8 Invesmen Managemen and Financial Innovaions, 4/2004 Term Srucure of Fuures Prices and Expeced Mean Reversion in Base Meal Prices Changyun Wang 1 Absrac We examine expeced mean reversion in six base meal prices using a unique LME daase consising of prices for fuures conracs wih mauriy daes spanning from 1 o 27 monhs. We documen significan evidence of mean reversion in spo prices for all hese meals, alhough he magniude of mean reversion differs across hese markes. We also find ha mean reversion in meal prices arises from a posiive covariaion beween spo prices and implied cash flow yields raher from a negaive correlaion beween spo prices and forward ineres raes. JEL Classificaion: G12; G13 Key words: Mean reversion; Base meals; LME; Cos-of-carry. I. Inroducion There is mouning evidence on mean reversion in financial asse prices (Fama and French, 1988a; Poerba and Summers, 1988; Sweeney, 2000). The resuling predicable paern in reurns has imporan implicaions for asse pricing and invesmen decisions. However, due o he non-availabiliy of reliable spo price daa, here has been relaively lile evidence on mean reversion in physical asse prices. In his aricle, we follow a similar mehodology o ha of Bessembinder e al. (1995) o examine wheher invesors anicipae mean reversion in six base meals, namely, primary aluminum, copper, nickel, lead, in, and zinc, using he LME (London Meal Exchange) price daa for fuures conracs wih delivery daes spanning from 1 o 27 monhs. Our sudy depars from he large body of lieraure on ex pos reurn predicabiliy ha examines he properies of markedeermined prices by focusing on wheher invesors expec mean reversion o occur in equilibrium. An advanage of using fuures price daa for conracs wih differen delivery daes is ha i allows us o ascerain he source of mean reversion, subjec o he assumpion ha he cos-of-carry relaion holds. This issue has been under debae in he exan asse pricing sudies (De Bond and Thaler, 1985; Fama and French, 1988a; Poerba and Summers, 1988) 2. We find evidence of expeced mean reversion in base meal prices for each marke we examine, alhough he magniude of mean reversion differs across hese markes. There exhibis he greaes degree of mean reversion in copper prices, while he leas mean reversion is presen in nickel prices. For example, 40% of a copper price shock is expeced o rever o is long-run mean over he subsequen year, on average; whereas only 10% of a nickel price shock is reversed. Moreover, we find ha he posiive covariaion beween spo prices and implied cash flow yields is he only source of mean reversion for all base meals. Evidence of mean reversion in hese meal prices is imporan for capial budgeing decisions, sorage decisions, and pracical rading (hedging) sraegies for worldwide meal producers, processors, and dealers, since he LME is a major world marke for non-ferrous meals, and LME prices form he basis for rading in physical meals hroughou he world 3. Bessembinder e al. (1995) examine mean reversion using observable erm srucure of fuures prices in broad US fuures markes, and find evidence of expeced mean reversion in each of 1 Ph.D., Deparmen of Finance and Accouning, School of Business, Naional Universiy of Singapore, Singapore. The auhor hanks Adrian Low for compuaional assisance. 2 There are wo major views in explaining mean reversion in financial asse prices. Some auhors argue ha he negaive auocorrelaions arise from a slowly decaying mean revering price componen, and hus may reflec ime varying equilibrium expeced reurns generaed by raional invesor behavior (Fama and French, 1988a). Oher auhors conend ha mean reversion in asse prices reflecs he emporary divergences of prices from he fundamenal value due o irraional rader behavior (De Bond and Thaler, 1985; Poerba and Summers, 1988). 3 Anoher major fuures marke for non-ferrous meals is he NYMEX in New York. However, NYMEX is confined o rading in copper and aluminum fuures conracs only, which are less liquid in erms of rading volume han hose on he LME.
