An Intervention Analysis on the Relationship between Futures Prices of Non-GM and GM Soybeans in China

Transcription

1 Journal of Busness & Economc Polcy Vol., No. 4; December 05 An Inervenon Analyss on he Relaonshp beween Fuures Prces of Non-GM and GM Soybeans n Chna Nanyng Wang Jack E. Houson The Unversy of Georga Ahens, GA 3060, USA Absrac In 004, he Dalan Commody Exchange (DCE) nroduced a separae fuures conrac for No. soybeans, whch ncludes GM soybeans. Wh hs change, he No. soybean fuures conrac defauled o a non-gm conrac. We defne he dfference beween he prces of non-gm and GM soybeans as he prce premum for non- GM soybeans. An nervenon analyss s used o es he effecs of he evens on he prce premum. We nvesgae hree evens wo conrac specfcaon changes n 005 and 00 and one gran law mplemenaon n 0 focusng on boh he drecon and sze of her mpacs. In concluson he conrac specfcaon changes dd affec he prce premum. I s also found he law ssue has permanenly ncreased he prce premum. Sudyng he marke response lnkages beween he wo soybean fuures markes s helpful for undersandng wheher he newly opened GM soybean fuures marke ransms prce nformaon effecvely. Keywords: Chna soybeans, GMO, non-gmo, Inervenon analyss, Impulse response funcon.. Inroducon In 00, Chna adoped a mandaory labelng polcy of genecally modfed (GM) food producs. Ths law mposed mandaory labelng for all GM food producs so ha consumers can denfy producs conanng genecally modfed organsms (GMOs). Chna also sared a new radng sysem n 00 n an effor o separae he radng of mpored GM soybeans from domescally produced non-gm soybeans. The sraegy of separang radng was nended by Chnese regulaors o proec domesc on-gmo producon, provde non-gm soybean growers a hgher sellng prce, and faclae markeng. Also n 00, L e al. (003) conduced a survey n Bejng ha revealed ha he wllngness o pay (WTP) for GM rce and GM soybean ol was posvely affeced by respondens percepons of her characerscs. These resuls mply ha, unlke Europe and Japan, here s a poenal marke for GM foods n Chna. However, recenly non-gm soybeans are wdely perceved o be healher han GM, such ha GM soybeans may no be perfec subsues for non-gm soybeans n eher consumpon or processng demand. A naural progresson for he prce dscovery process for a regulaed dfferenaed marke s he developmen of a fuures marke conrac. Thus, esablshng qualy specfcaons wh an deny-preserved marke, such as he Dalan Commody Exchange (DCE) GM soybean conrac, s mporan. The lack of a well-defned and lqud cash non-gmo soybean marke does no appear o hamper he developmen of he non-gmo fuures conrac. On December, 004, he DCE launched a new knd of more nclusve fuures conrac o ncorporae boh GM and non-gm soybeans, ha s, he SB# soybean conrac, whch made SB# a non-gm soybean conrac by defaul. SB# ams o connec Chna s and nernaonal soybean fuures markes and enhance he perceved mpacs of Chna s demands on nernaonal soybean markes. Ths conrac can be consdered as he frs publc fuures conrac for an deny-preserved (IP) crop n Chna. I also brough new challenges o Chna s soybean fuures markes research. Snce he nroducon of boech commodes n 996, farmers have rapdly adoped hs new echnology for producon, prmarly for soybeans, coon, and corn (Nelson, 00). In 03, GM feld area rose o a global oal of 74 mllon hecares. (GMO Compass). In erms of valuaon and prce changes, GM soybeans have a posve mpac on producer reurns (oupu), because here s a decrease n producon coss, easer managemen and hgher yelds. Chna has become he sxh larges producer of GM commodes, followng he Uned Saes, Brazl, Argenna, Inda and Canada (GMO Compass, 04).

