UFDC Home  Search all Groups  UF Institutional Repository  UF Institutional Repository  UF Theses & Dissertations   Help 
Material Information
Subjects
Notes
Record Information

Full Text 
THE A TIME CRUDE OIL SERIES SPOT ANALYSIS OF AND FUTURES MARKETS JING QUAN A DISSERTAT OF PARTIAL FOR THI ION PRESENTED TO THE GRADUATE S( THE UNIVERSITY OF FLORIDA IN FULFILLMENT OF THE REQUIREMENTS E DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1990 SCHOOL dedicate this dissertation to the faculty the Department of Economics at the University of Florida. ACKNOWLEDGEMENTS I deeply appreciate the encouragement, guidance, support and patience from Professor G.S. Maddala, Professor Toda, Professor L. Cheng and Professor J.S. Shonkwiler who served as my supervisory committee. I am also grateful to Mr . R. Duncan and Mr. Imran offering me a chance find this topic while was working World Bank. especially thank my parents and all other relatives in China for their continuous support understanding. TABLE OF CONTENTS Page ACKNOWLEDGEMENTS.......................... ............. ii ABSTRACT..................... ... .............. ........ vi CHAPTERS I INTRODUCTION............... ... ....... .... ..... 1 II SPOT PRICE AND FUTURES PRICES: THE........... 10 PRICE DISCOVERY OF CRUDE OIL FUTURES PRICES Introduction.................... .. ... ... 10 Data Analysis................. .............. 16 Using Crude Oil Futures of the Spot Price Predictors,...... Using Futures Prices to Update Data.. Combining Results and Futures Prices. * t. . * . Relationship Tests............... Integration Test............ Cointegration Test......... Causality Tests.................. Garbade and Silber Approach. Granger Causality Test..... Error Correction Model........... Summary.. ............ ......... .... 0...... * . ...... *. C ...... ...... *...... *......a * C ...... a..... * S . ..... ..* .. III FUTURES PRICES AND OPEC OIL SUPPLY.......... Introduction................ Data Analysis............... Simple Regressions.......... Specification Tests......... Integration Test....... Cointegration Test.... Garbade and Silber Approach. Granger Causality Test...... Summary.. ................... ...DO.. ...... ...... ...... ...... ...0.. .....B.Q. *.*.*...... * a . ......... ....O.... ...*.c.. *...*c..... *.....ca.s SPOT PRICES AND OPEC OIL SUPPLY.............. 108 Introduction............. Data Analysis........... Simple Regressions....... Specification Tests...... Integration Test.... Cointegration Test. The "SelfAdaptive" Model Summary.. ................ ....00"....0.0...... O..." .. .. .......... * e a a e *at*tt Ste.* * ... O... ...." ..... .. * t.. te... ...t e * C S *S et t etee ** t *5* * C S S C S S 0 C S S S S S C .".t"... t.......*. "....... ... .. .Q.... "te... .... C. .. ..e. *. S ..C C APPENDIX TO CHAPTER III........................... 140 BIBLIOGRAPHY. .. .. . . .. . . . 143 BIOGRAPHICAL SKETCH............ ..... .... ... .... 147 CONCLUSIONS. ......... .. ...... .. .... 138 Abstract of Presented in Partial the Degree of the Graduate School Fulfillment A TIME SERIES ANALYSIS THE CRUDE OIL SPOT AND FUTURES MARKETS Jing Quan December, 1990 Chairman: Major . G. Department Maddala : Economics This dissertation studies spot futures markets for crude oil. Its main objective is to study the relationship between these futures two m market contribution markets, this particular working dissertation the role crude market. better understanding crude markets. order investigate relationship between the two markets, two stage testing procedure is proposed in this study. First , the existence longrun relationship tested using cointegration Garbade tests. Silber Second, approach Granger are causality tests determine direction of causality. The first relationship investigated this study between crude oil spot and futures prices. This known "price discovery" role futures prices. Different assumptions about the futures market lead different Dissertation University Requirements of of Florida Doctor of Philosophy spot provide applied that price. future The prices market not efficiency provide information hypothesis, the other spot hand, assumes opposite conclusion. These two hypotheses are tested in this dissertation. It is found that spot price leads futures prices instead the futures price providing information on the spot price. Therefore, the hypothesis that futures prices have a price discovery role is rejected. Two additional relationships studied are those between OPEC supply and futures prices that between there same supply are cointegration and spot prices. relationships The hypotheses between that the variables are rejected both cases. case of OPEC supply spot prices, a dummy variables found that prices "selfadaptive" model with supply interruption is introduced to study the price behavior follow an adaptive process, . It that previous price information offers powerful influence current price. CHAPTER I INTRODUCTION The dramatic developments the world markets over past two decades have had farreaching effects world economy . Two price shocks and one price collapse exert powerful influence the global economy and international relations. Ramifications can seen world recessions , in the third world debt problems, attempts to resolve Arab Israeli differences in the expansion the U.S . military role Persian Gulf , and virulent boom bust cycles producing country the major producing regions in the Unit ed States. The relationship between and a country s economy has proved very significant. Needl ess say, the most significant event story the market over last twenty years was emerge nce of the Organization of Petroleum Exporting Country (OPEC), which has become the major force in determining the price of crude and the production policies of its members. addition , it revolutionized the contractual relationship with the international companies. companies petroleum sole market force (exploration, activities production, world transportation, refining, and marketing) has changed. The domestic pricing and energy resource development policies oilimporting countries also as well changed. their Massive energy transfers consumption wealth patterns from have major industrial countries OPEC member countries have occurred. Concern over balanceofpayment deficits recycling of oil money back to the economies of the industrial countries emerging has ix between itensified. New developed economic relationships industrial countries, are OPEC member countries, other less developed countries. These new relationships have been accompanied new era international diplomacy Obviously , energy civilization rests, is one of the foundations on which our and one for which there is no substitute. connection between consumption primary energy carriers and economic development is unmistakable. At present petroleum world's main energy carrier. Forecasts future crude prices, crude availability, OPEC's stability , and its price and production strategies all affect decisions developing oilimporting alternative countries costlier with energy respect resources. appreciate the relationship between oil and the economy, and world economy energy a good economic place to value begin. and volatility At today s prices, the the value primary energy production amounts about percent of the world economic output with accounting almost worth half three the times total the value. That value makes food energy grain production production (rice , wheat, and coarse grains). similar picture emerges world trade. total world exporters exports alone are a account most five about times 6 percent large value food grain exports. Even more important volatility value. 1970, when was still comparatively cheap and consumption growing rapidly, primary energy output was equal percent world GNP less than percent. In 1981, when price of oil reached peak, energy production constituted 10 percent, and percent, space world decade output. Swings necessarily set this motion magnitude Svast st ructural changes throughout the world economy. A second macroeconomic variable associated with inflation. inflation course, ratethe there are many monetary policy factors and in determining willingness trade inflation against unemployment However, one also has admit that the increase in oil price one the most important factors. Since an input in most industrial processes, heating, would difficult argue that 850 percent increase in the price of oil between October 1973 and January 1981 was maj or factor raising general price level. It was not only major, but crucial. Prior to the first oil shock, world prices were increasing at a rate of percent annually , and this was regarded as an aberration that would soon brought under control. Today doubledigit inflation is common in many countries , and it is not expected to be brought down to rates below the 1973 levels in the near future. Take 1979 1974, United States' example. prices Followin United economy a the c States during the )ilprice increased period 1973 shock enough restore quasirents capital goods their 19721973 levels. With only marginal increase in money wages, real wages fell enough pay increased bill. Moreover, imports increased very large amount, which removed barriers expansion production that might have resulted happened from the that physical inflation unavailability the energy. depreciation It also dollar prevented real price from rising face some minor upward adjustments the money price of oil. 1977 , the United States had more less recovered from first oilprice shock. Both employment and output were 19721973 was that prices moved faster after 1973. The latest example United States can also confirm relationship