Invesmen Managemen and Financial Innovaions, 4/2004 9 hese markes. Moreover, hey find sronger evidence of mean reversion in agriculural commodiies and crude oil markes han in financial asse markes. Fraser and McKaig (1999) follow he mehodology of Bessembinder e al. (1995) and explore UK fuures erm srucure daa. They find evidence of mean reversion for commodiies including aluminum, Bren, gas-oil, and copper. They also repor ha mean reversion in financial asse prices is presen in near raher han disan fuures conracs, which differs from he findings of Bessembinder e al. (1995). Alhough Fraser and McKaig examine base meals in heir sudy, only wo meals are included. Moreover, hey focus on he mean revering behavior in base meal prices for only wo disan mauriy daes (15 and 27 monhs). We supplemen he evidence on expeced mean reversion in base meal prices using he unique LME daase ha consiss of prices for fuures conracs wih mauriy daes spanning from 1 o 27 monhs. The unique rading pracice on he LME also offers several advanages in he sudy of mean reversion in equilibrium asse prices over a sandard fuures marke. Firs, reliable spo price daa are generally unavailable for mos physical asses, and herefore neares mauriy fuures price is used as a proxy for spo price (Fama and French, 1987; Bessembinder e al., 1995). The use of neares mauriy fuures prices can be problemaic due o ligh rading, possible fuures squeezes or corners, ec. (Fama and French, 1987) 1. This problem is resolved using he LME daa since boh physical meals and fuures conracs are raded a he same ime on he LME. Second, in a sandard fuures marke here are a limied number of conrac mauriy daes, and liquidiy is focused on a few shorer-erm conracs. As a resul, prices for longer-erm conracs are likely o be unavailable or unreliable. The LME s daily conrac srucure resuls in a large number of conracs wrien wih a fixed ime o mauriy up o 27 monhs, and he rading of longer-erm conracs shows subsanial liquidiy. II. The Cos-Of-Carry Theory and Expeced Mean Reversion The cos-of-carry heory explains he difference beween conemporaneous spo and commodiy fuures prices in erms of carrying charge ha is he ineres foregone ne of implied cash flow yields (convenience yields ne of warehousing coss, insurance coss, ec.) o a marginal commodiy holder (Working, 1949; Brennan, 1958) 2. The cos of carry relaion can be wrien as R CK K K P e K,, F, (1) where F (K) is he dae s fuures price for delivery of a commodiy on dae +K, P is he dae s spo price, R K, is he coninuously compounded ineres rae (per period), and C K, is he implied cash flow yield o a marginal commodiy owner, saed as a percenage of he asse price on a coninuously compounded basis. If he cos-of-carry relaion holds, mean reversion in equilibrium asse prices has wo major esable implicaions. Firs, a percenage change in he dae s fuures price is less han a percenage change in he dae s spo price. Pu differenly, he elasiciy of fuures prices wih respec o spo prices is less han uniy. Assume ha R K, and C K, are differeniable funcions of spo prices wih he firs derivaives of R and C respecively, hen he elasiciy of he dae s fuures price wih respec o he dae s spo price,, is given by 1 Anoher problem can arise if we use neares mauriy fuures price as a proxy for spo price o sudy expeced mean reversion phenomenon. Since fuures prices converge upon spo prices as ime approaches mauriy, a fuures price wih longer ime o mauriy is less accurae as a proxy for he curren spo price compared o a fuures price wih shorer ime o mauriy. Consequenly, he use of fuures price as a proxy for spo price would induce noise in he esimaion of erm slopes. 2 An alernaive heory of commodiy fuures prices is ha a fuures price conains wo componens: a forecas of he fuure spo price and an expeced risk premium (Dusak, 1973). Upon examining he explanaory power of hese wo compeing heories of fuures prices using daa from broader fuures markes, Fama and French (1987) conclude ha fuures prices appear o reliably respond o carrying cos variables, while here is less convincing evidence ha fuures prices conain risk premiums or power o forecas fuure spo prices.
