Exchange Rate Pass-Through and Its Relation to Market Power: Reinterpretation of the Degree of Exchange Rate Pass-through

Transcription

1 Exchange Rate Pass-Through and Its Relaton to Market Power: Renterpretaton of the Degree of Exchange Rate Pass-through MnKyoung Km, Guedae Cho, and Won W. Koo* Paper prepared for presentaton at the Amercan Agrcultural Economcs Assocaton Annual Meetng, Montreal, Canada, July 27-30, 2003 The authors, respectvely, a Research Assstant Professor and a Professor, Center for Agrcultural Polcy & Trade Studes, Department of Agrcultural and Appled Economcs, North Dakota State Unversty. mkm@ndsuext.nodak.edu; gcho@ndsuext.nodak.edu; wkoo@ndsuext.nodak.edu Copyrght 2003 by MnKyoung Km, Guedae Cho, and Won W. Koo. All rghts reserved. Readers may make verbatm copes of ths document for non-commercal purposes by any means, provded that ths copyrght notce appears on all such copes.

2 Abstract We propose a dfferent perspectve for nterpretaton of exchange rate pass-through: a relatvely lower (hgher) degree of pass-through mples a compettve (less compettve) market. Usng three dfferent wheat exportng countres, the Unted States, Canada, and Australa, and two mportng countres, Japan and Korea, we are not lkely to reject our hypothess. In the compettve market (Japan), the exportng countres determne ther prces based on changes n the competng country s prce, and as a result, there s a close-to-zero degree of exchange rate pass-through. However, a lower degree of prce competton and a sgnfcantly hgher degree of exchange rate pass-through are found n a less compettve market (Korea). Keywords: Contegraton Analyss, Exchange Rate Pass-through, Prce Competton, Wheat Market

3 Introducton There have been an ncreasng number of studes concernng prce adjustment behavors of traders n nternatonal commodty markets; n partcular, the ssue of prce adjustment of frms n response to changes n exchange rates has garnered attenton (Mann, 1986; Govannn, 1988; Kasa, 1992; Yang, 1995; Feenstra, et al., 1996). Many of the studes have found less-thancomplete exchange rate pass-through, suggestng reasons such as product characterstcs, trade barrers, market share, ntal sunk cost, and the degree of competton (Krugman, 1987; Dornbusch, 1988; Baldwn, 1988; Froot and Klemperer, 1989; Feenstra, 1989; Fsher, 1989; Menon, 1996). Imperfect competton has been frequently counted as a man reason for the low degree of exchange rate pass-through (Krugman, 1987). Because an exportng frm carefully observes and responds to competng frms behavor n a market, especally a compettve market, nformaton about the competng frms behavor should be assessed n the analyss of exchange rate passthrough to acheve a more specfc mplcaton of the ntensty of market competton, as ndcated n Goldberg and Knetter (1999) and Gross and Schmtt (2000). Thus, an emprcal analyss whch does not consder rval frms behavor s not consstent wth the basc concept of the Nash equlbrum, and the results are only suggestve of the nature and of mportance of mperfect competton. Two dstnct studes are found whch ncorporate the prce decsons of all compettors n a specfc market: Goldberg and Knetter (1999) and Gross and Schmtt (2000). Goldberg and Knetter developed a theoretcal model to estmate the ntensty of competton (or market power) n a specfc market. Usng exchange rates of compettors as cost shfters, they dentfed the resdual demand elastcty of a frm n a market that enables a measurement of the market power of the frm. Ther model does not, however, suggest a proper nterpretaton of exchange rate 1

4 pass-through regardng the ntensty of market competton. Ths s because t does not consder a frm s own exchange rates but compettors exchange rates and other cost shfters nstead. 1 Meanwhle, Gross and Schmtt utlzed the prce decsons of all compettors n an mported automoble market n Swtzerland. They found a relatvely low degree of exchange rate pass-through and attrbute the ncomplete pass-through manly to the mperfect competton n the market as t was depcted by the nfluental study, Krugman (1987). Cauton s, however, needed n construng ncomplete pass-through when there exsts competng frms n a market. Competng frms, whch export hghly substtutable products to a foregn country, take nto account compettors behavor, and as a result, ther markup prce cannot fully adjust to changes n the exchange rate. In ths case, t s dffcult to conclude that the low degree of exchange rate pass-through s manly due to mperfect market competton. Rather, t can lkely be attrbuted to perfect competton because the frm responds to the rval frms behavor. On the other hand, f the frm executes ts market power, then t does not consder other prces and determne ts prce based on the exchange rate changes. Thus, the degree of exchange rate pass-through s lkely hgher than the former case. We propose the followng hypothess. If a frm faces perfect market competton, ts prce s completely explaned by the prce of compettors n the market, and the degree of exchange rate pass-through should be almost zero. In contrast, f the frm faces mperfect market competton, the prces of compettors are not lkely to explan the frm s mport prce, and we observe almost full exchange rate pass-through. Therefore, greater explanatory power of other frms prces on a frm s mport prce and less exchange rate pass-through mply less market power. Less explanatory power held by other frms prces on a frm s mport prce and hgher exchange rate pass-through mply hgher market power. 2