between price and inflation. Consumer price index in the United States rose 0.4 percent in December 1989, for the year, it reached 4.6 percent, the highest since 1981. The Federal Reserve seems much concerned about fighting upcoming inflation. One of the biggest reasons behind the rise in the consumer price index was higher fuel prices resulting from December's freeze. The impact for January 1990 should be even more evident. The fear of inflation strongly influences the performance of the stock market. For instance, The biggest selloff in the New York stock market since the October 1989 "minicrash" stock market Friday, January 1990. The Jones average of 30 industrial tumbled 71.46 points to 2,689.21, largest loss since it plunged 190.58 in October 1989. One major causes were new inflation worries. Department of Labor reported that the producer price finished goods rose 0.7 percent in December index of exceeding advance estimates that measure inflationary pressures. index finished 1989 with increase percent, biggest in eight years. Finally , it should be mentioned that the oilprice shock initiated by OPEC helps some industrial countriesthose which uranium reserves and has contributed to England's ability to generate foreign exchange. From the above discussion, one can easily appreciate the importance of the world oil market. However since there are kinds markets oil, namely , the spot market futures market, in order to understand the oil markets, it is necessary to study both these markets, the relationships between them. The spot, cash , market need not market institutional sense word, simply arrangement between buyers and sellers that calls deliverythough perhaps consumptionof commodity the immediate future. case oil, Rotterdam Spot Market literally become household term; but although true that this district Holland blessed with extensive brokerage and oilstorage facilities, it bears only a passing resemblance to the conventional image of a market, since there physical locale, trading takes place means telecommunications. Similarly, the spot or cash price pertains immediate transfer of ownership of a commodity. assess course spot market helpful during the 1981, review 1970s, declined its the steadily recent price of history. After rising peaked at thereafter sharply a barrel barrel 1985, early 1980s. Indeed, after adjustment inflation, October 1987 price was below the price reached after the first shock 197374. The futures market is an arrangement that features paper transactions. Physical delivery the commodity occurs only a small minority cases. Strictly speaking, futures contract is a forward contract in that these contracts almost always refer particular month delivery; same time, however , a futures market is so organized that sales or purchases can be offset, the deliveries are unnecessary. In addition, futures contracts are bought and sold through an exchange, impersonally, with validity contracts guaranteed by the exchange. function smoothly, In order large numbers for this traders type of are re market quired in particular speculators, play major role generating flow contracts that permit other types market actors (such sellers purchasers physical commodities like oil) to avoid price risk. It should be made clear that these speculators, who are uninterested physical commodity, but have distinct ideas about the price of the future, are buying or selling contracts with intention of making an offsetting sale or purchase later for example, the offsetting purchase price less than that the original sale, the speculators will make a profit. . If, demand, weather influence, on. this thesis, three important relationships will studied: that between futures prices and spot pricesto test the role of crude oil futures prices in price discovery; that between futures prices OPEC supplyto test role OPEC supply formation expectation price the futures; finally, that between the spot prices and OPEC supplyto test influences sudden supply shocks the prices. The relationship between the futures price and spot price deliverable good maturity futures contract is one of the longstanding research issues. It is an important question, since resolution important implications futures market speculation, efficiency futures market , and effective implementation futures market hedges. The relationship between futures price supply been paid much attention before. However, important. Undoubtedly, futures markets , traders speculate on the supply available in the near future and determine how much they are willingtopay commodity. The market has long been sellers' market. Traders ' expectation of future prices depends heavily on the their expectations of future production. It is observed that once after annual meeting of OPEC member countries the new quota, crude futures prices will applied the market. All "oil crises" the past were due supply shocks. The spot pri ces respond almost immediately the supply . The opposite direction true. When price drops below certain point, OPEC will interrupt the price slide reducing supply and attempt to make the price rebounce. The res t of the thesi is organized follows . Chapter II studies the relationship between the futures price spot price. The relationship between the futures price OPEC supply is investigated Chapter . Chapter examines relationship between spot price and OPEC supply. And finally the conclusions are summarized in Chapter CHAPTER II SPOT PRICES AND FUTURES PRICES: THE PRICE DISCOVERY OF CRUDE OIL FUTURES PRICES Introduction Futures prices reflect opinions producers, consumers speculators about prices different commodities the markets at a later date. To be of value to hedgers, the futures price must respond quickly and accurately relevant new information. ability futures market process information been investigated many researchers. following two hypotheses have been the focus their study: random walk and market efficiency. The random walk hypothesis states that the most accurate forecast next period price today's price. There then correlation between price changes different days, information past prices useful forecasting future prices. Working (1934) and Kendall (1953) recognized that behavior prices very similar random walk. This suggests that speculators cannot use information present past prices trade profitably. This conclusion leads one definitions efficient market hypothesis (Jensen, 1978). The socalled linear process has been assumed in nearly all the tests. Many r early 1970s papers by Stevenson and Bear (1970) Leuthold (1972) Cargill Rausser (1975), however claimed that futures prices did exactly follow random walk, which implied that some information took more than one day be absorbed correctly into the prices. The efficient market hypothesis can be true when the random walk hypothesis invalid, provided that trading costs prevent complete exploitation of the information reflected relatively slowly in the prices. An efficient market processes information fairly, which means that information can equally easily accessed by different users. Efficiency ensures that a trader paying commission should consider the present prices to be the only relevant information in the history prices. Hence, corresponding to these two different assumptions, there are controversial results discussed in the literature in terms of the role of the futures markets. Of course, there little dispute about role the futures markets providing mechanism risk sharing. However, many conflicting views remain about their role as a mechanism predicting spot prices, price discovery, other words, studied efficiency the futures performance of the markets. live cattle Helmuth futures (1981) contracts and drew some rather strong conclusions: "Based on the theory efficient markets, existence the phenomenon predictable downward bias. Such a conclusion means that...the futures market fulfilling economic purpose providing hedging vehicle producers. (page 356) Helmuth' s conclusion methodology used study have been criticized later many researchers. their studies , Kolb and Gay (1983) wrote:. "the conclusion supported by this new methodology strongly opposes Helmuth' conclusion. In fact, it appears that futures contracts have exhibited exemplary price work behavior over the period examined. " (page Since issues market efficiency hedging effectiveness, and price discovery are interrelated sense that they are tested similar identical statistical techniques, this study designed distinguish one from other . Rather, puts the emphasis adequacy futures prices role price discovery. Price discovery process which new information affecting asset values becomes reflected in market prices. refers use futures prices pricing cash market transaction. There are three points associated with term "price discovery. " First, order "efficient" price predicting mechanism, futures markets do not have to be always right; it is enough that they are just right on average. That one of the definitions of efficiency is , in earn any more than fair competitive premium. Spectacular shortterm losses and gains, while surely allowed, would tend cancel each other over the long run. Second, other predictors could also be right on average, which brings up an interesting point. two predictors are both unbiased, then usual way to decide which one more "efficient" choose the one with the smaller variance. However, in choosing between price forecasts, a better rule might be to choose the one whose variance closer that price being predicted. one predictors futures market (whose efficiency is generally not questioned) the other econometric model with similar variance spot price series, then