Does Chinese OFDI really promote export?

Lin, Chih-Fan A icle Does Chinese OFDI eally p omo e expo ? China Finance and Economic Re iew P o ided in Coope a ion wi h: Sp inge Na u e Sugges ed Ci a ion: Lin, Chih-Fan (2016) : Does Chinese OFDI eally p omo e expo ?, China Finance and Economic Re iew, ISSN 2196-5633, Sp inge , Heidelbe g, Vol. 4, pp. 1-16, h ps://doi.o g/10.1186/s40589-016-0037-8 This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/194309 S anda d-Nu zungsbedingungen: Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen Zwecken und zum P i a geb auch gespeiche und kopie we den. Sie dü en die Dokumen e nich ü ö en liche ode komme zielle Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich machen, e eiben ode ande wei ig nu zen. So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen (insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en, gel en abweichend on diesen Nu zungsbedingungen die in de do genann en Lizenz gewäh en Nu zungs ech e. Te ms o use: Documen s in EconS o may be sa ed and copied o you pe sonal and schola ly pu poses. You a e no o copy documen s o public o comme cial pu poses, o exhibi he documen s publicly, o make hem publicly a ailable on he in e ne , o o dis ibu e o o he wise use he documen s in public. I he documen s ha e been made a ailable unde an Open Con en Licence (especially C ea i e Commons Licences), you may exe cise u he usage igh s as speci ied in he indica ed licence. h ps://c ea i ecommons.o g/licenses/by/4.0/ RESEARCH Open Access Does Chinese OFDI eally p omo e expo ? Chih-Fan Lin Co espondence: [email p o ec ed] The Wang Yanan Ins i u e o S udies in Economics, Xiamen Uni e si y, No. 422 o Siming Sou h Road, Siming Dis ic , Xiamen Ci y, Fujian P o ince, People’s Republic o China Abs ac Backg ound: Mos pape s ha e come o posi i e conclusions ega ding whe he Chinese ou wa d o eign di ec in es men eally p omo e expo . Me hods: This pape con ibu es by co ec ing model misspeci ica ion, w ong a iable selec ion and es ima ion me hods which a e p e alen in exis ing s udies and eexamine he ela ionship be ween ou wa d o eign di ec in es men and expo ade o China using panel da a om 2003 o 2014. Resul s: The eg essions indica e ha , on a e age, he poin es ima e o he elas ici y be ween Chinese ou wa d o eign di ec in es men and expo ade is a mos 0. 073, and i is no s a is ically signi ican . Sub-sample eg essions show ha Chinese in es men in de eloped economies sligh ly subs i u es expo while in es men in de eloping economies complemen s expo . Bu hese e ec s disappea when coun y-speci ic e ec s a e con olled. Yea -by-yea eg essions show ha he complemen a y e ec o OFDI on expo is on a s eady ise. Conclusions: Using co ec ly speci ied model wi h mo e app op ia e a iable selec ion and es ima ion me hods and ake in o accoun he quan i a i e di e ence be ween OFDI and expo , any “subs i u iona y”o “complemen a y”e ec is indeed negligible. Keywo ds: Ou wa d o eign di ec in es men , Expo , Complemen a ion e ec , Subs i u ion e ec Backg ound In oduc ion Wi h he con inuous g ow h o Chinese economy, Chinese ou wa d o eign di ec in es - men de elops apidly. F om he yea o 2003 when he “s a is ical bulle in o Chinese o - eign di ec ”was i s published o he yea o 2014, Chinese o eign di ec in es men low g ew om $28.5 billion o $1231.2 billion and he accumula ed s ock g ew om $299 bil- lion o $8826.4 billion, wi h he a e age annual g ow h a e being 40.8 and 36.0 %, espec - i ely. In 2014, Chinese o eign di ec in es men accoun ed o 9.1 and 3.4 % o global low and s ock, espec i ely, anks he hi d and eigh h in he wo ld; 18.5 housand Chin- ese in es o s ha e se up 29.7 housand en e p ises ha dis ibu e in 186 o eign coun ies ( e i o ies), which indica es ha China has g adually de eloped om a capi al a ac ion coun y o an in es men coun y. A he same ime, China has long become a majo expo ing economy since he 1990s. China’s o al expo s in 2014 amoun ed o $2.34 il- lion, anking i s in he wo ld. In he new backg ound ha China ac i ely p omo es he “One Bel and One Road”s a egy in neighbo ing coun ies, i is a g ea e de elopmen oppo uni y o Chinese o eign di ec in es men , which makes he e ec o o eign an d China Finance E co n o mi c R e i e w © The Au ho (s). 