5 The objectve of ths study s, therefore, to draw new nterpretaton of the exchange rate pass-through and ts relaton to the ntensty of market competton. By analyzng both the dynamc propertes of the exchange rate pass-through and the prcng decsons of all compettors, we can derve more accurate estmates of exchange rate pass-through and more thoughtful economc mplcatons of the nature of competton n a market. To test the hypothess, we analyze the Japanese and Korean wheat market and compare the results between the two countres for the followng reasons. Frst, Japan and Korea are the two largest consstent wheat-mportng countres n Asa and they mport from the same major exportng countres: the U.S., Canada, and Australa. Japan and Korea mport almost 6 mllon and 1.5 mllon tones per year, respectvely, from the Unted States, Canada, and Australa durng the perod of 1989 to Because only 10% of consumpton s suppled domestcally n Japan and there s no domestc producton n Korea, both countres have been heavly dependent on wheat mports. However, there are fundamental dfferences between the marketng systems of the two markets. Japanese wheat mports are operated by a state tradng agent, the Japanese Food Agency (JFA), whch regulates all wheat trade n Japan. On the other hand, due to a trade lberalzaton program, wheat mports n Korea have been unregulated snce 1990 even though there have been decreasng tarff rates (5% pror to 1993 and 3% to 1.8% for 1995 to 2004). Because of the smlarty of the producton statuses and the dscrepancy between the marketng systems, these two countres are suspected of havng dfferent types of market competton. For example, because of state tradng n Japan, they allocate mport quotas to the three exportng countres based on ther own wheat mport plan, suggestng that Japan may have monopsony power as ndcated n Love and Murnnngtyas (1992) and Carter, et al. (1999), though Carter, et al. dd not reject perfect competton. Accordng to Pck and Park (1991), Yang 3

6 and Lee (2001) and Cho, et al. (2002), the Unted States s, however, lkely to have monopoly power n the Korean wheat market. Ths study conssts of two parts. Frst, the model, whch examnes the degree of exchange rate pass-through and prce competton, s estmated by a contegraton analyss developed by Johansen and Juselus (1994) to accommodate the propertes that both prces and exchange rates are non-statonary and that the prces of the compettors are nterdependent and smultaneously determned regardless of the economc sze of countres. The analyss also has the advantage of resolvng the over-dentfcaton problem. To strengthen our hypothess, alternatve models, measurng market power, are estmated. For example, a model s developed to test market structure for the Japanese wheat market: monopsony versus compettve. However, to test mperfect market competton among the wheat mportng countres, the model developed by Knetter (1989) s also estmated. The organzaton of ths study s as follows. The next secton dscusses about nterpretaton of exchange rate pass-through n relaton to the ntensty of market competton by usng the resdual demand. Then emprcal models are dscussed, followed by the explanaton of the data and mportant prelmnary estmaton results. The emprcal results of our tests on the prce rvalry effect and exchange rate pass-through are addressed along wth the results of the market power tests. The fnal secton summarzes the prncpal fndngs and the mplcatons of the research. Interpretaton of Measures of Exchange Rate Pass-Through: A Dscusson A low degree of exchange rate pass-through s commonly beleved to be due to mperfect competton, or exstence of market power of an exportng frm n a foregn market. Under the assumpton of constant margnal cost, the export (mport) prce n foregn currency s often found to partally adjust to changes n exchange rates, whch s nterpreted as an ncomplete 4

7 mark-up adjustment of the frm, and thus devatons from perfect competton (Goldberg and Knetter, 1997). However, the low degree of exchange rate pass-through also occurs n a perfectly compettve market. In ths case, the frm s a prce taker and should sell ts products at the market prce f t wants to stay n the market even though exchange rates move unfavorably. Thus, almost zero exchange rate pass-through occurs. 2 These two seemngly contradctve vews are not, n fact, contradctory but matters of consderng other players nformaton n the market. For nstance, n the frst case, a partcular frm s behavor n the market s solely examned wthout regard to the competng frm s nformaton. However, n the later case, the behavor of competng frms s mplctly consdered through the market clearng prce. Therefore, t needs to be cautoned to draw the mplcaton of the degree of exchange rate pass-through relatng to market power when consderng competng frm s behavor n a market, whch s closer to a complete economc model. In ths secton, we propose a dfferent mplcaton of the degree of exchange rate passthrough regardng the ntensty of market competton usng two extreme cases of a market. However, our dea s easly generalzed from the cases. Consder a frm who exports ts products to a foregn market and competes wth other frms who export smlar products to the market. Usng the concept of the resdual demand used by Becker and Bresnahan (1988) and Goldberg and Knetter (1999), we can explan the relatonshp between the degree of exchange rate pass-through and the ntensty of market competton. The resdual demand curve s defned as the dfference between the market demand and the compettve frnge s supply curves, summarzng constrants placed on behavor by competng frms n the market. 5

8 Usng the relatonshp between the elastcty of the resdual demand curve and the frm s prce-cost margn, whch measures the frm s market power, t s easy to derve the relatonshp between pass-through and market power. The elastcty of the resdual demand determnes the sze of the prce-cost margn: (1) the flatter the demand curve, the lower the margn; (2) the steeper the curve, the hgher the margn. The resdual demand curve wll be completely flat, mplyng zero margn and mark-up, when a frm s completely constraned n the exercse of market power by the exstence of compettors. In ths case, the prce does not depend on ts own quantty but s completely determned by the prces of ts compettors, ndcatng zero exchange rate pass-through. Because the frm faces completely flat (or elastc) demand, changes n ts margnal cost do not affect ts mport prce, so there s no drect relatonshp between mport prce and exchange rate (Fgure 1). In contrast, f the frm completely exercses ts market power even when there exst compettors n the market, the frm changes ts prce based on ts own quantty and not the prce of ts rvals. If the market demand s almost perfectly nelastc, the frm has a capacty constrant, say at Q 0. Because the frm faces completely vertcal demand, changes n ts margnal cost are fully reflected n changes n ts prce, ndcatng changes n exchange rates drve the prce to change by the exact amount of the change n exchange rates (Fgure 2). 3 In realty, the resdual demand should be nether perfectly elastc nor nelastc. The dea we suggest through these two cases s that the degree of exchange rate pass-through s postvely related to the ntensty of market competton when consderng rvalry frm s behavor n the market. More specfcally, (1) f a frm faces an almost flat resdual demand (no market power; no markup), ts mport prce can be totally explaned by prce of other compettors n the market, and exchange rate pass-through should be almost zero; (2) f a frm faces an almost vertcal 6

