Tariff Cooperation in Free Trade Areas

Transcription

1 MPRA Munch Personal RePEc Archve Tarff Cooperaton n Free Trade Areas Joseph Ma and Andrey Stoyanov 6 February 2018 Onlne at MPRA Paper No , posted 7 February :49 UTC

2 Tar Cooperaton n Free Trade Areas Joseph Ma Andrey Stoyanov y Abstract Ths paper develops a model of endogenous trade polcy formaton to study the mpact of preferental trade agreements (PTA) on members external trade polces when members nternalze the ntra-bloc welfare e ects. Ths model s emprcally tested usng global trade data coverng 170 countres and 177 PTAs establshed between 1988 and Ths paper nds emprcal evdence of tar cooperaton between members of FTAs. Usng three d erent measures of poltcal relatons (the a nty scores from the UN General Assembly Votng Data, dyad allances data, and blateral events and nteractons data), we show that members wth good poltcal relaton cooperate more on external tar polcy after formaton of FTAs. On average, an ncrease n market share of PTA partners rms by one standard devaton s assocated wth about 3 percentage ponts ncrease n external tar n ndustres that matter for ntra-bloc members. 1 Introducton Do members n Free Trade Area (FTA) cooperate on ther external tar polcy? Ths queston s centrc to polcy makers who want to understand the welfare mplcatons of FTA for ts members. If FTA members cooperate on the level of ther external tar s, they would behave smlarly to those n Customs Unon (CU), and become more protectonst aganst outsders after the formaton of the trade agreement. 1 As a result, trade creaton wthn the trade bloc can be plausbly generated at the cost of trade loss from the outsders of the bloc. Theoretcal lterature have shown that FTA leads to further reducton n trade protecton among ts members but ths result s manly bult on the assumpton of Tongj Unversty, Department of Economcs and Fnance, School of Economcs and Management, 808/Block A Tongj Tower, 1 Zhangwu road, Shangha, Chna. E-mal: josephm@tongj.edu.cn y York Unversty, Department of Economcs, Faculty of Lberal Arts and Professonal Studes, 1084 Var Hall, 4700 Keele St., M3J 1P3, Toronto, Canada. E-mal: andreyst@yorku.ca 1 See Syropoulos (2002), Facchn et al. (2013), Kennan and Rezman (1990), Bagwell and Stager (1997), Ornelas (2005) and Freund and Ornelas (2010). 1

3 non-cooperatve polcy of FTA members. However, emprcally there has been very lttle known on ths matter. In ths paper, we attempt the emprcal study on ths matter. To do so, rst we construct a model of endogenous trade polcy formaton that ncorporates some of the channels, dent ed n prevous lterature, through whch FTA can a ect the external tar s of a FTA member. Ths base model provdes a testable predcton on how tar complementarty and tar revenue a act the choce of external tar polcy for FTA members through terms-of-trade e ects. Smlar results are found n Bond, Rezman, and Syropoulos (2004), Ornelas (2005b), and Estevadeordal, Freund, and Ornelas (2008). These results show that FTA nduces acceleraton on external tar lberalzaton among ts members, and ths result s consstent to the ndngs n prevous lterature about the buldng block e ect of FTA. Next, we extend ths model by ntroducng a poltcal factor that allows us to capture and d erentate the motve for tar cooperaton between FTA members. FTA members wll nternalze the e ect of ther external tar polcy on ther partner s welfare f they are concerned about each others. The nternalzaton takes the form of jontly optmzaton of external tar by ncorporatng the partner s welfare functon nto each other s objectve functons. We allow heterogenety n the degree of nternalzaton whch s weghted by ths poltcal factor. We use poltcal relaton between FTA partners as the measure for ths poltcal factor. Ths poltcal feature of the model shows that FTA members coordnate more on external tar settng when ther poltcal relatons are closer. It also mples that when a FTA s formed by close poltcal partners, ther external trade polcy tends to be more protectonst as what we would expect to nd among CU members. To test for tar cooperaton among FTA members, we formulate our emprcal spec caton based on the predctons of the equlbrum external tar polcy from the structural model. From the theoretcal model, we derve two equlbrum polces: () equlbrum external tar polcy under non-cooperatve setup; () equlbrum external tar polcy under cooperatve setup. We derve the d erence n external tar polcy from the equlbrum under these two setups and specfy our emprcal strategy to test the d erence as the e ect of tar cooperaton on external tar. The data we use for ths project come from several sources. Data on tar cover 170 d erent countres for the perod from 1988 to 2011 at HS 6-dgt level. Durng ths tme perod, a total of 177 free trade agreements have been establshed. Industral data n manufacturng sector coverng these countres durng the perods are at ISIC 3-dgt level. To approxmate poltcal relatons between 2

4 countres we employ three measures: () the a nty scores from the UN General Assembly Votng Data; () the formal allance status from the Correlates of War Formal Allance data; and () blateral poltcal events and nteractons from the Con ct and Peace Data Bank. 2 We then estmate our emprcal spec caton usng the data. Our results provde strong evdence for tar cooperaton among the FTA formed durng the perod from 1988 to On average, the external tar s on the rent-generatng ndustres for FTA partners lberalze 20% slower compared to other ndustres. The results also show that poltcal relaton plays an mportant role n determnng FTA partners external tar polcy. Comparng to the results wthout the poltcal factor as the benchmark, we nd an addton of 8% slower lberalzaton n external tar s on the rent-generatng ndustres among the FTA partners who share smlar economc and poltcal nterests, and an addton of 20% slower lberalzaton among those who have formal allance wth mltary defense treaty. Our work s related to the strand of lterature on the mpact of regonalsm on multlateralsm. As n Grossman and Helpman (1994) and Stoyanov (2009), RTA weakens the motvaton for tar protecton by creatng leakage n ths protectonst trade polcy so that reduces the ncentve for specal poltcally actve group lobbyng for protecton. Levy (1997) shows that blateral trade agreement may dsproportonally bene ts the countres medan voters, thus ncreasng the support aganst protecton lberalzaton. Krshna (1998) suggests that f RTA creates large gans to some powerful economc groups, reducton s external protecton becomes poltcally nfeasble. Our work s also related to the studes on ncentves to alter external tar s n RTAs. Kennan and Rezman (1990) show that n a three-country general equlbrum endowment economy, equlbrum external tar s are hgher when countres movng from FTA to CU. Rchardson (1993) nds that FTA members tend to lower external tar s to mtgate the negatve mpact from trade dverson generated by RTA. Bagwell and Stager (1997), Freund (2000), Bond, Rezman, and Syropoulos (2004), Ornelas (2005a, 2005b), and Sagg and Yldz (2010) llustrate how FTA nduces ncentve for multlateral trade through terms-of-trade e ect. Our paper s also related to Lmao (2007) s study. He looks at tar cooperaton of FTA partners n non-trade ssues. Preferental treatment s o ered to RTA partners n exchange for cooperaton on, for example, drug trade ssues. A reducton on external tar protecton would dmnsh the bene t the RTA partners from ths preferental treatment, so that t could erode the ncentve for the partners to mantan ther cooperaton on non-trade ssues. Ths paper s closely related to Ma and Stoyanov 2 Detal descrptons of the data can be found n Secton 4. 3