2016 Open Access This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0 In e na ional License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons license, and indica e i changes we e made. Lin China Finance and Economic Re iew (2016) 4:13 DOI 10.1186/s40589-016-0037-8 di ec in es men on expo ade an impo an issue. Di e en conclusions indica e di - e en policy implica ions: i he OFDI o China had a p omo ing e ec on expo s, i would be necessa y o ake po en ial ade ic ions and dispu e isks in o conside a ion and o p ope ly manage he expanding ade su plus and exchange p essu es on RMB. I he OFDI o China had a subs i u ion e ec on expo , i would be necessa y o assess he nega i e e ec o capi al expo and sh inking expo on domes ic economic g ow h and employmen in o de o a oid he epea ing mis ake o indus ial hollowing-ou caused by OFDI as some de eloped coun ies ha e expe ienced in his o y (Liu and Nie 2014). Classical li e a u es ha e ully discussed he e ec o OFDI on expo s: Based on he Hecksche -Ohlin in e na ional ade model, Mundell (1957) demons a ed ha i he wo economies ha e simila p oduc ion unc ions and consume demands, in e na ional di ec capi al in es men would equalize he ac o endowmen s uc u e and he ela i e p ices. Commodi ies o iginally p oduced and expo ed would now be p oduced in ano he coun y h ough he low o p oduc ion ac o s. In his sense, he low o p oduc ion ac o capi al is equal o ha o goods, which indica es ha expo would be eplaced. Howe e , Schmi z and Helmbe ge (1970) sugges ed ha he analysis amewo k was only sui able o in a- indus y ade o manu ac u ed goods. When de eloped coun ies in es ed in he de elop- ing coun ies wi h abundan na u al esou ces, he capi al coun y expo s machines, equip- men s, and o he capi al goods o he hos coun y while he hos coun y expo s c ude oil, i on o e, and o he p ima y p oduc s. This in es men - ade pa e n would inc ease in e - na ional ade. Mo e b oadly speaking, when he capi al coun y and hos coun y ha e e - ical di ision o labo , OFDI would p omo e he expo o in e media e goods o he capi al coun y (Helpman 1984). The heo y o p oduc li e cycle p o ides ano he pe spec i e: Ve - non (1966) poin ed ou ha in he ea ly li e s age o a p oduc , R&D and he expo o he p oduc we e ca ied ou in he domes ic coun ies.In hema u eanddeclinepe iod, ising p oduc ion cos s would make he a ional i ms ans e he p oduc ion p ocess o o he coun ies wi h low ac o cos h ough di ec in es men , which would lead o less expo . Kojima (1975) held a di e en iew and he sugges ed ha he ans e o “ma ginal indus- ies” ha ha e los he compa a i e ad an age would p omo e he expo o capi al goods and in e media e p oduc s o he hos coun y, which indica es ha FDI would inc ease he expo . A he same ime, he capi al coun y would impo manu ac u ed goods om he hos coun y. The o eign di ec in es men andg adien indus ial ans e wi h Japan ac - ing as he leading goose had a signi ican posi i e e ec on he in e na ional ade and eco- nomic g ow h in Eas Asia in he 1980s o 1990s o he las cen u y. The e a e o he li e a u es explo ing he ela ionship be ween FDI and expo om he pe spec i e o i m decision-making. Hi sch (1976) poin ed ou ha he mul i- na ional co po a ion had wo al e na i e ways o en e ing he o eign ma ke s: expo and FDI. Fi ms made a ional decisions based on domes ic and o eign p oduc ion cos s, anspo a ion cos s, sales cos s, and o he ac o s. He also hough ha i was a i ha pushes he expo cos s o i ms, which made he a i -jumping become one o he causes o FDI. Buckley and Casson (1981) conside ed he e ec s o ma ke demand size: Fo i ms, expo had a highe ma ginal cos due o in e na ional shipping, cus oms du ies, and o he ac o s, bu he ixed cos is low, which is sui able o mee ing smalle ma ke demands. Es ablishing ac o ies in o eign coun ies, on he o he hand, en ails highe ixed cos s bu lowe ma ginal cos s, which is sui able o mee ing la ge ma ke demands. When he o eign ma ke demand g ows apidly, he Lin China Finance and Economic Re iew (2016) 4:13 Page 2 o 16 i m would p e e he o eign di ec in es men ha has he ad an age o economy o scale. In he amewo k o cos analysis, o eign di ec in es men and expo ade a e mu ually subs i u ed. Howe e , cos is no he only ac o . Helpman (1984) conside ed ha wi h e- spec o he hos companies, he mul ina ional co po a ions end o ha e mo e ad anced p o- duc ion echnology, managemen , and R&D capabili y. Fo he main enance o hese unique ad an ages and