On the gravity equation of trade: a case of Germany

20 2016, XIX, 3 Ekonomie 20 DOI: 10.15240/ ul/001/2016-3-002 In oduc ion The g a i y model (o equa ion) o in e na ional ade is an economic analogy o New on’s Law o g a i y. The economic e sion o his law assumes ha in e na ional bila e al ade is di ec ly p opo ional o ‘sizes’ o ading economies, and indi ec ly p opo ional o hei dis ance. G a i y was in oduced in o economic heo y by Tinbe gen (1962), and la e his pionee wo k was ollowed by many o he s udies, see e.g. Ande son (1979; 2010), Be gs and (1985), Dea do (1998), Ande son and an Wincoop (2003), Helpman e al. (2008), o Be gs and and Egge (2011). A concise e iew o g a i y app oach can be ound e.g. in Ande son (2010) o Sal a ici (2014). Theo e ical explana ion o he g a i y equa ion o agg ega ed o disagg ega ed ade can be ound e.g. in Ande son (1979) o Chaney (2011). T ade in a eal wo ld is in l uenced no only by economic size and dis ance, bu also by ade ‘ ic ions’ ( ade cos s o ade ba ie s) such as bo de s among coun ies, a di e en language and cu ency, colonial ies, ee ade ag eemen s, e c., hese addi ional ac o s we e inco po a ed in o g a i y models as well, see e.g. Dea do (1998), Baie and Be gs and (2009), Ande son (2010), Be gs and and Egge (2011) o Sal a ici (2014). G a i y models heo e ically explain he ole o an economic size in bila e al ade l ows a any scale (coun ies, egions, e c.); hough he ole o a dis ance is no well unde s ood ye , see Disdie and Head (2008). Mo eo e , g a i y applies o o he socio-economic phenomena such as mig a ion o di ec o eign in es men s. Gene ally, acco ding o Ande son (1979) o Chaney (2011), g a i y models o ade can be conside ed he mos success ul empi ical ools in economics. The empi ical e idence o g a i y models is a he s ong, as a ious s udies epo he coe i cien o de e mina ion be ween 0.6 and 0.8. A me a-analysis o 1,467 es ima es in 103 pape s p o ided by Disdie and Head (2008) ound 1  in ela ion (1) (see below). Rema kably, he coe i cien γ has been s able (and close o one) o mo e han one cen u y. A heo e ical explana ion o his esul can be ound e.g. in Chaney (2011). G a i y model es ima ions a e usually ca ied ou o c oss-sec ional o panel da a. Howe e , coun ies all o e he wo ld o m a e y he e ogeneous sample. The e a e coun ies wi h cen u ies o indus ial adi ion and expo ( he USA, he UK, Ge many, e c.), coun ies ha expo agg essi ely in he las decades (Japan, Ko ea, China, e c.), and also a la ge numbe o de eloping coun ies whose expo s a e limi ed o a icles such as bananas o cocoa beans. Mo eo e , a ade is a p oduc o pa icula human ac ion, and people li ing unde di e en condi ions and egimes simply canno ac in a simila way. The e o e, one should no expec in e na ional ade o be uni e sally desc ibed o explained by one equa ion, model o a o mula. To mix such di e en ade pa ne s while looking o a gene al pa e n (as in physics) makes li le sense in economics. Hence, a mo e sensible app oach migh be he use o mo e homogenous se s o coun ies, such as de eloped (OECD) coun ies, La in-Ame ican coun ies, e c. This a o emen ioned app oach is ollowed in his pape , whe e he g a i y equa ion is used o model agg ega e expo ’s sha es o one coun y (Ge many) o i s ading pa ne s (impo ing coun ies). As he g a i y model is usually o mula