Quality Choice: Effects of Trade, Transportation Cost, and Relative Country Size. 1

Transcription

1 Qualty Choce: Effects of Trade, Transportaton Cost, and Relatve Country Sze. 1 (Prelmnary draft. Please, do not cte) Volodymyr Lugovskyy (Georga Insttute of Technology) Alexandre Skba (The Unversty of Wyomng) January 2008 We ntroduce a generalzed ceberg transportaton cost, whch combnes per-unt and ad-valorem components, nto a monopolstc competton model of trade wth endogenous qualty choce. In equlbrum, qualty ncreases n per-unt component of the ceberg and n frm-level export share. We derve an equlbrum where the export share and thus the qualty of exports decrease n the exporter country sze. Next we show emprcally on the sample of US mports that larger countres tend to export lower prced goods especally n those ndustres where transportaton cost has a larger per-unt component. Ths fndng rases a possblty that ncreasng returns play a larger role n nternatonal trade than prevously thought. Ths s because the customary approach that uses value of trade to dentfy the home market effects mght underestmate the effect of country sze because prce and quantty are effected n opposte drectons. Ths supposton s borne out by the data. Home market effect s present n about a quarter to a thrd more ndustres when volume of exports s measured by quantty rather than value. JEL F1, R1 1 We thank Davd Hummels and the partcpants of Purdue Brown Bag Semnar for helpful comments and suggestons.

2 I. Introducton The queston of nternatonal specalzaton s central to trade theory. Recently we learned that countres specalze wthn as well as across ndustres. Notably, Schott (2004) documents substantal varaton n product unt values across exporters to the US and relates t to per capta ncome of exporters. Hs results hghlght consderable nternatonal specalzaton even wthn the most narrowly defned product categores. Varaton n unt value s not accounted for by tradtonal model of trade wth product dfferentaton and ncreasng returns to scale. In those models specalzaton manfests tself as the home-market effect. Hanson and Xang (2004) show theoretcally and confrm emprcally that the strength of the home market effect promnently depends on the ad valorem sze of transportaton cost. The ad valorem ncdence of transportaton s however dfferent for goods of dfferent unt values. Hummels and Skba (2004) show that transportaton costs are less than one-to-one proportonal to the unt value of shpped goods and therefore lead to changes n relatve demands because hgher prced goods have lower ad valorem equvalent of transportaton costs (the Alchan-Allen effect). As a result countres tend to export hgher prced goods to more remote destnatons. In ths paper we combne nsghts from the prevous lterature to nvestgate whether the nature of the transportaton costs affects specalzaton n the models of trade n dfferentated goods produced under ncreasng returns to scale. We ntroduce a generalzed ceberg transportaton cost nto a monopolstc competton model of trade wth endogenous qualty choce. In dong so, we retan analytcal smplcty of the tradtonal ceberg assumpton but allow for both a per-unt and an ad valorem component of transportaton costs. The resultng model predcts a systematc negatve relaton between country sze and export prces. In equlbrum, qualty ncreases n per-unt component of the ceberg and 1

3 n frm s export share. Ths s because the frms that export more of ther output face a stronger ncentve to upgrade qualty as larger share of ther output ncurs transportaton cost. In the corner equlbrum wth one-way trade n the dfferentated sector, both the export share and qualty decrease n the exporter country sze. Intutvely, country sze matters relatvely less when transportaton costs are lower. So, f hgher prced goods face lower ad valorem transportaton cost, the locaton of the hgher transportaton cost ndustry s more senstve to the transportaton cost. Our theory can be anecdotally ntuted by notorous examples of small countres specalzng n hgh qualty goods such as Swss watches, Belgan chocolate, or Japanese photo lenses. In the emprcal exercse we frst estmate ndustry level transportaton cost functons to determne the degree to whch transportaton costs co-vary wth the product unt value. Next usng a sample of US mports we show emprcally that larger countres tend to export lower prced goods. The negatve effect of market sze on the prce of exports s stronger n those ndustres where transportaton costs have a larger per-unt component. Ths fndng rases a possblty that ncreasng returns play a larger role n explanng nternatonal trade than prevously thought. Ths s because the customary approach that uses value of trade to dentfy the home market effects mght underestmate the effect of country sze on exports because prce and quantty are effected n opposte drectons. Ths supposton s borne out by the data. Accordng to our estmates home market effect s present n about a quarter to a thrd more ndustres when volume of exports s measured by quantty rather than value. 2

