, the spot and futures markets are never in weak contango over the period given the convenience yield of 10%. This implies that investors are no longer holding CPO as inventories to obtain the convenience yield of 10% due to the immediate use of CPO in the short-run production. 5.2. Descriptive statistics for daily CPO spot and futures price changes: strong contango, weak contango and backwardation

4. Methodology

Table 1 presents the descriptive statistics for daily spot and futures price changes in strong contango, weak contango and backwardation periods, respectively. When the CPO spot and futures markets are in strong contango in which futures price exceeds the spot price, daily changes of futures price are positively skewed. However, trading in the spot market provides negative returns. In weak contango, daily changes of spot and futures prices in all cases are found to be positively skewed with positive mean values, suggesting that investors frequently gain from their arbitrage in both spot and futures markets. When both CPO markets are in backwardation, daily changes for CPO spot price are negatively skewed with positive mean values. This implies that the risk of having a left-tail event in the spot market has become more pronounced. In order to reduce production costs in the short run, investors frequently increase their demand for the physical stocks and trade the commodity in spot market with more caution. Meanwhile, daily changes of CPO futures prices are found to be positively skewed with negative values, suggesting that trades related to CPO futures during such period produce negative returns. In all sub-periods given the convenience yields, the futures market is observed to exhibit a higher positive skewness than the spot market. Furthermore, standard deviations indicate that daily futures price changes are more volatile than daily spot price changes. This means that daily changes of futures prices with positive skewness has relatively high volatility. Indeed, investors still prefer to trade futures as its price is positively skewed, despite the fact that volatility in the market is slightly higher than the spot market. The reason is that they may face difficulty in determining CPO price changes in particular when seasonal aspect is taken into account. This makes them encounter the problem in forming price expectation as available inventory for such commodity is not always guaranteed in the short run. As a result, they rely on futures prices to obtain the right signals of CPO price movement. For example, they require greater inventory from the futures market to buffer the

This study involves three-step analysis. The first step is to identify the strong contango, weak contango and backwardation periods based on the cost-of-carry theory (Eq. (1)). For this study, the CPO futures prices are obtained for delivery for 3-months forward (Ft , T )by assuming convenience yields of 0%, 1%, 2.5%, 5% and 10%, respectively. 2 Then, these futures prices are discounted and compared to spot prices. The second step is to evaluate the departure from randomness and weak-form informational efficiency for the respective spot and futures price changes. To achieve informational efficiency under the assumption of risk neutrality, the future price movement should be not foreseeable. Its variations are characterized by a random walk process because of the random nature of unpredictable events (Fama, 1965, 1970, 1991). The efficiency of spot and futures markets is tested for each sub-period by assuming convenience yields of 0%, 1%, 2.5%, 5% and 10%. The variance ratio (VR) is widely used in financial empirical studies to test the random walk hypothesis (Lo and Mackinlay, 1988; Poterba and Summers, 1988; Lo and Mackinlay, 1989; Wright, 2000). Using the VR, we assess the variance of price changes for spot and futures. If VR is higher than one, it indicates that price changes are persistent (long memory). If VR is lesser than one, it indicates that price changes are mean reverting (short memory). The third step is to compute and test correlation coefficients between spot and futures price changes (ri ) given convenience yields of 0%, 1%, 2.5%, 5% and 10% for each sub-period. To test the significance of individual correlation coefficient and difference between two correlation coefficients, Fisher r-to-z transformation is used to derive the ri distribution becomes the normal distribution. 2 Since spoilage of CPO depends on the seasonal production, storage costs of a commodity will vary from time to time. In this regard, this study assumes that the annual cost is 5%.

140

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

Fig. 2. Daily CPO spot and futures returns during strong contango, weak contango and backwardation periods, January 2000-July 2016. Source: Authors’ own identification.

fluctuations in production and consumption, consequently rising the futures volatility relative to spot volatility. In addition, kurtosis measures the peakedness or flatness of distribution of a series. Both series in all sub-periods have kurtosis values which are greater than three, indicating that both series have a high leptokurtic distribution relative to a normal distribution. As a result, Jarque-Bera (JB) test statistics provide the rejection of the null hypothesis of normal distribution at the 1% level of significance. This reflects the fact that existing volatility clustering over time, leading to a serial correlation with non-constant variation in both series.

price formation and vice versa.3 Prices often move in ways which are difficult to be explained rationally when there exists the effect of large influx of financial investors on efficiency of commodity market. Therefore, a non-random walk process cannot be a complete description of market price behavior because it ignores investors’ underreaction and overreaction to the new arrival of information. Table 2 presents the results of VRs as well as test statistics, Z (q ) for daily spot and futures price changes. When there is the sufficient futures price above the spot price (strong contango) by given the convenience yield from 0% to 10%, the VRs for the futures market provide a strong rejection of the null hypothesis of random walk (three rejections out of the four cases examined). This indicates that futures price changes are persistent due to illiquidity of futures trading. In the most of the cases examined for the spot