a oiding imi a ion o compe i o s, mul ina ional co po a ions may be unwill- ing o make ex e nal ma ke ansac ions, bu p e e in a- i m ade ha pu chases machin- e y equipmen and in e media e goods om hos coun y, which inc eases he expo o hos coun y. In addi ion, in he ade heo y o he e ogeneous i ms (Helpman e al. 2004), i is p oduc i i y ha de e mines he way ha i ms en e o eign ma ke s: i ms wi h highe p oduc i i y, la ge scale, and mo e abundan esou ces can pay a highe up on cos s, bea highe isk, and also end o en e o eign ma ke in he way o di ec in es men . On he con a y, i ms wi h lowe p oduc i i y would choose he expo ing mode wi h low ixed cos . Fo eign di ec in es men and expo ing a e wo kinds o al e na i e endogenous choices. The abo e heo e ical li e a u es ha e eached no consensus. Simila ly, among empi - ical li e a u es conce ning o he coun ies, some suppo s he complemen a ion e ec (Lipsey and Weiss 1981, 1984; Head and Ries 2001; Chen and Tang 2014) while o he s suppo he subs i u ion e ec (Ho s 1972; Belde bos and Sleuwaegen 1998; Obe ho e and P a e may 2012). In a wo d, he in es men in di e en coun ies, indus ies, de elopmen s ages, and o di e en mo i a ions shows di e en e ec s on expo s. Howe e , ob iously di e en om o eign li e a u es, domes ic esea ches a e s ik- ingly consis en , nea ly all o hem showing ha Chinese FDI signi ican ly p omo es ex- po s. Bu his pape inds ha hese conclusions a e doub ul due o he disad an ages o a iable selec ion, model se ing, and es ima ion me hods. The es ima ion o he po- en ial impac o FDI in he si ua ion o “One Bel and One Road”depends on he ac- cu acy and obus ness o empi ical esea ch. The e o e, he ma ginal con ibu ion o his pape is o co ec hese e o s and o eexamine he e ec o OFDI on expo s using panel da a om 2003 o 2014. The conclusion o his pape ejec s he common unde s anding ha Chinese FDI p omo es he expo s. Mo eo e , his pape also p o es ha di e en conclusions a e achie ed by co ec ing model misspeci ica ion, w ong a iable selec ion, and es ima ion me hods, no due o he use o a di e en sample. Conclusions and p oblems in he exis ing li e a u e Does Chinese FDI eally p omo e expo ? The exis ing li e a u es in China p o ide a i ma i e answe s. Based on Kojima’s heo y o “ma ginal indus y,”Cai and Liu (2004) asse ed ha he goal o Chinese in es men in less de eloped coun ies was o ans e labo -in ensi e indus ies ha would soon lose cos ad an age such as p ocess- ing manu ac u e, simple assembly p oduc s. They used ime se ies da a o ou wa d o eign di ec in es men , impo and expo o China om 1990 o 1999 o empi ical analysis and go he conclusion ha FDI p omo ed expo . Zhang (2005) used he co- in eg a ion eg ession, e o co ec ion model, and G ange causali y es o es ima e ime se ies da a o FDI and expo om 1982 o 2002 and go he conclusion ha FDI p omo ed expo . Zhang (2007) collec ed panel da a o FDI ha China in es ed in 72 coun ies om 2000 o 2004. Wi h he a e aging me hod, he acqui ed a sec ional da a- se o eg essions and go he conclusion ha FDI p omo e expo wi h he g a i y Lin China Finance and Economic Re iew (2016) 4:13 Page 3 o 16 model. Xiang (2009) used a panel co-in eg a ion and e o co ec ion model wi h he FDI da a ha China in es ed in 50 coun ies om 2000 o 2006, and he go he conclusion ha FDI can signi ican ly p omo e expo : he es ima ed elas ic coe icien is abou 0.9, which means 1 % inc ease o FDI would lead o 0.9 % inc ease o expo . Chen e al. (2010) used a single a iable eg ession speci ica ion and use he GLS, ixed e ec , and andom-e ec s model o he es ima ion using low and s ock da a ha China in es ed in 26 coun ies om 2003 o 2007 and he go he conclusion ha FDI p omo e expo . In addi ion, la es li e a u es such as Zhou and Niu (2012). Zhang (2012), Zhang and Huang (2013), Liu and Xie (2014), and Liu e al. (2014) used la ge samples o hei esea ch. Al hough he e a e many di e ences in model speci ica ion and es ima ion me hods, hese li e a u es basically go he same conclusion ha he OFDI o China p omo es i s expo s, despi e some docu- men ed “coun y he e ogenei y”. I e ain om u he discussions o sa e space. 