4 II. Theoretcal Framework 1. Model We want to explore how the relatve country-exporter sze and the nature of transportaton cost affect the qualty choce of exporters. For ths purpose, we extend the standard model of trade under monopolstc competton (Helpman and Krugman, 1985) to the case of multple dfferentated-product ndustres wth endogenous qualty choce and augmented transportaton cost. The major novelty s a generalzed ceberg transportaton cost whch ncludes both ad-valorem and per-unt components Preferences Lh and The world conssts of 2 countres, Home and Foregn, ndexed by h and f, wth populaton L f, respectvely. The preferences are symmetrc n Home and Foregn. For brevty, we wll set up the model only from Home s perspectve. Preferences of Home s representatve consumer are defned over a numerare good q 0 and S dfferentated sectors ndexed by s: 0 (1) U q 0 C s1 S s h h hs S s 1, s=0 where Chs s a composte of dfferentated varetes n sector s: 1 1 hs k hsk hsk 1, (2) C q where qhsk -- s Home s ndvdual consumpton of varety k n sector s; hsk -- s a qualty multpler of varety k n sector s; 1.2. Producton 3

5 Labor s the only factor of producton and s suppled nelastcally. Each consumer s endowed wth one unt of labor. The numerare sector s characterzed by perfect competton and constant returns to scale. One unt of labor can produce 1/w unts of the numerare n Home and 1 unt n Foregn. The numerare s traded at zero cost. We assume that the numerare sector s large enough for both countres to have strctly postve output of the numerare. The ntroducton of the numerare n the model smplfes the balance of trade calculaton and tes the wage to productvty n the numerare sector. Dfferentated varetes are produced by monopolstcally compettve frms, wth country -sector s specfc technologes. The producton requres a fxed number of workers s, and margnal labor requrement c s, whch s a functon of chosen qualty: (3) c sk 1 exp sk As Zs A, Z 0, s s where As and Z s are productvty parameters Augmented functonal form of the transportaton cost It s dffcult to fnd a realstc functonal form of the transportaton cost. If we focus on the vessel mode of transportaton, nternatonal shppng s a sophstcated technology whch ncludes both fxed and varable cost components. The fxed cost component s a functon of port nfrastructure, where the varable cost depends on dstance, volume of trade, and goodspecfc characterstcs (Hummels, Lugovskyy and Skba; 2008). Our goal s to suggest a functonal form whch would capture both ad-valorem and perunt components of transportaton cost, and, at the same tme, would preserve the analytcal 2 Ths type of unt labor functon was frst ntroduced by Flamm and Helpman (1987). A smlar functon n conjuncton wth CES preferences s used by Hummels and Klenow (2005). 4

6 smplcty of the ceberg. For ths purpose, we allow a part of transportaton cost to be ncurred n unts of labor. In partcular, we assume that to transport one unt of the fnal good z, t takes z 1 unts of the fnal good z, and f z unts of labor. Shppng ndustry s characterzed by perfect competton, zero profts, and prcng ts servces at margnal cost of shppng. The shppers are pad by producers wth the fnal good at the market prce. The labor used n shppng ndustry s rewarded at the market wage, snce labor s perfectly moble across all sectors. Under the above assumptons, the generalzed ceberg cost functon of delverng good k produced by Home s ndustry s to Foregn s: (4) fhsk f fh 1 fh. c hsk Note, that the magntude of ceberg s now a functon of the margnal cost of producton. Consequently t s good k - ndustry s specfc, and gven the functonal form of margnal cost (3) s endogenous on the qualty choce of the producer of good k. 1.4 Qualty n Autarky Equlbrum parameter Z: In autarky, the proft-maxmzng qualty level s equal to the correspondng productvty (5) *. 3 s Z s 1.5. Qualty n Trade Equlbrum Consder the proft functon of a frm producng varety k n Home s sector s: 3 Ths a partal case of equlbrum wth trade n whch the export share s equal to zero. The correspondng dervatons are gven n the next subsecton. 5