2 ISSN (Prn), (Onlne) Cener for Promong Ideas, USA Commercalzed GMO n Chna nclude B coon, delayed-rpenng omaoes, cucumber mosac vrus (CMV) ressan swee peppers, and color-alered peunas. However, as of hs wrng, no major GM gran or olseed crop, such as soybeans, corn, rce, or whea, has been approved for commercalzaon n Chna. Ths makes Chna he larges producng counry of non-genecally modfed soybeans. Soybeans are prmarly used as npus for Chnese food producs. Non-GM soybeans are mosly used for food and food producs. On he oher hand, mpored GM soybeans are manly used for vegeable ol, feed, and ndusral purposes. However, some raders may be purchasng non-gm soybeans for he same purpose as GM soybeans, snce here are no legal barrers on usng non-gm soybeans for ol or processng. Parcell (00) defnes he dfference beween he prces of non-gm and GM soybean fuures conracs as he prce premum for non-gm soybeans. The objecve of hs paper s o examne how effcenly hs prce premum for non-gm soybean fuures reacs o hree evens, ncludng wo conrac specfcaon changes and one legal ssue by denfyng he magnude and duraon of her mpacs. Inervenon analyss s used for hs purpose. Sudyng he marke lnkage beween he wo soybean fuures markes s helpful for undersandng wheher he newly opened GM soybean fuures marke ransms prce nformaon effecvely and effcenly. There have been some breaks ha may have nfluenced he prce relaonshps of he wo soybean fuures markes on DCE. Ths dscusson ncludes hree evens: () The DCE mplemened amendmens o he GM soybean conrac specfcaon o make ha conrac more nearly conform wh he nernaonal soybean rade sandards n 005; () The DCE made anoher conrac specfcaon change on boh non-gm and GM conracs o sharpen he dsncon beween non-gm and GM soybean fuures conracs and sablze he markes for non-gm and GM soybeans n 00; and () The Governmen of Chna ssued he Gran Law and an explanaory noce for he regulaon of GM producs on February, 0. We use an nervenon analyss frs suggesed by Box and Tao (965, 975) and furher developed by Larcker e al. (980), Enders e al. (99) and ohers. Inervenon analyss has advanages over he sandard even sudy mehod frs nroduced by Ball and Brown (968) and Fama e al. (969), snce allows he observed auocorrelaon n he model resduals o be removed, hus provdng mproved esmaes for relable sascal esng. Also, nervenon analyss provdes an mpulse response funcon o sudy he ransonal effecs followng an even.. Leraure Revew Inervenon mehodology was developed by fnancal economss o assess he performance of secures markes. Karagozoglu, Marell and Wang (003) esed how a change n he conrac sze of S & P 500 fuures conracs a he Chcago Mercanle Exchange affecs radng volumes afer he change s conduced. Smlarly, Thomakos e al. (008) analyzed he effecs of macroeconomc announcemens on reurns volales, covarance s and correlaons beween Eurodollar fuures and U.S. Treasury bond fuures and showed ha all hree reac o he nformaon conen of announcemens. Lle research has been underaken o assess he marke funconaly of deny-preserved crops, such as he GM soybean fuures markes. Parcell (00) descrbes hs new marke for non-gm soybean fuures a he Tokyo Gran Exchange (TGE) and compues he prce premum for non-gm soybean conracs. Aruga (0) examnes how effcenly he prce premum for non-gm soybeans a he TGEreac o an announcemen o change he conrac un, supplers, and expraon dae on he convenonal soybean fuures conrac. To dae, however, here s lle publshed on he workngs of GM soybean fuures markes n Chna. Wang and Ke (005) suded he effcency of he soybean fuures marke and concluded ha here s a long-erm equlbrum relaonshp beween he fuures prce (non- GMO) and cash prce for soybeans and he soybean fuures marke s weakly shor-erm effcen. Zheng e al. (0) esed he prce dscovery of he Chnese soybean fuures marke and ndcaed ha he Chnese non-gm soybean fuures marke s effcen, bu hey dd no analyze he GM soybean fuures. Our sudy would be he frs o analyze he relaonshp beween he non-gm and GM soybean fuures markes n Chna. The resul of hs sudy wll help undersand wheher he newly developed GM soybean fuures marke provdes valuable nformaon for s prce dscovery process. 3. Daa The daa are obaned from he DaaSream 5. provded by Thomson Reuers. The prce un s provded n Chnese Yen per merc one. Due o he lack of lqudy of frs nearby conracs, we consruc me seres of daly selemen prces of he hrd nearby conracs. When he fuures prce moves no he maury monh, we roll over he fuures prce o he nex maury monh. Only observaons ha have boh non-gm and GM prces on a gven day are used n he analyss.