1 Undoub edly, hese pape s p o ide solid ounda ions o s udying he ade e ec o OFDI. Howe e , upon e iewing hese exis ing empi ical li e a u es, he e a e a leas h ee disad an ages. The selec ion o he OFDI a iable The da a o OFDI in he cu en li e a u e is mainly om he “s a is ical yea book o China”and “s a is ical bulle in o Chinese o eign di ec in es men ”. They collec he low and s ock da a o FDI ha China in es ed in o he coun ies ( egions). The low da a eco ds he in es men amoun ha China in es s in o wi hd aws om he hos coun y. Zhang (2007), Chen e al. (2010), Zhang (2012), and Wang e al. (2014) s udied he e ec o OFDI on ade using low da a. Howe e , i can be seen ha si ua ions ha wi hd aw- ing o he loss o in es men which lead o he nega i e low OFDI a e no a e. I collec panel da a o Chinese OFDI om 2003 o 2014 om he s a is ical bulle in o Chinese o - eign di ec in es men ; among he e ec i e 1636 obse a ions o OFDI, he e a e as many as 124 obse a ions ha a e nega i e, which accoun s o 7.6 % in he o al sample. I I ake na u al loga i hm o he low da a, i is ine i able o dele e hese nega i e obse a- ions o o eplace hem wi h ze oes. The sample selec ion p oblem b ough by his non- andom censo ing is likely o cause biased esul s. Ano he p oblem is ha he OFDI low da a and ade da a may show a s ong posi i e co ela ion due o some shocks caused by exogenous ac o s ha may be di icul o quan- i y. Fo example, when poli ical leade s o China and hos coun ies signed s a egic co- ope a ion ag eemen s, bila e al economic and ade coope a ion would be p omo ed in he ollowing yea s, which leads o bo h apid g ow h o in es men low and ade. On he con a y, occasional de e io a ion o poli ical ela ions may lead o simul aneous sh inks o in es men lows and ade. In sho , OFDI and expo ha e he high posi i e co ela ion due o he exogenous poli ical impac , which makes he OFDI show a ce ain p omo ing ole in he eg ession wi h he expo as he explana o y a iable. Howe e , he endogenei y o simul anei y is essen ially due o he unobse able o unquan i iable ac o s. The ela- ionship is no he ocus o a en ion, so low da a o OFDI is no sui able o esea ch. In ac , when we s udy he long- e m e ec o p oduc ion and ope a ion b ough by OFDI, lagged da a o OFDI s ock da a would be he mos easonable choice. Fo example, Zhou and Niu (2012), Chih-Fan and Dong (2013), Zhang and Huang (2013), Liu and Xie (2014), and Wang and Xiang (2014) used he cu en s ock da a. Bu since he cu en Lin China Finance and Economic Re iew (2016) 4:13 Page 4 o 16 s ock da a is calcula ed based on cu en low da a, lagged s ock da a and in es men p o i . Since i includes in o ma ion conce ning he cu en low da a o OFDI, he p ob- lem o endogenous bias could no be a oided. In eali y, no ma e OFDI akes he o m o me ge and acquisi ion, he legal p ocedu e, si e cons uc ion, p ocu emen and ins al- la ion o p oduc ion equipmen , and ec ui men and aining o p oduc ion pe sonnel all need o consume a ce ain amoun o ime. Ins ead, lagged da a o OFDI s ock ully conside ed he equi ed ime lag ha in es men low ans o med in o ac ual p oduc ion capaci y. The e o e, his pape selec ed he lagged e ms o OFDI s ock as he co e explana o y a iable in empi ical analysis. 2 The pape also p o ides eg ession esul s using cu en low da a and s ock da a o compa ison. They a e based on he same se o sample so he di e ences be ween he conclusions a e solely a ibu able o di e en selec ions o co e explana o y a iables. Reg ession esul s e eal ha wi h he cu en low and s ock da a as he explana o y a iables, he es ima ed alue and signi icance o p omo ing e ec a e o e es ima ed while ha o subs i u ion e ec a e unde es ima ed. The lag o s ock da a has no p omo ing e ec on expo . 3 The speci ica ion o he empi ical model Limi ed by da a a ailabili y, ea lie s udies used he agg ega e ime se ies da a wi h he me hods o co-in eg a ion eg ession, e o co ec ion model (Cai and Liu 2004; Zhang 2005). La e s udies such as Xiang (2009) and Chen e al. (2010) used c oss-coun y panel da a; bu he idea o co-in eg a ion analysis is s ill ollowed. The p oblem wi h his kind o model speci ica ion is ha only OFDI is chosen o explaining impo and expo . The eg ession se ing o “single a iable” ails o ake o he ac o s in o consid- e a ion. Since China is a la ge coun y wi h apid economic g ow h and ac i ely pa ici- pa e in he globaliza ion o economic ac i i ies, i s OFDI and expo a e bo h on a ising end. F om a quan i a i e pe spec i e, lacking con ol a iables in eg essions will undoub edly each he conclusion ha hese wo a iables a e signi ican ly and posi i