10 Invesmen Managemen and Financial Innovaions, 4/2004 F ( K) P 1 K P ( R' C' ). (2) P F ( K) In equaion (2), mean reversion in spo prices requires o be less han one. Thus, R -C < 0. Second, here exiss a negaive relaion beween spo prices and he slope of fuures erm srucure, ha is, an increase in spo prices is associaed wih a decrease in he slope of fuures erm srucure, and vice versa. Le F (T) be he fuures price of he same asse for delivery on dae + T, where T > K, which can be wrien as R T, CT, F T P e. (3) T Using equaions (1) and (3), we can express he slope of dae s fuures erm srucure over delivery daes +K o +T, KT,, as LN[ F ( T) / F ( K)] T K KT, RKT, C, (4) KT, where R KT, and C KT, are he forward ineres rae and implied cash flow yield for he inerval from daes +K o +T respecively, which are given by R KT, T RT, K RK T K, and (5) T CT, K CK, CKT. (6), T K From equaion (4), mean reversion in equilibrium asse prices implies KT, RKT ' CKT ' 0. Any deeced mean reversion mus resul eiher from a negaive correlaion beween forward ineres raes and spo prices, i.e., R KT < 0, or from a posiive covariaion P beween implied cash flow yields and spo prices, i.e., C KT > 0, or boh. Therefore, he assumpion on he validaion of cos-of-carry relaion allows us o ascerain he source of mean reversion, which has been subjec o long-lasing debae in he exan asse pricing lieraure (Fama and French, 1988a; Porerba and Summers, 1988). However, his approach canno rule ou he possibiliy ha mean reversion resuls from marke inefficiency, or invesor irraionaliy. III. Daa and Mehodology This sudy analyzes prices of fuures conracs for six base meals raded on he LME primary aluminum, copper, in, lead, nickel, and zinc over he January 1993 January 1997 inerval. The unique daase consiss of daily spo and selemen prices for each meal fuures conrac for delivery daes of 1, 2, 3, 6, 9, 12, 15, and 27 monhs 1. These daa are obained from he LME. The sample period is chosen due o he nonavailabiliy of fuures prices daa for mos delivery daes. Two imporan LME rading pracices offer us an ineresing opporuniy o sudy expeced mean reversion in asse prices: firs, he LME s daily conrac srucure resuls in a large number of conracs for each meal on a paricular rading day (approximaely 81 conracs per meal each day) 2 ; and second, boh spo meals and fuures conracs are raded a he same ime on he LME. We also collec daa on London Iner-bank Offer mid raes (LIBOR) and Eurocurrency Serling mid raes over he same sample period from Daasream Inernaional. 1 The officially repored prices are confined o delivery periods of 3 monhs, 6 monhs, and 12 monhs only. The daa for he 27-monh selemen prices for he lead and in fuures are no available, while hose for he nickel fuures are only available from July 1995 onwards. 2 The LME conracs have daily promp daes up o hree monhs ahead, supplemened by a more limied number of longererm conracs up o 27 monhs ahead. This implies a new conrac is inroduced for every rading day.
Invesmen Managemen and Financial Innovaions, 4/2004 11 As discussed in Secion II, if spo prices are expeced o exhibi mean revering behavior, he elasiciy of fuures prices wih respec o spo prices is less han one, and declines wih he ime o mauriy. To es his hypohesis, we esimae he elasiciy of fuures prices wih respec o spo prices by regressing he firs difference of logarihmic fuures prices, LN[F (K)/F -1 (K)], on he firs difference of logarihmic spo prices, LN[P /P -1 ], i.e., i K) i i P LN K 0 1 1 ( ) P 1 i F ( i LN F, (7) where i indicaes he ih meal fuures, K denoes delivery of K periods ahead, and K = 1, 2, 3, 6, 9, 12, 15, and 27 monhs. If spo prices are mean revering, he esimaed elasiciy, i 1, is expeced o be less han uniy. Expeced mean reversion also implies a negaive relaion beween spo prices and he fuures erm slope. To deec expeced mean reversion in asse prices, we follow he mehodology of Bessembinder e al. (1995), and compue he fuures erm slope, KT, in equaion (4), on he basis of nearby ha indicaes relaive nearness o expiraion. For example, nearby 1 denoes ha he erm slope is compued by using he prices of conracs ha expire in 1 and 2 monhs, i.e., K=1 and T=2, nearby 2 indicaes ha he erm slope is compued by using he prices of conracs ha expire in 2 and 3 monhs, i.e., K=2 and T=3, ec. We regress he fuures erm slope on spo prices, ha is, i i i i i KT P. (8), 0 1 If spo prices are mean revering, he esimaed 1 i is expeced o be negaive. Since boh spo prices and fuures erm slopes are nonsaionary, equaion (8) is esimaed by using he firs difference 1. To allow for comparisons across he markes, each spo price series is scaled using is ime series mean. An advanage of assuming he validaion of he cos-of-carry relaion is ha i allows us o explicily examine he source of mean reversion. Since here are only wo componens in he erm slope in equaion (4): forward ineres rae and implied cash flow yield, here are wo cases in which mean reversion can occur in equilibrium: firs, implied cash flow yields are posiively associaed wih spo prices; and second, forward ineres raes covary negaively wih spo prices. Thus, we expec ha C KT > 0, or R KT, < 0, or boh. If KT < 0, along wih C KT > 0 and R KT = 0, expeced mean reversion resuls from he posiive comovemen beween implied cash flow yields and spo prices. If KT < 0, along wih R KT < 0 and C KT = 0, mean reversion arises from he negaive covariaion beween forward ineres raes and spo prices. To empirically es he source of mean reversion, we esimae he firs derivaives of implied cash flow yields and forward ineres raes wih respec o spo prices, C KT and R KT, by regressing changes in C KT, and R KT, on changes in spo prices respecively. We use daa on observable erm srucure of fuures prices o obain esimaes of implied cash flow yields using he following formula, CKT, LN[ F T F K R ( ) / ( )] KT, T K. (9) IV. Empirical Resuls Table 1 presens he resuls of esimaing equaion (7) for each meal and mauriy dae. The resuls indicae ha he esimaed elasiciy is posiive and significan a he 1% level for all hese meals and mauriy daes, indicaive of mean reversion in equilibrium spo prices. Moreover, he esimaed elasiciy decreases monoonically wih he ime o mauriy excep for he 3-monh 1 The resuls of uni roo ess ha are no repored (available upon reques) fail o rejec he exisence of uni roos for boh spo price and fuures erm slope series, wih a few excepions for fuures erm slopes, for example, erm slopes for nearby 1 primary aluminum, nearby 2 primary aluminum, copper, and nickel.