7 Q ES, (6) hsk hsk hsk phhskqhsk 1 EShsk p fhsk hs chskwhqhsk fhsk where p hhsk and p fhsk are the prce charged n Home and n Foregn, respectvely; Q hsk s the frm s total output; ES hsk s the share of exported quantty n the frm s output; fhsk s the generalzed ceberg transportaton cost defned by (4); hs and c hsk are the fxed and margnal unt labor requrements, respectvely; w h s the wage n Home. By solvng the proft maxmzng problem we can fnd the optmal qualty level: (7) Z 1ES hsk hs fhsk f fh c fhsk hsk 1, where f fh c fhsk hsk s the share of the per-unt component n the total ceberg cost. Note that f the share of exports s equal to zero, as n autarky, the result wll concde wth (5). The next two observatons are expressed as lemmas. Lemma 1: If and only f transportaton contans a per unt part, exporters produce hgher qualty goods than non-exporters. 4 Proof: Follows drectly from (7). Intutvely, as frms begn targetng export markets, they pay attenton to the magntude of trade cost. Endogenous trade cost generates an extra ncentve to nvest n hgher qualty: n addton to a hgher demand, hgher qualty now also decreases the ceberg transportaton cost. As a result producers choose hgher qualty compared to autarky level. 4 Note that ths predcton concdes wth Baldwn and Harrgan (2008) predcton. 6

8 Lemma 2: If and only f transportaton contans a per unt part, the qualty of exported goods s ncreasng n per unt part of transportaton cost and n the share of exports. Proof: Follows drectly from (7). From (7), the magntude of the (negatve) effect of qualty on the total ceberg s ncreasng n the share of the per-unt component n the total transportaton cost. Consequently, hgher share of per-unt transportaton cost ncreases the ncentve to nvest n hgher qualty. Also, from (6), export share ncreases the weght of transportaton cost n overall proft calculaton. Accordngly the more the producers are nvolved n exportng, the more they are motvated to nvest n hgher qualty Frm-Level Export Share We have establshed that the export share matters for qualty choce. The export share tself, though, s not an exogenous varable, and our next step s to dentfy ts determnants. To start wth, we have to dstngush between two scenaros possble for a gven sector s: ) nteror equlbrum characterzed by two-way trade; ) corner equlbrum n whch only one country produces and exports dfferentated varetes n sector s. As t wll be seen later, ths dstncton s crucal for determnng the frm-level export share both n terms of quanttatve and, what s more mportant, n terms of qualtatve results. In partcular, the relatve sze of the economy has no mpact on the export share n nteror equlbrum, but s the major determnant of export share n the case of corner equlbrum. 7

9 Frm Level Export Share n the Two-Way Trade Equlbrum As n Helpman and Krugman (1985) model, under two-way trade equlbrum, the export share of ndvdual frm s determned by the technologes avalable to each country, and the magntude of the trade costs, and s ndependent of the country szes Frm Level Export Share n the One-Way Trade Equlbrum Assume that the parameters of the model are such that Home s the sole producer and exporter of dfferentated varetes n sector s. Gven the Cobb-Douglas upper case utlty functon and the fact that ncome spent on the dfferentated varetes n sector s s equal to the value of these varetes at market prces, we start wth the followng equaltes: Q SE L N p 1 hs hs s F hs hfs hfs L w N Q SE p, s h h hfs hs hs hhs whch can be transformed nto the followng expressons after pluggng the equlbrum values of prce and quantty: L Ns w F L N 1 s F H, F from whch we derve the equlbrum value of export share: wlh Yh (8) SEhs 1 1 wl L Y Y h f h f. Thus n the corner equlbrum, the export share s decreasng n the relatve sze of the exporter. By pluggng ths result nto (7) we get (9) Y fhs Ahs exphs Zhs h Z hs 1 1 1, Yh Y f f fhs hs whch allows us to formulate and prove Lemma 3. 5 See equaton on page 207 for a correspondng formula. 8