3 Journal of Busness & Economc Polcy Vol., No. 4; December 05 A separae radng for GM soybean conracs sared on December, 004, and snce we use he hrd nearby conracs, he GM soybean fuures conracs exend back from January, 005. Table shows he deals of he conrac specfcaons for non-gm and GM soybeans. Fuures Premum Parcell (00) defnes he prce dfference beween he prces of non-gm and GM soybean fuures conracs as he prce premum for non-gm soybeans. We use he same defnon n hs sudy. We ake he dfference beween daly selemen prces of he hrd nearby non-gm and GM soybean fuures conracs as he prce premum. We frs es f here s srucural change n he premum seres. To examne hs, he Ba-Perron mulple srucural change es (Ba and Perron, 998) and Chow es are appled. Boh es resuls show ha wo breaks are he sascally adequae number of breaks for hs seres, whch are Ocober 3, 006, and Sepember 3, 0. The premum seres hus are spl no hree perods denfed by he above wo breaks. As seen n he Fgure, he prce premum for non-gm soybeans was posve from begnnng of he daase unl 00. Beween lae 00 o md 0, he prce for GM soybeans was surprsngly hgher han ha for he non-gm soybeans. Reasons for hs mgh be: () Durng ha perod, he concep of GM was no well known by Chnese consumers, and due o he hgher ol exracon rae of GM soybeans (GMO Compass), he processor would pay a premum for he GM soybeans; () The world soybean prce, whch ncluded large percenage of GM soybeans, ncreased dramacally afer he food crss n 006 and 007. (USDA) A he same me, he producon of non-gm soybeans n Chna could no mee domesc demand. Thus he mpored amoun of soybeans dd no decrease even hough he prce was hgher han her domesc non-gm soybeans. Sarng 03, he premum for non-gm soybeans becomes posve and remaned level unl he end of our daa perod. Ths could be he resul of he wdespread world conroversy of he safey ssue of GMOs. Descrpve sascs of he selemen prce of non- GM and GM soybeans, as well as he premums, are summarzed n Table. There are,365 observaons n he sample. The average daly premum s -5.5 CNY per merc on wh a sandard devaon of 56.7 Chnese Yuan. The average premum s posve n perod and n perod 3, bu negave n perod.ths sgnfcan change of values n premums reflecs he change of consumers audes. Even Descrpons There have been some dsrupons ha affeced he soybean fuures markes a he DCE and ha hese breaks may have nfluenced he prce relaonshps of he wo soybean fuures markes. Ths dscusson ncludes followng evens. Frs, he DCE mplemened amendmens o he GM soybean conrac specfcaon n 005. Ths change was nended o make Chna's GM soybean markes more closely conform wh he nernaonal soybean rade sandards, gvng prory o mpored soybeans. Several grade specfcaons changed here. For example, he new conrac specfcaon changes he ol exracon rae up o %. The new specfcaon sars from conracs raded n January 006, whch sared on Ocober 0, 005.Secondly, n 00, The DCE made anoher conrac specfcaon change on boh he non-gm and he GM conracs. The DCE was expecng ha he specfcaon change would sharpen he dsncon beween non-gm and GM soybean fuures conracs and sablze he markes for non-gm and GM soybeans. The deals of he specfcaon changes nclude he revsed qualy sandard and new mandaory requremen regardng new regsraons of sandard warrans for soybeans accordng o he new naonal labelng sandards. The packagng maerals, or accompanyng documens, should ndcae he produc name, caegory, grade, place of orgn, harves year and monh. The conrac usng he new specfcaon sars from he conracs raded n March00. Thrdly, on Februaryh, 0, he Governmen of Chna ssued he Gran Law and an explanaory noce for he regulaon of GM foods. I was he frs me ha GM food conrol laws had been made a he naonal level n Chna. The law saes ha: The scenfc research, expermen, producon, markeng and expor and mpor of genecally modfed gran seeds should comply wh relevan sae regulaon. No nsuon or ndvdual should apply genecally modfed echnology o major gran crops whou permsson. The law apples o grans, edble vegeable ol and olseeds. Ths mples ha he producon, rade and consumpon of unauhorzed genecally modfed gran and olseeds wll be banned n Chna. 4. Mehodology An nervenon analyss s used o es he effecs of he evens on he prce premum for non-gm soybeans n each sub-perod. We ulze he followng economerc ARMA model: 3