question arises whether forecast from one method, , using the futures price, could significantly improve the performance of the other predictor. Third based not , hedgers only (or speculators) on what are entering a they think will happen, futures market but also based what they fear will happen. hedger may think that prices will fall, but will fear such an event to the point of buying "insurance" 1.e., hedging futures contracts. they always did this, then enough speculator activity would push price "efficiently forecasted" level. interim shortterm bias reflecting this insurance premium would exist. stable longrun relationship between crude spot futures DOE/EIA prices (1986) issued a direction causality report about using (leadlag). Petroleum futures prices as the predictors of cash prices. The product studied the report distillate heating oil. Like many other studies on the theory of price discovery of futures prices , it employed standard econometric methods. Moreover, use any tests investigate longrun relationships between spot prices futures prices. However, relationship is critical in such studies. Before we get into practice test whether the futures prices can provide information forming spot prices , we should first make sure that a longrun relationship exists between two price series. If there is no longrun relationship at all between them, that the two price series do not follow same moving pattern over time, there then no point talking about how the futures prices can provide information about the formation spot prices. Ignoring first step can lead to spurious results. If futures price series and spot price series share same intertemporal characteristics, other words, they are not integrated same order, procedure that runs the regressions between these two series are incorrect due to the violation of the assumptions underlying the simple regression. Therefore it stable longrun relationship between the spot and futures prices. The second part hinges on the results from the first one. there no strong statistical evidence to show existence of such a relationship, the investigation comes to an end. If the results from the first part ascertain that such relationship well established, then the direction causality role (leadlag) futures will prices. tested this examine dissertation the discovery examine price discovery role futures prices the crude futures market.by following these For the first part, steps. several different methods, including the conventional methods used DOE/EIA report and some new techniques, exists stable will implemented longrun relationship test betwe whether en crude there oil futures prices spot price. Provided that relationship between them is ascertained, then I shall attempt to determine whether spot price leads futures prices vice versa. From the results presented, two points will become clear. First, there exists stable longrun relationship between onemonthahead crude oil futures price and the spot price. However , no stable longrun relationships can be found between the spot price and futures prices for more than one month ahead. Second, though a stable longrun relationship exists between spot price and onemonthahead futures The remainder of this chapter is organized as follows. first provide a brief overview of the data used in this study, indicating a relationship between the futures and cash markets "raw data" basis. Second present the conventional methods used DOE/EIA report (1986), namely, simple regression and combined regression or simulation. Even though results obtained from them are only suggestive can still get some intuitive perspective the movement price series. Third, the methods used to test the relationship between prices are described. Among them are integration test and the cointegration test. In order to test direction causality, the Garbade and Silber approach and the Granger causality test are implemented in the next two sections. Following this, we apply an error correction model investigate the dynamics long run relationship between spot price futures price. The chapter concludes with assessment usefulness futures prices forecasting the spot price. Data Analysis Futures trading has exploded since 1970 . Both the number of futures markets and the participation has increased rapidly since then. energy futures market, contracts petroleum products for delivery to 12 months in the future since March 1983. In less than ten years, this futures market has become the keystone of a new international pricing system that, time being least, has ended the price domination of the Organization of Petroleum Exchange Countries cartel. Due to many uncertain factors, such as the instability prices due to political influence , natural disasters accidents, wars, and intensified competition from other forms ideal substitutable tool energy risksharing. crude Based futures provide this, thus reasonable to say that the futures prices can be also serving as good predictors for the spot prices. The following preliminary figures and tables suggest that there close relationship between the spot and futures prices. prices Figure series 21 since shows 1984. the Two monthly things values are apparent the from spot it. First, crude oil prices have been declining since late1985; Second, this decline has been irregular. The fact that spot prices followed negative trend over this period important because it helps to explain why the futures prices follow a negative trend as well. The longer the length of the futures contracts, time smaller the futures prices were t+1+nmonthahead futures (that price generally less than the t+1monthahead price) which indicates that over time futures market has been more pessimistic J0 Figure . Crude r . *0t I* Oil Spot Prices iC c , ^^ futures prices would exceed spot prices by a small amount. reason is that, in order for the speculators to participate in futures market , they will expect to receive a premium for bearing a portion market risk due the fluctuation prices. Also, since these are nominal prices, would expected that futures prices should experience some upward pressure because of the inflation rate However, both of these upward biases can overstated because they could offset least other factors. First, cost storage, upper together bound with futures current price, spot since price, there acts would incentive to buy futures contracts which exceed the sum of the spot price the storage cost. Second, there may be strong expectations in the marketplace that prices are going to fall. This was case from 1984 through 1989. The spot pri ces are plotted against , 3 , 6 , and 9monthahead futures prices Figure through 25, respectively. One problem concerning the futures prices is the choice of the daily futures prices in a month which should be used to represent the future price for that month. If futures markets are efficient , then it should be true that all the contracts prices (1monthahead, 2monthahead, etc.) on any given day contain information known about the market that time. Averaging prices throughout the month which contains time series futures prices independent variable, he needs only to choose the most recent price current observation. Following this argument, only the futures prices available one reporting per month need to be used to represent a monthly value in a time series. For most of this study, the price reported on the 15th each month will used the 15th day is a trading not, then the closest adjacent day from 15th which is a trading day will be used. Picking midmonth values less prone to bias than first or endofmonth ones since large institutional investors often use computer trading programs close their position certain months, giving rise to extreme fluctuations prices of futures contracts on these days. Table 21 provides the mean, standard deviation, maximum, and minimum for the following prices: spot price and futures prices , 6 , and 9monthahead contracts. monthahead contracted is not used because it is very closely correlated with the 1month and 3monthahead ones. The 4 , 5 , and 8 month contracts are examined either because they are found to be closely correlated with month contracts. In this analysis, the actual series used are lagged number months futures contracts. instance, December 1985 1mo nth futures price that . If , 6 contract observed in November is used for predicting (or being correlated with) cash prices one month later. Cross correlations for these series are shown in Table 22. As shown, the mean of crude oil spot price from January 1984 through July 1989 $19.33 barrel, while mean futures prices , 3 , 6 and 9month ahead are slightly larger than mean spot price, expected. The table also shows that the correlation between these series falls off from 0.95 1month ahead , progressively declining to 0 9month clearly that ahead. the ] Figures 1monthahead through futures 2.5 a price Ilso demonstrate closely mimics actual spot price, while correlation diminishes gradually 9month graphs. graphs also show more. First late1985, futures prices falsely months signaled a large drop ahead in the spot price, futures prices are, and the more larger the gap between futures spot ces. Second, futures prices repeated this false signal late 1988 , though smaller degree. Third , the futures prices did correct such wrong signals after certain period time, although the more months ahead futures prices took longer time to adjust. Another preliminary way to view the futures markets is to calculate ratio of futures prices divided by spot prices the same month. Under ideal conditions, r. ) I 1 : . I. i .39 14 I ~1 A\ 7 A 193a I'.ont hOut Figure Oil Spot Futures Prices S. I.. SCot 1i87 ';aa 4 7 9 I  i .7'^ ^ */I 7  7 3.ontnOut Futures  oot Figure Oil Spot Futures Prices 1994 19i8 1997 1988 1985 32 T 30 4 E ; * ? ^ f I N ~ 7 ' V 3 1S87 rfutures Figure Oil Spot Futures rices 22 20 1s  ^ I 144 12 13T a *^ "TTf r i 1984 1995  :oot I 92 N .4, 1, '1 * *\ ~  '7 'i <7! 