5.3. Weak-form market efficiency of daily CPO spot and futures markets: strong contango, weak contango and backwardation To detect a random walk process, the VRs which associate with intervals q = 2, 4, 8 and 16 for respective spot and futures price changes are estimated in each sub-period. The VR test exploits the fact that the variance of the increments of a random walk is a linear function of q, where it is generally not consistent with the stochastic behavior of daily series. Therefore, the non-rejection of a random walk process does not necessarily imply the efficiency of

3 A random walk process is too stringent to serve as a characteristic of information efficiency, implying that price changes are serially independent with a constant probability distribution across time. If the random walk hypothesis is based on the theory of efficiency, a random walk process happens in the short term and the efficient market happens in the long term (LeRoy, 1973; Lucas, 1978; Lo and MacKinlay, 1988; Lo and MacKinlay, 2002).

141

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

Table 1 Descriptive statistics for daily CPO spot and futures price changes, January 2000 - July 2016.

0% convenience yield: Strong contango Spot Futures Weak contango Spot Futures Backwardation Spot Futures 1% convenience yield: Strong contango Spot Futures Weak contango Spot Futures Backwardation Spot Futures 2.5% convenience yield: Strong contango Spot Futures Weak contango Spot Futures Backwardation Spot Futures 5% convenience yield: Strong contango Spot Futures Weak contango Spot Futures Backwardation Spot Futures 10% convenience yield: Strong contango Spot Futures Weak contango Spot Futures Backwardation Spot Futures

Obs

Mean

Std. Dev

Max

Min

Skewness

Kurtosis

JB

197 197

−0.0015 0.01

0.0187 0.0467

0.078 0.3540

−0.0469 −0.0503

0.3587 4.3579

3.9887 27.1511

12.2468*** 5411.28***

303 303

0.0013 0.0066

0.0138 0.0306

0.0932 0.297

−0.0379 −0.0383

1.4695 5.7099

10.3317 46.724

787.703*** 25782.8***

3983 3983

0.0002 0.0002

0.0169 0.0399

0.0992 0.7388

−0.1104 −0.5325

−0.072 1.086972

7.8229 105.290

3863.7*** 1737247***

215 215

−0.0011 0.011

0.0186 0.046

0.078 0.354

−0.0469 −0.0503

0.3269 4.285

3.9414 26.6533

11.7683*** 5669.909***

285 285

0.0011 0.0059

0.0135 0.03

0.0932 0.297

−0.0365 −0.0383

1.6378 6.0584

11.2846 52.0204

942.447*** 30279.04***

3474 3474

0.0002 −0.001

0.017 0.0401

0.0992 0.7388

−0.1104 −0.5325

−0.1661 0.6206

7.9376 113.41

3544.99*** 1764545***

244 244

−0.0002 0.0107

0.0191 0.0437

0.0932 0.354

−0.0469 −0.0503

0.699191 4.4408

5.554 28.988

86.2206*** 7668.487***

256 256

0.0005 0.0056

0.0122 0.031

0.051 0.297

−0.0365 −0.0383

0.7306 6.0957

5.7194 51.088

101.6554*** 26251.41***

3474 3474

0.0002 −0.001

0.017 0.0401

0.0992 0.7388

−0.1104 −0.5325

−0.1661 0.6206

7.938 113.40

3544.992*** 1764545***

332 332

−0.0001 0.0109

0.0174 0.0445

0.0932 0.3540

−0.0469 −0.0503

0.743 4.4809

6.3124 27.9694

182.32*** 9735.7***

168 168

0.0007 0.0026

0.0127 0.017

0.0452 0.1298

−0.0365 −0.0383

0.6369 2.909

5.1841 22.078

44.751*** 2784.8***

3474 3474

0.0002 −0.001

0.017 0.0401

0.0992 0.7388

−0.1104 −0.5325

−0.16613 0.6206

7.9376 113.403

3544.99*** 176455***

500 500

0.0002 0.0081

0.0159 0.0378

0.0932 0.3540

−0.0469 −0.0503

0.7245 5.1282

6.5604 37.5126

307.841*** 27006.5***

NA NA

NA NA

NA NA

NA NA

NA NA

NA NA

NA NA

NA NA

3348 3348

0.0001 −0.0011

0.0172 0.0407

0.0992 0.7388

−0.1104 −0.5325

−0.1778 0.6185

7.8123 110.951

3248.254*** 1625854***

Notes: CPO denotes as crude palm oil. Convenience yields during the period of strong contango are assumed to be 0%, 1%, 2.5%, 5% and 10%, respectively. *** denotes as the null hypothesis of normal distribution is rejected at the 1% level.