ed in a mul iplica i e o m, i is log- ans o med in o a linea equa ion and coe i cien s o a model a e es ima ed by an app op ia e eg ession me hod. A p oblem o a co ec es ima ion o a g a i y model is a b oadly discussed issue, see e.g. Heckman (1979), Sil a and Ten ey o (2006), ON THE GRAVITY EQUATION OF TRADE: A CASE OF GERMANY Jiří Mazu ek EM_3_2016.indd 20EM_3_2016.indd 20 8.9.2016 14:10:578.9.2016 14:10:57 21 3, XIX, 2016 Economics 21 Ma ínez-Za zoso e al. (2007), Helpman e al. (2008), Baie and Be gs and (2009), Egge (2010), He e a and Baleix (2010) o He e a (2013). Howe e , he e is no a consensus on wha he mos app op ia e eg ession es ima ion is in he case when assump ions o O dina y Leas Squa es (OLS) me hod a e iola ed, in pa icula , when a signi i can he e oscedas ici y is p esen . Unde such ci cums ances he Pseudo Poisson Maximum Likelihood (PPML) me hod, he Nonlinea Leas Squa es (NLS) me hod o he Feasible Gene alized Leas Squa es (FGLS) me hods we e p oposed by Sil a and Ten ey o (2006) o Ma ínez-Za zoso e al. (2007). Ne e heless, i assump ions o OLS a e sa is i ed, i is he BLUE (Bes Linea Unbiased Es ima o ) me hod. The aim o he a icle is o examine how he agg ega ed g a i y models i he la es expo da a o Ge many bo h wi hou and wi h ade ic ions. Ge many was selec ed o his s udy because i is he 3 d la ges wo ld expo e (behind he USA and China), and i anks among he mos de eloped coun ies o he wo ld. I is sui ably loca ed in he middle o he con inen su ounded by many ading pa ne s in di e en dis ances; and las , bu no leas , he da a o Ge many can be conside ed accu a e and eliable. The pape is o ganized as ollows: in sec ion 1 Ge many’s expo is b ie l y discussed, in sec ion 2 he da a is desc ibed, in sec ions 3 and 4 agg ega e g a i y models wi hou and wi h ic ions a e p esen ed along wi h hei esul s. Conclusions ollow a he end o he a icle. 1. Ge many’s Expo Ge many belongs among coun ies wi h a posi i e balance o in e na ional ade in he long e m. In 2014, he balance (expo minus impo ) eached 217 billion Eu os acco ding o S a is isches Bundesam (2015b). Commodi y s uc u e o expo is a he s able in he las decade. The mos impo an expo i ems a e anspo a ion (ca s, ca pa s, engines and ae ial echnology), ins umen s o elec ical enginee ing, elecommunica ions echnology, o i ce echnology and machine y; see S a is isches Bundesam (2015a). Figu e 1 p o ides mo e de ailed commodi y s uc u e o expo in 2014. The mos impo an ade pa ne s o Ge many a e F ance, USA, UK, Ne he lands and China (see also Appendix A). In 2014, 58% o Ge many’s expo wen o EU, 17% o Asia, 12% o Ame icas, 10% o non-EU Eu opean coun ies, 2% o A ica and he es (0.3%) o Aus alia and Oceania acco ding o S a is isches Bundesam (2015b). Ge many is a ede al pa liamen a y epublic consis ing o 16 cons i uen s a es. In 2014, he s onges expo ing s a es we e Baden- Wü embe g (16% o o al Ge many’s expo ), No h Rhine-Wes phalia (15.9%), Ba a ia (14.9%) and Lowe Saxony (6.9%). A ound 85% o expo was a comple ed p oduc ion, 5% semi- i nished p oduc ion, and a ound 1% accoun ed o aw ma e ial. Mo e de ailed da a including a s uc u e o expo by indi idual s a es can be ound in S a is isches Bundesam (2015b). Fig. 1: The commodi y s uc u e o Ge many expo in 2014 (in billion Eu os) Sou ce: S a is isches Bundesam (2015a) EM_3_2016.indd 21EM_3_2016.indd 21 8.9.2016 14:10:578.9.2016 14:10:57 22 2016, XIX, 3 Ekonomie 22 2. Me hod and Da a Ge many’s expo is examined ia agg ega e g a i y equa ions wi h and wi hou ic ions. These equa ions (models) a e in oduced and desc ibed in de ail in he ollowing wo sec ions. As g a i y equa ions ha e a mul iplica i e o m, hey we e linea ized by he loga i hmic ans o ma ion. Then, linea eg ession was pe o med by he mos sui able es ima ion me hod. Fo he empi ical in es iga ion he ollowing da a was used:  Impo ing pa ne s’ sha es (in %) o Ge many’s expo o he yea 2013 we e ob ained om S a is isches Bundesam (2014). The da a is p o ided in he o m o a anking o ading pa ne s in he descending o de . Fo his s udy he lis o all coun ies was unca ed and includes 66 main pa ne s ( om F ance o Qa a ) wi h an indi idual sha e o impo om Ge many a leas equal o 0.10%, as he ela i e e o o alues lowe han 0.10% would be inapp op ia ely high (see a discussion a he end o sec ion 4). These 66 coun ies comp ise 97.6% o Ge many’s expo .  Dis ances be ween Ge many and impo ing coun ies (in kilome e s) we e ob ained om a dis ance calcula o a Timeandda e (2013). The dis ance be ween wo coun ies was de i ned as an ai dis ance be ween hei capi al ci ies.  GDP (PPP) in billions USD o impo ing coun ies we e e ie ed om he In e na ional Mone a y Fund (2013).  O he coun ies’ da a include hei cu ency, a membe ship o Schengen a ea and an exis ence o join bo de s wi h Ge many. The dependen a iable in he models was impo ing pa ne s’ sha es (in %), o he a iables we e conside ed independen . All da a is p o ided in Appendix A. I should be no ed ha expo sha es and GDP (PPP in billion dolla s) could be a subjec o la e e isions. 3. F ic ionless G a i y Model, Resul s and Discussion The s anda d g a i y model (equa ion) o agg ega e in e na ional ade usually akes he ollowing o m (Chaney, 2011): ij ij ij GDP GDP Tk d    , (1) whe e Tij is a ade om a coun y i o a coun y j, GDPi deno es g oss domes ic p oduc , k is a posi i e coe i cien , and dij is a geog aphic dis ance o bo h coun ies. Ande son (2010) assumes ha supply Yi o a coun y i is a ac ed by a demand Ej o a coun y j, whe e dij deno es a dis ance o bo h coun ies, and p oposes he ollowing ic ionless and agg ega e g a i y model o ade: 2 iJ ij ij YE Td  (2) Also, in some al e na i e g a i y models an income pe capi a (along wi h a popula ion) o coun ies is used ins ead o supply and demand, and an e o e m is added on he igh hand side o equa ions; see Ande son (1979): ij i j i j ij ij MYYNNdU      (3) In (3) ij M is he dolla l ow o a gi en good om a coun y i o a coun y j, i Y and j Y a e incomes in bo h coun ies, i N and j N a e hei popula ions, and ij U is a log-no mally dis ibu ed e o e m wi h (ln ) 0 ij EU. In he las decades, mo e sophis ica ed models o disagg ega ed goods and wi h ade ic ions we e de eloped, see e.g. Ande son (2010) o Sal a ici (2014). In his pape he ollowing agg ega e ic ionless g a i y model o Ge many’s expo sha es is conside ed:   i i iDIST GDP kE  (4) In (4) i E deno es a sha e o Ge many’s expo (in %) o a coun y i, i GDP is a g oss domes ic p oduc o an impo ing coun y i, i DIST is a dis ance be ween coun y i and Ge many, and α, β, and k a e coe i cien s. Rela ion (4) simply s a es ha an expo ises when an impo ing coun y is close