ely co ela ed, which is mos p obably caused by se e e omi ed a iable bias. Wha is mo e, China only began la ge scale o OFDI a e 2005; be o e hen, China’s annual OFDI low was less han en billion and he s ock was only ens o billions, which is di icul o o m an impac on expo wi h illion-dolla -le el in an economic sense. The e o e, his pape conduc s empi ical analysis wi hin he amewo k o he g a i y model, which has been he mos popula wo kho se in he ield o in e na ional ade. I includes he majo ac o s o in e na ional ade o imp o e he explana o y powe o he model as well as o weaken he impac o missing a iables. The esul s show ha he maximum elas ici y o OFDI on expo in he ull sample eg ession is 0.073 while in he subsample is 0.12 a mos . Taking he huge magni ude gap be ween China OFDI s ock and expo ade in o conside a ion, he ange o he elas ici y coe icien is mo e easonable compa ed o hose be ween 0.3 and 0.9 ound in p e ious pape s such as Xiang (2009), Zhang (2012), and Liu and Xie (2014). The choice o model es ima ion me hod The combina ion o c oss-coun y panel da a and he g a i y model has been popula in la es esea ch (Zhou and Niu 2012; Hu and Qiao 2013; Chih-Fan and Dong 2013; Lin China Finance and Economic Re iew (2016) 4:13 Page 5 o 16 Zhang and Huang 2013; Wang e al. 2014; Wang and Xiang 2014). The mos commonly used es ima ion me hod o panel da a is he ixed-e ec s model. I his me hod was used, ime-in a ian explana o y a iables such as geog aphical dis ance, whe he he wo coun ies a e bo de s o ha e he common language, would be classi- ied as indi idual ixed e ec s and a e consequen ly uniden i iable, which is a di icul p oblem in he g a i y model. Exis ing li e a u es p o ide h ee possible solu ions: he i s is he pooled OLS o andom-e ec s model. These me hods a e used o eg ess o ob aining he es ima ed coe icien s o all a iables, bu pooled OLS gi es up he con ol o indi idual he e o- genei y, and whe he he andom-e ec s model is sui able is unce ain. 4 Bo h o hem canno gua an ee he consis ency o es ima ions. The second is he a ying-coe icien model. This me hod is used by ecen li e a u es and go he conclusion ha some “coun y he e ogenei y”exis in he e ec o Chinese OFDI on expo ade (Zhang 2012; Liu and Xie 2014). This me hod is gene ally applicable o he long panel da a (namely ha he annual numbe Tis g ea e han he numbe o indi iduals N). The Chinese OFDI da a is a ypical sho panel, which indica es ha he me hod is no sui able. The hi d is he dynamic panel da a model es ima ed using sys em GMM me hod (Liu e al. 2014). Howe e , wha he dynamic model es ima es is he sho - e m equilib ium ela ionship bu no he long- e m impac o OFDI on expo s. Mo eo e , he explana o y a iable o Chinese GDP is missed in he g a i y model o Liu e al. (2014). Since he OFDI s ock o China is posi i ely co ela ed wi h GDP, he conclusion ha OFDI has a posi i e e ec on expo ade is likely a “ alse imp es- sion”caused by omi ed a iable bias. 5 The empi ical analysis in his pape uses mul iple speci ica ions o ob ain obus conclusions: I use a pooled OLS model and con ol o bila e al GDP, geog aphical dis ance, lagged OFDI s ock as well as he e ogenei ies ha may in luence ade in heg a i ymodel.Ialsousea ixed-e ec s model o ob ain a wi hin-g oup consis - en es ima o , bu I would no be able o ob ain he es ima ed coe icien s o ime-in a ian a iables such as geog aphical dis ance. Second, in o de o alle ia e he endogenei y bias in he g a i y model caused by he log-linea ized o da a ha iola es he condi ions o consis en es ima ion, I use he Poisson pseudo max- imum likelihood es ima o sugges ed by Sil a and Ten ey o (2006). The esul s show ha he conclusion in his pape is e y obus . Me hods In he ield o in e na ional ade, he g a i y model o iginally p oposed by Tin- be gen (1962) and de eloped by Ande son (1979), Be gs and (1985) has become he s anda d pa adigm o empi ical esea ch. Al hough he g a i y model was i s p oposed as an empi ical equa ion, i has solid heo e ical ounda ions. Some o m o he g a i y model can be de i ed om he neo-classical in e - na ional ade heo y based on ac o endowmen s uc u e and compa a i e ad an age o om he new ade heo y based on monopolis ic compe i ion. Acco ding o he e iew a icle o Ande son and an Wincoop (2004), a e in oducing ac o s such as pe capi a income, ade ba ie s, exchange a es, bo de