5 (2015) s study, whch analyses the e ect of CUSFTA on Canadan external trade polcy. In ths study, they categorze Canadan ndustres n quartles ranked by the rent generated to US, and they nd that the ndustres whch create the least rent to US have exhbted fastest lberalzaton n external tar s. The paper s organzed as the follows. In the next secton, we present the model of endogenous trade polcy formaton and derve the equlbrum external tar polcy under tar cooperaton. In Secton 3, we present the emprcal spec caton and the nterpretatons of the key varables. In Secton 4, we dscuss the data used for ths project. In Secton 5, we present the emprcal results and the dscusson of the results. Secton 6 concludes. 2 The Theory In ths secton, we study the channels through whch d erent degrees of poltcal relaton can a ect a country s choce on ts external tar s. We present a monopolstc competton model wth d erentated products and free market entry, and derve the equlbrum trade polcy for a country whch faces heterogeneous poltcal a nty to her partners wth preferental trade agreements (PTA). The theortcal model wll lay foundatons for our emprcal spec catons. To dentfy tar cooperaton wthn a FTA, our strategy s to ncorporate a measurement for poltcal relatons among FTA partners as the weghts for the mportance of the partners natonal welfares that are taken nto consderaton nto government s decson of the formaton of ts external trade polcy. Then derve and compare the equlbrum trade polcy wth and wthout tar cooperaton. Consder a model wth a Home country, H, tradng wth R partner countres wth PTA and F countres wthout PTA, hereafter ndexed by H, f and r respectvely. Note that f = 1; :::; F and r = 1; :::; R, and country j 2 fh; f; rg. All countres produce and trade N + 1 goods, wth the rst good beng a numerare, traded at no costs and produced by perfectly compettve rms. For all other ndustry the number of rms n each country j s xed and equal to n j, and each rm produces a dstnct varety of a good. All rms n ndustry are assumed to be symmetrc wthn country j, therefore they share the same demand functon, producton technology and charge the same prce. There are (n H + P F f=1 n f + P R r=1 n r) varetes n ndustry avalable to consumers n country H for H 6= f and H 6= r. Suppose that the preferences of a representatve agent n country j can be denoted by a quaslnear 4

6 utlty functon wth a constant elastcty of substtuton for varetes n ndustry nx U = X 0 + ln X ; =1 nx a = 1 (1) =1 where s the fracton of total expendture the agent spends on ndustry goods. X 0 s consumpton of the numerare good. X s the sub-utlty derved from the consumpton of d erentated product produced at home and abroad, and assumed to take the non-symmetrc CES form, whch can be represented by 0 X H d 1 1 H x H 1 1 r x r + r=1 f=1 R + X n r d n fj d 1 1 f x f 1 A 1 (2) where > 1 s the elastcty of substtuton among varetes of good, d j denotes the taste parameter for from country j, and x j s the demand for produced n country j. Maxmzng equaton (1) subject to standard budget constrant, we obtan the demand functon and aggregate prce ndex for the d erentated product : x j = a d j p 8 j 2 fh; f; rg (3) R 1 P H d H p 1 + X n r d r p 1 n f d f p 1 A : (4) j P 1 H r + r=1 f=1 f Home country government sets two types of ad valorem tar s - preferental tar, r, on mports from country r and MFN tar, f, on mports from country f. The pro t functon facng d erent rms n Home country can be wrtten as: H = (p H c H ) q H (5) j = ((1 j )p j c j ) q j 8 j 2 ff; rg where q j s the quantty suppled and c j s the margnal costs of producton. Assumng the number of rms s large enough that an ndvdual rm s decson on prcng has no mpact on the aggregate prce ndex P, each rm takes the prce ndex as gven. Knowng the demand functon, each rm 5

7 maxmzes pro t by settng ts prce: p H = 1 c H, p j = 1 ( 1) (1 j ) c j 8 j 2 ff; rg (6) where j s the ad valorem tar collected by the Home country government. Each rm sets ts prce by a mark-up over ts margnal cost. Substtutng equaton (6) nto (5), the pro t functons can be wrtten n a convenent form: H = 1 p H x H (7) j = (1 j ) 1 p j q j 8 j 2 ff; rg: The next step towards dervng the optmal trade polcy on external tar s to set up government s objectve functon. One ssue s that we do not know the form of the objectve functon. Partcpaton n any trade agreement s afterall a poltcal decson therefore ths functon can take varous forms dependng on the objectves of governments. 3 socally desrable for ther domestc welfare. For now, we assume that governments objectves are Let the objectve functon, G, consst of the sum of consumer surplus from consumpton of d erentated goods (CS), tar revenue (T R), and pro ts of domestc rms ( H ): where and G = CS( j ) + T R( j ) + CS( j ) = U(X 0 ; X ; j ) T R() = nx n H H ( j ) (8) =1 nx =1 H;F;R X j=1 p j n j x j ; (9) nx XF;R j n j x j : (10) =1 j=1 2.1 Tar cooperaton n FTA One dstncton between members of FTA and CU s that FTA members mantan autonomous external tar polces whle CU members jontly set and share common external tar s. Ths subtle d erence creates the tendency for members of CUs to coordnate and adopt hgher external tar s than those of 3 Grossman and Helpman (1994), (1995), Krshna (1998), and Stoyanov (2009) show that specal nterest groups n uence governments objectves. Lmao (2007) shows that RTAs can be used by governments as motves to nduce partner country to cooperate n non-trade areas. 6