changes in weak contango follow a non-random walk process. One possible explanation is that investors still mostly focus on the futures market because it provides a convenience for them to hedge the risk from inflation. Although the results of VRs indicate a randomness of spot price changes in weak contango, it does not imply that the spot market is efficient. This may due to investors in the market under react to the news as they anticipate a squeeze on actual physical CPO in the short run, causing them are willing to pay a high premium for having a preference for actually holding physical CPO. This is indeed the case when fundamental factors do not solely influence spot prices during weak contango period. During the backwardation period, the price of a futures contract that lowers than the expected spot price at the contract maturity will induce investors to buy futures and sell spot. Given convenience yields of 0%, 1% 2.5%, 5% and 10%, the estimated VRs for daily futures price changes that less than one provide three

market, no rejection of the null hypothesis is found. This implies that investors are less concern in allocating their inventories based on the movement of spot prices. Indeed, what investors would like to do during the strong contango period is they expect the availability of physical fruit for CPO in the future being subject to seasonal factors and future inventories are not physically available for sale today. This makes the delivery of futures contracts will be deferred. In order to compensate their perceived risk, they are usually focusing on the futures market to alter their expectation about forthcoming CPO price trends. The reason to support this finding is futures prices rise and fall are in line with deviations from an inflationary expectation. Such explanation is supported by the estimated positive VRs, indicating that the futures market shows inclination always towards persistence instead of the spot market. As surprised, the results reveal evidence that futures price 142

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

Notes: Convenience yields during the period of strong contango are assumed to be 0%, 1%, 2.5%, 5% and 10%, respectively. Variance ratios (VR(q )) is defined as VR = σc2(q )/σa2(q ), where σc2(q ) is an unbiased estimator of 1/q of the variance of the qth

Table 2 Results of variance ratio VR(q) and variance-ratio test statistic Z(q) for market efficiency, January 2000 - July 2016.

difference of the daily price change and σa2(q ) is an unbiased estimator of the variance of

Number q of base observations forming variance ratios 2

4

8

16

1.1 (0.52) 1.91 (4.07) ***

1.28 (1.68)* 2.61 (4.47) ***

1.2 (0.98) 3.46 (19.83) ***

1.08 (0.57) 1.43 (1.001)

1.14 (0.84) 1.68 (12.3) ***

0.9 (−0.56) 1.74 (7.41) ***

1.03 (0.21) 1.75 (10.22) ***

1.01 (0.19)

1.07 (1.09)

1.14 (2.47)**

0.74 (−4.29) ***

0.72 (−5.76) ***

1.32 (5.14) *** 0.74 (−4.03) ***

1.09 (0.51)

1.25 (1.62)

1.24 (1.28)

1.96 (5.41) ***

2.75 (19.73) ***

3.58 (6.02) ***

0% convenience yield: Strong contango Spot 1.01 (0.08) Futures 1.38 (1.06) Weak contango Spot Futures Backwardation Spot Futures

0.82 (−0.96) 1% convenience yield: Strong contango Spot 0.99 (−0.01) Futures 1.41 (1.2) Weak contango Spot Futures

1.01 (0.02) 1.28 (1.13)

1.02 (0.09) 1.43 (5.62) ***

1.06 (0.40) 1.44 (3.85) ***

0.92 (−0.45) 1.44 (1.8)*

Backwardation Spot

1.01 (0.15)

1.06 (0.86)

1.11 (1.92)*

0.74 (−4.29) ***

0.71 (−5.80) ***

1.25 (4.11) *** 0.73 (−4.17) ***

0.95 (−0.27)

0.96 (−0.27)

0.87 (−0.73)

1.87 (5.08) ***

2.47 (20.00) ***

2.8 (14.9)***

Futures

0.82 (−0.96) 2.5% convenience yield: Strong contango Spot 0.99 (−0.05) Futures 1.3 (1.14) Weak contango Spot Futures

1.04 (0.31) 1.23 (0.91)

1.04 (0.22) 1.35 (4.82) ***

1.03 (0.19) 1.35 (6.25) ***

0.81 (−1.28) 1.38 (5.96) ***

Backwardation Spot

1.01 (0.15)

1.06 (0.87)

1.11 (1.85)*

0.82 (−0.96) 5% convenience yield: Strong contango Spot 0.99 (−0.1) Futures 1.37 (1.44)

0.74 (−4.3) ***

0.71 (−5.81) ***

1.25 (4.04) *** 0.73 (−4.15) ***

0.97 (−0.2) 2.05 (5.48) ***

0.95 (−0.36) 2.87 (12.03) ***

0.93 (−0.49) 3.31 (27.11) ***

Futures

Weak contango Spot Futures

1.10 (0.56) 1.30 (1.02)

1.22 (1.36) 1.64 (4.09) ***

0.99 (−0.03) 1.70 (3.36) ***

0.93 (−0.39) 1.61 (2.54) ***

Backwardation Spot

1.01 (0.15)

1.06 (0.86)

1.11 (1.91)*

0.74 (−4.29) ***

0.71 (−5.81) ***

1.25 (4.11) *** 0.73 (−4.17) ***

0.95 (−0.38)