and/ o iche . I should be no ed ha i absolu e alues o expo s (e.g. in billions o USD) we e conside ed in (4) ins ead o ela i e expo s, hen only he coe i cien k would change. EM_3_2016.indd 22EM_3_2016.indd 22 8.9.2016 14:10:578.9.2016 14:10:57 23 3, XIX, 2016 Economics 23 Fo he eg ession analysis he ela ion (4) is e o mula ed in he ollowing way: jjj E k GDP DIST    (5) G a i y equa ion is ans o med loga i hmically (all a iables a e posi i e) which yields: ln ln ln ln jjj E k GDP DIST    (6) The co ela ion ma ix o a iables in he model (6) is shown in Table 1. As i can be seen, co ela ion coe i cien s we e no pa icula ly high. Mul icollinea i y in he model (6) was examined ia he Va iance In l a ion Fac o (VIF), 2 1/(1 ) ii VIF R , whe e 2 i R is he p opo ion o a iance in he i- h independen a iable associa ed wi h o he independen a iables in a model, see O’B ian (2007). A ule o humb s a es ha o alues o VIF la ge han 10 mul icollinea i y o a model can be conside ed a se ious p oblem. In he model (6) VIF o bo h explana o y a iables was only a ound 1.3. Fo he eg ession model (6) he da a om Appendix A was used. The eg ession was pe o med ia s a is ical so wa e G e l. Residuals we e examined o exogenei y, no mali y and he e oscedas ici y. All assump ions o OLS we e sa is i ed wi h an excep ion o he e oscedas ici y associa ed wi h he loga i hm o GDP, whe e he null hypo hesis (homoscedas ici y) could be ejec ed by Whi e’s es a p = 0.04 le el. The e o e, G e l’s buil - in OLS wi h he co ec ed he e oscedas ici y me hod (which inco po a es weigh ed leas squa es me hod) was used o he es ima ion. Resul s a e epo ed in Table 2. As i can be seen om Table 2, bo h eg esso s (logs o GDP and dis ance) we e ound signi i can a 0.01 le el. As expec ed, eg ession coe i cien o loga i hm o dis ance is nega i e, and he coe i cien o loga i hm o GDP is posi i e, bo h coe i cien s a e close o 1, which is in acco d wi h o he s udies’ i ndings. The adjus ed coe i cien o de e mina ion R2 = 0.742, which is wi hin a ange o 0.6-0.8 was ound in o he simila s udies as well. Ln(expo ) Ln(Dis ) Ln(GDP) Ln(expo ) 1 -0.291 0.525 Ln(Dis ) 1 0.506 Ln(GDP) 1 Sou ce: own Reg esso /me hod OLS wi h c. h. Cons . 1.030 (0.554)* Ln(Dis ) -0.903 (0.081)*** Ln(GDP) 0.896 (0.071)*** No. o obse . 66 Sou ce: own No es: S anda d e o s in b acke s. * signi i can a 10%; ** signi i can a 5%; *** signi i can a 1% Tab. 1: The co ela ion ma ix o a iables om he model (6) Tab. 2: The ic ionless g a i y model – es ima ion esul s EM_3_2016.indd 23EM_3_2016.indd 23 8.9.2016 14:10:578.9.2016 14:10:57 24 2016, XIX, 3 Ekonomie 24 4. G a i y Model wi h F ic ions, Resul s and Discussion F ic ions o any kind (bo de s, a di e en language o cu ency, colonial ies, legisla i e, cul u e, eligion, e c.), which in l uence eal ade, can be inco po a ed in o he g a i y models as well. In his pape he ollowing ade ic ions as explana o y a iables we e examined: Adjacency (A): he exis ence o na ional bo de s is conside ed one o he mos impo an ic ions in in e na ional ade. Acco ding o Ande son (2010), c oss-bo de ade is ypically educed by a ac o o 1/20 o 1/3 o i s po en ial alue. Howe e , when a ade is ca ied ou be ween coun ies wi hou common bo de , i is educed e en mo e signi i can ly. Cu ency (C): di e en cu ency migh diminish ade olumes due o exchange