s, common language dummies, ee ade ag eemen s, and colonial ela ionship in o he Lin China Finance and Economic Re iew (2016) 4:13 Page 6 o 16 g a i y model, i can explain mo e han 80 % o in e na ional ade lows. I has become one o he mos success ul empi ical models in economics esea ch. In his pape , he ade g a i y model is used as an empi ical amewo k o es ima e he expo e ec o OFDI: lnEXPi; ¼β1lnGDPCHN; þβ2lnGDPi; þβ3lndis iþβ4lnOFDIs ocki; ‐n þXθþω þεi; ð1Þ In he equa ion, Chinese expo o he coun y iEXP is explained by he GDP o China, he GDP o coun y i, and he geog aphical dis ance, which a e he p oxies o commodi y supply, demand capaci y, and anspo a ion cos s, espec i ely. They a e he basic a iables in he g a i y model. The key explana o y a iable is he lagged e m o OFDI s ock ha China has in es ed in coun y i. As men ioned ea lie , his can a oid he omi ed bias p oblem o eg ession es ima ion caused by he simul anei y be ween he OFDI low and ade. A he same ime, he lagged e m o OFDI akes he ime lag o in es men s o ming ac ual p oduc ion capaci ies in he hos coun y in o conside a ion. In his pape , I conside he model wi h lag phases o in es men s ock (i.e., n=1,2,3.). Subsequen eg essions show he esul s wi h he cu en low and cu en s ock as he co e explana o y a iables, which di ec ly assesses he impac o di e en co e explana o y a iables on empi ical esul s. O he con ol a iable Xand he ac o s ha cap u e coun- y he e ogenei ies a e ully con olled, which includes a neighbo ing dummy, a common language dummy (i.e., he o icial language is Chinese o English 6 ), a ee ade ag eemen (FTA) wi h China dummy, OECD membe coun y dummy, and WTO membe dummy. In addi ion, con inen dummies (Asia, Eu ope, A ica, No h Ame ica, La in Ame ica, and Oceania) and law o igin dummies (Anglo-Ame ican Legal Sys em, F ench ci il law, Ge man law, and No dic law) a e also con olled in he eg essions. Ω is he yea ixed e ec , which can con ol he collision o global mac oeconomic luc ua ions on China’s expo o o he coun ies. In o de o ge a obus empi ical conclusion, based on he mixed eg ession model (POLS), he dis u bance e m can be decomposed: ε i, =μ i + i, , he empi ical model can be ew i en as a ixed-e ec s model (FE): lnEXPi; ¼β1lnGDPCHN; þβ2lnGDPi; þβ3lndis iþβ4lnOFDIs ocki; ‐n þZγþω þμiþ i; ð2Þ In he model, all ime-in a ian dummies a e classi ied as coun y ixed e ec s μ i, ,so hey a e no longe iden i iable. Bu using he ixed-e ec s model allows me o ob ain he “wi hin-g oup”consis en es ima o s o o he a iables in he g a i y model. They can be compa ed wi h he esul s o pooled OLS eg essions o ob ain a obus conclusion. The da ase used in his pape is a panel o Chinese OFDI and expo ade om 2003 o 2014. The da a o OFDI is om “The S a is ical Bulle in o Fo eign Di ec In es men o China”, he da a o expo is om “The China S a is ic Yea book”o a i- ous yea s. The GDP o China and o he coun ies a e om he wo ld de elopmen indica o da abase o he Wo ld Bank. O he da a, such as geog aphic dis ance om China, language, law o igins, and o he coun y cha ac e is ics a e om he CIA Wo ld Fac book and he F ance In e na ional Economics Resea ch Cen e (CEPII) da abase. Be o e unning he eg essions, da a was p ocessed as ollows: Lin China Finance and Economic Re iew (2016) 4:13 Page 7 o 16 (1)Hos coun ies wi h in es men s ock less han $100 million and con inuous obse a ions less han 5 yea s a e d opped om he sample, which can a oid he a enua ion bias caused by measu emen e o p oblem. (2)In es men s o ypical o sho e inancial cen e s, Hong Kong, and Macau a e dele ed. Basically, hese in es men s se “shell”companies o capi al ans e , ax a oidance, and e en some o he illegal ac i i ies. E en ually, hese in es men s low o o he coun ies o e en home, which in ol es no ac ual economic ac i i ies such as employmen , p oduc ion, and ope a ion in he hos coun y (Wang 2013). They would ha e no e ec on he expo ade so keeping hese obse a ions would lead o alse conclusions. (3)Taking na u al loga i hm o he con inuous a iables. This can elimina e he p oblem o he e oskedas ici y and ex eme alue o a iables wi h la ge alue, such as ade, GDP, and geog aphical dis ance. The eg ession coe icien s ob ained by logged da a on bo h sides o he equa ion can be in e p e ed as he elas ici y coe icien s, which ha e clea economic meanings. Wi h he p ocedu es desc ibed abo e, panel