Bilateral Trade Flows and Nontraded Goods

Transcription

1 Blateral Trade Flos and Nontraded Goods Yh-mng Ln Department of Appled Economcs, Natonal Chay Unversty, Taan, R.O.C. Emal: Abstract Ths paper develops a monopolstc competton model th nontraded goods hch provdes an explanaton for hy the real volume of trade s much loer than predcted by Helpman and Krugman's (1985) model. Furthermore, t explans the phenomenon that the volume of trade among hgh-ncome countres s relatvely larger than the volume of trade beteen hgh-ncome and lo-ncome countres. We also derve a testable gravty equaton from ths model. A sample of 1995 ncludng 118 countres s examned. Our results sho that evdence from the data s consstent th the predcton of ths model; further, and the goodness-of-ft ncreases as nontraded goods are consdered. JEL Classfcaton: F10, F12 Keyords: Blateral Trade Flos; Gravty Equaton; Monopolstc Competton Assstant Professor, Department of Appled Economcs, Natonal Chay Unversty. 580 Snmn Road., Chay, 600, Taan. Emal: yxl173@mal.ncyu.edu.t. Resdual errors are mne alone.

2 1. Introducton In the last fe decades, the most mportant development n the theory of nternatonal trade s the monopolstc competton model. Helpman and Krugman (1985) proposed a model n hch monopolstcally compettve frms produce dfferentated goods usng an ncreasng return to scale technology (IRS) and all ndvduals have dentcal homothetc preferences and a love for varety. Ther model provdes an explanaton for the phenomenon of large volumes of trade among smlar countres th a factor-proportons ve of ntersectoral trade flos, hch could not be explaned by the tradtonal Heckscher-Ohln (HO) theorem. In Helpman and Krugman's model, t s assumed that the economy has free trade, balanced trade, no transport cost, all tradeable goods, and dentcal producton technology across countres. The monopolstc competton model yelds the follong equaton to predct the volume of blateral trade, M YY s Y, (1) Y here M s the mports of country from country, Y ( Y ) s the gross domestc product (GDP) of country (), Y s the total orld ncome, and Y s Y s the share of country n total orld ncome. Equaton (1) means that the blateral trade flos are postvely related to the product of countres' GDPs, hch s the smplest form of gravty equaton. Hoever, the volume of trade n the real orld s much less than the amount 1

3 predcted by equaton (1). For example, the volume of trade n the orld s about 5,214 bllon U.S. dollars, hch s much loer than the predcted number, 25,033 bllon US dollars n Obvously, the blateral trade flos are overestmated by Krugman and Helpman's model. Many possble factors could reduce the blateral trade flos. For example, hgh transport cost could decrease the volume of trade. Most countres do not have completely free trade. Further, not all goods n the real orld are tradeable such as servces, educaton, and housng. All of these facts ll reduce the volume of trade. Hoever, most of the semnal studes do pay attenton to the nfluence of transport cost. The gravty equaton, n general, shos that the volume of blateral trade s not only postvely related to both ncomes, but also negatvely related to the dstance beteen the to trade partners, n hch the dstance s used as a proxy varable for transport costs. The study of the gravty equaton s probably the most successful emprcal ork n nternatonal trade. The model as frst ntroduced by Tnbergen (1962) and Anderson (1979) as the frst to derve the gravty equaton theoretcally. Anderson (1979) assumed that preferences are Cobb-Douglas (or CES) and the goods are dfferentated by countres of orgn, hch s called the Armngton assumpton. Bergstrand (1985) used CES preferences over Armngton-dfferentated goods to develop a general 2

4 equlbrum model of orld trade, thus yeldng a reduced-form gravty equaton for blateral trade nvolvng prce ndces. In ths paper, the assumpton that all goods are tradable s relaxed. I ncorporate nontraded goods n a monopolstc competton model. Ths model tres to provde an explanaton of hy the real volume of trade s much loer than predcted. The ntuton s qute smple. It s because the countres do not have so many goods that are tradable n the real orld. An estmable gravty equaton can be derved from ths model. Moreover, snce ths model ncorporates nonhomothetc preferences, t also can explan the phenomenon that the volumes of trade among the ndustralzed countres are relatve large to the volumes of trade beteen developed and less-developed countres, hch as proposed by Lnder (1961) and Markusen (1986). Markusen (1986) proposed a nonhomothetc model to explan the dfference beteen the volume of W-E trade and the volume of N-S trade. Unfortunately, hs model does not offer a testable model to predct the volume of trade from hs model. Hoever, n the extenson of hs paper, Hunter and Markusen (1988) proposed a nonhomothetc model and estmated a lnear expendture system (LES) to sho that the demand s nonhomothetc. The remander of ths paper s organzed as follos: n Secton I, e set up the model and derve the gravty equaton th nontraded goods. Secton II descrbes the data sets used n ths study. The emprcal results are gven n Secton III. They strongly 3