a es unce ain y; i may equi e addi ional ansac ion cos s as well. Loca ion in he Schengen a ea (LISA): he Schengen a ea, es ablished in 1995, abolished in e nal bo de con ols and allowed ee ans e o people wi hin he a ea. This could a ec ade olumes wi hin he Schengen a o ably. All hese ade ic ion a iables a e dummy (bina y) a iables wi h alues 0 o 1, see Table 3. Also se e al o he ade ic ions we e conside ed a he beginning o he s udy, such as democ acy index o impo ing coun ies, leng h o common bo de s wi h Ge many, geog aphic loca ion o impo ing coun ies (no jus hei dis ance), o he ac whe he anspo a ion o goods is managed on a land o a sea, bu p elimina y esul s showed hese a iables we e no s a is ically signi i can in he examined models, o we e highly co ela ed wi h o he independen a iables, so hey we e elimina ed om he model. Ne e heless, hough only h ee a iables associa ed wi h ade ic ions we e le in he model, i s explana o y powe was e y high (see esul s a he end o his sec ion). The ic ion g a i y model has he ollowing o m: (%) iii E k GDP DIST     () ii i EXP C A LISA     (7) Log- ans o m o (7) yields: l n( ) ln ln ln iii E k GDP DIST     ii i CALISA   (8) The co ela ion ma ix o all eg esso s in (8) is p o ided in Table 4. Va iance In l a ion Fac o (VIF) o all explana o y a iables was ound lowe han 4, wi h a maximum alue o 3.24 o LISA. The e o e, mul icollinea i y o he model (8) did no cons i u e a p oblem. Again, assump ions ega ding he use o OLS we e examined, wi h he esul OLS is an app op ia e es ima ion me hod wi h an excep ion o he e oscedas ici y o ln(GDP), so OLS wi h co ec ed he e oscedas ici y in G e l was pe o med again. Es ima ion esul s a e shown in Table 5. All a iables associa ed wi h ade ic ions we e ound s a is ically signi i can , hough only adjacency was ound s a is ically signi i can a 0.01 le el. Mo eo e , eg ession coe i cien s o all ic ion a iables we e ound nega i e as expec ed. Acco ding o coe i cien s’ alues, he mos impo an ade ic ion is adjacency – coun ies no bo de ing wi h Ge many impo signi i can ly less (by a ac o o 1.6) han Ge many’s neighbo s. Adjus ed coe i cien o de e mina ion o he model is e y high, R2 = 0.92, which indica es he model is app op ia e and possess high explana ion powe . T ade ic ion Ac onym Values The same cu ency C yes: 0, no: 1 Sha ed bo de s wi h Ge many A yes: 0, no: 1 Loca ion in he Schengen a ea LISA yes: 0, no: 1 Sou ce: own Tab. 3: Selec ed ade ic ions EM_3_2016.indd 24EM_3_2016.indd 24 8.9.2016 14:10:588.9.2016 14:10:58 25 3, XIX, 2016 Economics 25 A guably, he mos impo an de e minan s o expo olumes (o sha es) a e dis ance be ween ading coun ies and a alue o GDP o impo ing coun ies (coun ies’ weal h). Howe e , when expo is low, in l uence o o he ac o s may be no negligible. Fo example, one la ge go e nmen o p i a e con ac (such as opening o Bosch’s subsidia y in Kenya du ing 2014), may change he expo da a subs an ially. This can be easily illus a ed by he S a is ische Bundesam (2013; 2014) da a on Ge many’s expo s om 2012 and 2013. The change in expo olumes be ween yea s 2012 and 2013 o he op en impo e s om Ge many ( om F ance o Belgium) was abou 2% on a e age, bu o coun ies anked om he 101 h o he 110 h place ( om El Sal ado o Zambia) his yea - o-yea change amoun ed o 46% on a e age. A ques ion a