12 Invesmen Managemen and Financial Innovaions, 4/2004 zinc fuures. However, he degree of mean reversion differs across hese meals. There exhibis a greaer degree of mean reversion in copper and lead prices, while he speed of mean revering is slower in nickel prices. For he copper marke, he poin esimae for he 12-monh fuures is 0.597, implying ha 40.3% of a copper price shock is reversed over he subsequen year. On he conrary, he esimae for he nickel marke is 0.902, suggesing ha only abou 10% of a nickel price shock is reversed over he subsequen year. Elasiciies of disan fuures price changes wih respec o spo price changes Table 1 Primary aluminum Copper Nickel Lead Tin Zinc 1-monh 0.9790 (7.78)*** 2-monh 0.9625 (11.72)*** 3-monh 0.9472 (12.88)*** 6-monh 0.8890 (15.22)*** 9-monh 0.8348 (19.90)*** 12- monh 15- monh 27- monh 0.7833 (22.81)*** 0.7436 (18.45)*** 0.5945 (20.28)*** 0.9061 (16.47)*** 0.8517 (20.32)*** 0.7937 (24.27)*** 0.7005 (27.73)*** 0.6403 (30.48)*** 0.5974 (32.21)*** 0.5685 (31.73)*** 0.5261 (30.01)*** 0.9825 (4.07)*** 0.9758 (5.50)*** 0.9533 (9.82)*** 0.9446 (8.90)*** 0.9279 (11.50)*** 0.9021 (13.20)*** 0.8948 (13.04)*** 0.8554 (9.40)*** 0.8703 (4.82)*** 0.8276 (6.39)*** 0.8202 (16.91)*** 0.7016 (20.87)*** 0.6691 (21.77)*** 0.6383 (22.47)*** 0.6641 (21.13)*** 0.9327 (6.66)*** 0.9168 (8.32)*** 0.9009 (12.92)*** 0.8624 (13.23)*** 0.8417 (14.39)*** 0.8231 (14.99)*** 0.8215 (15.52)*** 0.9412 6.46)*** 0.9251 (8.32)*** 0.9600 (10.81)*** 0.8744 (11.96)*** 0.8410 (15.14)*** 0.8091 (17.35)*** This able repors he poin esimaes of elasiciies (equaion (4)). The figures in parenheses are - saisics, which are for he hypohesis ha he slope coefficien esimae is less han one, and are correced for heeroskedasiciy and auocorrelaion based on he Newey-Wes (1987). Price series are deflaed using heir ime series means. ***, **, and * denoe significan a he 1%, 5% and 10% level. The resuls of esimaing equaion (8) are repored in Table 2. The slope coefficien esimae is he poin esimae of he firs derivaive, KT,. The evidence shows ha, consisen wih he hypohesis of mean reversion in asse prices, he esimaed slope coefficien is negaive and significan for all excep he nearby 6 in fuures (posiive and insignifican) and he nearby 6 lead and zinc fuures. Thus, an increase in spo prices is associaed wih a decrease in he slope of fuures erm srucure, and vice versa. Noe also ha he magniude of esimaed coefficiens decreases wih he ime o expiraion. However, he coefficien esimae for copper ends o decline faser in magniude han ha for he oher meals. This suggess sronger mean reversion anicipaed by copper raders han raders in he oher meal markes, in line wih he previous resul. Our findings of expeced mean reversion are consisen wih hose of Bessembinder e al. (1995) and Fraser and Mckaig (1995). These auhors documen evidence of mean reversion in asse prices in he US and UK respecively. Tables 3 and 4 presen evidence on he source of mean reversion. The risk free rae is proxied by he London Iner-bank Offer mid rae (LIBOR). Table 3 repors he esimae of he derivaive of implied cash flow yields wih respec o spo prices, C KT. The esimae of C KT is posiive and significan for all hese meals, suggesing ha he implied cash flow yield sensiiviy of spo prices is a reliable source of mean reversion in spo prices. The resuls of regressing changes in forward ineres raes on changes in spo prices are repored in Table 4. The signs of slope coefficien esimaes are no uniformly negaive, and insignifican. Therefore, i does no appear ha he covariaion beween spo prices and forward ineres raes induces mean reversion in meal prices.