10 Lemma 3: In the corner on-way trade equlbrum, the qualty level s decreasng n the GDP of the exporter. Ths effect s ncreasng n the per-unt share of the transportaton cost. Proof: Follows drectly from (9) Home Market Effect Revsted Startng wth Krugman (1980) and Helpman and Krugman (1985) the lterature on Home Market effect focuses on the case of nteror equlbrum n whch, ceters parbus, larger country exports more varetes than smaller country. We consder a dfferent angle of ths ssue. Let us combne all dfferentated sectors n one ndustry and assume a perfectly asymmetrc dstrbuton of productvtes between Home and Foregn. (as, e.g., n Fsher, Dornbush, and Samuelson 1977). The condtons defnng the ndustres n whch ndustres we should observe corner equlbrum are smlar to the condtons derved by Helpmand and Krugman (1985). It s possble to show that a larger country wll have more sectors n whch t s a sole exporter than a smaller country. Thus, at ndustry level, the home market effect can stem not only form the fact that larger country produces more varetes n sectors wth nteror equlbrum, but also from the fact that larger country s a sole exporter n more sectors than a small country s. 6 If the Home Market Effect stems even partally form the second channel, the magntude of the Home Market Effect should be greater when we compare the quanttes exported than when we compare the values exported, snce smaller country wll produce hgher qualty goods and charge hgher prces for ts exports. 6 The proof of ths statement s avalable upon request form the authors. 9

11 III. Emprcs 1. Data Our data cover the US mports from all exporters worldwde, measured at 10 dgt level of the Harmonzed Classfcaton System ( categores) n Denote a 10-dgt commodty category by k. For commodty k mported form country the data nclude the f.o.b. value, V k, quantty, Q k, and shppng charge, S k. We compute the unt value of commodty k from country as p V / Q, and the correspondng freght rate as f S / Q k k k k k k. We refer to HS 10 category as product, or good. For GDP and GDP per capta data we used the World Development Indcators data Estmatng ndustry transportaton cost functons Theory ndcates that relatve delvered prce and hence the qualty of exports crucally depend on the specfc part of transportaton cost. In order to test whether transportaton cost s ceberg and to determne the degree to whch transportaton cost devates from ceberg, we estmate a transportaton cost functon that allows for flexble relaton between prce and transportaton cost. For every ndustry K the followng transportaton cost functon s estmated usng a subsample of US mports restrcted to sngle shpments. ln f ln p ln D s 0K 1K s 2K s where f s per unt freght charge to delver a unt of shpment s from country p s unt value of good n shpment s mported from country K HS 2 dgt ndustry 7 For trade data, we used the US Census Imports of Merchandse. 8 Data source: World Bank. 10

12 D dstance to Ths form of transportaton cost functon s commonly used n trade lterature, see for example Hummels (1999), Hummels, Skba (2004). Our prmary nterest les n the prce elastcty of transportaton cost, 1K. If 1K 0 transportaton cost does not depend on prce. Ths s the case of per unt transportaton cost. On the other extreme, f 1K 1 transportaton cost s proportonal to the value of the shpped good. Summary of the estmaton results can be seen n Fgure 1. Tradtonal ceberg transportaton cost s rejected for all 97 HS 2-dgt ndustres. There s also a substantal degree of varaton across ndustres n the degree of devaton from the standard ceberg. 3. Effect of sze on the prce of exports Next we nvestgate the relaton between exporter s sze and the prce of exports. Snce transportaton s not ceberg for all ndustres, we expect sze to lower the prce of exports. Ths correlaton alone would be consstent wth our theory but mght also arse through some other channel. A more persuasve case n favor of our explanaton could be made f the negatve effect of market sze on the prce of exports were stronger for ndustres where transportaton cost s less lke ceberg. In order to formally test the varyng strength of the sze-prce connecton we use the followng emprcal specfcaton: ln P lny T lny T ln C D K K 3 K 4 5 K where K P measure of exporter s prce n ndustry K (average prce or Fsher s deal prce ndex) Y measure of exporter s market sze (nomnal GDP or market potental) 11