8 FTAs. 4 Tar coordnatons among CUs have been crtczed because they could hnder the vablty of multlateral free trade. However, t s not yet clear whether tar coordnaton exsts among FTAs because government can adjust external tar rates to accommodate the mpact from the changes n preferental polcy and reassure ts objectves. 5 Next, we look at the case when a government of FTA members s concerned about the welfare of other members of the FTA, t ncorporates the welfare of other members nto ts objectve functon (8), t becomes: nx RX G = CS( f ) + T R( f ) + n H H ( f ) + r W r ; (11) =1 r j=1 where W r s the welfare of partner country r and r 2 [0; 1] measures the degree of concern country H to W r. Snce the external polcy of H wll a ect the pro ts of exportng rms of r n H and not a ect the consumer surplus and tar revenue of partner r, we can smplfy expresson (11) as the followng: nx nx RX G = CS( f ) + T R( f ) + n H H ( f ) + r n r r ( f ); (12) =1 =1 r=1 where n r r s the total pro ts of rms from partner r n ndustry n H. Government of H nternalzes the pro ts of rms from partner r (essentally partner s welfare) nto consderaton when optmzng ts objectve functon, and the larger s r the hgher degree of nternalzaton takes place for ts partner r s welfare n H s objectve functon. By choosng the level of external tar rate f, H maxmzes ts object n (12), and after collectng terms, we obtan the followng equlbrum external tar polcy: where s H = nrprxr a f f = 1 s H + ( 1) RX r=1 r s r + 1 RX r (1 r ) s r (13) s the market share of domestc rms n ndustry and s r = nrprxr a r=1 s market share of rms n ndustry from partner country r n H. On the left hand sde of equaton (13) s external tar rate f multpled by f, whch s the mport demand elastcty for goods mported from outsders. In (13), s H and s r are postvely related to external tar f. The postve relaton between s H and f suggests that f the market share of domestc rms n ndustry s low after the formaton of FTA, t s optmal to have a low external tar rate for. One explanaton for ths 4 See Kennan and Rezman (1990) and Facchn et al. (2013) among others for ratonales for hgher external tar s under CUs. Estevadeordal et al. (2008) nd tar complementarty only n FTAs. 5 See for example of optmal external tar s settngs Kennan and Rezman (1990), Rchardson (1993), Y (1996), Bagwell and Stager (1999), Cadot et al. (1999), Freund (2000) and more recently Ornelas (2005a, 2005b), (2007), Facchn et al. (2009). 7

9 postve relaton s the protecton leakage e ect. If s H s small n H, protecton from hgh f wll be an ne cent rent-generator for domestc rms, nstead, part of the rents would go to the partners rms. Because protecton for rents to domestc rms becomes less e ectve when s H s small, H would be better o to shft some of the rents from FTA partners back to CS by lowerng f. The second term n (13) re ects the e ect of tar complementarty such as n Rchardson (1993). If preferental tar s are low - large preferental margns, t s optmal for government to lower the correspondng external tar s. The complementary e ect s stronger wth larger market share of partner rms and closer substtude of product. Intutvely, a drop n ntra-bloc tar nduces a shft of mports from outsders, who have comparatve advantage n producton of, to less e cent partner r thanks to preferental tar treatment. Thus, the trade dverson reduces overall welfare. To mtgate the welfare lost from the dstorton of trade pattern, governments can lower external tar s to redrect some of the mports back to ther orgnal sources. 6 The last term n (13) ndcates the e ect from tar cooperaton. The postve relaton between f and s r suggests that, f H s concerned about partner r s welfare, t s optmal to coordnate wth hgh external tar for ndustry f the market shares of rms from partner r n ndustry s large after the formaton of FTA. The ntensty for tar cooperaton s strong f r s hgh. Reducng the external tar n would nduce competton comng from external rms and thus reduce the rents of the partner s rms. For the case when r = 0, t mples no tar cooperaton between FTA members. The welfare functon of FTA partner W r n (11) does not enter the objectve functon of H government. In such case, the external tar polcy n equlbrum becomes ROW ROW = 1 s H + ( 1) RX r s r : (14) r=1 3 Emprcal spec caton Polcy response often may not be nstantaneous. To allow for polcy response delays, we ntroduce tme dmenson to both (13) and (14). Next, we move to estmate tar cooperaton among FTA partners. Frst, suppose H and r form a trade agreement at tme t. If H coordnates ts external tar for r nto 6 Other researchers also provde d erent explanatons for tar complementarty, for example Bagwell and Stager (1999) n terms of trade motvatons; Grossman and Helpman (1994) and Ornelas (2005a) n poltcal economy factors; Sagg and Yldz (2010) n endowment models wth endogenous trade agreements. 8

10 perod t + 1, from (13), we have f f;t+1 = 1 s H;t+1 + ( 1) RX r=1 r;t+1 s r;t RX r (1 r;t+1 ) s r;t+1 : (15) Comparng (15) to the equlbrum external tar polcy represent by (14) n perod t, and tme d erencng gves us the followng: r=1 f 1 f ;t = 1 1 s H;t + ( 1) RX r=1 1 r;t s r;t + 1 RX r (1 r;t ) s r;t (16) where 1 represent tme d erence by 1 perod. Note that the last term on the rght hand sde of (16) s the market share from r n level, for whch, one nterpretaton can be that t captures the mportance of ndustry to partner r. Equaton (16) outlnes the two forces drvng external tar s n opposte drectons - tar complementarty and tar coordnaton. Wth our man goal focused on testng tar coordnaton among FTAs, and motvated by (16), our emprcal framework s the followng: r=1 1 f 1 f ;t = s H;t + 2 RX 1 r;t s r;t + 3 R X r=1 r=1 r;t (1 r;t ) s r;t + ;t (17) where s r;t s measured by the value of mports of good from partner r over the total domestc spendng on at tme t and s H;t s the domestc output over the total domestc spendng on at t. Equaton (17) assembles the relatonshp between external tar and the key explanatory varables from the model nto an econometrc form. Note that for the FTAs where r;t mmedately reduce to zero, for followng years r;t become zero. In such case, the term P R r j=1 r j;ts rj;t dsappears and P R r j=1 r j;t 1 rj;t srj;t becomes P R r j=1 r j;t s rj;t. In prevous lterature, the role of poltcal a nty s often overlooked when nvestgatng the relatonshps between preferental tar and MFN tar for ndustres traded wth FTA partners and outsders. 7 The e ects of preferental tar from d erent FTA partners on MFN tar are treated homogeneously. However, relatons of countres are heterogeneous and thus there s a strong reason to beleve the e ect of tar cooperaton on external tar must be d erent by country f poltcal relatonshp between partner countres matters when a country sets up ts external polces. One novelty of ths study s that we estmate the e ect 7 For example, Lmao s (2006) approach s to assgn a dummy varable that equals one for ndustres US mports from ts PTA partner and compare the change n MFN tar s n these ndustres to those that do not mport from PTA partners post- and pre- Uruguay Round. 9