0.96 (−0.3)

0.96 (−0.29)

1.96 (2.99) ***

2.43 (22.59) ***

2.54 (16.49) ***

NA NA

NA NA

NA NA

NA NA

1.01 (0.14)

1.06 (0.89)

1.11 (1.89)*

0.82 (−0.95)

0.74 (−4.24) ***

0.71 (−5.68) ***

1.25 (3.93) *** 0.73 (−4.22) ***

Futures

0.82 (−0.96) 10% convenience yield: Strong contango Spot 0.99 (−0.06) Futures 1.37 (1.52) Weak contango Spot Futures Backwardation Spot Futures

the first difference of the daily price change. The heteroscedasticity-consistent standard normal test-statistic, Z (q ) values are reported in the parentheses. The significance of test statistics indicates that the null hypothesis of VR(q ) equals to one is rejected. ***, ** and * denote as significant at the 1%, 5% and 10% levels, respectively.

rejections of the null hypothesis of a random walk out of the four cases examined. This indicates that futures price changes are mean reverting, where any shock produces a temporary effect on the subsequent price movement. This non-random walk process is often attributed to investors who are prone to overreaction as the futures market generate negative mean returns (as shown in Table 1). Since available inventory for CPO is not always guarantee in the short run, they consider looking towards investing in the futures market during such period to determine a right signal on the movement of CPO prices. As a result, their behavior causes futures prices to change more than is justified by the news. During the same period, the estimated VRs for daily spot price changes are found to be more than one, providing that two rejections of the null hypothesis with interval q = 8 and 16. It is evident that spot volatility persists over time is due to prolonged effect of any shock. This explains that investors expect lower spot prices in the future as compared to the present is due to demand on such perishable commodity cannot be met out of current production. This situation typically raises an option value of holding the physical inventories, causing them to lock and reflect the present spot prices by adjusting to a shock. 5.4. Correlation coefficients between daily CPO spot and futures price changes: strong contango, weak contango and backwardation The confidence intervals are used to obtain a range of plausible values for unknown population correlation between daily changes in spot and futures prices. Table 3 shows that intervals with probabilities of 90%, 95% and 99% in the majority of cases for all sub-periods do not contain zero. This indicates that the correlation is significantly different from zero. Positive lower and upper estimated correlations of both series demonstrate that changes of both daily spot and futures prices move in tandem across time during all sub-periods. Assuming convenience yields equal to 0%, 1%, 2.5%, 5% and 10%, we further examine whether correlations between daily spot and futures price changes during two sub-periods are significantly different. The testing on a significance of two independent correlations is constructed using the Fisher-z transformation. The results are summarized in Table 4. For a comparison between correlations of both series in strong contango and weak contango periods, the resulting negative values for test statistics indicate that correlations in strong contango period are significantly weaker than weak contango period. Assuming convenience yields equal to 0%, 1% and 2.5%, the testing on a comparison between correlations in weak contango and backwardation periods provides negative values for test statistics. This indicates that correlations during weak contango period are weaker than backwardation period. Assuming the convenience yield is 5%, correlations in both sub-periods are found to be insignificantly different. Lastly, the positive test statistics indicate that correlations in the backwardation period are significantly stronger than those in the strong contango period given all convenience yields. The intervals with different levels of confidence and testing on two correlations using different levels of significance support that daily spot and futures price changes are highly correlated during 143

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

Table 3 Confidence intervals for correlation coefficient of daily CPO spot and futures price changes, January 2000 - July 2016. r

0% convenience yield: Strong contango

0.1330

Weak contango

0.3103

Backwardation

0.5502

1% convenience yield: Strong contango

0.1368

Weak contango

0.3242

Backwardation

0.5502

2.5% convenience yield: Strong contango

0.1263

Weak contango

0.373

Backwardation

0.5502

5% convenience yield: Strong contango

0.1598

Weak contango

0.5826

Backwardation

0.5502

10% convenience yield: Strong contango

0.2032

Weak contango

NA

Backwardation

0.5534

z

90% confidence interval

95% confidence interval

99% confidence interval

Lower

Upper

Lower

Upper

Lower

Upper

0.0116

0.2464

−0.005

0.2679

−0.04

0.3087

0.1339

0.3932

0.1251

0.4087

0.1070

0.4381

0.2550

0.5693

0.2538

0.5729

0.2514

0.58

0.0179

0.2452

0.0022

0.2657

−0.030

0.3048

0.1396

0.4087

0.1308

0.4244

0.1126

0.4542

0.255

0.5693

0.2538

0.5729

0.2514

0.58

0.0154

0.2285

0.0005

0.248

−0.03

0.2851

0.1614

0.4582

0.152

0.4739

0.1355

0.5036

0.255

0.5693

0.2538

0.5729

0.2514

0.58

0.0486

0.2465

0.0371

0.263

0.0137

0.2945

0.2426

0.6607

0.2362

0.6745

0.2229

0.7

0.255

0.5693

0.2538

0.5729

0.2514

0.58

0.2061 (0.0449) NA

0.0857

0.2726

0.0774

0.2858

0.0608

0.3111

NA

NA

NA

NA

NA

NA

0.6232 (0.017)