13 C vector of specfc cost shfters (GDP per capta, human captal, physcal captal, labor) D dstance to T K vector of varables that reflect dstortons to relatve prces ntroduced by transportaton of goods from ndustry K The prce of country s exports can be smply measured as a trade weghted average. Such approach would however blend dfferences n prce levels wth dfferences n composton of exports. Therefore, a prce ndex could offer dstnct advantages. We modfy Hummels and Klenow s (2002) use of Fsher prce ndex to construct a separate ndex for exporter n ndustry K. P K pjkqjk pjkqwjk kxjk kxjk pwjkqjk pwjkqwjk kxjk kxjk where j US X jk set of categores k mported from to the US p Wjk average prce of US mports n category k q Wjk total quantty of US mports n category k We use nomnal GDP and market potental as two alternatve measures of the market sze. Market potental augments country s GDP to nclude dstance weghted GDP of ts tradng partners. Followng the Fujta, Krugman, and Venables (1999), and Hanson and Xang (2004), we calculate the market potental as: MP j GDP j D j GDP A 12

14 where GDP nomnal GDP model. 9 D j dstance from between and j A land surface area of country We set equal 0.92 usng Hummels s estmate of the dstance coeffcent n the gravty The results of estmaton are presented n tables 1 and 2. Table 1 reports estmated coeffcents for pooled regresson, and Table 2 reports estmaton results for smlar specfcatons wth ndustry fxed effects. Industry fxed effects control for unobservable ndustry specfc characterstcs but also remove some of the useful varaton n the transportaton cost parameters. The effect of sze on prce s unversally negatve when statstcally sgnfcant. The effect of sze on the Fsher s deal prce ndex s consstently weaker than on the average prce ndcatng that the change n composton of exports plays a role n the lowerng of the exports prce wth exporter s sze. Not only larger exporters tend to export lower prced products they also tend to export more of the lower prced exports. Consstently wth the theoretcal predctons ndustres whose transportaton costs are closer to ceberg (farther from 0 and closer to 1) exhbt a weaker negatve relaton between market sze and prce of exports. Ths s evdenced by the postve coeffcent on the ( ˆ 1 K sze) nteracton term. 4. Implcatons for dentfcaton of the home market effects It s customary n the lterature on the home market effect, for example Feenstra, Markusen, and Rose (1998), Davs and Wensten (1999), Hanson and Xang (2004), to rely on value to dentfy the relaton between sze of the domestc market and volume of exports. Value 9 Settng equal 0.92 s also consstent wth Hanson and Xang (2004) 13

15 of trade s more commonly avalable and usually more precsely measured than quantty. In addton, standard model of trade n dfferentated products based on Krugman s framework do not allow for prce varaton. So prce and quantty can be used nterchangeably. However f the prce reducng effect descrbed n ths paper s suffcently pervasve, the effect of market sze on the value of exports s lkely to be smaller than the effect on quantty. Usng values of trade could lead to underestmatng of the strength and extent of the home market effect. In order to gauge whether the prce channel plays an mportant role we wll estmate the effect of home market sze separately on prce and quantty for all HS 2-dgt ndustres across exporters to the US n ln 3 4 PQ Y C D t K K K K K K K K Q Y ln C D t K 0K 1K 2K 3K 4K K K where Y measure of exporter s market sze (nomnal GDP or market potental) C vector of specfc cost shfters D dstance to t K trade weghted average tarff rate on ndustry K exports from country The home market effect s defned by more than one-to-one relaton between exporter s sze and volume of exports. For every ndustry we perform a one sded t test of the null hypothess that the coeffcent on sze s smaller or equal to unty: H0: 1K 1 and H0 : 1K 1. The results of ths exercse are summarzed n Table 3. Consstently wth prevous fndngs home market effect s more pronounced n specfcatons that use market potental rather than GDP to measure the sze of domestc market. The home market effect s clearly more pronounced when quantty rather than prce s used to measure the volume of trade. 14