11 of preferental tar on external tar s under a poltcal economc perspectve by d erentatng the mportance of FTA partners welfare to government H so that the mpact on ts external polcy s weghted d erently by how close the poltcal relatons are ts partners. In equaton (17), r s a partner country spec c poltcal parameter that weghts the mportance of the market shares from r n H. The dea s that, f partner country r s poltcally close to H, r s hgh and share of partner rms s r;t matters more n the relaton to f ;t ceters parbus, we would expect a more sgn cant of 3 n the estmaton of (17). To test whether or not poltcal relaton r matters for tar cooperaton, we estmate two sets of results on (17) - one treatng all r homogeneously (e.g. r = 1 for all r), and another allowng r to vary. Consder four scenaros - a par FTA partners can have hgh or low r pre- or post- FTA formaton. Fgure 1 llustrates the level of MFN tar of H for non-members before and after the formaton of FTA. External tar, f ;t, on country f s at f ;t 0 ( L r j;t0 ; H r j;t0 ) from tme t 0. Suppose H and r forms a FTA at tme t 1, the external tar on f can ether go up or down or reman unchanged. 8 The average MFN tar rate s 9:6% for the perod from 1988 to 2011, and have decreased by 0:35% per year. Fgure 1 shows ths general downward trend of global MFN tar s. The dea s that f Home and country r j are close poltcal partners, Home s concerned about r j s welfare and adopts a more protectve external polcy after the formaton of FTA. In such case, f s hgher than t would otherwse be n post PTA perod t 1. For the country par, poltcal a nty set ( H r j;t0 ; H r j;t1 ) and ( L r j;t0 ; H r j;t1 ), f ;t+1 wll be at f Hgh ;t 1. External tar s lower at f Low ;t 1 f Hgh ;t 1 and f Low ;t 1 for ( H r j;t0 ; L rj;t 1 ) and ( L r j;t0 ; L rj;t 1 ). The d erence between s the stumblng e ect on external lberalzaton comng from tar cooperaton based on ther level of poltcal a nty rj;t. More mportantly, a statstcally sgn cant rj;t would provde a possble answer to the puzzle why some nd stumblng block e ect n FTAs whle others nd buldng block e ect. If tar cooperaton exsts, we would be more lkely to nd t n ndustres that mport under PTA. We expect there s no tar cooperaton for ndustres that have no mports from PTA partners because the hgher external tar rate does not ncrease the pro ts of partner countres rms, and t s not n the nterest of the Home government too. We expect the e ect of cooperaton wll be stronger f the sze of market share of partners rms n s large. 9 The the e ect of PTA on external tar would 8 WTO member conutres have legally bound commtments on tar rates, whch act as the celngs on tar s. Governments can adjust tar rates upward only f the appled rates are lower than the bound rates. 9 Usng mport share of PTA partner rms n Home may not be the best way to measure the mportance of ndustry to partner country as some suggest the mportance of ndustry n Home for PTA partner country should be measured by the export share of n partner country to Home. However, we argue ths s a reasonable measure for capturng tar cooperaton because f Home s concerned about welfare of partner country, t cares the most for large ndustres n the 10

12 probably be undermned because the dummy varable forgoes the sze e ect. The key varable of nterest s P R r=1 r ;t (1 r;t ) s r;t because we are nterested to know how f ;t responds n ndustres that are bene cal to r. For capturng the potental further delays n polcy responds, we nclude 6 more lag perods of P R r=1 r ;t (1 r;t ) s r;t n (17). Changes n preferental polcy wll force partners rms to adjust ther producton and management, and the process lkely take tme. Any changes n Home s market condtons comng from the mpact of partners rms adjustments are more lkely to emerge n later perods. We expect government s response through external polcy to any market condton changes wll not happen smultaneously so we lag all explanatory varables to one perod tme. After mplementng the lag perods, the estmaton equaton becomes Y ;t = X 1 ;t X 2 ;t X 3 ;t X 3 ;t 2 + ::: + 8 X 3 ;t 6 + ;t (18) where Y ;t = X 3 ;t 1 f 1 f ;t, X1 ;t 1 = 1s H;t = s H;t s H;t 1, X;t 2 1 = P R r=1 P R r;ts r;t r=1 r;t 1s r;t 1, 1 = P R r=1 r ;t 1 (1 r;t 1 ) s r;t 1 and X 3 ;t 2 = P R r=1 r ;t 2 (1 r;t 2 ) s r;t 2. Also note that tme d erence n equaton (18) can also help remove any country-ndustry xed e ects. 3.1 Estmaton ssues Endogenety s a serous concern when estmatng the e ect of market share on external tar polcy because the level of external tar can reversely a ect the market shares of both domestc and foregn rms. Snce we are nterested n establshng the causal relatonshp, a good nstrument for the market shares of PTA partners n Home becomes very mportant. To do ths we construct an nstrument varable for traded goods from PTA partners followng the methodology detaled n Frankel and Romer (1999). We predct trade ows between countres usng geographc characterstcs from gravty equaton. Ths s a vald approach because trade ows between countres are hghly correlated to ther geographcal characterstcs, whch are unlkely correlated to any trade polces. We regress the values of mports n log for every ndustry on dyadc gravty varables. We perform the followng: ln(m j ) = a 0 + a 1 ln D j + a 2 ln N + a 3 ln N j + a 4 ln A + a 5 ln A j (19) Home market. +a 6 (L + L j ) + a 7 B + a 8 B ln D j + a 9 B ln A + a 10 B ln A j +a 11 B ln N + a 12 B ln N j + a 14 B (L + L j ) : 11

13 In equaton (19), D j s the dstance between country and j, N and A are the populaton and area n country repectvely, and L ndcates whether country s landlocked. The tted values of equaton (19) are used as our nstrument varable (IV ) for mports of Home from partner r j, denoted by mport_v1 n Table 1. We construct three addtonal IV to ensure results wll be not senstve to the way we construct the IV. For the second IV, we extend equaton (19) to nclude controlng for hgh dmensonal country-year xed e ects for country and j followng the estmaton procedure n Martyn et al. (2006). The tted values are denoted by mport_v2 n Table 1. The thrd IV and the fourth IV, nclude controls for ndustry-year xed e ects and country-ndustry-year xed e ects respectvely and are denoted by mport_v3 and mport_v4. 4 Data Ths paper uses data from several d erent sources. Tar data comes from UN Comtrade Database, coverng 177 free trade agreements nvolved 170 d erent countres for the tme perod from 1988 to 2011, avalable at the 6-dgt HS class caton level. Tar rates are aggregated to ISIC 3-dgt level based on Product Concordance from World Integrated Trade Soluton (WITS). The data on output, mports, exports and mport demand elastctes f are obtaned from Ncta, and Olarreaga (2006). Domestc share, s H;t, s constructed by the value of domestc output n ndustry over the total domestc spendng on. Smlarly, the share of PTA partner, s rj;t, s the rato of the value of mports from partner r j to domestc spendng n. The mssng values n output data s mputed wth the tted value from regressng output data to ndustry ndex. The data on elastcty of substtuton,, s obtaned from Broda and Wensten (2006) at ISIC 3-dgt level. The data on geographcal characterstcs s from Mayer and Zgnago (2011). Gravty equaton varables come from Head and Mayer (2013). We employ several approaches to measure blateral poltcal relaton rj;t. For our rst measure, we use the a nty score ndex constructed by Voeten (2013) based on the Unted Natons General Assembly Votng Data (UNGAVD). The data records votng nformaton on General Assembly resolutons for each UN member country. UN members can approve, abstan, or dsapprove of each resoluton. Based on the votng data, the dyadc a nty scores s constructed as the share of smlar votes between country and country j over the total of country s votes. The a nty score ndex s often used to measure the degree of smlarty n the economc and geopoltcal nterests of a par of countres (Alesna and Dollar, 2000). For example, the average magntude of the a nty score ndex 12