1987  Soot Futures Figure Oil Spot Futures Prices 1984 1985 19g6 ?^si 7 9*.tantr.Out standard deviations for these calculations. Not surprisingly, ratios 1monthahead and 3monthahead are closest , and as expected, ratio of 1monthahead futures small est standard deviation. Figures 26 through 29 show the ratios of futures prices spot prices corresponding time periods. obvious that the more months ahead the futures prices are, larger the fluctuation is. There is a slightly downward shift in the graphs indicating that over time the futures market has been pessimistic about crude prices than actually warranted. Spot prices were falling, the futures prices expected them fall even more during this period. Usinq Crude Oil Futures Prices Predictors of The Spot Price this section, results of a number of alternative methods test relationship between crude oil futures prices spot price are presented analyzed. The specifications include simple regression update current data series, and a simulation which uses formulas to "combine" regression results and known futures prices. These two methods are used EIA/DOE report (1986). The advantage these methods simplicity , even though usefulness results limited. Despite their shortcomings, these 1 4 *i 4 7 methods halo travel 1 tho potential Orosoective A.. A L Lt t&1 L. .1 V t~. JLt t L^. A. I/^ t A. A .. __ h t.. V .. t' ..f La .*_ .5.. hJ ._ A V F  %.F I.. .JL y 1 LtrnthCut Futtres 2 1.9  1.9  1.7  1.6  1.5  1,4  1.3  1,2 1 1 1  0 9 0.8  07.1 0.6 1 03 4 1984 1985 198e6 1987 r \ \, V f / 1 1is8 Figure Price Ratio: Futures/Spot 939 3AtMot nOut Futures t12  1  o09 0 a  1984 1985 1986 1987 1988 '= Figure Price Ratio Futures/Spot 6MLcntnOut ;utures 1985 1988 6987 1988 1989 Figure Price Ratio: Futures/Spot 1984 0.9  0,8  0.4  Figure Ratio Futures/Spot negative, suggesting intuitively that these two variables are related at all, then there is no need to go any further. Usina Futures Prices to Update Current Data The first method that the futures data could be used to update data series those cases when official data current calendar month few months. To determine futures prices could be used this manner four estimations are attempted. The crude oil prices reported by IECCM of World Bank are regressed the New York Mercantile Exchange futures prices for crude oil using 1month, 3month, 6month and 9month ahead contract price. Thus to fill in missing data points the following equations are estimated: = a + 3 FPt 21) FP,6 83 FPt9 (24) where = crude oil FPt,J price at time t; = jmonthahead futures prices for delivery at time = regression coefficients. futures price delivery the current month. Tabli gives the results of the estimation . When equations are estimated and corrected expected autocorrelation , the 6 month ahead futures crude prices are seen significant predictors spot price prices . Thus can appears used that the fill 1 and the 6 month missing data ahead contract values; When the step ahead pri ces are used, the estimates are statistically significant . Two interesting points here are worth mentioning. Firs , while the 3 month ahead futures price no significant effect on the spot price 6month ahead futures price does . One possible explanation is that there are coincidentally period. two Inflationary competing forces expectations work and risk over time premium would both tend increase distant futures ces , while long run negative trend spot pri ces would work ress futures prices . If there been a strong upward trend, then the 3month ahead futures price might have had a significant role. Second, while both 6month ahead futures prices are Thi significant, the coeffi may the cents have opposite timing people signs however expectation formation. is obvious that takes a while before people absorb new information. When crude price ses unexpe ctedly month, people may expect to drop late. Combining Model Results and Futures Prices the above discussion futures prices were viewed potential inputs energy models. Along with other energy related data , decision makers would also have knowledge about futures prices when they make price forecast . Now, however, futures markets will viewed competing forecasting mechanism . The question we are asking whether the futures prices obtained from regression model add anything to the prices combined attempt forecast answer based the question estimated studying results futures prices. Two estimated and . In crude both ces cases are reported used here: below, those 1month from ahead futures prices make a significant contribution to the combined forecast, but the results with more than 1month ahead futures prices are ambiguous. Simulations from are used to generate estimation results . First, two equations are estimated, each form: t = a + f SMPt where the price estimated from one the model and 23). 34 associated forecast adds anything to the combined forecast in a statistically significant manner. The test of significance for the coefficient combined is also forecasts. a test the e the unbiasedness estimate statistically different from zero, then no bias is indicated. In contrast, it is statistically different from zero, then either an upward or a downward bias is present. The results are given in Table 25. From tables, we can see that 1monthahead futures prices consistently make statistically significant contributions the combined process (Case Case However, the 1monthahead futures prices generate a bias when they are combined with (23) (Case The bias combinations would have been serious problem there had not been be slightly "risk consistent trend, higher than premium" since futures spot prices due Interestingly, the prices should presence of 3monthahead futures prices turn out to be significant in Case which combines Strangely enough 9monthahead futures prices are also significant in the model which combines with (23) (Case However, since sign bias not consistent across two models when more distant futures prices were used. Therefore, again that the general weakness predictive ability of futures prices more than 1 month out is Relationship Tests The results the previous section assume stationarity spot futures prices. Even though we may get some intuitive idea of the price performance, the basic assumptions underlying the simple regression are no longer valid, prices are nonstationary when the . The results obtained in the sections above are useful only if both spot prices and futures prices are stationary levels. However, this not case, the results of the of the simple regression would be spurious misleading. Therefore, necessary test stationarity and the nature of the stochastic movement of the prices. Hence, will employ the integration cointegration tests below to take these aspects into account. Integration Test Many macroeconomic variables are often found nonstationary levels. empirical fact that these variables appear integrated order (stationary the first differences), or to possess unit roots. Integration tests are designed help evaluate the nature nonstationarity of economic variables. In other words, whether they are stationary in levels or in differences of some order. Suppose there are two economic variables which are nonstationary question ask whether they move of two economic variables. In the example here, we try to link the crude oil spot price Pt and the futures price and test the relationship between them. stable relationship exists, then it may be possible to make quantitative inferences about crude price from the observations futures prices. will implement integration test examine stationarity of the crude oil spot price and futures price. form must longrun share equilibrium same relationship, intertemporal the characteristics. variables That they must integrated same order. The dynamic property single series can be described by how often needs difference achieve timeinvariant linear properties and provide a stationary process. A series that has at least invariant mean and variance and whose autocorrelation "short memory" called I(0) , "integrated order zero" . If, on the other hand, a series needs to be difference times to become I(0), said to be integrated of order denoted I(A) The order of integration is inferred by testing for unit roots. The most widely applied unit root tests are: CRDW: DurbinWatson Test of Sargan and Bhagrava (1983); DickeyFuller test; these tests the null hypothesis that series are 1(1) , that  I(1). The three statistics used in this analysis are as follows CRDW: I(1), CRDW .532 at 99% level; pet. + vt I(1) , negative statistic lower than and (95%) or 4.07 (99%) ; ADF + Sl=1YiAet ~ I(1) negative and statistic lower than 3 or 3.37 (99%) ; where et are the residuals from X regression (CRDW) and n is selected to be large enough ensure that the residuals are white noise . A statistically significant and negative sign coefficients 3 signifies that changes can be reversed over time, and level is stable. The SarganBhagrava (CRDW), DickeyFuller (DF) Augmented DickeyFuller (ADF) test statistics are reported in Table 26 and 27. The results in table 26 show that all series are nonstationary level significance. Therefore, the rates of change of all the variables are tested : Xt = a + et =  + same level integrated significance. first order, Thus, that the I(1) prices This are implies that the levels of crude oil spot price and each four futures prices show