Invesmen Managemen and Financial Innovaions, 4/2004 13 The slope of fuures erm srucure and he level of spo prices Table 2 Primary aluminum Copper Nickel Lead Tin Zinc Nearby 1-0.3834 (-9.85)*** Nearby 2-0.3571 (-10.38)*** Nearby 3-0.3447 (-10.81)*** Nearby 4-0.3280 (-18.32)*** Nearby 5-0.3163 (-15.69)*** Nearby 6-0.2466 (-4.55)*** Nearby 7-0.2153 (-8.12)*** -1.1583 (-16.79)*** -1.0314 (-17.25)*** -0.6239 (-18.25)*** -0.4007 (-18.35)*** -0.2755 (-16.44)*** -0.1845 (-10.23)*** -0.0678 (-6.47)*** -0.1606 (-8.96)*** -0.1533 (-4.93)*** -0.1447 (-5.37)*** -0.0978 (-9.17)*** -0.0938 (-9.17)*** -0.0991 (-2.28)** -0.0143 (-2.00)** -0.7715 (-12.51)*** -0.6937 (-9.42)*** -0.4695 (-10.24)*** -0.1924 (-8.47)*** -0.1639 (-6.39)*** 0.0837 (1.38) -0.5101 (-12.29)*** -0.4877 (-9.60)*** -0.3052 (-8.18)*** -0.1279 (-9.54)*** -0.1146 (-8.05)*** -0.0224 (-0.32) -0.6131 (-9.34)*** -0.5986 (-5.54)*** -0.5084 (-8.27)*** -0.2007 (-11.24)*** -0.1906 (-10.89)*** 0.0446 (1.47) -0.0970 (-8.42)*** This able repors he esimaed coefficiens 1 i in equaion (7). 'Nearby 1' denoes resuls obained using prices of 1-monh and 2-monh conracs o compue he slope. 'Nearby 2' denoes resuls obained using he prices of 2-monh and 3-monh conracs. 'Nearby 3' denoes he resuls obained using he prices of 3-monh and 6-monh conracs, ec. To faciliae he comparisons of coefficien esimaes, each price series has been divided by is own mean. Regressions are esimaed in firs differences due o he non-saionary daa series. The figures in parenheses are -saisics, which are for he hypohesis ha he coefficien esimae is differen from zero, and are correced for heeroskedasiciy and auocorrelaion based on he Newey-Wes (1987). *** and ** denoe significan a he 1% and 5% level. Implied cash flow yields and he level of spo prices Table 3 Primary aluminum Copper Nickel Lead Tin Zinc Nearby 1 0.3829 (9.79)*** 1.1594 (16.81)*** 0.1612 (9.11)*** 0.7743 (12.54)*** 0.2366 (13.25)*** 0.6175 (9.44)*** Nearby 2 0.3779 (6.39)*** 1.0350 (17.27)*** 0.1515 (7.01)*** 0.6938 (10.42)*** 0.2109 (6.48)*** 0.6005 (6.55)*** Nearby 3 0.3409 (10.58)*** 0.6215 (18.30)*** 0.1341 (5.30)*** 0.6693 (10.22)*** 0.1489 (5.76)*** 0.5579 (8.21)*** Nearby 4 0.3340 (18.11)*** 0.3999 (18.06)*** 0.0967 (8.75)*** 0.1932 (9.44)*** 0.1302 (9.12)*** 0.2024 (9.88)*** Nearby 5 0.3126 (14.85)*** 0.2756 (15.63)*** 0.0954 (8.34)*** 0.1023 (7.02)*** 0.1103 (7.26)*** 0.1836 (9.84)*** Nearby 6 0.2911 (11.34)*** 0.2523 (14.56)*** 0.0873 (8.21)*** 0.1004 (4.23)*** 0.1015 (6.98)*** 0.1656 (8.33)*** Nearby 7 0.2607 (9.85)*** 0.2109 (9.63)*** 0.0678 (6.96)*** 0.1589 (8.12)*** This able repors he esimaed coefficien obained by regressing implied cash flow yields on spo prices. The implied cash flow yield is compued as CKT, RKT, LN F T / F K / T K. 'Nearby 1' denoes resuls obained using prices of 1-monh and 2-monh conracs o compue he slope. 'Nearby 2' denoes resuls obained using he prices of 2-monh and 3-monh conracs. 'Nearby 3' denoes he resuls obained using he prices of 3-monh and 6-monh conracs, ec. LIBOR mid rae is used as a proxy