5 support the predcton of ths model n a sample of 118 countres n Secton IV concludes. 2. The Model In ths secton, the model s set up. Consder an open economy hch has balanced trade, no transport costs, and th dentcal producton technologes. There are m countres and to types of goods, tradeable goods x goods x k k and nontraded goods z. The tradeable are dfferentated manufactured goods hch are produced th producton technologes th ncreasng returns to scale. The nontraded good s a homogeneous commodty and s produced th a constant returns to scale (CRS) technology. It s assumed that all ndvduals consume tradeable goods as ell as nontraded goods. Consumers have the follong dentcal nonhomothetc preferences, hch are gven by U 1 x k k u, z 0 1, (2) here u s a strctly concave functon. All ndvduals maxmze ther utlty subect to ther ncome. Snce the subutlty functon for dfferentated goods s homogeneous of degree one, e can use a to-stage budgetng procedure to solve ths utlty maxmzaton problem. The consumer's problem can be rertten as Max, z U u (3) 4

6 s.t P p z z I. here I s ndvdual ncome, x 1 k k s the quantty ndex for dfferentated goods, P 1 1 k pk s the prce ndex for, pz and pk for good z and x k, respectvely. If e consder as the numerare, the utlty can be consdered as a quas-lnear utlty functon. Accordng to the property of quas-lnear utlty functon, the consumpton of z s constant hch s determned by u p, (4) z z P for all consumers f the ncome s bg enough. There s no ncome effect for z. Increasng ndvdual ncome does not change the quantty of demand for good z at all, and all the extra ncome goes entrely to the consumpton of dfferentated goods. Let z z satsfy equaton (4) hch means that z s the demand quantty of notraded good for every consumer n every country and thus s ndependent of ndvdual ncome and the prces of the dfferentated commodtes. Furthermore, t s also assumed that the ndvdual's ncome n every country s bgger than z. Due to ths specal property of the nonhomothetc preference, the producton of nontraded good for country s Z n z and s produced frst n country, here n s the populaton of country. In addton, the dfferentated manufactured goods are produced and freely traded. Just lke the mperfect competton model proposed by Helpman and Krugman (1985) 5

7 and Helpman (1987), a number of frms produce one dfferentated commodty n a monopolstcally compettve market. Also, all frms are equpped th dentcal IRS technology, and free entry leads to zero proft n equlbrum. In ths model, the volume of trade s dfferent from the result of Helpman and Krugman (1985) but smlar. Snce ths property of nonhomothetc preference s love for varety for dfferentated goods, each country ll demand all foregn varetes accordng to the country's share of orld value of dfferentated goods. Therefore, the value of dfferentated goods that country mports from, denoted M, s X X M, (5) X here X s the value of a dfferentated good produced n country, and X X s the orld output of dfferentated goods. Let Y X Z denote the GDP of country. Rearrangng equaton (5), M can be rertten as M Y Z Y Z Y n z Y n z. Y Z X (6) Takng natural logarthms of both sdes of equaton (6), t follos that here ln ln M ln X ln Y n z Y n z X (7) X 2 YY Yn z Y n z n n z ln ln, s a constant. In order to derve an estmable gravty equaton, t s necessary to lnearze the last 6

8 term of equaton (7). Applyng the frst order Taylor seres approxmaton at z 0, t yelds the follong estmable gravty equaton ln n Y n Y M lnx lny Y z, Y Y (8) here y c ln 1 Y 1 Y ln Y Y Z, s per capta ncome of country, s the dsturbance term. Comparng the ne gravty equaton to the conventonal gravty equaton, there s a ne tem, 1 y 1 y z, n ths model. It shos that the blateral trade flo s not only related to the GDP of both countres, but also related to the demand for nontraded goods and the per capta ncomes n both countres. Accordng to the above equaton, blateral trade flo decreases f the consumpton of nontraded goods ncreases or the per capta ncome decreases. It explans that the volume of trade beteen to rch countres s larger than that beteen to loer-ncome countres. 3. Data There are three databases used n ths study. The orld blateral trade flos data s orgnally obtaned from the CD-ROM World Trade Flos, , th Producton and Tarff Data. The data used to test ths model s from the year of There are 182 countres or regons n ths database. In Feenstra (2000), t ndcated that the man source for blateral trade data s the Unted Natonal Statstcal Offce. Another 7

9 data set used n the study s the GDP and per capta ncome data hch s donloaded from Harvard Unversty, CID (Center for Internatonal Development). The GDP and per capta ncome are PPP-adusted. In Gallup and Sachs (1999), t ndcated that most of the GDP and per capta ncome data are from the World Bank. For countres hch are mssng n the World Bank, the data s obtaned from CIA (1996, 1997). In order to compare ths nontraded goods model and the conventonal gravty model, a dstance data set s also employed. The dstance data set s donloaded from Purdue Unversty. Ths data set contans 137 countres and t provdes the dstance beteen captal ctes n klometers. Combnng the above three data sets, there are 118 countres employed n ths paper. 4. Emprcal Results In ths secton, the gravty equatons derved from the nonhomothetc model are estmated by usng the above data set. Based on equaton (8), the gravty equaton can be specfed as the follong to forms 1 1 ln ln M Y Y z, ln 2 1 y y (9) here, and z are the coeffcents to be estmated. Theoretcally, ˆ s expected to be negatve, z should be postve, and ˆ 1 and ˆ 2 should be around 1. In order to demonstrate the advantage of ths ne model, the conventonal gravty equaton s also to be estmated, hch s 8