ises, whe he such da a can be conside ed mo e han jus a noise. In his s udy only he da a o op 66 coun ies wi h a leas 0.10% sha e o Ge many’s expo was examined, bu i s o al sha e o expo is 97.6%. Remaining 173 coun ies sum up o only 2.4% o Ge many’s expo , bu , i used, hey would o m a majo i y o he da ase . Tha is he main eason why hey we e le ou o his s udy. Howe e , i a e aged da a o e some longe pe iod ( i e o en yea s) demons a e mo e s abili y, hen i migh be possible o include hese coun ies as well. Conclusions In his s udy Ge many’s la es agg ega e expo sha es wi h he use o a g a i y equa ion wi hou and wi h ade ic ions we e examined. This is a sligh ly di e en app oach om a s anda d me hodology whe e ade olumes a e s udied wi h he use o c oss-sec ional o panel da a. Also, only da a o coun ies wi h impo sha es om Ge many exceeding 0.10% we e employed, as coun ies wi h lowe ade olumes a e suscep ible o la ge yea - o- yea l uc ua ions, which a ec he es ima ion by g a i y equa ion nega i ely. The use o a e aged da a o e longe pe iods migh be mo e app op ia e as elimina ion o smoo hing Ln(DIST) Ln(GDP) A C LISA Ln(DIST) 1 0.506 0.545 0.493 0.726 Ln(GDP) 1 0.077 0.239 0.364 A 1 0.334 0.562 C 1 0.655 LISA 1 Sou ce: own Reg esso s OLS wi h c. h. cons . 0.416 (0.494) Ln(Dis ) -0.695 (0.088)*** Ln(GDP) 0.877 (0.040)*** A -0.495 (0.135)*** C -0.351 (0.135)** LISA -0.328 (0.171)* No. o obse . 66 Sou ce: own No es: S anda d e o s in b acke s. * signi i can a 10%; ** signi i can a 5%; *** signi i can a 1% Tab. 4: Co ela ion o explana o y a iables in he model (8) Tab. 5: Es ima ion esul s o he model (8) EM_3_2016.indd 25EM_3_2016.indd 25 8.9.2016 14:10:588.9.2016 14:10:58 26 2016, XIX, 3 Ekonomie 26 o da a l uc ua ions may esul in much be e s a is ical pe o mance o g a i y models in gene al. One o main i ndings o his s udy is ha he ic ionless g a i y model is e y success ul in i ing he da a wi h adjus ed coe i cien o de e mina ion R2 equal o 0.74. As expec ed, expo sha es we e ound ( oughly) di ec ly p opo ional o a GDP o an impo e and nega i ely p opo ional o impo e s’ dis ance. Mo e in e es ing i ndings conce n he g a i y model wi h ic ions, namely adjacency, cu ency and loca ion in he Schengen a ea. This model i ed he da a e en be e han ic ionless model, wi h he adjus ed coe i cien o de e mina ion R2 as high as 0.92. All ic ions we e ound s a is ically signi i can a 0.10 le el, and hei eg ession coe i cien s we e ound nega i e, which means hey we e ac o s con ibu ing o he ade dec ease indeed, wi h (no )adjacency as he mos impo an ade ba ie i sel . The bo de e ec , which diminishes ade subs an ially e en in cases whe e no bo de s a e physically p esen (as in he EU), is s ill conside ed puzzling, see Ande son (2010). One possible explana ion, somewha o e looked in he li e a u e, migh es in in o ma ion de i ciency. To ade, in o ma ion abou demand and supply o pa icula goods mus be a ailable o bo h po en ial ade pa ne s. Bu subjec s o ade (Ge many’s expo e s, o example) a e be e in o med abou si ua ion a hei home ma ke in Ge many han abou he si ua ion a neighbo ing ma ke s ( o example in Belgium), because hey p ima ily acqui e in o ma ion h ough home Ge man media (TV, newspape s, In e ne , e c., and also h ough pe sonal con ac ). Howe e , nea bo de s wi h Belgium Ge man expo e s