14 Invesmen Managemen and Financial Innovaions, 4/2004 for risk free rae. In order o faciliae he comparison of coefficien esimaes, each price has been divided by is own ime series mean. Regression is esimaed in firs differences due o he non-saionary daa series. The figures in parenheses are -saisics, which are for he hypohesis ha he coefficien esimae is differen from zero, and are correced for heeroskedasiciy and auocorrelaion based on he Newey-Wes (1987). *** and ** denoe significan a he 1% and 5% level. Forward ineres raes and he level of spo prices Table 4 Primary aluminum Copper Nickel Lead Tin Zinc Nearby 1-0.0002 (-0.07) Nearby 2 0.0007 (0.19) Nearby 3-0.0040 (-1.03) Nearby 4 0.0058 (139) Nearby 5-0.0038 (-0.64) Nearby 6 0.0033 (0.69) Nearby 7-0.0058 (-1.02) 0.0013 (0.56) 0.0013 (0.63) 0.0023 (1.04) 0.0019 (0.65) 0.0046 (1.39) 0.0037 (1.16) 0.0018 (0.63) -0.0006 (-0.22) 0.0028 (0.70) -0.0021 (-0.48) -0.0029 (-0.92) -0.0007 (-0.26) -0.0004 (-0.12) -0.0023 (-0.59) -0.0009 (-0.20) -0.0008 (-0.23) -0.0010 (-0.31) 0.0010 (0.32) 0.0032 (0.75) 0.0017 (0.35) 0.0001 (0.02) 0.0016 (0.38) -0.0045 (-1.02) -0.0045 (-0.75) -0.0078 (-1.18) -0.0009 (-0.11) -0.0010 (-0.66) -0.0030 (-0.78) -0.0098 (-0.38) 0.0059 (0.96) -0.003 (-0.29) -0.0002 (-0.81) -0.0032 (-0.87) This able repors esimaed coefficien obained by regressing forward ineres raes on spo prices. 'Nearby 1' denoes resuls obained using prices of 1-monh and 2-monh conracs o compue he slope. 'Nearby 2' denoes resuls obained using he prices of 2-monh and 3-monh conracs. 'Nearby 3' denoes he resuls obained using he prices of 3-monh and 6-monh conracs, ec. LIBOR mid rae is used as a proxy for he risk free rae. In order o faciliae he comparison of coefficien esimaes, each price has been divided by is own ime series mean. Regression is esimaed in firs differences due o he non-saionary daa series. The figures in parenheses are -saisics, which are for he hypohesis ha he coefficien esimae is differen from zero, and are correced for heeroskedasiciy and auocorrelaion based on he Newey-Wes (1987). *** and ** denoe significan a he 1% and 5% level. Our resuls show ha i is C KT raher han R KT ha is he source of mean reversion in base meal prices, consisen wih hose of Fraser and McKaig (1999) who find no evidence of negaive coefficien esimaes for he relevan variables in he forward ineres rae regression for aluminum and copper. However, Bessembinder e al. (1995) show ha boh C KT and R KT induce mean reversion in precious meals (silver, gold, and plainum). In conjuncion wih he evidence in Tables 1 and 2, he coefficien esimae for he recen nearby copper fuures is noiceably larger in magniude han ha for oher meal fuures. This is likely due o he alleged Sumiomo copper manipulaions occurred on he LME 1. Fuures manipulaions usually resul in a siuaion where shor-run supply becomes subsanially more inelasic han long-run supply, or more volaile convenience yields o marginal copper holders (backwardaions) (Fama and French, 1988b). Therefore, when spo prices increase, here exiss a larger increase in implied cash flow yields, resuling in sronger mean reversion in prices. 1 I is alleged ha Sumiomo Corporaion inermienly manipulaed he world copper marke mainly operaed on he LME copper fuures marke over he period of 1993-1996, which caused noice and resuled in inquiries iniiaed by he Securiies and Invesmen Board (SIB) and he Commodiy Fuures Trading Commission (CFTC) a he end of 1995. The manipulaion scheme failed due o fierce aack by a hedge fund in June 1996, resuling in copper prices falling from $2,700/onne o below $2,000/onne. For an analysis of he effec of silver manipulaions on he forward/fuures spo relaions, see Fama and French (1988b).