10 M lny lny. ln 2 1 (10) Ordnary Least Square (OLS) estmaton s employed to estmate the above regressons. The estmaton results are shon n Table 1. There are 13,806 observatons n ths sample. In Table 1, t shos all of the estmated coeffcents n equaton (9) and (10) are sgnfcant at 1% sgnfcant level. The most mportant s that all of the sgns of the estmated coeffcents concde th our expectaton. In partcular, ẑ s around $1,053 n ths model, hch means that all ndvduals consume at least about $1,053 goods produced by ther country every year. Comparng to Equaton (10), t can be found that the estmated coeffcents of GDPs n Equaton (9) are sgnfcant closer to unty than n Equaton (10) hch means that ths ne model has a theoretcal advantage. Ths nontraded goods gravty equaton s more consstent th the data than the conventonal gravty model. The 2 R of Equaton (9) s about 0.60, hch s bgger than the 2 R n Equaton (10). That shos that the goodness-of-ft of ths ne model s better than the conventonal model. Next, I examne the performance of ths model hen a dstance term exsts and compare the tradtonal gravty equaton th a dstance term. The results are also gven n Table 1 here Equaton (9' ) and Equaton (10' ) are Equaton (9) and Equaton (10) th dstance term, respectvely. The s the coeffcent of the dstance term n each model. Just lke the results above, the sgns of estmated coeffcents are as expected and 9

11 sgnfcant. We also can fnd that the estmated coeffcents of GDPs are also much closer to unty than the result of Equaton (10' ). Table 1: Estmaton Results Coeffcent Equaton (9) Equaton (10) Equaton (9 ) Equaton (10 ) 1 2 z (0.317) 1.51 (0.017) 1.14 (0.017) (47.272) (0.235) 1.68 (0.015) 1.32 (0.015) (0.416) 1.52 (0.016) 1.15 (0.016) (44.606) (0.034) (0.361) 1.71 (0.014) 1.34 (0.014) (0.034) 2 R Notes: Standard errors are gven n parentheses. 5. Conclusons In ths paper, e tred to provde an alternatve explanaton of hy the real volume of trade s much less than the volume predcted by Helpman and Krugman's model. We consder nontraded goods as an mportant mpact on blateral trade flos. Furthermore, ths model also explans hy the volume of trade beteen N-S s much less than the volume among developed countres. We developed a model th nontraded goods and nonhomothetc preferences and derved an estmable gravty equaton. The model tres to make a connecton to brdge the gap beteen theory and emprcal ork on the role of nontraded goods. Ths model 10

12 fnds that blateral trade flos are not only related to the GDP of to countres, but also related to the per capta ncome of the to countres. The ndvdual consumpton level of nontraded goods can be estmated, hch s also related to blateral trade flos. In the emprcal ork of ths study, e employed a sample of 1995 data th 118 countres to test ths model. The emprcal results are consstent th the expectatons from ths model. References Anderson, J., 1979, A Theoretcal Foundaton for the Gravty Equaton, Amercan Economc Reve, 69, Bergstrand, J., 1985, The Gravty Equaton n Internatonal Trade: Some Mcroeconomc Foundatons and Emprcal Evdence, Reve of Economcs and Statstcs, 67, CIA, 1996, The World Factbook, Washngton, D.C. Central Intellgence Agency. CIA, 1997, The World Factbook, html. Feenstra, R., 2000, World Trade Flos, , th Producton and Tarff Data, Workng Paper, Unversty of Calforna, Davs. Gallup, J. and J. Sachs, 1999, Geography and Economc Development, Workng Paper, 11

13 Harvard Unversty. Helpman, E., 1987, Imperfect Competton and Internatonal Trade: Evdence from Fourteen Industral Countres, Journal of the Japanese and Internatonal Economes, 1, Helpman, E. and P. Krugman, 1985, Increasng Returns, Imperfect Competton, and Foregn Trade, Cambrdge: MIT Press. Hunter, L., 1991, The Contrbuton of Nonhomothetc Preferences to Trade, Journal of Internatonal Economcs, 30, Hunter, L. and J. Markusen, 1988, Per Capta Income as a Bass for Trade. In R. Feenstra (eds.), Emprcal Methods for Internatonal Trade, Cambrdge: MIT Press. Lnder, S., 1961, An Essay on Trade and Transformaton, Ne York: Wley and Sons. Markusen, J., 1986, Explanng the Volume of Trade: an Eclectc Approach, Amercan Economc Reve, 76, Tnbergen, J., 1962, Shapng the World Economy: Suggestons for an Internatonal Economc Polcy. Ne York: Tenteth Century Fund. 12