4 ISSN (Prn), (Onlne) Cener for Promong Ideas, USA Pr eemum ( ) Premum EVENT () j j 0 j j where Premum s he emum dsurbance; EVENT s an even dummy varable; and, Pr n perod ; s a consan; () s a me rend; s a normal.. d. j j and are he coeffcens o be esmaed. We consder fve nervenon funcons n hs sudy. As presened n Fgure, n all fve models, EVENT akes he value of 0 before even day, and on he even day. The value of EVENT beyond even day depends on he chosen nervenon funcon. In model he nervenon funcon represens a pure jump, where he even dummy remans equal o uny unl he end of he sub-sample perod. The pure jump nervenon funcon arguably models he effec of he even as a consan permanen change o he premum whn he perod. Model s an mpulse funcon ha bes characerzes a purely emporary nervenon for one monh afer he even. Model 3 hrough model 5 are prolonged mpulse funcons ha assume ha he nervenon wll reman o be uny for one monh and begn o decay and reachng zero afer 80 days, 05 days and 40 days for models 3, 4, and 5, respecvely. Equaon can be expressed as: B L)Premum ( ) ( L) EVENT ( () where B (L) and (L) are polynomals n he lag operaor L. The coeffcens of B(L) are he auoregressve (AR) componens, and he coeffcens of (L) are he movng average (MA) componens of he auoregressve movng average (ARMA) model. The coeffcen s of specal neres o he analyss, as provdes he nformaon abou he mpac of he even on he performance of he dfference beween prce of non-gm and GM soybeans. An augmened Dckey Fuller es was performed on premum seres o ensure ha hese hree sub-seres dd no conan a un roo. Sequenal -ess begnnng wh lag were ulzed o deermne he approprae number of lags for he un roo es (Ng and Perron 995). The hree evens whn each perod are assumed o be exogenous srucural breaks for he premum seres. The un roo hypohess was rejeced a he less han % level for he frs wo perods; however, was no rejeced for he hrd perod. The absence of a un roo means ha he effec of he frs wo evens wll evenually de ou, bu no for he hrd perod case. We hus add he rend n he ARMA model for he hrd perod. The esmaon procedure was conduced usng he sandard Box Jenkns mehod. In choosng among alernave plausble ARMA models, he lowes Akake Informaon Creron mehod was ulzed. Dagnosc checkng was performed by plong he resduals and he correlogram of resduals squared o nsure ha hey are characerzed by a whe nose process. Also, he auoregressve heeroskedascy (ARCH) Lagrange mulpler es was performed and resuled n on-sgnfcan sascs, whch mples he absence of he ARCH effec. 5. Resuls The bes fng model for hese hree perods s an ARMA (, ) model. I can be wren as: Premum ( ) Premum EVENT (3) j0 j j The emprcal resuls of he effecs of hese hree evens on he premum for non-gm soybeans for all fve models are repored n Table 3. I presens maxmum lkelhood esmaes of he nervenon analyss of daly premum for non-gm soybeans n he Dalan fuures marke usng ARMA (, ) models. To accoun for he global fnancal crss, we creae a varable, CRISIS, whch equals uny beween Sepember 5, 008 and June 30, 009 n perod (Glber, 00). The sascally sgnfcan coeffcens of he even dummes represen he nal, or mpac, effecs of he evens. In he frs perod, he coeffcens ndcae he nal ncrease of 5.6 o 9.6 CNY per merc on per day for model hrough 5. To provde he economc sense of he ncrease n he premum performance, we compare hs number wh he average premum per day before he even dae: represens a 35.6% o 67.5% ncrease n premum. In he second perod, he nal effec s a decrease of 33.6 o 58.3 CNY per merc on per day, whch represens a decrease of 38.% o 66.% n he premum. 4