16 IV. Conclusons and Dscusson We fnd that when a good s exported by many countres the average prce of exports s lower for larger countres, although not unversally. We show how ths lnk between the sze and average prce can arse from the nteracton between the Alchan-Allen effect and the home market effect. The effect of sze on prce s stronger when transportaton cost covares less wth the unt value of the shpped good. The emprcal results are consstent wth our theoretcal predctons. Not only larger countres tend to shp lower prced goods but also ths negatve relaton s stronger when transportaton cost s less ad valorem. There can be potentally other reasons of why large countres have lower export prces. In partcular, large countres mght choose a hgh fxed cost - low varable cost technology due to larger sze of the domestc market. Alternatvely, mporters mght consder a country of orgn as addtonal factor of dfferentaton whch mght force producers form large countres to charge lower export markups due to hgher competton wth smlar varetes. Unfortunately wth the data n hands, we are not able to test what s exactly the mechansm whch lowers the export prces for large countres. In terms of theoretcal contrbuton, our message s that even f we assume endogenous technologcal choce and dfferences n markups, larger countres have ncentves to specalze n lower qualty. In ths paper we do not model a relaton between elastcty of substtuton and qualty. The elastctes of substtuton for hgh and low qualtes are dentcal. Ths assumpton of convenence s not always nnocuous because t s the nteracton between the ceberg transportaton cost and elastcty that determne exstence and relatve strength of the home market effect for hgh and low qualty varetes. 15

17 References Robert C. Feenstra & James A. Markusen & Andrew K. Rose, "Undertstandng the Home Market Effect and the Gravty Equaton: The Role of Dfferentatng Goods," NBER Workng Papers 6804, Natonal Bureau of Economc Research Fujta, Masahsa; Krugman, Paul and Venables, Anthony J. The spatal economy: Ctes, regons, and nternatonal trade. Cambrdge, MA: MIT Press, Hanson and Xang (2004): The Home-Market Effect and Blateral Trade Patterns, The Amercan Economc Revew, Vol. 94, No. 4. (Sep., 2004), pp Hummels, Davd (2001) Toward a Geography of Trade Costs, Purdue Unversty, mmeo Hummels, Davd, Peter J. Klenow, "The Varety and Qualty of a Naton's Trade," NBER Workng Papers 8712, Natonal Bureau of Economc Research, Hummels, Davd and Skba, Alexandre (2004): "Shppng the Good Apples Out? An Emprcal Confrmaton of the Alchan-Allen Conjecture," Journal of Poltcal Economy, Unversty of Chcago Press, vol. 112(6), pages , December. Krugman, Paul, and Elhanan Helpman (1985). Market Structure and Foregn Trade. The MIT Press. Hummels, Davd, Lugovskyy, Volodymyr, and Skba, Alexandre (2008), Journal of Development Economcs Schott, Peter (2004): Across-Product versus Wthn-Product Specalzaton n Internatonal Trade. Quarterly Journal of Economcs 119(2): (May 2004). 16