14 for US and ts major poltcal alles (Canada, France, Israel, UK, and Australa) ranges from 0:58 to 0:86 for the perod between 2000 and 2012, and the score ranges from 0:17 to 0:21 for countres such as Iraq, Afghanstan, Cuba, Indonesa, and Chna durng these perods. Common poltcal nterests between countres s an mportant ndcator of good nternatonal relatons, yet not a perfect one. As Voeten (2013) ponts out, some countres may have share smlar votng patterns n the UN on global matters but have poor poltcal relatons (e.g. Inda and Pakstan). Therefore, t s necessary to have addtonal poltcal relaton proxes to complement UN a nty scores. Our second measure of poltcal relatons s the formal allance status between two countres. We retreve allances data from the Correlates of War Formal Allance (COWFA) v4.1 data set rst constructed by Small and Snger (1969) and mantaned by Gbler and Press (2009). Ths dataset covers the perods from 1835 to The COWFA dvdes country s allances nto 3 categores - defense pact, neutralty (non-aggresson) treaty, and entente agreement. We use an ndcator varable to classfy countres are n good poltcal relaton f there s defense pact, whch s the hghest level of mltary commtment among these three classes and t requres ntense poltcal cooperaton. For the thrd measure of poltcal relatons, we use the frequences of blateral events and nteractons from the Con ct and Peace Data Bank (COPDAB), whch records actons of approxmately 135 countres toward one another on a daly bass. The majorty of the nternatonal events that nvolve countres are related to poltcal relatons (37:6%), economoc relatons (27:5%), mltary and strategc relatons (14:9%), and cultural and scent c relatons (10%). The blateral nteractons are categorzed as cooperatve, neutural or uncooperatve. The events are recorded for the perod from 1948 to 1978, a bt over 20 years pror to the rst year of our data set, and the relevance of these events for the current poltcal relatons s certanly a concern. However, we argue that the nternatonal relatons between countres have remaned relatvely stable snce the end of the Second World War and the frequency of dplomatc nteractons from 1950s to 1970s can stll be nformatve of the current nternatonal relatons. 10 In order to solate the role of poltcal relatons from other n uences to the frequency of blateral events, such as the relatve sze of two countres, we rst regress the number of dplomatc cooperatve nteractons on the log of populaton, GDP, geographc area, and the log of dstance from one another usng the full sample of country-pars. 11 Because the dependent varable s a count varable wth a 10 Ths may not be the case for relatonshp wth countres from the former Sovet Block, so we excluded them from ths analyss. 11 These varables are retreved from Research and Expertse on the World Economy (CEPII). 13

15 large dsperson, the model s estmated by negatve bnomal regresson wth country-year xed e ects. We use the resduals from the regresson as our measure of poltcal relatons, whch s essentally the frequency of blateral events purged from the scale e ect. Fgure 2 plots poltcal nteractons ndex aganst the a nty scores for US and other countres n our sample. It reveals that the frequency of dplomatc nteractons s postvely correlated wth a nty scores, suggestng that the two measures capture smlar aspects of poltcal relatons. Table 1 summarzes the varables of nterests n ths study. The average MFN tar n our sample s 7.3% durng the perod from 1988 to The average reducton of MFN tar s about 0.4% each year durng ths perod. The shares of domestc output over domestc consumptons are just over 60% and shrnks at about 1% per year on average for countres n our sample. The poltcal relaton proxes from UN a nty scores, defense pack, poltcal nteractons are summarzed n UN, Defense, and Interacton respectvely. 5 Results In ths secton, we wll provde results from estmatng spec caton (18) and dscuss ther economc mplcatons. Frst, we want to learn whether n general countres are concerned wth the welfare of ther PTA partners. One way s to test f countres o er protectonst trade polcy to umbrella the ndustres that are mportant to ther PTA partners. We begn wth estmatng spec caton (18) wthout ncludng any poltcal relaton proxy. Ths estmaton shows how external tar of each country H responds n the ndustres whch have presence from ther PTA partners rms. If there s tar coordnaton between PTA members, we would expect external tar s to be relatvely hgher n these ndustres or they are lberalzed slower than they would otherwse be. Lnear regresson results are provded n Table 2. Column (1) reports the result on how changes n external tar respond n ndustres wth market presence of partners rms for lagged 6 perods. Results n column (2) and column (3) are adjusted by elastcty of substtuton and mport demand elastcty respectvely. Results n column (4) are adjusted by both types of elastcty. All these results are controlled for country-ndustry xed e ects. Results n Table 2 show that there are postve assocatons between changes n external tar s and market presence of PTA partner countres rms. The e ects are stronger n the 2 nd and the 3 rd lagged perods wth coe cents statstcally sgn cant at 0:01. The results suggest that an ncrease n market share of PTA partners rms by one standard devaton s assocated wth about 3 percentage 14