9 resdual demand (monopolstc power wth almost perfectly nelastc demand; hgh markup), prces of other compettors n the market cannot explan the frms mport prce, and the effect of exchange rates s fully transmtted to the prce. Thus, there s full exchange rate pass-through. Therefore, less explanatory power of other frms prce on a frm s mport prce and hgher exchange rate pass-through mply the frm s hgher market power and hgher markup. In contrast, hgher explanatory power of other frms prce on the frm s mport prce and less exchange rate pass-through mply less market power and markup of the frm Emprcal Model Specfcaton for Exchange Rate Pass-Through Recently Goldberg and Knetter (1999) developed a one-equaton model whch s a functon of resdual demand to estmate the nterestng summary statstcs of the demand for a specfc frm. Ths model resolves the problem of estmatng all the underlyng parameters of demand, cost, and conduct varables n the system of equatons, whch requres publcly naccessble specfc frm-level data n the nternatonal agrcultural market and the specfc functonal forms of the equatons. An advantage of usng the one-equaton model s that t captures the jont mpact of the enttes of nterest on market power through the elastcty of the resdual demand curve. In ths secton, we brefly descrbe ther model and modfy t to examne the degree of own exchange rate pass-though to mport prces regardless of an endogenety problem. The model emprcally utlzed n our study corresponds wth a recently developed tme-seres model by Gross and Schmtt (2000), who used a sub-game perfect equlbrum model. Consder the case that there are N compettors sellng smlar goods n a sngle mportng country. In ths case, each exporter,, faces the followng demand system: P and P D( Q, P, Y, δ ) where j. (1) = j Q are the prce and quantty, respectvely, of the export good, for a specfc frm ; P s a vector of the prces of export goods for all other compettors n a market; Y s ncome of the j 7

10 mportng country; and δ s a vector of the coeffcents to be estmated. To avod potental dentfcaton problem due to the smultanety of the varables, Goldberg and Knetter consdered the supply relaton for all other compettors n the market: P = C ( Q, W, Z, Γ ) D ( Q, Y, δ) Qθ, (2) 1 1 where Q s the vector of all frms quanttes; C ( ) s the margnal cost of the frm; D 1 s the partal dervatve of the demand functon wth respect to export quantty; prces pad by frm ; 1 W s a vector of factor Z are other varables that shft costs; Γ are unknown parameters; and θ s a conduct parameter. Wth (1) and (2) for all other compettors n the market, we have the followng reduced form of an equaton: P = P * ( Q, W, Z, Γ, θ, Y, δ), (3) N N N N where N denotes the unon of all compettors. Put (3) nto (1), we have the followng resdual demand: * P = D( Q, P (), Y, δ ), where j or (4) j R P = D ( Q, W, Z, Γ, θ, Y, δ), (5) N N N N where D R represents the resdual demand curve. The elastcty of R D wth respect to Q mples the market power of the frm over product s prce, takng nto account the adjustment of all other frms prces and quanttes. Regardng (4) and (5), several ssues should be addressed. Frst, n many cases of nternatonal agrcultural markets, t s not plausble to collect especally hgh frequency level data (quarterly, monthly), and hence, t s not possble to drectly estmate the resdual demand elastcty. However, because the objectve of ths study s to examne the degree of exchange rate pass-through and ts relaton to market power, we would nterpret the market ntegraton by nvestgatng the relatonshp between exchange rates and Q, 8

11 mport prces as explaned n the prevous secton. In fact, exchange rates, as a cost shfter, explan most varaton n quantty Q as explaned n Goldberg and Knetter (1999). Second, because the prces of compettors are affected by Q, whch s an endogenous varable, Goldberg and Knetter (1999) used (5) nstead of (4). If the dentfcaton problem matters when Equaton (4) s estmated, t can be drectly resolved wth a dfferent econometrc method, whch does not suffer from an endogenous problem. Then Equaton (4) becomes the tme-seres model, whch concdes wth the one used by Gross and Schmtt, as follows: * P = D( S, P, Y, δ ), where j (6) j S s the exchange rate between export and mport countres currency. An advantage of usng (6) s that t examnes the exchange rate pass-through by consderng relevant nformaton for all compettors. Cauton should be needed to nterpret the degree of exchange rate pass-through: the lower the degree the closer to perfect competton, whle the hgher the degree the more market power the frm executes. In addton, all the varables, except the exchange rate and ncome of the destnaton market, are determned smultaneously, whch s more consstent wth the concept of a non-cooperate smultaneous game. Therefore, a stochastc (or tme-seres) analyss must be approprate. Exchange rates and ncome are treated as exogenous varables, affectng mport prces but not nfluenced by the prces n return. 4 The last ssue s about substtutablty and product dfferentaton: substtutablty between products wll be hgher wthn a hghly dfferentated product-group. As Bodnar, et al. (2002) proved, the mpact of hgher substtutablty between products s to moderate the exchange rate pass-through nto the mport prces. The ncrease n substtutablty rases the elastcty of demand faced by the frm, and as a result, smaller prce changes are necessary to acheve the new proft-maxmzng level of sales n the foregn market. Thus, f product 9

12 dfferentaton s not taken nto account n a model, the mplcaton of exchange rate passthrough s lkely msleadng. Accordng to the Internatonal Grans Councl s World Gran Statstcs, the Unted States exports four dfferent types of wheat to Japan: dark northern sprng (14%), hard red wnter (13%), hard red wnter ordnary (12%), and soft whte wheat, whle Canada and Australa export hard red sprng wheat (13.5%) and whte wheat, respectvely. 5 Each of these types of wheat can be treated as a dfferentated product that competes wth other types of wheat that have a proten, mplyng smlar purpose of usage. For example, the U.S. dark northern sprng wheat (14%) mght compete wth Canadan hard red sprng wheat (13.5), and the U.S. whte wheat mght compete wth Australan whte wheat. Thus, we dfferentated wheat products nto two groups based on smlar attrbutes: the frst (the U.S. dark northern sprng wheat and Canadan hard red sprng wheat) as hard red wheat (hw) and the second (the U.S. whte wheat and Australan whte wheat) as whte wheat (ww) that ams at Udon noodle markets, assumng that t s hghly substtutable between exportng countres wthn a dfferentated product-group but not lkely substtutable between the groups due to dfferent qualtes, makng dfferent purpose of usage. The hgher the substtutablty among exportng countres, the more competton among the countres. To accommodate the dverse wheat types, Equaton (6) must be modfed as follows: v *v P = D( S, P, Y, δ ) (7) j where v denotes the respectve wheat types: hw and ww. 6 An advantage of usng (7) s a drect nvestgaton of the relatonshp wthn wheat types as well as between exportng countres. Emprcal Model Specfcaton for Measurement of Market Power 7 To examne the ntensty of market competton n the Japanese wheat market, monopsony versus compettve market, we develop a smple model, whch s a dual representaton of the one n Knetter (1989). 10