da e o 155 coun ies om 2004 o 2014 is acqui ed, which con ains 33 de eloped coun ies, and 122 de eloping coun ies. Fo he es ima ion me hod o model, Eqs. (1) and (2) a e log-linea ized om he o iginal o m o g a i y model: EXPi; ¼αGDPβ1 CHN; GDPβ2 i; dis β3 iOFDIs ockβ4 i; −nXθηi; Among he a iables, η i, is he independen e o e m which sa is ies he condi ion: E(η i, | GDP CHN, ,GDP i, , dis i , OFDI s ock i, −n ,X)=1 We can ob ain: EEXP i; GDPjCHN; ;GDPi; ;dis i;OFDIs ocki; −n;X  ¼αGDPβ1 CHN; GDPβ2 i; dis β3 i OFDIs ockβ4 i; −nXθHowe e , a e he log-linea iza ion, we can ge E(1nη i, )≠1n E(η i, )= 0 acco ding o he Jensen inequali y. The e o e, he e o e m ε i, , i, in Eqs. (1) and (2) canno sa is y he classical assump ion o ze o mean. In ac , E(1nη i, ) depends no only on E(η i, ) bu also co ela es wi h highe o de dis ibu ions o η i, . I he highe o de momen s o η i, is co ela ed wi h he explana o y a iables, OLS es ima ion would su e om endogenous bias. 7 The e o e, his pape uses he Poisson pseudo-maximum likelihood es ima ion p o- posed by Sil a and Ten ey o (2006). 8 Hao and Ma (2012) used he Mon e Ca lo simula- ions o compa e he esul s o PPML and o he commonly used es ima ion me hods o he g a i y model (OLS, NLS, Tobi ). They ound ha he es ima ion e o o PPML is he smalles and he es ima ion o ac o s such as FTA is mo e conse a i e and p uden compa ed wi h OLS. Resul s and discussion The es ima ion esul s o ull sample Table 1 shows he eg ession esul s o ull sample. Models (1), (2), and (3) show he esul s o pooled OLS, wi h he one o h ee lag o OFDI as he explana o y a iables espec i ely. Models (4), (5), and (6) add he indi idual ixed e ec s o he hos Lin China Finance and Economic Re iew (2016) 4:13 Page 8 o 16 5 Since he da ase used in his pape has a la ge o e lap wi h he sample o Liu e al. (2014), his conjec u e can be e i ied. Using he same model speci ica ion and he sys em GMM es ima ion me hod, we ind ha i he a iable o Chinese GDP we e con olled in he model o ade g a i y, he s ock o OFDI would no be signi ican . I he a iable o Chinese’s GDP is delibe a ely omi ed, he s ock o OFDI is signi ican ly posi i e and he es ima ion coe icien s is simila wi h hei esul s. Limi ed by space, he esul is no shown in he empi ical pa . The empi ical da a and STATA codes a e a ailable om au ho upon eques . 6 Al hough English is no he o icial language o China, he e a e lo s o Chinese s udying English, so i has a g ea in luence on China. 7 A his poin , he a iance o he η i, : Va (η i, )= (GDP CHN, , GDP i, , dis i , OFDI s ock i, −n ,X)≠cons an 8 Con ined in his pape , he de ails o PPML me hods a e no shown, eade s can see he pape o Sil a and Ten ey o (2006), (The Re iew o Economics and S a is ics). Chinese li e a u e such as Sun Lin and Ni Kaka (2013), Tian Wei e al. (2013) used his me hod. 9 The in e p e a ion o ixed e ec model is di e en om ha o he pooled OLS model. The es ima ed coe icien e lec s he wi hin-g oup e ec , which means ha 1 % inc ease o Chinese OFDI s ock o one speci ic hos coun y con ibu es o se e al pe cen age inc ease o expo . 10 In Table 1, he coe icien s o Chinese and hos coun y’s GDP a e signi ican ly posi- i e. The coe icien o geog aphic dis ance is signi ican ly nega i e, which is in line wi h expec a ion. The signs o o he con ol a iables a e simila wi h he exis ing li e a u e. Limi ed by he space, his pape does no show he coe icien s and -s a is ic. 11 Cons ained by space limi a ions, he es ima ion o sub-sample does no p o ide he eg ession esul s ha a e based on he w ong a iable selec ion, model speci ica- ion, and es ima ion me hod. Acknowledgemen s This pape is suppo ed by he Na ional Social Science Fund o China (G an : 15CJL025) and he NSE Special Resea ch Fund g an ed by he Cen e o New S uc u al Economics o Peking Uni e si y. Compe ing in e es s The au ho decla es ha he has no compe ing in e es s. Recei ed: 12 Augus 2016 Accep ed: 12 Oc obe 2016 Re e ences Cai R, Liu Q (2004) A e he Chinese in e na ional di ec in es men and ade complemen a y? The empi ical analysis based on he ma ginal indus y heo y. Wo ld Econ S udy 8:64–70. Chen L, Yang Z, Hou Z (2010) Expo d i en o expo subs i u ion?