16 ponts ncrease n external tar n ndustry for the 3 rd perod after formaton of PTA. We report the panel structure regresson results from estmatng spec caton (18) wth d erent poltcal relaton proxes n Table 3. Column (1) provdes the results unweght by any poltcal relaton measures as the benchmark. For estmaton results n Column (2) to (4), the market share from PTA partners rms are weghted by poltcal relaton proxes. For example, when s rj;t s weghted by defense treaty Defense;r (n Column 3), t estmates the e ect of those ndustres mports from country H s poltcal allances. The estmaton results relaxng the elastcty are reported n column (5) - (8). If close poltcal relaton between countres nduces hgher co-operaton on external trade polcy, we would expect the e ect of weghted s rj;t on changes n external tar to be stronger. Comparng to the benchmark results n Column (1), we nd results weghted by poltcal relaton proxes are stronger especally for the 2 nd and 3 rd perods nto PTA. For example, results n Column (2) suggests that those ndustres wth rms from PTA partner countres whch share smlar geopoltcal and economc nterests enjoy about 50% more tar protecton at 3 rd year nto PTA. We nd smlar e ects when s rj;t s weghted by Defense;r and Interacton;r. Next, we report the IV estmaton results for Imports from PTA partners. Frst, we estmate the benchmark model usng unweght s rj;t and whch s nstrumented by four d erent IV s whch dscussed n the prevous secton. Results are reported ntable 4 Column (1) - (4). Column (5) - (8) report the estmaton results excludng the elastcty. In Table 5, we show the IV estmaton results for spec caton (18) usng three poltcal relaton proxes. The results from IV estmators are robust and consstent to the prevous ndngs. 6 Concluson In ths paper, we develop an endogenous trade polcy formaton model n a monopolstc competton framework wth d erentated products and free market entry. We derve the equlbrum trade polcy that ncorporates a poltcal factor to captures tar cooperaton between FTA member countres. Usng ths model, we show that multlateral lberalzaton s slower f FTA partners are concerned about the welfare of other members when settng ther external trade polcy. We test ths the predcton of the model usng comprehensve trade and ndustry data, and nd evdence that supports the tar cooperatve hypothess. In addton, we nd that the closer are the two FTA partners n poltcal relatons the more lkely these two members coordnate ther external trade polcy to bene t the natonal welfare of the other. 15

17 References Alesna, A., and D. Dollar (2000): Who Gves Foregn Ad to Whom and Why?, Journal of Economc Growth, 5(1), Andrews, M., T. Schank, and R. Upward (2006): Practcal Fxed-E ects Estmaton Methods for the Three-Way Error Components Model, Stata Journal, 6(4), 461. Bagwell, K., and R. W. Stager (1997): Multlateral Tar Cooperaton durng the Formaton of Free Trade Areas, Internatonal Economc Revew, 38(2), pp Bond, E. W., R. G. Rezman, and C. Syropoulos (2004): A Strategc and Welfare Theoretc Analyss of Free Trade Areas, Journal of Internatonal Economcs, 64(1), Broda, C., J. Greenfeld, and D. Wensten (2006): From Groundnuts to Globalzaton: A Structural Estmate of Trade and Growth, Dscusson paper, Natonal Bureau of Economc Research. Cadot, O., J. De Melo, and M. Olarreaga (1999): Regonal Integraton and Lobbyng for Tar s Aganst Nonmembers, Internatonal Economc Revew, 40(3), Estevadeordal, A., C. Freund, and E. Ornelas (2008): Does Regonalsm A ect Trade Lberalzaton Toward Nonmembers?, The Quarterly Journal of Economcs, 123(4), Facchn, G., P. Slva, and G. Wllmann (2013): The Customs Unon ssue: Why do we observe so few of them?, Journal of Internatonal Economcs, 90(1), Feenstra, R. C. (1994): New Product Varetes and the Measurement of Internatonal Prces, Amercan Economc Revew, pp Frankel, J. A., and D. Romer (1999): Does Trade Cause Growth?, Amercan Economc Revew, pp Freund, C. (2000): Multlateralsm and the Endogenous Formaton of Preferental Trade Agreements, Journal of Internatonal Economcs, 52(2), Freund, C., and E. Ornelas (2010): Regonal trade agreements, Annu. Rev. Econ., 2(1), Gbler, D. M., and C. Press (2009): Internatonal Mltary Allances, CQ Press Washngton, DC. 16

18 Grossman, G. M., and E. Helpman (1994): Protecton for Sale, Amercan Economc Revew, 84(4), (1995): The Poltcs of Free-Trade Agreements, Amercan Economc Revew, 85(4), Head, K., and T. Mayer (2013): Gravty Equatons: Workhorse, Toolkt, and Cookbook, Handbook of Internatonal Economcs, 4, Kennan, J., and R. Rezman (1990): Optmal Tar Equlbra wth Customs Unons, Canadan Journal of Economcs, pp Krshna, P. (1998): Regonalsm And Multlateralsm: A Poltcal Economy Approach, The Quarterly Journal of Economcs, 113(1), Levy, P. I. (1997): A Poltcal-Economc Analyss of Free-Trade Agreements, Amercan Economc Revew, pp Lmao, N. (2007): Are Preferental Trade Agreements wth Non-trade Objectves a Stumblng Block for Multlateral Lberalzaton?, Revew of Economc Studes, 74, Ma, J., and A. Stoyanov (2015): The e ect of the Canada-US Free Trade Agreement on Canadan Multlateral Trade Lberalzaton, Canadan Journal of Economcs, 48(3). Mayer, T., and S. Zgnago (2011): Notes on CEPIIŠs Dstances Measures: The GeoDst Database, Workng Papers , CEPII. Ncta, A., and M. Olarreaga (2007): Trade, Producton, and Protecton Database, , The World Bank Economc Revew, 21(1), Ornelas, E. (2005a): Endogenous Free Trade Agreements and the Multlateral Tradng System, Journal of Internatonal Economcs, 67(2), (2005b): Trade Creatng Free Trade Areas and the Undermnng of Multlateralsm, European Economc Revew, 49(7), (2007): Exchangng Market Access at the Outsders Expense: the Case of Customs Unons, Canadan Journal of Economcs, 40(1), Rchardson, M. (1993): Endogenous Protecton and Trade Dverson, Journal of Internatonal Economcs, 34(3-4),

19 Sagg, K., and H. M. Yldz (2010): Blateralsm, Multlateralsm, and the Quest for Global Free Trade, Journal of Internatonal Economcs, 81(1), Small, M., and J. D. Snger (1969): Formal Allances, : An Extenson of the Basc Data, Journal of Peace Research, 6(3), Stoyanov, A. (2009): Trade Polcy of A Free Trade Agreement n the Presence of Foregn Lobbyng, Journal of Internatonal Economcs, 77(1), Syropoulos, C. (2002): On Tar Preferences and Delegaton Decsons n Customs Unon: a Heckscher-Ohlne Approach, The Economc Journal, 112(481), Voeten, E. (2013): Data and Analyses of Votng n the Unted Natons General Assembly, Routledge Handbook of Internatonal Organzaton. London: Routledge, pp Y, S.-S. (1996): Endogenous Formaton of Customs Unons Under Imperfect Competton: Open Regonalsm s Good, Journal of Internatonal Economcs, 41(1),