can acqui e in o ma ion om Belgian sou ces, and hus could be, a leas pa ially, in o med abou i s ma ke . This in o ma ion acquisi ion is e en mo e educed when he e is no bo de be ween bo h coun ies (no adjacency), as in o ma ion can be sha ed only indi ec ly (by In e ne , sa elli e TV) o by (no so o en) pe sonal con ac . Tha is why ade wi h a o eign subjec is less likely. This pape was suppo ed by he Minis y o Educa ion, You h and Spo s Czech Republic wi hin he Ins i u ional Suppo o Long- e m De elopmen o a Resea ch O ganiza ion in 2015. Re e ences Ande son, J. E. (2010). The G a i y model (NBER Wo king Pape No. 16576). Camb idge, MA: Na ional Bu eau o Economic Resea ch. Ande son, J. E. (1979). A heo e ical ounda ion o he g a i y equa ion. Ame ican Economic Re iew, 69(1), 106-116. Ande son, J. E., an Wincoop, E. (2003), G a i y wi h G a i as: A solu ion o he Bo de Puzzle. Ame ican Economic Re iew, 93(1), 170-192. doi:10.3386/w8079. Baie , S. L., & Be gs and, J. H. (2009). Bonus Ve us OLS: A Simple Me hod o App oxima ing In e na ional T ade-Cos E ec s using he G a i y Equa ion. Jou nal o In e na ional Economics, 77(1), 77-85. doi:10.1016/j.jin eco.2008.10.004. Be gs and, J. H. (1985). The G a i y Equa ion in In e na ional T ade: Some Mic oeconomic Founda ions and Empi ical E idence. Re iew o Economics and S a is ics, 67(3), 474-481. Be gs and, J. H., Egge , P. (2011). G a i y Equa ions and Economic F ic ions in he Wo ld Economy. Palg a e Handbook o In e na ional T ade, 532-570. Chaney, T. (2011). The g a i y equa ion in in e na ional ade: an explana ion (NBER Wo king Pape s No. 19285). Camb idge, MA: Na ional Bu eau o Economic Resea ch. Dea do , A. (1998). De e minan s o Bila e al T ade: Does G a i y Wo k in a Neoclassical Wo ld? The Regionaliza ion o he Wo ld Economy. 7-32. Disdie , A., Head, K. (2008). The puzzling pe sis ence o he dis ance e ec on bila e al ade. Re iew o Economics and S a is ics, 90(1), 37-41. doi:10.1162/ es .90.1.37. Egge , P. (2000). A No e on he p ope econome ic speci i ca ion o he g a i y equa ion. Economic Le e s, 66(1), 25-31. doi:10.1016/S0165-1765(99)00183-4. Heckman, J. J. (1979). Sample selec ion bias as a speci i ca ion e o . Econome ica, 47(1), 153-61. doi:10.2307/1912352. Helpman, E., Meli z, M. J., & Rubins ein, Y. (2008). Es ima ing T ade Flows: T ading Pa ne s and T ading Volumes. Qua e ly Jou nal o Economics, 123(2), 441-487. doi:10.1162/qjec.2008.123.2.441. He e a, E. G. (2013). Compa ing al e na i e me hods o es ima e g a i y models o bila e al ade. Empi ical Economics, 44(3), 1087-1111. doi:10.1007/s00181-012-0576-2. EM_3_2016.indd 26EM_3_2016.indd 26 8.9.2016 14:10:588.9.2016 14:10:58 27 3, XIX, 2016 Economics 27 He e a, E. G., & Baleix, J. M. (2010). A e es ima ion echniques neu al o es ima e g a i y equa ions? An applica ion o he impac o EMU on hi d coun ies’ expo . Re ie ed Feb ua y 4, 2014, om h p://www.ub.edu /jei/ pape s/GOMEZ-MILGRAM.pd . In e na ional Mone a y Fund. (2013). IMF Da a. Re ie ed Oc obe 12, 2013, om h p:// www.im .o g. Ma ínez-Za zoso I., Nowak-Lehmann, F., & Vollme , S. (2007). The log o g a i y e isi ed (CEGE Discussion Pape No. 64). Gö ingen: Uni e si y o Gö ingen. O’B ian, R. M. (2007). A Cau ion Rega ding Rules o Thumb o Va iance In l a ion Fac o . Quali y and Quan i y, 41(5), 673-690. doi:10.1007/s11135-006-9018-6. Sal