13 Suppose that the Japanese Food Agency (JFA) s the only purchaser of wheat from the exportng countres, t recognzes that the amount of wheat t demands wll nfluence the prce that t has to pay for t, and t sells wheat to the Japanese domestc market. Then, ths relatonshp can be summarzed as ts proft maxmzaton problem as follows: max p d f( x) w( x) x, (8) x where p d s the Japanese wheat domestc prce, f(x) s the wheat producton functon for the Japanese domestc market, and w(x) s a supply functon such that f the JFA wants to purchase x unts of wheat t must pay a prce of w(x) to the exportng countres. Then, solvng the frst order condton for (8) s represented as wheat mport prce, w MC ε = ε + 1, (9) where ε s the elastcty of the supply functon and s postve, and MC represents the margnal cost. Because the markdown ( εε+ 1) s less than one, w must be less than the margnal cost. Thus, f the JFA executes ts market power, the mport prce (w) must be less than a market prce. If the Japanese market s compettve, then ε =, meanng that the elastcty does not vary across the exportng countres (flat supply functon). As a result, the mport prce (w) must be equal to MC. follows: In order to estmate (9), the cross-sectonal-tme-seres model as n Knetter (1989) s as ln w w w w = α + δ + γ ln s + ε (10) t t t t w w where αt s the tme effect, δ s the country effect, and εt s the error term. As proposed n Knetter (1989) and Pck and Park (1991), we can use (10) to dstngush the types of market competton. Frst, n a compettve market, mport prces are the same for all exportng 11

14 w countres because there s no country effect, ndcatng δ = 0. In the compettve market, w changes n the blateral exchange rates do not affect the mport prces, so that γ = 0. Second, n an mperfectly compettve market, the mport prces must be dfferent across the exportng countres, and changes n the exchange rates are fully passed-through to the mport prces (the w w supply elastcty,ε, vares over the exportng countres). Thus, δ 0 and γ 0. To examne the market power that the exportng countres may have across the mportng countres, Japan and Korea, we adopt the model developed by Knetter. The model s as follows: ln p p p p t t t t = α + δ + γ s + ε (11) where p s wheat mport prce. The nterpretaton about market competton s the same as above and more detals are n the Knetter (1989) and Pck and Park (1991). Data Quarterly data are used between 1980:I and 2000:IV for Japan and 1993:I and 1999:III for Korea. 8 Three varables are needed: mport prces from respectve exportng countres, the Unted States (u), Canada (c), and Australa (a); nomnal exchange rates between respectve exportng countres and mportng countres; and mportng countres ncome. For Japan, wheat mport prces from respectve exportng countres (p uj, p cj, p aj n respectve groups) are average quarterly prces based on the monthly prces, drawn from the Internatonal Grans Councl s World Gran Statstcs. For the hard red wheat group, the prces of U.S. dark northern sprng wheat and of Canadan hard red sprng wheat are collected. For the whte wheat group, the prces of U.S. whte wheat and Australan whte wheat are collected. The prces are quoted at nearest shpment poston and denoted n Japanese yen per ton. For Korea, however, we are not able to dfferentate wheat products mported to the Korean market because of lmted data avalablty. An average prce of all wheat types s used for the Korean market 12

15 (p uk, p ck, p ak ), whch s provded by Monthly Report of Trade Statstcs by Korea Natonal Statstcs Offce. The monthly prce seres provded by World Gran Statstcs have a small number of mssng observatons. To approxmate the mssng observaton, the super smoothng s used (Fredman, 1984). There were more than three consecutve mssng observatons and more than eghteen data ponts mssng out of two hundred forty n monthly prces. It should be noted that such a procedure mght not ntroduce the potental for bases and nconsstences n the estmaton and subsequent statstcal nferences. Nomnal exchange rates (s u, s c, s a ) are defned as the wheat exportng countres currences per respectve mportng countres currences, Japanese Yen and Korean Won. Exchange rates are provded by the Economc Research Servce (ERS) n the U.S. Department of Agrculture (USDA). Real GDP data for both Japan and Korea are collected from Internatonal Fnancal Statstcs. Note that both exchange rates are assumed and tested to be exogenous n determnng mport prces and that all the varables are converted nto logarthms. Intal Investgaton We test nonstatonarty of the ndvdual tme seres usng the Augmented Dckey-Fuller test (ADF) and the Phlps-Perron test (PP) wth ntercept and trend to avod possble problems caused by heteroskedastcty n the varables. In addton, the statonarty test based on the LR 2 test ( χ ) s conducted due to low power of dstngushng slow mean revertng from nonstatonarty. 9 The results of these tests ndcate that the varables contan unt root for both Japan and Korea data, meanng that these varables are not mean revertng but become I(1), statonary, at a 95% sgnfcance level. 13