— he ma ginal indus y s a egy es o Chinese OFDI. Finance T ade Econ 2:78–85. Hu B, Qiao J (2013) The ade e ec o China’s OFDI—based on dynamic panel model sys em GMM me hod. Econ Manag 4:11–19. Hao J, Ma H (2012) The new de elopmen o g a i y model and he es o Chinese o eign ade. J Quan Tech Econ 10:52–68. Jiang G, Jiang D (2014) Expo e ec o Chinese OFDI, Econ Res J 5:160–173. Chih-Fan L, Dong Q (2013) The expo e ec o Chinese OFDI: inhibi ion o p omo ion based on he empi ical s udy o panel da a o 21 coun ies. Commen s P od Ind 5:152–160. Liu H, Nie F (2014) Hollow e ec o Chinese OFDI o manu ac u ing indus y. China Ind Econ 4:83–96. Liu X, Gou Q, Lu F (2014) The impac analysis o Chinese OFDI on expo . Wo king Pape Econ Res, WP672 Liu Z, Xie R (2014) The empi ical analysis o coun y e ec o Chinese OFDI on ASEAN. Wo ld Econ S ud 6:80–86. Mao Q, Xu J (2014) Does Chinese OFDI p omo e o inhibi he en e p ise expo . J Quan Tech Econ 9:3–21. Qiao J, Hu B (2015) How FDI a ec expo —based on he ma ching o manu ac u ing en e p ises. J In T ade 4:126–136. Wang B (2013) The misunde s ood o icial da a— he eal OFDI o China. In Econ Re 1:61–74. Lin China Finance and Economic Re iew (2016) 4:13 Page 15 o 16 Wang S, Tian T, Xie R (2014) The ade e ec esea ch o Chinese FDI. Wo ld Econ S ud 10:80–86. Wang S, Xiang J (2014) C ea ion o subs i u ion— he mechanism esea ch o Chinese FDI on ade. Wo ld Econ Res 6:66–72. Xiang B (2009) The ade e ec esea ch o Chinese FDI—based on he co-in eg a ion analysis o panel da a, Finance T ade Econ 4:77–82. Zhang C (2012) The ade e ec esea ch o OFDI, J Quan Tech Econ 6:74–85. Zhang J, Huang P (2013) Replacemen o expo s o p omo ing expo s— he esea ch on he e ec o FDI on expo . J In T ade 3:95–103. Zhang R (2005) The analysis o he ela ionship be ween OFDI and in e na ional ade. Wo ld Econ S ud 3:23–27. Zhang Y (2007) The ela ionship be ween OFDI and ade: complemen a y o al e na i e. J In T ade 6:87–93. Zhou X, Niu R (2012) Chinese OFDI and he ade e ec —based on he empi ical esea ch on panel g a i y model. In Econ T ade Res 5:69–81. Ande son JE, an Wincoop E (2004) T ade cos s. J Econ Li 42(3):691–751 Ande son JE (1979) A heo e ical ounda ion o he g a i y equa ion. Am Econ Re 69(1):106–116 Belde bos R, Sleuwaegen L (1998) Ta i jumping DFI and expo subs i u ion: Japanese elec onics i ms in Eu ope. In J Ind O gan 16(5):601–638 Be gs and JH (1985) The g a i y equa ion in in e na ional ade: some mic oeconomic ounda ions and empi ical e idence. Re Econ S a 67(3):474–481 Buckley PJ, Casson M (1981) The op imal iming o a o eign di ec in es men . Econ J 91(361):75–87 Chen W, Tang H (2014) The d agon is lying wes : mic o-le el e idence o Chinese ou wa d di ec in es men . Asian De Re 31(2):109–140 Head K, Ries J (2001) O e seas in es men and i m expo s. Re In Econ 9(1):108–122 Helpman E, Meli z MJ, Yeaple SR (2004) Expo e sus FDI wi h he e ogeneous i ms. Am Econ Re 94(1):300–316 Helpman E (1984) A simple heo y o in e na ional ade wi h mul ina ional co po a ions. J Pol Econ 92(3):451–471 Hi sch S (1976) An in e na ional ade and in es men heo y o he i m. Ox o d Econ Pap 28(2):258–270 Ho s T (1972) Fi m and indus y de e minan s o he decision o in es ab oad: an empi ical s udy. Re Econ S a 54(3):258–266 Kojima K (1975) In e na ional ade and o eign in es men : subs i u es o complemen s. Hi o subashi J Econ 16(1):1–12 Lipsey RE, Weiss MY (1981) Fo eign p oduc ion and expo s in manu ac u ing indus ies. Re Econ S a 63(4):488–494 Lipsey RE, Weiss MY (1984) Fo eign p oduc ion and expo s o indi idual i ms. Re Econ S a 66(2):304–308 Mundell RA (1957) In e na ional ade and ac o mobili y. Am Econ Re 47(3):321–335 Obe ho e H, P a e may M (2012) FDI e sus expo s: mul iple hos coun ies and empi ical e idence. Wo ld Econ 35(3):316–330 Schmi z A, Helmbe ge P (1970) Fac o mobili y and in e na ional ade: he case o complemen a i y. Am Econ Re 60(4):761–767 Sil a S, Ten ey o S (2006) The log o g a i y. Re Econ S a 88(4):641–658 Tinbe gen J (1962) An analysis o wo ld ade lows in shaping he wo ld economy. Twen ie h Cen u y Fund, New Yo k Ve non R (1966) In e na ional in es men and in e na ional ade in he p oduc cycle. Q J Econ 80(2):190–207 Submi you manusc ip o a jou nal and benefi om: 7 Con enien online submission 7 Rigo ous pee e iew 7 Immedia e publica ion on accep ance 7 Open access: a icles eely a ailable online 7 High isibili y wi hin he fi eld 7 Re aining he copy igh o you a icle Submi you nex manusc ip a 7 sp inge open.com Lin China Finance and Economic Re iew (2016) 4:13 Page 16 o 16