Invesmen Managemen and Financial Innovaions, 4/2004 15 V. Conclusions In his paper, we examine mean reversion in equilibrium base meal prices using he erm srucure of fuures prices a which conracs are raded on he LME. We find evidence of mean reversion in each of he six base meals primary aluminum, copper, nickel, lead, in, and zinc. The esimaed elasiciy of disan fuures prices wih respec o spo prices is less han uniy for all meals and delivery daes. We also find a negaive relaion beween spo prices and fuures erm slope, alhough he magniude of mean reversion differs across hese meals. There exhibis he greaes degree of mean reversion in copper prices, wih 40% of a price shock o be reversed in he subsequen year, on average; whereas he leas degree of mean reversion presens in nickel prices, wih only abou 10% of a price shock o be reversed. If he no arbirage cos-of-carry relaion holds, expeced mean reversion can arise from he covariaion beween spo prices and implied cash flow yields, and/or he correlaion beween spo prices and forward ineres raes. Our empirical evidence shows ha, similar o he findings in Fraser and McKaig (1999), mean reversion in all base meal prices sems solely from he implied cash flow yield sensiiviy of spo prices. References 1. Bessembinder, H., Coughenour, J. F., Seguin, J. P. and Smoller, M. M. (1995) Mean reversion in equilibrium asse prices: Evidence from he fuures erm srucure, Journal of Finance, 50, 361-375. 2. Brennan, M. J. (1958) The supply of sorage, American Economic Review, 48, 50-72. 3. De Bond, W. F. and Thaler, R. (1985) Does he sock marke overreac? Journal of Finance, 40, 793-805. 4. Dusak, K. (1973) Fuures Trading and Invesor Reurns: An Invesigaion of Commodiy Marke Risk Premiums, Journal of Poliical Economy, 81, pp. 1387-1406. 5. Fama, E. F. and French, K. R. (1987) Commodiy fuures prices: Some evidence on 6. forecas power, premiums, and he heory of sorage, Journal of Business, 60, 55-73. 7. Fama, E. F. and French, K. F. (1988a) Permanen and emporary componens of sock prices, Journal of Poliical Economy, 96, 246-273. 8. Fama, E. F. and French, K. F. (1988b) Business Cycles and he Behavior of Meals Prices, Journal of Finance, 43, 1075-1093. 9. Fraser, P. and McKaig, A. J. (1999) Do invesors expec mean reversion in asse prices? Journal of Business Finance and Accouning, 26, 57-81. 10. Heaney, R. (1998) A es of he cos-of-carry relaionship using he London Meal Exchange lead conrac, Journal of Fuures Markes, 18, 177-200. 11. Poerba, J. and Summers, L. (1988) Mean reversion in sock prices: Evidence and implicaions, Journal of Financial Economics, 22, 27-59. 12. Swanson, N. R., Zeng T., and Kocagil, A. E. (1996) The probabiliy of mean reversion in equilibrium asse prices and reurns, Working Paper, Pennsylvania Sae Universiy. 13. Sweeney, R. J. (2000) Mean reversion in G-10 nominal exchange raes, Working paper, Georgeown Universiy. 14. Telser, L.G. (1958) Fuures rading and he sorage of coon and whea, Journal of Poliical Economy, 66, 233-255. 15. Theil, H. (1966) Applied Economic Forecasing, Volume 4. Amserdam: Norh-Holland. 16. Working, H. (1949) The heory of he price of sorage, American Economic Review, 12, 1254-1262.