0.2559

0.5727

0.2547

0.5764

0.2522

0.5835

0.1338 (0.0718) 0.3209 (0.0577) 0.6186 (0.017) 0.1376 (0.0687) 0.3363 (0.0595) 0.6186 (0.017) 0.127 (0.0644) 0.3919 (0.062) 0.6186 (0.017) 0.1612 (0.0551) 0.6664 (0.0779) 0.6186 (0.017)

Notes: CPO denotes as crude palm oil. Convenience yields during the period of strong contango are assumed to be 0%, 1%, 2.5%, 5% and 10%, respectively. r denotes as a correlation coefficient between daily spot and futures price changes. Based on Fisher r-to-Z transformation, z values are computed using Zi = (1/2)ln [(1+ri )/(1−ri )] to derive ri distribution to become

1/(n − 3) and reported into parentheses. The lower and upper confidence interval limits for z are computed ⎡ ⎛ ⎞ ⎤⎡ ⎛ ⎞ ⎤ ⎢ 2⎜Z L ⎟ ⎥ ⎢ 2⎜Z L ⎟ ⎥ = Zi + Z(∝/2)σzi , respectively. The lower and upper confidence interval limits for r are computed using rLi = ⎢e ⎝ i⎠−1⎥/⎢e ⎝ i⎠+1⎥ and ⎣ ⎦⎣ ⎦

a normal distribution. Standard deviations (SD) are computed using σzi = using ZLi = Zi − Z(∝/2)σzi and ZUi

⎡ ⎛ ⎞ ⎤ ⎢ 2⎜ZU ⎟ ⎥ rUi = [e2(ZU )−1]/⎢e ⎝ i⎠+1⎥ , respectively. ⎣ ⎦

as compared to the strong contango period. This explains that investors who are concern about the convenience yield tend to hold CPO as inventories on hand since they have a high expectation on the greater short-run production. During the backwardation period, they are less sensitive towards convenience yields as the correlation of both series is found to remain no much changes given an increasing convenience yield from 1% to 10%. This finding suggests that their convenience yields of holding physical inventories insufficiently cover production costs in the short run, so that their physical inventories are readily available for immediate use in the production instead of holding it on hand.

the backwardation period, but less correlated during the weak contango period and the least correlated during the strong contango period. Such finding differs from Tilton et al. (2011) who contend that spot and futures prices in strong contango are closely correlated in the case of copper due to the investor demand drive up the futures price to sufficiently cover storage costs. For the case of CPO in strong contango, our finding validates that the availability of the physical fruit in the future period is not always guaranteed due to seasonality and spoilage. This lens to support that a cause of spoilage in the physical fruit makes futures traders encounter difficulty in selling CPO. Due to this, both price changes are found to be weakly correlated in strong contango, but correlation of both series is not entirely constant due to the convenience yield of holding physical inventories. As observed in Fig. 3, given a rising convenience yield from 0% to 1%, correlation of both series is being slightly increased. Assuming the convenience yield of 2.5%, such correlation is found to decrease subsequently. The further increase of convenience yield towards 10%, correlation of both price changes is found to rise. Given a rising convenience yield from 0% to 5%, correlation during the weak contango period is found to increase most rapidly

6. Conclusions Distinct from the theoretical exposition by Tilton et al. (2011) that the investor demand exists when spot and futures prices for a commodity are closely correlated during the strong contango period due to an inter-temporal arbitrage. This study extends their paper in two aspects. First, this study takes into account the variance of the increments of a random walk in detecting the weak-informational efficiency of the futures market as it is related 144

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

measure the degree of weak-form efficiency for the spot and futures markets, respectively. The degree of weak-form informational efficiency is further taken into account to ascertain whether it can be related to the degree of correlation between spot and futures price changes. Our results show that the investor demand on CPO differs from the findings of Tilton et al. (2011) and Gulley and Tilton (2014) in terms of two aspects: First, spot and futures prices are highly correlated during backwardation period. Second, the investor demand on the futures market is highly correlated with spot and futures prices during the backwardation period, but lesser during the weak contango period, and the least correlated during the strong contango period. Third, it is found that the degree of efficiency in the futures market is related to the degree of correlation between spot and futures price changes. Higher degrees of futures efficiency are linked to higher correlations between spot and futures markets and vice versa. As a commodity, CPO is susceptible to seasonality due to natural growing cycle. During the backwardation period, investors anticipate insufficient supply of CPO inventories for the short-term production. To meet the production, investors tend to over react the new arrival of information in the futures market by taking a long position. Their fast reaction to the current information vis-àvis prior information leads to a high speed of information arrival in reflecting the current price movement. The estimated variance ratio shows that CPO futures price changes are not seriously deviated from their average values. In this regard, the futures market during such period is considered to be the most efficient because its prices incorporate and reflect relevant information immediately. This provides a greater role of investor demand. Consequently, price changes in the futures market are highly correlated with price changes in the spot market. In contrast, during the strong contango period, investors adjust their decisions by buying CPO in the spot market and selling it in the futures market. However, due to seasonality and spoilage, the availability of physical fruit in the future is not always guaranteed. Consequently, price changes in the CPO futures market are less correlated with price changes in the spot market. There is a lesser role of investor demand in influencing the changes in CPO prices. The empirical results show that the prolonged existence of informational arrival in the futures market as its variance ratio is the highest, reducing the degree of efficiency in the futures market. Lastly, during the weak contango period, the degree of efficiency in the futures market and correlation between price changes in spot and futures markets is in between the above two sub-periods.