similar temporal properties. Therefore, the level of the spot price and futures prices are expected to be statistically linked over the longrun, and that the ratios of spot price and futures prices are expected to be constant over time the longterm. Cointecration Test After establishing that crude oil spot price and futures prices are integrated , both of the first order , the next step is to examine whether they are also cointegrated. The idea of this test that individual economic time series can integrated of order one, 1(1), but certain linear combinations of the series may be stationary, i.e., I(0). That implies that linear combination individually nonstationary economic variables may be stationary . Formally, two variables are said cointegrated there exists constant such that C Q, integrated order zero I(0). then stationary with positive, finite spectrum zero frequency. This rather special condition, because implies that both series individually have extremely important longrun components. However, similar the one applied test for integration followed . In first stage , the coefficient C is estimated by OLS; the second stage, the resulting series FPt. tested I(0) rather than I(1). The same procedure employed in the section above, is used here to test Zt. Table 28 gives results of the test. It is clear , from the table below, that the only futures price variable with statistics CRDW, ADF being significant percent confidence level the 1month ahead futures prices. None the other futures price variables have significant ADF values . Therefore, can be concluded that only 1monthahead futures price cointegrated with the spot price; the hypotheses about other futures prices, however , cannot be accepted at the 95 percent significance level. The findings that the spot price and 1monthahead futures price crude are I(1) and certain liner combination of them is I(0) have clearly demonstrated a long run relationship between the spot and futures prices remaining question now is what direction of "causation" this relationship follows, that which price leads, which price follows. attempt to answer this question in the next section. = p Causality Tests been ascertained above that stable longrun relationship exists between the spot price and month ahead futures prices. The next problem is to determine the direction of causation between the spot price and the futures price. this section the Garbade and Silber approach and the Granger causality test procedure will be employed investigate the direction of causality. Since a longrun relationship is only found between the spot price and 1monthahead futures price, denote 1monthahead futures price later on in this chapter , unless otherwise indicated. The Garbade and Silber ADDroach essence the price discovery function futures markets hinges changed on whether new futures prices information changed reflected cash first prices. Garbade and Silber approach provides a framework for analyzing whether one market is dominant in terms of information flows and price discovery . The model can be applied to the crude oil markets, with one minor modification. The prices, both spot and futures, are often reported on a monthly basis while this only lengthens the transaction period, the fundamental characters of the monthly data are identical to those of daily reported price. ,t1 (26) where St is the logarithm of the spot price for month t and Ft the logarithm the 1monthahead futures price for the same month. The constant terms, and af, have been added to equation (26) to capture any secular price trends in the data and any persistent differences between spot price and futures price attributable different quotation conventions. coefficient fs and f3 reflect the influence of the lagged price from one market the current price the other market. Since spot futures market prices are same commodity, one would anticipate that both and f are non negative. The ratio ps/(p can used examine price discovery relationship between the two markets. ratio equals unity that 40f = 0), convergence of spot futures prices occurs because spot price always moves towards the futures prices. This is an extreme case where the spot market "pure satellite" the futures market. ratio s/ + Pf) equals zero that fs = 0), then futures prices always adjust towards spot prices and futures market pure satellite. Intermediate values between Zero   . one imply mutual adjustments and feedback Equation (26) can be solved function  St1 to yield: t St where a  as'  Ps' and et = ut  Vt . As can be observed convergence equation between 27) spot reflects futures the prices. speed small, relatively little difference between futures and spot prices month persi month Prices will therefore converge quickly. We rearrange equations (26) algebraically as follows, that they can be estimated via ordinary S ] + least squares: The estimates 6 obtained with ordinary least squares are summarized in Table (29) . p /(Cps + ,f) represents the relative contribution futures market price discovery process, and 6 measures the rate of the convergence of futures and spot prices. Two results have emerged. First, the estimate of 3, is significantly positive. This finding shows that of Ft. 6 (Ft1  St1) = (,f tS, + et F F * t c1 wa v positive nor significant, suggesting that little feedback of pricing information occurs from the futures market to the spot market. The derived estimate of the parameter 6 indicates that less than half differential between futures spot prices month arbitrage persists undertaken rather month quickly This to bring implies about that price convergence between two markets. Granqer Causality Test The leadlag relationship can also examined through Granger's causality test, which states that the stationary linear combination of levels must Grangercause the change in at least one of the cointegrated variables. is well known that causality sense of Granger from a variable 1 to variable 2 can arise two reasons. The variable may cause in the commonlanguage sense, or instead the variable may anticipate forecast former case intervention which changes the stochastic process for z2 will change while latter case intervention which changes the stochastic process for will change the behavior The Granger's test is designed to examine whether two time series are generated separately from each other . When they are generated separately, one time series provides no information characterizing the other. Following Granger (1969), the simple causal model liPti 1 1 t i1 l JF , (210) where denotes crude spot price and 1month ahead futures price, and are taken uncorrelated whitenoise series. some of $ are zero , then said cause . Similarly some are zero , then said cause events there feedback both relationship occur, between then said t and . The that usual Ftest can be applied to test the null hypothesis that does cause .e. , = 0 for , or F t does cause , izJ = 0 Equation 210) are estimated least squares method under restrictions . = 0 J respectively. Table summarizes the results. From table, clear that the restrictions imposed equation 29) are statistically significant at 81 confidence level. contrast those equation (210) are statistically significant According definition Granger causality , thi suggests that spot price causing the future price, but the futures price is not causing il n ( i ti .e. = s = 2 futures prices not play important role price discovery and the spot market always leads the futures market. The Error Correction Model Now that long run relationship between crude spot price 1nonthahead futures price and direction of the causality have been established, we are in a position to study the dynamics of the stochastic movement prices. Since the crude oil spot price and 1monthahead futures price are cointegrated. An error correction model can be used to examine more closely the relationship between these variables. The error correction model (ECM) cointegrated variables commonly interpreted reflecting partial adjustment one variable another. Campbell Shiller (1988) show error correction models may also ari when one variable forecasts another . The concepts of cointegration and error correction are closely related. Indeed, It has been proven that two variables which are cointegrated have an error correction model representation. An error correction model for two variables relates the changes in the variables to lagged changes and a lagged linear combination of levels. The linear combination of levels which enters the error correction model just that combination which is stationary levels. __ correction model can thought description stochastic process by which the variables eliminate or correct the equilibrium error. Let us consider the crude oil spot price Pt and 1month ahead futures price FPt know that both them are integrated of order one and therefore they are cointegrated, their linear combination integrated order zero. Therefore residuals they from are the said cointegrated cointegrated. regression Define  C FPt the equilibrium error which enters the error correction model below:  YZt1 + y=1" 6 APt '~ =1 T"i ti + s.n l AFPt.J i j + jS lAFPt S+sEs 0 AFP t ]j=1 2J t (211) (212) where Y ' 6i and 8. are coefficients, et and et are assumed to be white noi ses. equations have been estimated descending order of generality. The results are given the Table 211 below. model. Two main First, conclusions error follow correction from the error correction term statistically very significant in equation (211), but neither the value of its coefficient nor the tstatistics of the equation significant. (212) equilibrium. suggests that disequilibrium distance .e. AFPt Y2Zt Since the value of Zt