5 Journal of Busness & Economc Polcy Vol., No. 4; December 05 As for he hrd perod, he even has an nal effec of an ncrease of 6.9% o 30.7% n he premum. As one can see, he resuls are heavly nfluenced by he choce of he nervenon funcon. Ths llusraes he mporance of he nervenon funcon chosen for he analyss. The long-run effec esmaon requres judgmen n model selecon. Que lkely, prolonged mpulse models, such as models 3, 4, 5, wh he decayng funcon would be approprae n he case of he frs wo evens, as he exogenous effecs would dsspae over me and he premum would begn o move back o her orgnal paerns. However, hs requres arbrarly seng he even dummy o zero a some pon of me afer he aack whle he even could sll be a sgnfcan facor n he premum. Some raders n he soybean fuures marke may sll consder he conrac specfcaon change and he law ssue effec of he GM producs when hey perform n he soybean fuures marke. Hence, he even dummy ha says equal o uny hrough he end of he sample perod s a reasonable modelng assumpon. Based on hs judgmen, we ulze model o esmae he long-run effec of he hree evens and he mpulse response funcons. The longrun effec of he evens can be assessed by calculang he change n he long-run mean of he premum seres n model. The long-run effec (LRE) of nervenon s gven by he followng equaon: LRE (4) ( ) where and are AR erm coeffcens of ARMA(,) model presened n equaon 3. Afer subsung he coeffcens n Eq. 3.4, we fnd ha LRE equals o 3., negave 54.5 and 5.8 CNY, respecvely, n each perod usng model. The LRE yelds much larger economc sgnfcance han he nal effec n he frs perod. The magnude of he mpac s much smaller han he cumulave change of he premum allegedly caused by he even. However, he LRE of he las wo perods s very smlar n magnude o her nal effec, suggesng ha almos all of he premum change can be arbued o he even n he las wo perods. Impulse Response Funcon One of he advanages of he nervenon analyss s ha he model can provde researchers wh addonal nformaon, such as he ransonal effecs of an even. As mpled by he un roo es, he effec of he even of he frs wo perods wll evenually de ou and he daly decrease wll dampen and evenually dsappear, bu no for he hrd perod. The reducon rae of daly losses ha are arbuable o he even can be provded by he mpulse response funcon. Usng a lag operaor we rewre Eq. 3 as: ( L L ) Premum EVENT j (5) j j0 and Premum ( EVENT ) j (6) j ( L L ) j0 Nex, we subsue wh ( L L ) ( L)( L) where and are characersc roos of he polynomal B ( L) 0. Wh he characersc roos, he ARMA (, ) model can be nvered o oban he mpulse response funcon. Premum j 0 EVENT (7) j j Equaon 8 s an mpulse response funcon. By dfferenang Eq. 3.7 and updang by perods, one can race he response of he premum's performance o he even: d Premum ( )( ) (8) devent Snce n he lm,, he LRE of he nervenon: 5

6 ISSN (Prn), (Onlne) Cener for Promong Ideas, USA LRE (9) )( ) ( ) ( Equaon 9 can be ulzed o calculae he effec of he even n a predeermned perod of me afer he occurrence. For nsance, f an even happens n perod, one can expec he decrease n daly premum n perod +3 by: d Pr emum 3 3 ( )( 3 ) (0) devent Where reflecs he drec mpac of he premum performance and he followng erms reflec he effec of he even mulpled by he effec of Premum, Premum and Pr emum, respecvely. Fgure 3 shows he mpulse response of he premum's performance o he hree evens ulzng he esmaes of model, where he vercal bars represen he rajecory of he IRF and he lnes are he smoohed rend usng movng averages mehod. For model, he characersc roos of he polynomal B ( L) 0, and are esmaed o be and negave , 0.78 and negave 0.975, and negave.6783, respecvely, for hree perods. The area above he curve represens he cumulave effec on he premum. Snce he absolue values of boh and are less han uny n he frs wo perods, he relave mpac on he premum performance s decreasng wh me and reaches zero afer 330 days and 0 days, respecvely. However, he relave mpac of he ssue of law keeps a level of 0 CNY per merc on, snce he absolue value of s more han uny. 6. Conclusons As he larges soybean mporer, Chna s