16 Two statonary and one statonary relatonshps among the prces are dentfed n Japan (one relaton for each group) and Korea, respectvely, based on the maxmum egenvalue test and the trace tests at a 95% sgnfcance level. Because the tests are asymptotc dstrbutons and can be an naccurate approxmaton to the small sample dstrbuton, these contegraton relatonshps are confrmed by both the egenvalue of the companon matrx and the long-run speed adjustment n the VECM. The normalzed vald contegratng vectors explaned by the long-run coeffcents n the VECM are as follows: for the whte wheat group n whch the Unted States and Australa compete, uwj awj p = 0.751p τ, (12) where p uwj and p awj are U.S. and Australa whte wheat mport prces n Japan, respectvely, and τ s a tme trend; and for the hard red wheat group n whch the Unted States and Canada compete, p uhj chj = p, (13) where p uhj and p chj are U.S. and Canada hard red wheat prces n Japan, respectvely. Both U.S. prces (p uwj and p uwj ) are postvely related to the pres of respectve competng countres n each wheat group. The fndng of contegraton mples that U.S. prces and the prces of respectve competng countres tend to move proportonal to each other over long-horzons and do not drft too far. In the Korean wheat market, the hypothess of one contegratng vector cannot be rejected at the 95% sgnfcance level based on the lkelhood rato statstc that s 6.46 wth χ 2 (5). The normalzed contegratng vectors are as follows: ak uk ck p = 0.361p p τ, (14) where p ak, p uk, and p ck are Australan, U.S. and Canadan wheat mport prces n Korea, respectvely. The contegratng vector (14) dentfes a stable tme-trend relatonshp among 14

17 prces and ndcate that Australan prce (p ak ) s assocated wth postve movement n U.S. prce (p uk ) and Canadan prce (p ck ). For both Japanese and Korean markets, the co-movement of competng prces ndcates that exportng countres take account of the other partes prces to determne ther prces n the long-run. Although t s too early to determne, the larger coeffcents of the long-run relatonshp n the Japanese wheat market presage of more prce rvalry among compettors n the Japanese wheat market than n the Korean market. Accordng to Gross and Schmtt (2000) and Feenstra, et al. (1996), the stable long-run relatonshp among prces can be nterpreted that the wheat exporters may am at long-run market shares n the wheat markets. Results We dvde the dscusson of the results n two parts. Frst, we examne the degree of exchange rate pass-through through short-run dynamc analyss usng the vector error correcton model (VECM) and long-run mpulse response analyss usng the vector movng average representaton (VMAR). Second, we nvestgate the ntensty of market competton n respectve markets to strengthen our hypothess: low degree of exchange rate pass-through mples close-to perfect competton whle hgh degree of pass-through mples mperfect competton. Prce Rvalry Relatonshp and Exchange Rate Pass-Through Short-run Dynamcs Usng the dentfed parameters of the contegraton relaton ( and ), the short-run relatonshp parameters n the VECM are estmated. Snce the exchange rates are weakly exogenous, the model s re-estmated condtonal on the varable. By removng nsgnfcant coeffcents of the varables based on a lkelhood rato test, the model s estmated by usng full- 15

18 nformaton maxmum lkelhood estmaton (FIML). The results of short-run exchange rate pass-through are presented n Tables 1 and 2. Hgher prce competton among exportng countres s found n Japanese and less competton n Korean markets. In the Japanese wheat market (Table 1), all coeffcents of prces are postve for both wheat groups, ndcatng they have an ncentve to follow respectve compettor s prce changes. The prce rvalry relatonshps between the competng countres ndcate that whte wheat s hghly substtutable between the Unted States and Australa, as s hard red wheat between the Unted States and Canada, suggestng that prce competton occurs when wheat s hghly substtutable between exportng countres. For example, U.S. whte wheat prce responds to a 100% ncrease n Australan prce by 21.3% to 41.1% over three quarters, whle Australan whte wheat prce reacts to a smlar ncrease n U.S. prce by 24.0% to 34.4% over two quarters. For hard red wheat, U.S. prce responds to Canadan prce change by 17.2% to 43.3%, whle Canadan prce reacts to U.S. prce changes by 22.5% to 35.9%. In the Korean wheat market (Table 2), however, prce competton among exporters s not as compettve as n the Japanese wheat market. Although the coeffcents of prces are all postve, the degree of competton s less than t s n Japanese wheat market. In addton, the speed of response s at most two quarters. For example, U.S. wheat prce only responds to changes n Australan prce by 12.5% to 18.2%, and Canadan wheat prce reacts to U.S. and Australan prces by 19.1% and 15.9%, respectvely. Only Australan prce s ndcatve of sgnfcant prce competton wth the two other exporters over two quarters, but the response szes are less than 20.0%. We recognze that ths fndng s partally due to lmts of data avalablty: wheat products are not dfferentated as n the Japanese market, and a short sample tme perod s used. 16

19 However, the dfferent magntudes of the prce rvalry relatonshps n Japan and Korea s, at least, suggestve of the level of prce competton between exportng countres. To strengthen ths fndng, we further examne the ntensty of market competton and fnd consstency between the two results, whch s explaned n the later secton. When takng compettor s behavor nto account, the exchange rate pass-through s expected to partally penetrate to mport prces. The coeffcents of changes n each exportng country s own exchange rates are all negatve n the respectve markets, ndcatng that an apprecaton of an exportng country s currency relatve to the mportng country s currency leads to an ncrease n the correspondng prce, represented n the mportng country s currency. In the Japanese wheat market, a sgnfcantly low degree of exchange rate pass-through s found for both wheat groups as shown n Table 1. For example, U.S. whte wheat prce, represented n Japanese yen, responds postvely 7.4% n frst quarter delayed to 100% apprecaton of own currency relatve to Japanese yen, and the mpact decays to 4.3% over three quarters, whle own currency apprecaton rases Australan prce by 10.3% to 9.8% over two quarters. For hard red wheat, the relatve apprecaton of own currency ncreases U.S. prce, represented n Japanese yen, by 3.2% to 5.2% and Canadan prce by 4.7% to 2.5% over three quarters. These levels of exchange rate pass-through are, however, much lower than the level of pass-through n the Korean wheat market (Table 2). U.S. wheat prce reacts postvely 40.5% to 31.6% to relatve apprecatons of own currency, Canadan prce responds by 44.8% to 26.0%, whle Australan prce responds by more than 30.0%. Long-run Impulse Response To verfy the consstency of our fndng over tme, the exchange rate pass-through and prce rvalry relatonshp are examned n the long-run, and the results are presented n Tables 3 17