measures the deviation from ir =6 1 + s iAP. fluctuation of the crude spot price in the real world . In contrast, futures market deviates ess away from equilibrium "risk equatic and sharing" role much thus more futures better thai stable. Thi markets. n equation justifies Second, 11) , overall, terms the stati (unadjusted stical and adjusted), significance the the model. standard Thi error throws , and light lead lag questi on. The changes futures price are well explained changes the lagged spot prices futures pri ces Moreover , while the lagged spot prices play important role equations 12) , the changes lagged futures prices are significant equation 11) Therefore spot as in the market previous dominates section the , we have futures found market, again but that vice versa. Having analyzed dynamics the price movement, next estion about forecast stability illustration, both equation 11) and 12) are estimated by using the monthly data from 1984 to 1988, excluding first seven months 1989 . The results , provided Table 13), are identi to those obtained when the whole sample used The forecast values of changes in both spot futures prices are summarized in Table 13) . The variance of the difference between actual predicted values from equation 12) equation (212) doing much better than the other one erms of prediction . Above all, we conclude that equation should be used to capture dynamic interaction between these prices. Summary Thi chapter been divided into two related parts. First , it has investigated whether there exists a stable long run relationship between crude spot price futures prices . 1 , 3 9month ahead futures prices have been used carry out test. Both conventional cointegration test methods have been performed . We find that while there exists a stable long run relationship between spot price monthahead futures price , no such relationships exists between spot price other futures pric es. Second, a dynamic regression testing scheme Granger causality test have been applied further investigate price lead lag relationships between the spot 1month ahead futures price. The results from both tests leads us to conclude that crude spot price generally leads futures price incorporating new pricing information, that crude spot market always dominates futures markets These results suggest that the crude futures Table Selected Statistics 21 for Crude Oil Spot and Futures Price Price Series Mean S.D. Max. Min. Spot Futures 19.33 6.24 8.72 1Monthahead 3Monthahead 6Monthahead 9Monthahead 21.39 21.04 20.93 21.05 5.98 5.96 5.95 5.85 31.35 30.73 30.73 30.62 11.04 10.58 10.86 10.92 Table CrossCorrelations Spot and Futures for Crude Oj Price Series Spot 1Month 3Month 6Month 9Month Spot 1.00 0.95 0.85 0.69 0.62 1Monthahead 1.00 0.87 0.70 0.70 3Monthahead 1.00 0.85 0.73 6Monthahead 1.00 0.89 9Monthahead 1.00 Table Selected Statistics Futures 23 for Crude Oil Spot and Price Ratios Ratio to Spot Price of: Mean S.D. Max. Min. Futures: 1Monthahead 3Monthahead 6Monthahead 9Monthahead 0.43 0.91 0.37 0.63 0.63 2.53 2.94 Table Regression Oil Coefficients Futures Prices 6 and as Predictors 9Month Ahead Crude of Spot Prices Variables Equations (2.1) (2.2) (2.3) (2.4) Constant 10.16* 13.23* 18.18* 15.06* (3.09) (4.07) (5.12) (4.16) FPi 0.31* (3.49) FP,3 0.11 (0.98) FP,6 0.29* (2.32) FP9 0.02 (0.17) 1stOrder Autocorrelation 0.94 0.95 0.96 0.94 R 0.96 0.95 0.94 0.93 R2 0.96 0.95 0.94 0.93 DW 1.73 1.15 1.03 1.13 Std. Error 1.28 1.40 1.38 1.47 Estimation Technique CochraneOrcutt Notes: : tStatistic in parentheses. * denotes that coefficients the 95 percent or higher : The notations are for all are confidence tables significant level. this dissertation, not otherwise indicated. Table Regression Coefficients 25 for Equations Combining Crude Oil Price with Futures Prices Regression Coefficients Case a p, 1 1 0.62 0.55* 0.43* (0.80) (4.20) (3.08) 2 0.03 0.89* 0.10* (0.04) (26.10) (3.23) 3 0.42 0.94* 0.03 (0.57) (30.11) (1.30) 4 0.71 0.95* 0.01 (0,90) (30.25) (0.44) 5 1.63 0.49* 0.54* (2.13) (4.12) (4.54) 6 0.38 0.92* 0.09 (0.45) (10.35) (1.09) 7 0.09 1.00* 0.01 (0.11) (15.89) (0.16) 8 1.19 1.06* 0.11* (1.50) (26.44) (3.64) Key: Case 1: (21) and 1monthahead futures prices Case 2: (21) and 3monthahead futures prices Case 3 (21) and 6monthahead futures prices Case 4: (21) and 9monthahead futures prices Case 5 (23) and 1monthahead futures prices Case 6 (23) and 3monthahead futures prices Case 7: (23) and 6monthahead futures prices Case 8: (23) and 9monthahead futures prices Table Results of Integration Tests for Crude Oil Spot and Futures Prices Series CRDW DF ADF (crit. value (crit. value (crit. value 99%: 0.532) 99%: 4.07) 99%: 3.77) P 0.050 1.37 1.44 FP1 0.094 1.62 1.72 FP3 0.077 1.52 1.72 FP6 0.060 1.36 1.69 FP9 0.564 1.03 1.49 Table Results of Differenced Integration Tests Crude Oil Spot and First Order Futures Prices. Series CRDW ADF (crit . value (crit. value (crit . value 99%: 0.532) 99%: 4.07) : 3.77) 4.99 6.61 6.69 6.55 6.67 4.64 4.90 4.80 4.24 4.03 Table Results of Cointegration Tests for Crude Oil Spot and Futures Prices. Series CRDW DF ADF (crit. value (crit. value (crit. value 99%: 0.532) 99%: 4.07) 95%: 3.17) Case 1 1.507 9.59 3.45 Case 2 1.173 5.32 2.84 Case 3 0.964 4.47 2.87 Case 4 0.119 4.70 2.80 Key: Case 1: Spot Price vs 1MonthAhead Futures Price Case 2: Spot Price vs 3MonthAhead Futures Price Case 3: Spot Price vs 6MonthAhead Futures Price Case 4: Spot Price vs 9MonthAhead Futures Price Table 29 Estimated Coefficients Futures and Spot Prices Garbade and Silber's Approach Parameter Period 1984M1 1989M7 Ps 0.033 (0.29) f 0.684 (7.92) os!(oof) 0.oo** 6 0.40 Notes: Silber the s procedure calculated value requires both f3ps 0.05, but and Pf are Garbade positive. Table 210 Causality On Crude Oil and Futures Prices Dependent Variable Independent Pt F illltl illlillli Constant 0.41 .44) .11 .80) .15 .82) 1.13) .05) .10 0.72) .38 0.88 0.88 S.E. likelihood 143.70 164.49 Significance of restriction Table 211 Error Correction Model Equation (2.6.1) (Sample 1984M1 1989M7). = 1.243Zt.i + 1.688APt i (1.93) (2.71)  0.074APt.2 (0.31)  0.300APt3 (1.22) 0. 456AFPt.,1  0.123AP.2, + 0.035APt.3 (1.73) (0.69) (0.34) = 0.30 R2 = 0.20 S.E. = 1.30 D.W. 1.93 Equation (2.6.2) (Sample 1984M1 1989M7) AFPtI = 0.078Z.1 + 1.198AP + 0.382AP_2 + 0.138AP.3 t1 .1 X2 t3 (0.15) (2.42) (2.03) (0.70) 0.433AFPt.,1 (2.06) 0.129APt2,i (0.92)  0.153APt_3 (1.88) = 0.74 R2 = 0.70 S.E. = 1.03 D.W. = 2.05 Table 212 Error Correction Models and Forecasting Stability Equation (2.6.1) (Sample 1984M1 1988M12) = 0.243ZtI + 0.729APt.1  0.108APt_2 + 0.534APt3 (1.29) (3.33) (0.38) (1.74) 0.160AFPt , (0.69) 0.299APt.2,1 (1.42) 0.002APt3 (0.22) = 0.28 R2 0.18 S.E. 1.34 D.W. 1.91 Equation (2.6.2) (Sample 1984M1 1988M12) AFPtl1 = 0.143Zti + 1.012APt1 + 0.340APt2 0.024APt3 (1.00) (6.10) (1.57) (0.102) 0.292AFPt.,I (1.66)  0.030APt_,i (0.19)  0.143APt3 (1.66) = 0.75 R2 = 0.72 S.E. = 1.03 D.W. = 2.10 Table 213 ECM Predicted Values of Changes in Spot And Future Prices Spot Price Futures Price Time Actual Pred. diff. Actual Pred. diff. CHAPTER III FUTURES PRICES AND OPEC CRUDE OIL SUPPLY Introduction The relationship between futures prices OPEC crude supply is not as clear as the one between futures prices and crude oil spot price. And little research has been done in this field yet. Naturally, different assumptions will lead to different conclusions. instance, crude futures prices follow random walk, there will be no relationship at all between OPEC supply and futures prices. As a result, nothing can be used to predict the futures prices. contrary, traders are assumed to have rational expectations then there is a definite relationship between futures prices OPEC crude supply. can argued that speculators their futures prices based on their expectations of the market near future. demand crucial factors undoubtedly play very markets important like role supply process trader's expectations. The purpose this chapter investigate relationship between OPEC supply and futures prices. It will become clear that even though some relationship between elementary methods , these variabi are not expected stationary . Thi suggests that the results obtained elementary methods can wrong. the last Chapter , we shall first try to det ermine a stable relationship exists. Unfortunately most of the researchers on price discovery have ignored this step went to the test of causality directly However, results stabi obtained relationship Garbade does and exist, Silber then approach Granger causality test nay incorrect conclusions The rest this chapter is organic zed as follows . First, study some prior information about data using some preliminary methods such stati stical analy S1S , graphic illustrations and simple regressions Second , the results from simple regression are presented next . The character stics time series under consideration will tested following section with integration cointegration tests . Even though futures prices and OPEC supply are found cointegrated, which suggests that there long run stable relationship between them the Garbade Silber approach Granger causality test are used anyway the next two sections tc show one can get leading results finally , the conclu sons are ese nted bri efly Data Analysis analysis simple include regressions. graphs, basic Figure statistical shows the analysis, monthly values OPEC crude production series since 1984. Several things are apparent from First , the OPEC crude production has had an upward trend. This may explain that the crude price been declining from peak so called "oil crisis" and has recovered that level. It also helps to explain why, over time, futures market has been pessimistic about crude prices that the crude futures prices follow a negative trend as well. Traders in the market expect that the upward trend OPEC production will continue. Secondly, decline has been irregular there is no recognizable pattern follow. Table 31 provides the mean, standard deviation, maximum, and minimum for the following variables: OPEC oil production futures prices 9 monthahead contracts. Again, futures prices picked are the same as those stated in the last chapter , that is the 2monthahead is very closely correlated with the 1month and 3monthahead ones and the 4 8monthahead correlated with contracts and 9 are month found be closely contracts. Cross correlations for these series are given Table 32. shown table , the mean OPEC crude production from January 1984 through July 1989 17.83 o 18  17 16 15 1I 1984 ! 