Table 4 Results for testing difference between two correlation coefficients of daily CPO spot and futures price changes, January 2000 - July 2016.

0% convenience yield: Strong contango vs Weak contango vs Backwardation vs 1% convenience yield: Strong contango vs Weak contango vs Backwardation vs 2.5% convenience yield: Strong contango vs Weak contango vs Backwardation vs 5% convenience yield: Strong contango vs Weak contango vs Backwardation vs 10% convenience yield: Strong contango vs Weak contango vs Backwardation vs

Standard error

Test statistic

Weak contango Backwardation Strong contango

0.0921 0.0602 0.0738

−2.0304** −4.9476*** 6.5713***

Weak contango Backwardation Strong contango

0.0909 0.0619 0.0707

−2.1862** −4.5586*** 6.7989***

Weak contango Backwardation Strong contango

0.09 0.0651 0.0666

−2.9431*** −3.4814*** 7.3801***

Weak contango Backwardation Strong contango

0.0954 0.0797 0.0577

−5.2959*** 0.6001 7.9291***

Weak contango Backwardation Strong contango

NA NA 0.0481

NA NA 8.6764***

Notes: Convenience yields during the period of strong contango are assumed to be 0%, 1%, 2.5%, 5% and 10%, respectively. Standard deviations are computed using σ(zi − zj ) = [1/(ni −3)] + [1/(nj −3)] . Test statistics are computed using [zi − zj ]/σ(zi − zj ) . ⎛ ⎞ The null hypothesis of no difference between two population correlations ⎜ρi = ρj, i ≠ j ⎟ ⎝ ⎠ is rejected if the test statistic value is greater than the upper bound critical value from a standard normal distribution or the test statistic value is lesser than the lower bound critical value from a standard normal distribution. *** denotes as the null hypothesis is rejected at the 1% significance level.

to the spot-futures relation. If the futures market is more efficient, its price changes tend to be highly correlated with the changes in the spot price. Second, this study follows the analysis by Gulley and Tilton (2014) who compare the degree of correlation during the strong contango, weak contango and backwardation periods in the copper futures. From a survey of the literature, there is a lack of literature that links the weak-form informational efficiency to spot-futures relation for a soft commodity. Hence, this study attempts to examine the investor demand in the case of Malaysian CPO futures based on the above authors’ theoretical exposition in the context of weakform market efficiency for CPO spot and futures markets. Following their line of research, a variance-ratio test is incorporated to

Fig. 3. Correlation coefficients between daily changes in CPO spot and futures prices across convenience yields of 0%, 1%, 2.5%, 5% and 10%, January 2000-July 2016.