hgh demand means ha many foregn growers canno gnore prce sgnals from Chna when makng mporan producon and markeng decsons. Ths paper examned how effcenly he DCE non-gm and GM soybean fuures markes reac o wo conrac specfcaon changes and one law ssue by esng he nfluence on he prce premum for non-gm soybeans. We mplemen nervenon analyss o en years of daly prces on soybean fuures conracs o analyze he paern of he marke responses o hree major evens (he conrac specfcaon changes n 005 and 00, and he gran law ssue n 0), of whch effecs are consdered o perss for a long perod of me raher han a one-day jump. The consequences of hese evens on he prce premum were capured by an ARMA model. Resuls show ha premum response o each of hese hree evens s sascally sgnfcan, and he duraons are dfferen for each even. The range for change of premum s negave 60 o posve 70 percenage pons, wh he mpac of he conrac specfcaon change n 00 beng he larges. The resuls revealed ha he prce premum for non-gm soybean fuures conracs changed subsanally afer evens. Among he hree evens, he mpac of gran law ssue on premum s permanen n our sample perod. In concluson, he conrac specfcaon change from he DCE for he soybean fuures conrac dd affec he prce premum beween he GM and non-gm soybean fuures conracs. Therefore, hese wo cases of changes can be consdered as successful. Hence, here was an nformaonal effcency n he marke. I s also found from he sudy ha he effec of he legal ssue dd no dsappear for he prce premum for non-gm soybeans. I permanenly rased he prce premum for non-gm soybean. The dspue of GM foods nvolves consumers, farmers, boechnology companes, governmenal regulaors. However, hs dd no deer he developmen of he GM fuures marke n Chna. The fac ha he non-gm and GM soybeans fuures markes are effcen can provde governmen planners more evdence and confdence o help he sar of he fuures radng for oher commodes. For nernaonal soybean growers, raders and processors, an effcen DCE GM soybean fuures marke wll generae sronger neres n parcpang n Chnese fuures radng asa mechansm o hedge nernaonal ransacons and agans varaons n her local markes, whch may arse from he growng Chnese demand whch lead growng mpored GM soybeans. 6

7 Journal of Busness & Economc Polcy Vol., No. 4; December 05 Table. Summary of he conrac specfcaon a he DCE SB # (Non-GM) SB # (GM) Dae Tradng Began 998 Dec h, 004 Conrac Un 0 merc ons Tradng Hours 9:00-:30 a.m, :30-3:00 pm. Bejng Tme, Monday-Frday Conrac Monh Jan, Mar, May, July, Sep, Nov Prce Quoaon CNY/MT Las Tradng Day 0h radng day of he delvery monh Las Delvery Day 3rd day afer he las radng day of he delvery monh Sandard Grade No. 3 Yellow; GM soybeans are no permed o be delvered Impored GM soybeans Delvery Pons The warehouses apponed by he DCE Source: DCE04 Table. Summary Sascs N Mean Sandard devaon Mnmum Maxmum Perod Non_GM Soybean Prce GM Soybean Prce Premum Perod Non_GM Soybean Prce GM Soybean Prce Premum Perod 3 Non_GM Soybean Prce GM Soybean Prce Premum Whole Perod Non_GM Soybean Prce GM Soybean Prce Premum Source: Daa sream 5. 7

8 ISSN (Prn), (Onlne) Cener for Promong Ideas, USA Table 3. The mpac of he evens on Premum of non-gm soybean n chna Perod Model Model Model 3 Model 4 Model 5 Consan 44.35a a 56.9a a 56.8a (9.897) (7.04) (6.86) (6.868) (7.68) Premum a -0.08a -0.0a -0.9a -0.3a (0.07) (0.030) (0.08) (0.08) (0.08) Premum a 0.9a 0.6a 0.7a 0.3a (0.08) (0.00) (0.09) (0.09) (0.00) ε- 0.94a 0.908a 0.99a 0.99a 0.97a (0.03) (0.033) (0.09) (0.09) (0.030) EVENT 5.593a 4.8a.93b 0.594a 9.574a (3.6) (.847) (8.88) (9.49) (.57) Adj. R-sq Inal effec 35.6% 55.30% 5.37% 47.03% 67.54% LRE LRE(%) 5.76% 6.9% 58.58% 5.57% 76.60% Perod Model Model Model 3 Model 4 Model 5 Consan b a a b b (54.639) (6.35) (6.590) (6.884) (6.99) Crss 6.789a a a a a (.35) (3.007) (.950) (.89) (.893) Premum a a a a a (0.06) (0.057) (0.057) (0.057) (0.057) Premum a a a a a (0.059) (0.055) (0.055) (0.054) (0.054) ε a a a a a (0.63) (0.05) (0.05) (0.05) (0.05) EVENT c a a a a (7.49) (.) (6.459) (4.57) (4.808) Adj. R-sq Inal effec -6.07% -66.9% -38.