1985 199S 1987 nA r 1988 Figure OPEC Crude Oil Production 1929 coefficients between OPEC crude supply and futures price series are negative. This result, which surprising, tells that the OPEC supply has negative effect futures production prices, lower are the that more futures prices. This can be easily explained by the fact that when there is excess supply price will tend decline. Second, unlike relationship between spot price and futures prices shown the Table 22, which shows that the correlation between spot futures prices falls from 0.95 1month ahead, progressively declining to 0.62 for 9month ahead, cross correlations increase from 0.465 (in absolute value) for the 1monthahead to 0.531 for the 3monthahead and 6monthahead then drop to 0 .528 for 9monthahead futures. The correlation between OPEC supply , 9monthahead futures prices larger than that between 1monthahead futures price. The 1 , 6 and 9monthahead crude oil futures prices are plotted against OPEC production in Figure 32 through 35, respectively . They demonstrate some interesting movements of futures prices and the OPEC production since 1984. crude oil First, futures prices move oppositely to that of the OPEC production. Generally speaking, when there is an increase the production, there is a drop in the futures prices. Second, , 6 K!> IN N x 1984 195S 1986 1987 1988  OPEC Of SucoD t y .utures Figure Oil Supply Futures ces m9a *Monthuut production are very big. The futures prices falsely signaled trend oil c crisis, supply people . This were e is due expecting the the fact that prices after in the futures to increase. From Figure 31. we can see clearly that during this mid1984 and late 1985 periods, the OPEC crude oil production such kept negative declining, trend too. the Therefore production traders expected continue anti cipated another shortage supply . Second, after mid 1986, futures prices have been following OPEC production much more closely. This because during this time, investors realized the mistake in their expectation adjusted according the new information. There another reason that the futures markets are getting more mature and the new technology has been adopted so that people the market can receive and respond the new information more quickly prices declined . Third, a great deal. after mid1985, This was due futures the continuous increase in the OPEC crude oil production. Since mid1985, OPEC faced strong competition from nonmember countries. order to maintain the market share, OPEC has been increasing its production steadily. This exerted a downward pressure on the futures markets. Observing this fact the traders became pessimistic about the market. Fourth, from the figures, it can seen easily that the 6monthahead futures prices 7YA I 11 v< ' r /'^ ( \ 7. \ r, , 4 19B5 1986 9g7 198a  OPEC 01 Stcoty 3MoflntOut Figure Oil Supply Futures Prices 1984 7 ^ ' 7' NflN 197 7 w/ 7 C/ ^ 19g8 1989  OPEC O05 5uooly 6MonthOut Futures Figure Supply Futures ces 1995 198s ,Tr\ r j ' A <7 t ^H 1985 19865 1987 1988 1989  CPEC O0 Suc ly tuturs Figure Oil Supply Futures Prices 1984 9'onthuut Another calculate preliminary ratio of review futures the prices relationship OPEC production of crude same month. Table 33 presents mean ratios and standard deviations for these calculations. Again, the lagged response effects of supply is verified. The ratios of 3 , 6 and 9monthahead futures prices to the OPEC supply have values closest to unity . This suggests that 3 , 6 and 9 monthahead futures prices move more closely along with OPEC However production , by looking than 1monthahead standard futures deviations, prices one can also notice that relationship between the OPEC production 9month ahead futures prices have higher variance than ones between the supply and 1monthahead futures prices, price because OPEC ratio production 1monthahead has smallest futures standard deviation. This time span during which people adjust their expectations. The longer the time span, the more likely some new events are to happen during this period , and more likely that investors readjust their expectations. Figure 36 through show the ratios of the OPEC crude production futures prices the corresponding time periods. There is a slightly upward shift the graphs indicating that over time the futures market been pessimistic about crude oil prices. obvious that prior , 6 reached OPEC production. The futures prices over adjusted to the increase in the OPEC production. These "over shoots" have had an impact on the futures markets. They were the turning points when the futures market switched from being optimistic pessimistic. The market never returned level before 1986. After the overreaction, both investors and the market adjusted. The market was becoming more mature and traders could information using respond modern more instantaneously technology. futures new prices then followed the OPEC production more closely much closer to unity . The ratios are , even though there are some fluctuations. Simple Recressions this section, simple regressions are used investigate relationship between OPEC crude production and futures prices are presented and analyzed. Thus following equations are estimated: FPt,l (31 p St (32) = a + 3 St (33) FPt.9 8 St (34) where = the OPEC crude oil production at time = a + = a + lMonthOut Futures A S'N \ A. A1 i . .T ..r t I .. .. I T I I i 1rtl + f. a a ... 1 ..T. ." " 1985 1986 19ia 1988 1989 Figure 3.6 Oil Supply /Futures Prices 1984 3MontnhOut Futtraes 1985ga 198 987 1988 1989 Figure Supply /Futures Prices 1984 6MonthOut Puttures A /V 2! 1997 1989 1989 Figure 3.8 Supply /Futures Prices 1.7 t 1 1984 1985 1986 / 9mrontnout futtres prices 1. 1 0.6  0.5  0.4 0.3 0.2  0. 1  1984 1985 Figure OPEC Oil Supply /Futures 1986 1987 1988 1989 Table 34 gives the results of this estimation. From this preliminary table, can obtain following conclusions. First, the supply variables are statistically very significant in all four equations the corresponding coefficients are negative standard expected. errors, Second, we can conclude in terms that of Rsquared the equations with , 6 and 9monthahead futures prices performed better than the one with 1monthahead futures prices. This again verifies the lagged response effects of supply on people's expectations futures prices. Basically, however, four equations may perform poorly. This is because of the fact that the time series consideration nonstatinary. under basic assumptions of simple regression are violated . Therefore, results from this method can be "spurious" the following sections of this chapter, and misleading. the nonstationarity of time series variables will tested. Specification Tests Integration Test Many macroeconomic variables such price, supply, demand Furthermore exhibit these variables nonstatinonarity often appear to levels. integrated of order one (stationary in the first differences) , or to possess other words, whether they are stationary levels differences some order. The cointegration test aims detecting whether there exists a stable relationship between the level of two economic variables. In the example here, link the OPEC crude production and futures prices and test the relationship between them. a relationship, If there exists then it may be possible that some quantitative effect of the futures prices on the OPEC crude oil production can made. The theory integration test method was discussed Chapter Two, hence details will repeated here. Table report SarganBhagrava (CRDW) DickeyFuller (DF) Augmented DickeyFuller (ADF) test statist iCS. Table 35 shows that for all the series their level unit root hypothesis rejected at 99% level of significance. This suggests that even though some results are obtained previous section, they are not valid. order to get these variables to satisfy the normal conditions of simple regression, we need to difference all the variables see whether the difference series has a unit root. From Table 36 it is apparent that all of the difference variables reject hypothesis unit root same level significance. Thus , all the variables in level are 1(1) . This implies that levels of the OPEC crude oil production and prices are expected to be statistically run, linked over the long and that the ratios of spot price and futures prices will be constant over time in the longterm. Cointeoration Test After investigating the fact that both the OPEC crude oil production futures prices are integrated first order the next step is to examine whether they are also co integrated. The idea behind the cointegration test follows. Suppose two economic variables show nonstationarity they share same character the nonstationary stochastic movement? More precisely while individual economic time series can integrated order one, I(1) , certain linear combinations of the series can be stationary or I(0) clear from table