20 and 4. The tables report the long-run mpulse response functon n the VMAR for a untary change of shock gven to each varable. The sgnfcance of each entry, represented by coeffcents n bold, ndcates that the shock gven to each varable exhbts a permanent effect on each prce. A postve and remarkable symmetry n the sze of long-run response of prces ndcates that there exsts the prce rvalry relatonshp for hghly substtutable wheat between exportng countres n the Japanese market: the Unted States and Australa for whte wheat and the Unted State and Canada for hard red wheat (Table 3). For example, 100% ncreases n respectve U.S. and Australan whte wheat prces causes the compettor s prce to ncrease by around 25%, whle the Unted States and Canada follow a 100% ncrease n ther compettor s hard red wheat prces by ncreasng ther prce around 30%. Smaller magntudes of prce rvalry relatonshp are found n the Korean wheat market, as found n the short-run (Table 4). The Unted States s more perceptve about Australan prce: t ncreases ts prce by 22.6% n reacton to a 100% ncrease n Australan prce, whle Australa responds by 17.7% to U.S. prce change. Canadan and Australan prces symmetrcally react 16.7% and 15.1% to each other. However, as n the short-run, a change n Canadan prce does not nfluence U.S. wheat prce n the Korean market: a 100% ncrease n Canadan prce causes U.S. prce to ncrease by only1.20%. In the long-run, dfferent szes of the degree of pass-through are found n the two markets, as found n the short-run. A sgnfcantly low degree of exchange rate pass-through s found n the Japanese wheat market (Table 3), whle a much hgher degree of pass-through s found n the Korean market (Table 4). For example, as shown n Table 3, 100% apprecaton of each exportng country s currency relatve to Japanese yen ncreases U.S. and Australan whte wheat prce, represented n Japanese yen by 4.6% and 8.7%, respectvely, and U.S. and Canadan 18

21 hard red wheat prce by only 5.4% and 4.5%, respectvely, n the Japanese wheat market. Meanwhle, around 33% of exchange rate pass-through to Korean wheat mport prces, represented n Korean won, occurs n response to 100% apprecaton of each exportng country s currency relatve to Korean won. The dscrepancy between the two markets mples that our hypothess s not lkely to be rejected, and as a result, t s possble for the exportng countres to fnd ther optmal prces by takng nto account changes n competng countres prces, slowng the exchange rate passthrough to the prces when the market s compettve lke the Japanese wheat market n ths study. Exportng countres are, however, less concerned about ther competng countres prces when a market s less compettve, and as a result hgher exchange rate pass-through occurrng, as n the Korean wheat market. Intensty of Market Competton For the argument above to be properly nterpreted n relaton to exchange rate passthrough and market competton, we further nvestgate the ntensty of market competton n respectve markets. Models (10) and (11) are estmated usng OLS wth dummy varables as n Knetter (1989) and Pck and Park (1991). Tables 5 and 6 report the estmates of country (δ ) and exchange rate effects (γ ) for the two markets. In the Japanese wheat market (Table 5), nether δ nor γ are not sgnfcant at 95% sgnfcance level, so we cannot reject the null hypothess of perfect market competton for the sample perod. Usng Model (10), we are not able to fnd Japanese monopsony power for purchasng wheat, whch supports our hypothess of hgh prce rvalry relatonshp, slowng the exchange rate pass-through to mport prces when the market s close to perfectly compettve. Although drect comparson s not lkely plausble because of utlzaton of dfferent data set and model, our result s contrary to one found n Love and Murnnngtyas (1992), who found 19

22 monopsony power n the Japanese wheat market, but partly supports Carter, et al. (1999), who dd not reject perfect market competton nor monopoly power n the Japanese wheat market. Table 6 shows the lkely results of monopoly power of the exportng countres across the markets. Because the prce data for Korea s an average prce of all wheat types, the same prce data for Korea are used over the two groups of wheat types accordng to Japanese wheat mports, so that the results are only suggestve of prce rvalry and exchange rate pass-through relaton. In contrast to prevous studes that dd not dfferentate wheat types, we found that exportng countres that can dscrmnate prce across the destnaton markets are dfferent across wheat types. For whte wheat, the null hypothess of perfect market competton should be rejected for Australa, ndcatng that Australa has a certan level of market power relatve to the Unted States n the Korean wheat market. Meanwhle, the Unted States s found to have a certan level of market power for hard red wheat by dscrmnatng prce across destnaton markets, Japan and Korea, relatve to Canada. Conclusons We nvestgate the prce decson behavor of wheat exportng countres, the Unted States, Canada, and Australa, n response to changes n exchanges rates when the competng frms export hghly substtutable wheat products to Japanese and Korean wheat markets. The Japanese and Korean markets are selected and the results for the two countres are compared because the two countres are the largest wheat mporters n Asa and they mport from the same three countres, but the marketng systems n the two countres are dfferent: state trade n Japan versus prvatzed trade n Korea. We found wheat type specfc prce competton n the Japanese market for both short- and long-run: the Unted States versus Australa n the case of whte wheat and the Unted States versus Canada for hard red wheat. Because each wheat type s hghly substtutable between the 20