145

Resources Policy 53 (2017) 135–146

Y.-H. Go, W.-Y. Lau

Liu, X., 2009. Testing market efficiency of crude palm oil futures to European participants. In 1Proceedings of the 13th EAAE Seminar A resilient European food industry and food chain in a challenging world, Chania, Crete, Greece. http:// ageconsearch.umn.edu/bitstream/58085/2/Liu. pdf (Accessed on June 22, 2011). Liu, X., Filler, G., Odening, M., 2013. Testing for speculative bubbles in agricultural commodity prices: a regime switching approach. Agric. Financ. Rev. 73 (1), 179–200. Lo, A.W., MacKinlay, A.C., 1988. Stock market prices do not follow random walks: evidence from a simple specification test. Rev. Financ. Stud. 1 (1), 41–66. Lo, A.W., MacKinlay, A.C., 1989. The size and power ofthe variance ratio test in finite samples: a Monte Carlo investigation. J. Econ. 40 (2), 203–238. Lo, A.W., MacKinlay, A.C., 2002. A Non-Random Walk Down Wall Street. Princeton University Press, Princeton. Loayza, N.V., Ranciere, R., Servén, L., Ventura, J., 2007. Macroeconomic volatility and welfare in developing countries: an introduction. World Bank Econ. Rev. 21 (3), 343–357. Lucas, R.E., Jr, 1978. Asset prices in an exchange economy. Econometrica 46 (6), 1429–1445. Lucey, B.M., O'Connor, F.A., 2013. Do bubbles occur in the gold price? An investigation of gold lease rates and Markov Switching models. Borsa Istanb. Rev. 13 (3), 53–63. Ma, C.K., Mercer, J.M., Walker, M.A., 1992. Rolling over futures contracts: a note. J. Futures Mark. 12 (2), 203–217. Mahalik, M.K., Acharya, D., Babu, M.S., 2014. Price discovery and volatility spillovers in futures and spot commodity markets: some Indian evidence. J. Adv. Manag. Res. 11 (2), 211–226. Martin, L., Garcia, P., 1981. The price-forecasting performance of futures markets for live cattle and hogs: a disaggregated analysis. Am. J. Agric. Econ. 63 (2), 209–215. McKenzie, A.M., Holt, M.T., 2002. Market efficiency in agricultural futures markets. Appl. Econ. 34 (12), 1519–1532. Miffre, J., Brooks, C., 2013. Do long-short speculators destabilize commodity futures markets? Int. Rev. Financ. Anal. 30, 230–240. Ministry of Agricultural and Agro-Based Malaysia, 2014. 〈http://www.kada.gov.my/en/ web/guest/dasar-agro-makanan-negara〉, Accessed on 10 August 2015. Poterba, J.M., Summers, L.H., 1988. Mean reversion in stock prices: evidence and implications. J. Financ. Econ. 22 (1), 27–59. Oellermann, C.M., Brorsen, B.W., Farris, P.L., 1989. Price discovery for feeder cattle. J. Futures Mark. 9 (2), 113–121. Östensson, O., 2011. Comment: investor demand and spot commodity prices. Resour. Policy 36 (4), 372–374. Östensson, O., 2012. The 2008 commodity price boom: did speculation play a role? Miner. Econ. 25 (1), 17–28. Pindyck, R., 2001. The dynamics of commodity spot and futures market: a primer. Energy J. 22 (3), 1–29. Sanders, D.R., Irwin, S.H., 2010. A speculative bubble in commodity futures prices? Cross‐sectional evidence. Agric. Econ. 41 (1), 25–32. Sanders, D.R., Irwin, S.H., 2011a. New evidence on the impact of index funds in US grain futures markets. Canadian. J. Agric. Econ. 59 (4), 519–532. Sanders, D.R., Irwin, S.H., 2011b. The impact of index funds in commodity futures markets: a systems approach. J. Altern. Invest. 14 (1), 40–49. Sanders, D.R., Irwin, S.H., Merrin, R.P., 2010. The adequacy of speculation in agricultural futures markets: too much of a good thing? Appl. Econ. Perspect. Policy 32 (1), 77–94. Schroeder, T.C., Goodwin, B.K., 1991. Price discovery and cointegration for live hogs. J. Futures Mark. 11 (6), 685–696. Silvapulle, P., Moosa, I.A., 1999. The relationship between spot and futures price: evidence from the crude oil market. J. Futures Mark. 19 (2), 175–193. Stevens, J., de Lamirande, P., 2014. Testing the efficiency of the futures market for crude oil in the presence of a structural break. Appl. Econ. 46 (33), 4053–4059. Tilton, J.E., Humphreys, D., Radetzki, M., 2011. Investor demand and spot commodity prices. Resour. Policy 36 (3), 187–195. Tomek, W.G., Gray, R.W., 1970. Temporal relationships among prices on commodity futures markets: their allocative and stabilizing roles. Am. J. Agric. Econ. 52 (3), 372–380. Westerlund, J., Narayan, P., 2013. Testing the efficient market hypothesis in conditionally heteroskedastic futures markets. J. Futures Mark. 33 (11), 1024–1045. Will, M.G., Prehn, S., Pies, I., Glauben, T., 2015. Is financial speculation with agricultural commodities harmful or helpful? A literature review of empirical research. J. Altern. Invest. 18 (3), 84–102. Working, H., 1953. Futures trading and hedging. Am. Econ. Rev. 43 (3), 314–343. Wright, J.H., 2000. Alternative variance-ratio tests using ranks and signs. J. Bus. Econ. Stat. 18 (1), 1–9. Zhang, Y.J., 2013. Speculative trading and WTI crude oil futures price movement: an empirical analysis. Appl. Energy 107, 394–402.