below, that statistics CRDW ADF is significant at 95 percent confidence level. This implies that no linear combination of the I(1) OPEC crude production and futures prices is possibly stationary and they do not share the same character of the nonstationary stochastic movement. Therefore, can conclude that futures prices are cointegrated with the OPEC crude production level. From above analysis, conclude that there stable longrun relationship between the OPEC Crude researchers make mistakes here. The next sections are designed illuminate these mistakes . It known that OPEC crude production and futures prices are co integrated , which suggests that there long run relationship between them. disregard this fact purpose , and ahead to implement the popular methods used by many researchers, Garbade and Silber approach, and Granger causality t As it will become clear , controversial results are obtained : one method suggests that futures prices res pond the OPEC crude production; other , on the other hand , suggests that futu res ces Granger cause OPEC production The Garbade Silber s Approach We present Garbade Silber approach which leads following model time sern behavior t1.IJ where tIS logarithm the OPEC crude production month t and the logarithm of the jmonth ahead futures prices deliv in the same month. The constant terms indicate trends data and persistent ' as I  i other variable. The ratio Pf/( + 3) can be used to examine relationship between variables. ratio equals unity that s = 0), then implies that futures prices always moves in response to the OPEC crude oil production. ratio / f/(s P f) equals zero that = 0), then the OPEC oil production always adjusts to the futures pr ices. in response An intermediate value between zero and imply ies mutual interaction between the two variables. Equation can be solved t as function  St1 t1  St  St1) where a  a = 1  P = ut  s' As can be observed equation (36) , reflects the speed convergence between two variables. If 6 is small , only a small fraction the difference between month persists month this case, will converge quickly. rearrange equation (35) algebraically following way , so that it can be estimated via ordinary least squares: F 1, cSS ^C 1 fl /* The estimates obtained from this equation using ordinary least squares are summarized in Table (38) through Table (311) . Also given in these tables are the ratio f#/({s + 9Pf) which show the relative contribution of the OPEC crude production formation futures prices, which measures the rate of the convergence of these two kinds of variables. These tables reveal the following points. First, cases, estimates /3's are significantly positive . This finding leads conclude that OPEC crude oil production leads the futures prices in incorporating new information. Second, the estimates are also significantly positive, though much smaller than Pf's,, which suggests smaller feedback from the futures market OPEC crude oil production. Third, The derived estimates of 6's shown in the tables indicate that only a small amount of difference between futures prices and the OPEC production in month t1 persists to month t. This can also be explained by the lagged response effect of production. In this month, the traders are unable to absorb the difference in the relationship between the OPEC oil production and the futures prices which happened last month. Little the difference persists from last month to this month. However, it may have to be taken into account several months later. Fourth expectations futures prices. This last point strengthens the first point that not only does OPEC crude production lead futures prices , but also influence on the futures prices significant. Granger Causality Test Ignoring the finding that there long run relationship between the OPEC crude production futures prices which was based on the cointegration test previous sections , let implement Granger test to oil supply futures prices anyway According to Granger (1969) , the simple causal mod t = = z t i =1 1 t + =1 i2Ft where denotes OPEC crude production futures prices, , 3 and 9monthahead . 1it are taken uncorrelated white noise series. t is said cause t if some are zero. Similarly t is said cause some s are zero. both these events occur , there feedback relationship between hypothesis and The that usual t does not Ftest cause applied to test corresponding null to 81 J = 0 + use restricted least squares method estimate equation and imposing the restrictions respectively . Table 12) through Table 15) summarize estimation results. The following results emerge. First, unclear what kind of relationships exists between the OPEC crude production and the futures prices. The restrictions imposed on the equations are stati stically significant for different variables For 1monthahead futures price , it seems that the OPEC oil production is Granger caused futures ces . The significance level about cent. For , and 6month ahead futures prices however , the situation is just the opposite futures ces are Granger caused OPEC production. The significance level about per cent. For 9monthahead futures ces , the restrictions on both OPEC production and the futures prices are not states tically significant, result which suggests that neither OPEC production Granger causes the futures ces nor futures ces Granger cause OPEC production. Compared with results presented the two previous sections , the outcomes obtained thi section are obviously different The results obtained there with the Garbade Silber strong approach effect indicate that futures the pri OPEC ces. oil The production outcomes has thi between 1monthahead futures prices and that between the 6month ahead futures prices are exactly the opposite . The contradiction findings due the nonexis tence stable long run relationship between OPEC production futures prices, question that needs answered first step of the proper "two step" testing procedure long run relationship between variable . If we ignore ste and ahead investigate causal relation ships then is not surprise at all that conflicting results are obtained. Summarv chapter , we have studied relation ships between OPEC crude production futures prices Different methods simple have methods been used such to examine graphic relationship simple regression suggest that variable there . The an apparent coefficients relationship regressing between. the ese futures prices on the OPEC production is negative which implies that the OPEC production has downward pressure futures prices. can also be easily seen graphs. With the exception an unstable period , there was a drop futures ces when production increased. The integration cointegration tests were then used the OPEC production were integrated order one, suggesting that levels of both OPEC production and futures prices were unstable, however, their first order differences are stable . This finding prompted us to employ the cointegration test examine the hypothesis whether such nons tationarity the same order was generated a similar mechanism . The cointegration test rejected thi hypothesis. means that the nonstationarity of the OPEC production and that futures prices are generated same mechani . There fore we can conclude that there no long run stable relationship between these variable es. The last sections of thi chapter showed the mi stakes that can arise fact that there no long run stable relationship between two variabi ignored. Most research ers have made this mistake . They usually did test the exi stence of the long run relationship before they went estigate nature relationship with methods such Garbade and Silber approach or the Granger causality test. When longrun relationship does exist, results about relationship are meaning ess. These sections confirm this argument We showed conflicting outcomes could arise about relationship between the futures ces the OPEC production when different methods were used. The results from the Garbade silber approach causality test suggested a lack of a consistent relationship. For the 1monthahead futures price, the OPEC production was Granger caused by futures price. For , and 6monthahead futures prices, however, situation was reversed: futures prices are Granger caused by the OPEC production. Table Selected Statistics OPEC Crude Oil Supply Futures Price Series Mean S.D. Max. Min. Supply 17.83 1.82 22.46 14.46 Futures 1Monthahead 21.39 5.98 31.35 11.04 3Monthahead 21.04 5.96 30.73 10.58 6Monthahead 20.93 5.95 30.73 10.86 9Monthahead 21.05 5.85 30.62 10.92 Table 32 CrossCorrelations OPEC Crude Oil Supply and Futures Price Variable Supply 1Month 3Month 6Month 9Month Supply 1.000 0.465 0.531 0.531 0.528 1Monthahead 1.000 0.872 0.700 0.700 3Monthahead 1.000 0.854 0.732 6Monthahead 1.000 0.889 9Monthahead 1.000 Table 33 Selected Statistics for OPEC Crude Oil Supply and Futures Price Ratios Ratio to Supply: Mean S.D. Max. Min. Futures 1Monthahead 1.23 0.42 1.94 0.57 3Monthahead 1.21 0.43 1.97 0.62 6Monthahead 1.21 0.43 2.01 0.66 9Monthahead 1.21 0.43 2.04 0.66 Table Regression Coefficients Crude Oil Futures Prices on OPEC Crude oil 9Step Ahead Production Variable Dependent variable F1 F3 F6 F9 Constant 48.91* 57.28* 50.42* 49.86* (7.53) (8.32) (8.18) (8.02) S 1.54* 1.67* 1.66* 1.61 (4.24) (4.93) (4.81) (4.66) 1stOrder 0.90 0.88 0.88 0.90 Autocorrelation R2 0.22 0.28 0.28 0.28 R2 0.20 0.27 0.27 0.27 DW 0.17 0.18 0.20 0.16 Std. Error 5.37 5.09 5.08 5.01 Table Results of Oil Integration Tests Supply and Futures for OPEC Crude Prices. Series CRDW DF ADF (crit. value (crit. value (crit. value 99%: 0.532) 99%: 4.07) 99%: 3.77) S 0.231 1.64 1.76 FP1 0.094 1.62 1.72 FP3 0.077 1.52 1.72 FP6 0.060 1. 36 1.69 FP9 0.564 1.03 1.49 