20 7 Appendx Some of the dervatons: The response of prce and quantty to change n the tar H = ( 1) x = P 1 rj = ( 1) x r fj = x fj 1 f + ( 1) x f j 1 pfj n fj d fj = P P 1 f Response of consumer surplus to changes n external tar : Consumer surplus s s fj CS() = V (Y; P ) Y = nx a + =1 =1 nx a ln a P = nx a (ln a 1 ln P ) = a = = X P 1 f Response of tar revenue to change n external tar : s fj R() P n P F;R =1 j=1 jp j n j x j a s fj + a (1 f s fj + ( 1)X f s fj 1 f + RX ( X rj s rj r j=1 = a s fj F X a ( 1) 1 f f s fj ( 1)X f s fj + RX ( X rj s rj r j=1 = a F X s fj a ( 1) 1 f f s fj + ( f s fj + ( RX X rj s rj r j=1 = a (1 f ) a ( 1) (1 f f +( 1 f P f s fj +( RX X rj s rj r j=1 19

21 = 2 X F 41 f ( 1) f + ( 1) f s fj + ( 1) Response of rms pro ts to change n external tar : 3 RX rj s rj 5 r P n =1 P R r j=1 n r j rj = 1 RX (1 rj )s rj r P n P 2 3 F =1 f n =1 f = X 4 1 f (1 f ) s fj (1 f ) = X 4 1 f 1 (1 f s fj A5 (20) Import demand fj f x fj = f 1 f ( 1) s fj 3 5 j F j = 1 f fj f x fj = ( 1) s fj p fj x fj f j = 2 p fj x fj 4 x fj 1 f + ( 1) x f j P = + ( 1) s fj P 1 f s fj 3 5 ( 1)(1 f ) c fj p fj x fj j F j fj p fj x fj = ( 1) s fj 20

22 Table 1. Summary statstcs of varables of nterest. Varable Descrpton Mean Standard devaton Mn Max N τ ROW MFN tarff Δ t-1 τ ROW Change n MFN tarff τ PTA Preferental tarff Δ t-1 τ PTA Change n preferental tarff ε j Import demand elastcty σ j elastcty of substtuton S H,t Home ndustry share Δ t-1 S H Change n Home ndustry share S r,t Partner ndustry market share Δ t-1 S r,t Change n partner ndustry share ln(mport) log of mports ln(mport_iv 1 ) log of mports (nstrument 1) ln(mport_iv 2 ) log of mports (nstrument 2) ln(mport_iv 3 ) log of mports (nstrument 3) ln(mport_iv 4 ) log of mports (nstrument 4) ф UN UN affnty ndex ф Defense Defense treaty ndcator ф Interacton Countres nteractons ndex Note: τ ROW and τ PTA are MFN tarff and preferental tarff respectvely, aggregated from 6-dgt HS classfcaton to 3- dgt ISIC classfcaton. S r,t s the rato of output to domestc consumpton of. Imports are recorded at 6-dgt HS classfcaton. The correlatons between mports and ts nstrument varables mport_iv 1 - mport_iv 4 are 0.505, 0.829, 0.83, and respectvely. ф UN s UN affnty score. ф Defense =1 when two countres s wth defense pack. ф Interacton s the measure of poltcal and economc nteractons between countres purged from scale effects.

23 Table 2. Lnear regresson results of the effects of poltcal relaton on external tarffs. (1) (2) (3) (4) (5) (6) ΔS H,t (0.250) (0.212) (0.328) (0.335) (0.416) (0.567) r (1-τ r,t-1 ) S r,t * 0.049* (0.097) (0.091) (0.143) (0.142) (0.185) (0.164) r (1-τ r,t-2 )S r,t *** 0.201*** 0.527** 0.263** 0.198*** 0.338*** (0.000) (0.000) (0.026) (0.021) (0.003) (0.000) r (1-τ r,t-3 )S r,t *** 0.154*** 0.631*** 0.322*** 0.104*** 0.230*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) r (1-τ r,t-4 )S r,t ** 0.021** 0.114* 0.056* * (0.036) (0.035) (0.055) (0.053) (0.296) (0.095) r (1-τ r,t-5 )S r,t ** 0.034** 0.187** 0.093** * (0.030) (0.029) (0.049) (0.048) (0.122) (0.088) r (1-τ r,t-6 )S r,t * 0.032* 0.179* 0.089* 0.010* (0.066) (0.065) (0.085) (0.083) (0.067) (0.198) Country-ndustry fxed effects Yes Yes Yes Yes Yes Industry-year fxed effects Yes Year fxed effects Yes Yes Yes R No. of Obs F statstcs The dependent varable s Δτ j,t n column (1), (5), (6,), and [ σ j /(σ j -1)]Δτ j,t n column (2), and ε j Δτ j,t n column (3), and ε j [σ j /(σ 1)]Δτ j,t n column (4). ΔS H,t s 1-perod tme dfference n the rato of domestc output to domestc consumpton of. S r,t s the rato of mport from partner r to domestc consumpton of at tme t. Column (1) - (4) nclude country-ndustry fxed effect. Column (5) ncludes ndustry-year fxed effects. Columne (6) ncludes country-ndustry and year fxed effects. p-values, based on standard errors clustered at country-ndustry level, are reported n parentheses. * p<0.10, ** p<0.05, *** p<0.01

24 Table 3. Estmaton results for the effects of poltcal relaton on external tarffs n panel structure. (1) (2) (3) (4) (5) (6) (7) (8) ф r =1 ф UN,r ф Defense,r ф Interacton,r ф r =1 ф UN,r ф Defense,r ф Interacton,r ΔS H,t (0.300) (0.303) (0.317) (0.302) (0.537) (0.541) (0.549) (0.538) r ф r (1-τ r,t-1 ) S r,t *** (0.112) (0.140) (0.000) (0.114) (0.134) (0.114) (0.100) (0.139) r ф r (1-τ r,t-2 )S r,t ** 0.274** 0.736*** 0.436** 0.338*** 0.478*** 0.890*** 0.427*** (0.011) (0.028) (0.000) (0.011) (0.000) (0.000) (0.000) (0.000) r ф r (1-τ r,t-3 )S r,t *** 0.487*** 0.441*** 0.420*** 0.230*** 0.345*** 0.221*** 0.268*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) r ф r (1-τ r,t-4 )S r,t ** 0.076** 0.095*** 0.051* 0.020* (0.037) (0.046) (0.000) (0.067) (0.072) (0.102) (0.464) (0.110) r ф r (1-τ r,t-5 )S r,t ** 0.138** 0.157*** 0.084* 0.028* 0.068* (0.033) (0.043) (0.000) (0.061) (0.066) (0.060) (0.528) (0.117) r ф r (1-τ r,t-6 )S r,t * *** (0.062) (0.134) (0.000) (0.107) (0.166) (0.250) (0.676) (0.259) Country-ndustry fxed effects Yes Yes Yes Yes Yes Yes Yes Yes Year fxed effects Yes Yes Yes Yes F-statstcs Rho R No. of Obs The dependent s ε j [σ j /(σ j -1)]Δτ j,t n column (1) - (4) and Δτ j,t n column (5) - (8). Column (1) and (5) report estmates not weghted by proxy of poltcal relaton. S H,t s 1-perod tme dfference n the rato of domestc output to domestc consumpton of. S r,t-1 s the rato of mport from partner r to domestc consumpton of at tme t-1. ф r s the poltcal relaton proxy. Estmates n column (2) and (6) are weghted by poltcal affnty scores from UN Assembly Votngs. Estmates n column (3) and (7) are weghted by defense treaty dummy. Estmates n column (4) and (8) are weghted by dyad poltcal nteractons. p-values, based on standard errors clustered at country-ndustry level, are reported n parentheses. * p<0.10, ** p<0.05, *** p<0.01