18 Table 1. Effect of market sze on export prce, pooled specfcatons. Dep. varable Average prce Fsher's deal prce ndex Average prce Fsher's deal prce ndex Measure of sze GDP GDP Market potental Market potental (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Sze [13.98]** [6.90]** [3.41]** [11.76]** [5.71]** [3.40]** [8.47]** [5.03]** [1.09] [5.38]** [4.22]** [1.06] Transportaton cost parameters ˆ 0 K sze [0.31] [1.32] [3.42]** [2.82]** [0.17] [1.66] [2.38]* [1.90] ˆ 1 K sze [5.38]** [2.23]* [5.38]** [2.96]** [4.27]** [2.05]* [4.52]** [2.35]* ˆ 0 K [0.88] [0.39] [3.81]** [2.89]** [0.96] [0.68] [2.61]** [1.98]* ˆ 1 K [6.67]** [2.95]** [5.88]** [3.15]** [5.33]** [2.80]** [4.93]** [2.54]* Dstance [5.58]** [5.58]** [5.00]** [3.79]** [3.70]** [2.69]** [3.75]** [3.89]** [5.68]** [1.91] [1.94] [3.33]** Cost controls GDP p.c [8.48]** [8.80]** [5.83]** [7.74]** [7.63]** [5.25]** [4.40]** [5.26]** [6.48]** [4.32]** [4.45]** [5.17]** Human cap [3.28]** [4.43]** [3.38]** [4.48]** Captal [3.25]** [1.66] [4.35]** [2.53]* Labor [4.12]** [3.48]** [2.38]* [1.15] N obs R sq Notes: 1) Absolute value of t statstcs n brackets. 2) * sgnfcant at 5%; ** sgnfcant at 1%. 3) All varables except ˆ 0 K and ˆ 1K are n logarthms. 4) ˆ 0 K and ˆ 1K are obtan by estmatng transportaton cost functon ln fk 0K 1K ln pjk 2K ln D jk for every HS 2 dgt category. 17

19 Table 2. Effect of market sze on export prce, specfcatons wth ndustry dummes. Dep. varable Average prce Fsher's deal prce ndex Average prce Fsher's deal prce ndex Measure of sze GDP GDP Market potental Market potental (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Sze [12.85]** [3.75]** [2.43]* [1.41] [7.13]** [5.00]** [8.23]** [2.63]** [0.67] [4.27]** [5.90]** [3.90]** Transportaton cost parameters: ˆ 0 K sze [0.91] [1.13] [0.86] [0.69] [0.74] [1.88] [0.58] [0.00] ˆ 1 K sze [1.66] [0.10] [8.16]** [6.11]** [1.38] [0.06] [7.11]** [4.90]** Dstance [6.05]** [6.04]** [6.29]** [2.55]* [2.50]* [2.86]** [4.33]** [4.34]** [6.80]** [1.72] [1.79] [2.96]** Cost controls: GDP p.c [12.84]** [12.82]** [7.62]** [9.52]** [9.43]** [6.01]** [9.76]** [9.78]** [9.66]** [10.84]** [10.96]** [7.58]** Human cap [3.84]** [4.56]** [3.52]** [4.37]** Captal [4.80]** [1.11] [6.05]** [1.48] Labor [5.17]** [3.45]** [3.82]** [1.71] N obs R sq Notes: 1) Absolute value of t statstcs n brackets. 2) * sgnfcant at 5%; ** sgnfcant at 1%. 3) All specfcatons nclude HS 2 dgt ndustry dummes. 4) All varables except ˆ 0 K and ˆ 1K are n logarthms. 5) ˆ 0 K and ˆ 1K are obtan by estmatng transportaton cost functon ln fk 0K 1K ln pjk 2K ln D jk for every HS 2 dgt category. 18

20 Table 3. Number of HS 2-dgt ndustres exhbtng the home market effect. Measure of market sze Market potental GDP Dependent varable Value Quantty % Value Quantty % Sgnfcance level: Notes: 1) there are 97 HS 2-dgt ndustres. 2) Sgnfcance level refers to the p-value of the one sded t test of null that the coeffcent on exporter s sze s less or equal to unty n the followng two equatons : PQ Y ln C D t K K K K K K K K Q Y ln C D t K 0K 1K 2K 3K 4K K K 19

21 Prce elastcty of transportaton cost, ˆ 1 K ˆ 1 ˆ 0 - per unt transportaton cost) ad valorem, 1 ( 1K K Fgure 1. Dstrbuton of ndustres by the prce elastcty of transportaton cost 20