% % 50.5% LRE LRE(%) -6.9% % % -55.3% % Perod 3 Model Model Model 3 Model 4 Model 5 Consan a a a a a (38.68) (8.705) (30.763) (3.095) (33.908).908 a.993 a.994 a.995 a.997 a (0.4) (0.39) (0.33) (0.34) (0.38) Premum a a a a a (0.87) (0.80) (0.77) (0.83) (0.76) Premum a a a a a (0.8) (0.75) (0.7) (0.78) (0.7) ε a a a a a (0.66) (0.59) (0.56) (0.6) (0.55) EVENT a 9.8 a a 39.7 a a (7.350) (5.564) (3.979) (3.707) (6.065) Adj. R-sq Inal effec 30.65% 6.87% 6.0%.7% 9.95% LRE LRE(%) 30.5% 6.79% 5.99%.60% 9.8% Noe: a Sascal sgnfcance a he % level b Sascal sgnfcance a he 5% level c Sascal sgnfcance a he 0% level 8

9 Journal of Busness & Economc Polcy Vol., No. 4; December 05 Noe: The prces for he non-gm and GM soybeans are gven n Chnese yen and are m of soybeans. Fgure. Prce premum for non-gm soybeans (prce dfference beween he non-gm and GM soybean fuure conrac) Panel A: Model Panel B: Model PanelC: Model 3 Panel D: Model 4 Panel D: Model 5 Fgure. Inervenon models. Panel A hrough E llusrae nervenon funcons ulzed n he analyss. 9

10 ISSN (Prn), (Onlne) Cener for Promong Ideas, USA Panel A Panel B Panel C Fgure 3. Impulse response funcon of he premum for non-gm soybean prce afer even n 005 (Panel A), 00 (Panel B) and 0 (Panel C) References Aruga, K. (0). Marke effcency n he non-genecally modfed soybean fuures marke. Journal of Agrculural & Food Indusral Organzaon, 9(), -. Box, G. E.,& Tao, G. C. (965). A change n level of a non-saonary meseres. Bomerka, 5(), 8-9. Ball, R.,& Brown, P. (968). An emprcal evaluaon of accounng ncome numbers. Journal of accounng research, 6(), Box, G. E., & Tao, G. C. (975). Inervenon analyss wh applcaons o economc and envronmenal problems. Journal of he Amercan Sascal Assocaon, 70(349), Ba, J.,&Perron, P. (998). Esmang and esng lnear models wh mulple srucural changes. Economerca, 66(), Enders, W., Sandler, T., & Parse, G. F. (99). An economerc analyss of he mpac of errorsm on oursm. Kyklos, 45(4), Fama, E. F., Fsher, L., Jensen, M. C., & Roll, R. (969). The adjusmen of sock prces o new nformaon. Inernaonal economc revew, 0(), -. Glber, C. L., & Morgan, C. W. (00). Food prce volaly. Phlosophcal Transacons of he Royal Socey B: Bologcal Scences, 365(554), Global GM planng 03, 04. GMO Compass. [Onlne] Avalable: hp:// Genecally engneered plans for food and feed. 05. U.S. Food and Drug Admnsraon. [Onlne] Avalable: hp:// rch/boechnology/defaul.hm. Karagozoglu, A. K., Marell, T. F., & Wang, G. H. (003). The spl of he S&P 500 fuures conrac: Effecs on lqudy and marke dynamcs. Revew of Quanave Fnance and Accounng, (4), Larcker, D. F., Gordon, L. A., & Pnches, G. E. (980). Tesng for marke effcency: a comparson of he cumulave average resdual mehodology and nervenon analyss. Journal of Fnancal and Quanave Analyss, 5(0),

11 Journal of Busness & Economc Polcy Vol., No. 4; December 05 L, Q., Curs, K. R., McCluskey, J. J., & Wahl, T. I. (003). Consumer audes oward genecally modfed foods n Bejng, Chna. The Journal of Agro boechnology Managemen and Economcs, 5(4), Ng, S., & Perron, P. (995). Un roo ess n ARMA models wh daa-dependen mehods for he selecon of he runcaon lag. Journal of he Amercan Sascal Assocaon, 90(49), Nelson, G. C.,& Pno, A. De. (00). Genecally modfed organsms n agrculure: economcs and polcs. San Dego: Academc Press. Parcell, J. L. (00). An nal look a he Tokyo Gran Exchange non-gmo soybean conrac. Journal of Agrbusness, 9(), Thomakos, D. D., Wang, T., Wu, J., & Chuderewcz, R. P. (008). Macroeconomc announcemens, nraday covarance srucure and asymmery n he neres rae fuures reurn. Journal of Fuures Markes, 8(9), Wang, H. H., & Ke, B. (005). Effcency ess of agrculural commody fuures markes n Chna. Ausralan Journal of Agrculural and Resource Economcs, 49(), 5-4. Zheng, S., Xu, P., Foser, K., & Wang, Z. (0). Prce dscovery n he Chnese soybean fuures marke: New evdence abou non-gmo soybean radng. Journal of Chnese Economcs, (), 3-3.