25 Table 4. Estmaton results for the effects of on external tarffs usng nstrument varables. (1) (2) (3) (4) (5) (6) (7) (8) Iv 1 Iv 2 Iv 3 Iv 4 Iv 1 Iv 2 Iv 3 Iv 4 r (1-τ r,t-1 ) S r,t *** 0.769** 0.645** 0.602** 1.034*** 0.465*** 0.410** 0.397** (0.000) (0.020) (0.026) (0.025) (0.000) (0.007) (0.011) (0.013) r (1-τ r,t-2 )S r,t *** 0.288*** 0.258*** 0.237*** 0.479*** 0.474*** 0.462*** 0.457*** (0.004) (0.000) (0.001) (0.001) (0.000) (0.000) (0.000) (0.000) r (1-τ r,t-3 )S r,t *** 0.429*** 0.382*** 0.372*** 0.413*** 0.359*** 0.331*** 0.326*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) r (1-τ r,t-4 )S r,t *** 0.381*** 0.302*** 0.281*** 0.669*** 0.271*** 0.224*** 0.213*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) r (1-τ r,t-5 )S r,t *** 0.485*** 0.396*** 0.393*** 0.162** 0.287*** 0.236*** 0.231*** (0.000) (0.000) (0.000) (0.000) (0.022) (0.000) (0.000) (0.000) r (1-τ r,t-6 )S r,t *** 0.311*** 0.248*** 0.252*** 0.865*** 0.202*** 0.168*** 0.168*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) ΔS H,t (0.333) (0.286) (0.289) (0.289) (0.276) (0.218) (0.217) (0.217) Country-ndustry fxed effects Yes Yes Yes Yes Yes Yes Yes Yes LM statstc Cragg-donald wald Hansen J. statstcs F-statstcs No. of Obs The dependent s ε j [σ j /(σ j -1)]Δτ j,t n column (1) - (4) and Δτ j,t n column (5) - (8). ΔS H,t s 1-perod tme dfference n the rato of domestc output to domestc consumpton of. S r,t-1 s the rato of mport from partner r to domestc consumpton of at tme t-1. Column (1) and (4) are nstrumented wth Iv 1. Column (2) and (6) are nstrumented wth Iv 2. Column (3) and (7) are nstrumented wth Iv 3. Column (4) and (8) are nstrumented wth Iv 4. p-values, based on standard errors clustered at country-ndustry level, are reported n parentheses. LM statstc reports for underdentfcaton test. Wald F. statstc reports for weak dentfcaton test. Hansen J statstc s zero for the case of exact dentfcaton. * p<0.10, ** p<0.05, *** p<0.01

26 Table 5. Estmaton results for the effects of poltcal relaton on external tarffs usng nstrument varables. ф UN,r ф Defense,r ф Interacton,r (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) IV1 IV2 IV3 IV4 IV1 IV2 IV3 IV4 IV1 IV2 IV3 IV4 r ф r (1-τ r,t-1 ) S r,t *** 2.140*** 1.868*** 1.599*** 0.963*** 0.293*** 0.203*** 0.191*** 2.319*** 0.779** 0.631** 0.579** (0.000) (0.000) (0.001) (0.000) (0.000) (0.003) (0.009) (0.005) (0.000) (0.033) (0.046) (0.045) r ф r (1-τ r,t-2 )S r,t *** 0.348*** 0.320*** 0.308*** 1.100*** 1.104*** 1.092*** 1.121*** 1.105*** 0.438*** 0.401*** 0.357*** (0.006) (0.000) (0.000) (0.001) (0.003) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) r ф r (1-τ r,t-3 )S r,t *** 0.531*** 0.485*** 0.483*** 2.797** 0.587*** 0.497*** 0.505*** 0.936*** 0.524*** 0.465*** 0.468*** (0.000) (0.000) (0.000) (0.000) (0.043) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) r ф r (1-τ r,t-4 )S r,t *** 0.723*** 0.618*** 0.531*** 0.990*** 0.362*** 0.237*** 0.222*** 1.134*** 0.398*** 0.308*** 0.287*** (0.005) (0.000) (0.000) (0.000) (0.004) (0.000) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) r ф r (1-τ r,t-5 )S r,t *** 0.868*** 0.751*** 0.726*** 1.292*** 0.443*** 0.291*** 0.292*** 1.477*** 0.500*** 0.399*** 0.392*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) r ф r (1-τ r,t-6 )S r,t *** 0.283*** 0.247*** 0.246*** 0.624*** 0.385*** 0.237*** 0.259*** 1.241*** 0.343*** 0.263*** 0.269*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) ΔS H,t (0.337) (0.291) (0.293) (0.293) (0.298) (0.318) (0.317) (0.318) (0.346) (0.285) (0.288) (0.288) Country-ndustry fxed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes LM statstc Cragg-Donald Wald No. of Obs The dependent varable s ε j [σ j /(σ j -1)]Δτ j,t. ΔS H,t s 1-perod tme dfference n the rato of domestc output to domestc consumpton of. S r,t-1 s the rato of mport from partner r to domestc consumpton of at tme t-1. Estmates n column (1) - (4) are weghted by poltcal affnty scores from UN Assembly Votngs. Estmates n column (5) - (8) are weghted by defense treaty dummy. Estmates n column (9) - (12) are weghted by dyad poltcal nteractons. p-values, based on standard errors clustered at country-ndustry level, are reported n parentheses. LM statstc reports for underdentfcaton test. Wald F. statstc reports for weak dentfcaton test. Hansen J statstc s zero for the case of exact dentfcaton. * p<0.10, ** p<0.05, *** p<0.01