a ici, L. (2014). The g a i y model in in e na ional ade (AGRODEP Technical No es No. TN-04). Washing on, DC: In e na ional Food Policy Resea ch Ins i u e. San os Sil a, J. M., & Ten ey o, S. (2006). The Log o G a i y. The Re iew o Economics and S a is ics, 88(4), 641-658. doi:10.1162/ es .88.4.641. S a is isches Bundesam . (2013). Fo eign T ade – Ranking o Ge many’s ading pa ne s in o eign ade. Re ie ed Ap il 28, 2014, om h ps://www.des a is.de/EN/Fac sFigu es/ Na ional EconomyEn i onmen /Fo eignT ade/ T adingPa ne s. S a is isches Bundesam . (2014). Fo eign T ade – Ranking o Ge many’s ading pa ne s in o eign ade. Re ie ed Ap il 22, 2014, om h ps://www.des a is.de/EN/Fac sFigu es/ Na ional EconomyEn i onmen /Fo eignT ade/ T adingPa ne s. S a is isches Bundesam . (2015a). The main Ge man expo p oduc : mo o ehicles. Re ie ed Decembe 5, 2015, om h ps://www.des a is.de/EN/Fac sFigu es/ Na ionalEconomyEn i onmen /Fo eignT ade/ T adingGoods/Cu en .h ml. S a is isches Bundesam . (2015b). S a is isches Jah buch 2015. Re ie ed Decembe 6, 2015, om h ps://www.des a is.de/DE/Publika ionen/ S a is ischesJah buch/S a is ischesJah buch2015. pd ?__blob=publica ionFile. Timeandda e. (2013). Calcula o s and Time s. Re ie ed Oc obe 24, 2013, om h p:// www. imeandda e.com/wo ldclock/dis ance.h ml. Tinbe gen, J. (1962). Shaping he Wo ld Economy: Sugges ions o an In e na ional Economic Policy. New Yo k: The Twen ie h Cen u y Fund. Mg . Jiří Mazu ek, Ph.D. Silesian Uni e si y in Opa a School o Business Adminis a ion in Ka iná Depa men o Ma hema ical Me hods in Economics [email p o ec ed] EM_3_2016.indd 27EM_3_2016.indd 27 8.9.2016 14:10:598.9.2016 14:10:59 28 2016, XIX, 3 Ekonomie 28 Coun y Expo sha e (%) Dis . (km) GDP PPP (bil. USD) C LISA A 001 F ance 9.146 879 2,534.5 0 0 0 002 Uni ed S a es 8.174 6,727 16,768.1 1 1 1 003 Uni ed Kingdom 6.906 932 2,320.4 1 1 1 004 Ne he lands 6.492 577 780.3 0 0 0 005 China 6.121 7,377 16,149.1 1 1 1 006 Aus ia 5.148 523 376.7 0 0 0 007 I aly 4.871 1,183 2,035.4 0 0 1 008 Swi ze land 4.293 752 432.0 1 0 0 009 Poland 3.885 520 896.8 1 0 0 010 Belgium 3.882 651 455.0 0 0 0 011 Russian Fede a ion 3.275 1,616 3,491.6 1 1 1 012 Spain 2.868 1,870 1,488.8 0 0 1 013 Czech Republic 2.843 280 287.6 1 0 0 014 Tu key 1.955 2,042 1,443.5 1 1 1 015 Sweden 1.894 813 418.2 1 0 1 016 Hunga y 1.601 691 229.6 1 0 1 017 Japan 1.562 8,940 4,667.6 1 1 1 018 Denma k 1.449 356 240.9 1 0 0 019 Ko ea Rep. 1.322 8,150 1,697.0 1 1 1 020 B azil 1.033 9,573 3,012.8 1 1 1 021 Slo akia 0.973 554 144.0 0 0 1 022 Uni ed A ab Emi a es 0.906 4,641 570.6 1 1 1 023 Romania 0.882 1,297 371.2 1 1 1 024 Saudi A abia 0.844 4,175 1,553.1 1 1 1 025 India 0.837 5,793 6,776.0 1 1 1 026 Mexico 0.818 9,741 2,058.9 1 1 1 027 Canada 0.807 6,146 1,518.4 1 1 1 028 Aus alia 0.785 16,062 1,052.6 1 1 1 029 Sou h A ica 0.780 8,789 662.6 1 1 1 030 No way 0.750 840 328.0 1 0 1 031 Finland 0.747 1,109 218.3 0 0 1 032 Po ugal 0.582 2,315 268.8 0 0 1 033 Singapo e 0.577 9,928 425.3 1 1 1 034 Taiwan 0.538 8,971 970.9 1 1 1 035 Hong Kong 0.514 8,767 382.5 1 1 1 036 Luxembou g 0.507 601 48.5 0 0 0 037 I eland 0.500 1,320 213.3 0 1 1 038 Uk aine 0.492 1,210 392.5 1 1 1 Appendix A: The da a o he g a i y model wi h ic ions – Pa 1 EM_3_2016.indd 28EM_3_2016.indd 28 8.9.2016 14:10:598.9.2016 14:10:59