23 exportng countres n the compettve market, the respectve country determnes ts prce based on changes n competng country s prce, and as a result, a close-to-zero degree of exchange rate pass-through occurs. In the case of the Korean wheat market, however, less degree of prce competton and sgnfcantly hgher degree of exchange rate pass-through are found, ndcatng that the Korean market s less compettve than the Japanese market. Estmaton on the ntensty of market competton s conducted to strengthen our fndngs above: the Japanese wheat market s found to be compettve rather than monopsonstc, whle Australa and the Unted States are found to have market power for respectve wheat types, whte wheat and hard red wheat, n the Korean wheat market. Overall, ths study may contrbute to a better understandng of the degree of exchange rate pass-through by ncorporatng substtutablty between wheat products and competng frms prces, as ndcated n Goldberg and Knetter (1999) and Gross and Schmtt (2000), and ts relaton to the ntensty of market competton. In contrast to the prevalng belef that low degree of pass-through s due to mperfect market competton, we conclude that lower the degree of exchange rate pass-through, the hgher the market competton, when prce nterdependency among exportng countres matters. In a hghly compettve market such as the Japanese wheat market, the prce of the exportng frm s more lkely to be explaned by the prces of compettors n the market, slowng down the exchange rate pass-through. On the other hand, n a less compettve market such as the Korean market, the prces of compettors are less lkely to explan the frm s prce, so the degree of exchange rate pass-through s hgher than the Japanese market. 21

24 Table 1. Short-run Dynamcs for Japanese Wheat Market k pt 1 p k t 2 p k t 3 st 1 st 2 st 3 yt 1 yt 2 yt 3 ecm Whte Wheat p u (1.67) p a (2.96) (1.98) (1.96) (3.51) (-1.90) (-1.83) (-2.26) (-1.95) (-1.92) (2.01) (1.80) (3.12) (2.56) Hard Red Wheat p u (1.97) p c (2.02) k pt 1 p k t 2 p k t 3 st 1 st 2 st 3 yt 1 yt 2 yt (1.691) (2.57) (3.07) (-2.27) (-2.22) (-1.95) (-2.23) (-2.02) (-2.18) (1.83) (1.743) ecm (1.76) Note that k denotes competng countres, and s denotes own exchange rates. ecm n respectve wheat group ndcates error correcton model (12) and (13), respectvely. Table 2. Short-run Dynamcs for Korean Wheat Market c pt 1 p c t 2 p a t 1 p a t 2 st 1 st 2 yt 1 yt (3.00) (3.25) ecm P u (3.64) (2.033) (-2.07) (-2.304) (2.70) u pt 1 p u t 2 p a t 1 p a t 2 st 1 st 2 yt 1 yt 2 ecm1 P c (4.48) (2.78) (-2.38) (-1.79) (2.63) (1.83) (4.01) u pt 1 p u t 2 p c t 1 p c t 2 st 1 st 2 yt 1 yt 2 ecm P a (2.21) (1.849) (2.92) (2.39) (-2.392) (-2.172) (3.05) 22

25 Table 3. Long-run Impulse Response n Japanese Wheat Market Shocks gven to p u p a s y p u Whte (2.716) (2.849) (-2.610) (-0.169) Response Wheat p a (1.998) (2.674) (-2.312) (1.096) Shocks gven to p u p c s y p u Hard (2.261) (2.718) (-2.716) (3.144) Red Response Wheat p c (2.509) (1.139) (-2.463) (1.302) Note that s denotes own exchange rates. Sgnfcant responses to own prces are not n bold. Table 4. Long-run Impulse Response n Korean Wheat Market Shocks gven to p u p c p a s k y p u (1.423) (0.565) (2.557) (-2.244) (1.111) p c Response (2.532) (1.553) (2.929) (2.553) (1.988) p a (2.495) (1.971) (1.218) (-2.614) (1.490) k Denotes respectve countres. Japanese Monopsony Power lnw t α t δ γ s t ε t = (15) Table 5. Japanese Monopsony Power U.S. Whte Wheat Australa U.S. Hard Red Wheat Canada Numbers n parentheses are t-statstcs. δ w (-0.393) (0.408) (0.073) (-0.054) γ w (-1.393) (1.090) (0.836) (-1.069) 23

26 Table 6. Market (Monopoly) Power Between Japanese and Korean Wheat Market p p δ γ Japan (-0.027) (1.330) U.S Korea Whte (-0.207) (-1.330) Wheat Japan (1.568) (-1.189) Australa Korea (1.893) (-3.643) Japan (1.226) (0.891) U.S Korea Hard Red (2.404) (-8.358) Wheat Japan (-0.913) (-1.092) Canada Korea (1.269) (-0.713) Numbers n parentheses are t-statstcs. 24

27 Fgure 1 (a) Fgure 2 (a) MC(S ) P MC(S ) MC(S 0 ) P MC(S 0 ) P 0 Q Q 0 Q Fgure 1 (b) Fgure 2 (b) P P P 2 P 0 P 0 S S 0 S S 0 S S Fgure 1 (a) explans that changes n margnal cost (MC(S 0 ) MC(S )), whch s the functon of exchange rates (S), does not affect the prce when the frm faces the elastc resdual demand curve. Fgure 1 (b) explans the relatonshp between the exchange rates and the prce, whch s drven from Fgure 1 (a) through the margnal cost movement. In the case of the flat resdual demand, the relatonshp between the exchange rates and the prce s depcted as horzontal, mplyng zero exchange rate pass-through. Meanwhle, Fgure 2 (a) explans that changes n margnal cost (MC(S 0 ) MC(S )) are fully reflected n the prce changes (P 0 P 1 ) when the frm fully exercse ts market power, facng nelastc demand curve. Fgure 2 (b) shows one-toone relatonshp between the exchange rate and the prce when the frm exercses the market power. The relatonshp s depcted as 45 degree lne, mplyng full exchange rate pass-through. 25