References Adämmer, P., Bohl, M.T., 2015. Speculative bubbles in agricultural prices. Q. Rev. Econ. Financ. 55, 67–76. Bohl, M.T., Javed, F., Stephan, P.M., 2013. Do commodity index traders destabilize agricultural futures prices? Appl. Econ. Q. 59 (2), 125–148. Bos, J.W., Van der Molen, M., 2012. A bitter brew? Futures speculation and commodity prices. METEOR Research Memorandum (RM/12/044). METEOR, Maastricht University School of Business and Economics, Maastricht, 1–39. Brooks, C., Prokopczuk, M., Wu, Y., 2015. Booms and busts in commodity markets: bubbles or fundamentals? J. Futures Mark. 35 (10), 916–938. Capelle-Blancard, G., Coulibaly, D., 2011. Index trading and agricultural commodity prices: a panel Granger causality analysis. Int. Econ. 126, 51–71. Caporale, G.M., Ciferri, D., Girardi, A., 2014. Time‐varying spot and futures oil price dynamics. Scott. J. Polit. Econ. 61 (1), 78–97. Chang, C.P., Lee, C.C., 2015. Do oil spot and futures prices move together? Energy Econ. 50, 379–390. Charles, A., Darné, O., 2009. The efficiency of the crude oil markets: evidence from variance ratio tests. Energy Policy 37 (11), 4267–4272. Charles, A., Darné, O., Kim, J.H., 2015. Will precious metals shine? A market efficiency perspective. Int. Rev. Financ. Anal. 41, 284–291. Emekter, R., Jirasakuldech, B., Went, P., 2012. Rational speculative bubbles and commodities markets: application of duration dependence test. Appl. Financ. Econ. 22 (7), 581–596. Etienne, X.L., Irwin, S.H., Garcia, P., 2014. Bubbles in food commodity markets: four decades of evidence. J. Int. Money Financ. 42, 129–155. Fama, E., 1965. The behavior of stock market prices. J. Bus. 38 (1), 34–105. Fama, E., 1970. Efficient capital markets: a review of theory and empirical work. J. Financ. 25 (1), 387–417. Fama, E.F., 1991. Efficient capital markets: II. J. Financ. 46 (5), 1575–1617. Fattouh, B., Kilian, L., Mahadeva, L., 2013. The role of speculation in oil markets: what have we learned so far? Energy J. 34 (3), 7–33. Fernandez, V., 2015. Spot and futures markets linkages: does contango differ from backwardation? J. Futures Mark., 1–22. Gain Report, 2014. Malaysian Biofuels Annual 2014, Report Number: MY4011. Garbade, K.D., Silber, W.L., 1983. Futures contracts on commodities with multiple varieties: an analysis of premiums and discounts. J. Bus. 56 (3), 249–272. Gilbert, C.L., 2010. How to understand high food prices. J. Agric. Econ. 61 (2), 398–425. Gulley, A., Tilton, J.E., 2014. The relationship between spot and futures prices: an empirical analysis. Resour. Policy 41, 109–112. Guilleminot, B., Ohana, J.J., Ohana, S., 2014. The interaction of speculators and index investors in agricultural derivatives markets. Agric. Econ. 45 (6), 767–792. Haase, M., Zimmermann, Y.S., Zimmermann, H., 2016. The impact of speculation on commodity futures markets: a review of the findings of 100 empirical studies. J. Commod. Mark.. Huchet, N., Fam, P.G., 2015. The role of speculation in international futures markets on commodity prices. Res. Int. Bus. Financ. 37, 49–65. Inoue, T., Hamori, S., 2014. Market efficiency of commodity futures in India. Appl. Econ. Lett. 21 (8), 522–527. Irwin, S.H., Sanders, D.R., Merrin, R.P., 2009. Devil or angel? The role of speculation in the recent commodity price boom (and bust). J. Agric. Appl. Econ. 41 (2), 377–391. Irwin, S.H., Sanders, D.R., 2011. Index funds, financialization, and commodity futures markets. Appl. Econ. Perspect. Policy 33, 1–31. Irwin, S.H., Sanders, D.R., 2012a. Testing the Masters Hypothesis in commodity futures markets. Energy Econ. 34 (1), 256–269. Irwin, S.H., Sanders, D.R., 2012b. Financialization and structural change in commodity futures markets. J. Agric. Appl. Econ. 44 (3), 371–396. Kaufmann, R.K., Ullman, B., 2009. Oil prices, speculation, and fundamentals: interpreting causal relations among spot and futures prices. Energy Econ. 31 (4), 550–558. Keynes, J.M., 1930. A Treatise on Money: In 2 Volumes. Macmillan & Company, 142–144. Kim, A., 2015. Does futures speculation destabilize commodity markets? J. Futures Mark. 35 (8), 696–714. Kocagil, A.E., 1997. Does futures speculation stabilize spot prices? Evidence from metals markets. Appl. Financ. Econ. 7 (1), 115–125. Kofi, T.A., 1973. A framework for comparing the efficiency of futures markets. Am. J. Agric. Econ. 55 (4), 584–594. Kristoufek, L., Vosvrda, M., 2014. Commodity futures and market efficiency. Energy Econ. 42, 50–57. LeRoy, S.F., 1973. Risk aversion and the martingale property of stock prices. Int. Econ. Rev. 14 (2), 436–446. Leuthold, R.M., 1974. The price performance on the futures market of a nonstorable commodity: live beef cattle. Am. J. Agric. Econ. 56 (2), 271–279.

146