28 References Baldwn, P. "Hysteress n Import Prces: The Beachlead Effect." The Amercan Economc Revew 78(1988): Bodnar, G. M., B. Dumas, and R. C. Marston. "Pass-through and Exposure." Journal of Fnance 57(2002): Carter, C. A., D. MacLaren, and A. Ylmaz. "How Compettve s the World Wheat Market?" Workng Paper No Departement of Agrcultural and Resource Economcs, Unv. of Calforna Davs. Cho, G., H. Jn, and W. Koo. "Measurng the Market Power of the U.S. Wheat Exporters n Asan Countres: An Issue about Adjustment of Nomnal Exchange Rate When Usng as a Cost Shfter.". Amercan Agrcultural Economcs Assocaton Annual Meetng, July 28-31, Dornbusch, R. "Doubts About the McKnnon Standard." Journal of Economc Perspectve 2(1988): Feenstra, R. C. "Symmetrc Pass-Through of Tarffs and Exchange Rates Under Imperfect Competton: An Emprcal Test." Journal of Internatonal Economcs 27(1989): Feenstra, R. C., J. E. Gagnon, and M. M. Knetter. "Market Share and Exchange Rate Pass- Through n World Automoble Trade." Journal of Internatonal Economcs 40(1996): Fsher, E. "A Model of Exchange Rate Pass-Through." Journal of Internatonal Economcs 26(1989): Fredman, J. H. "A varable span smoother." Tech. Rep. 5. Laboratory for Computatonal Statstcs, Dept. of Statstcs, Stanford Unv., CA. Froot, K. A., and P. D. Klemperer. "Exchange Rate Pas-Through When Market Share Matters." The Amercan Economc Revew 79, no. 4(1989): Govannn, A. "Exchange Rates and Traded Goods Prces." Journal of Internatonal Economcs 24(1988): Goldberg, P., and M. Knetter. "Measurng the Intensty of Competton n Export Markets." J. Internatonal Economcs 47(1999): Gross, D. M., and N. Schmtt. "Exchange Rate Pass-through and Dynamc Olgopoly: an Emprcal Investgaton." Journal of Internatonal Economcs 52(2000): Johansen, S., and K. Juselus. "Identfcaton of the Long-run and the Short-run Structure: An

29 Applcaton to the ISLM Model." Journal of Econometrcs 63(1994): Johansen, S., and K. Juselus. "Testng Structural Hypotheses n a Multvarate Contegraton Analyss of the PPP and the UIP for UK." Journal of Econometrcs 53(1992): Juselus, K., and R. MacDonald. "Interest Rate and Prce Lnkages between the USA and Japan: Evdence from the Post-Bretton Woods Perod." Unpublshed Manuscrpt,(2000b). Juselus, K., and R. MacDonald. "Internatonal Partes between Germany and the Unted States: A Jont Modelng Approach." Unpublshed paper,(2000a). Kasa, K. "Adjustment Costs and Prcng-to-Market Theory and Evdence." Journal of Internatonal Economcs 32(1992): Km, M., and W. Koo. "How Dfferently Agrcultural and Industral Sectors Respond to Exchange Rate Fluctuaton?" Agrbusness and Appled Economcs Report No Center for Agrcultural Polcy and Trade Studes North Dakota State Unversty, Knetter, M. M. "Prce Dscrmnaton by U.S. and German Exporters." The Amercan Economc Revew 79, no. 1(1989): Krugman, P. (1987) Prcng to Market When the Exchange Rate Changes., ed. S. Arndt, and J. Rchardson. Cambrdge, MA, MIT Press, pp. pp Love, H. A., and E. Murnnngtyas. "Measurng the Degree of Market Power Exerted by Goverment Trade Agences." Amer. J. Agr. Econ. 74(1992): Mann, C. "Prces, Proft Margns, and Exchange Rates.". Federal Reserve Bulletn, June, Menon, J. "The Degree and Determnants of Exchange Rate Pass-Through: Market Structure, Non-Tarff Barrers and Multnatonal Corporatons." The Economc Journal 106, no. 435(1996): Pck, D. H., and T. A. Park. "The Compettve Structure of U.S. Agrcultural Exports. " Amer. J. Agr. Econ. 73(1991): Yang, S., and W. Lee. "Exporters' Market Power n Agrcultural Import Markets In Korea.". Amercan Agrcultural Economcs Assocaton Annual Meetng, August 5-8, 2001.

30 Notes 1 Own exchange rate s defned as the rato between an exportng frm s currency and an mportng country s currency, whle compettors exchange rates are defned as the rato between rvalry-exportng frms currency and mportng country s currency as defned n Goldberg and Knetter (1999). 2 Then there are two the questons: The frst one why they stay n the market even though there s no gan. There are some answers. For nstance, t stays because the market share s mportant (Froot and Klemperer, 1988); because ntal nvestment costs are bg (Baldwn and Krugman: 1988); because of the capacty constrant, smply do dumpng. Then, the next queston s how they can stay. There are some answers; the frm has some markup n dfferent markets; n the case of the agrcultural trade, domestc subsdy. 3 Because the elastcty of exchange rate pass-through, the effect of the exchange rate on the mport prce n foregn currency, should be n a lmted range ( 0 ε 1 ) Bodnar, G. M., B. Dumas, and R. C. Marston. "Pass-through and Exposure." Journal of Fnance 57(2002): , the relatonshp between the exchange rate and the prce should le between zero degree and 45 degree lne as shown n Fgures 1 and 2. 4 Ths s an approprate assumpton, specfcally for exchange rates. The exchange rate s found to be weakly exogenous n the U.S. agrcultural sector (Km and Koo, 2002) 5 Numbers n parentheses represent the level of proten. 6 Because of data avalablty, t s not possble to dfferentate wheat Korea mports from those three countres, and as a result, product dfferentaton s not consdered n the Model (7). 7 If quantty data s avalable, then the resdual demand model (5) should generate better estmates of market competton. However, because of data lmtaton n the two markets, alternatve models testng market power should be utlzed. 8 We lmt our data startng 1980 because (1) there are many mssng observatons of Australan whte wheat mport prces and nconsstency n U.S. dark northern sprng wheat before 1975; (2) quarterly GDP data are avalable from 1980; and (3) the U.S. dollar has sgnfcantly apprecated relatve to other currences durng the last two decades, whch mght affect wheat trade wth the mportng countes. In addton, we would lke to thank Prof. Seung-Ryong Yang and hs former student Won-Jn Lee at the Department of Food and Resource Economcs n Korea Unversty for provdng Korean prce data. 9 See Johansen and Juselus (1992), Juselus and MacDonald (2000b), and Km and Koo (2002) for the LR test of Unt Root.