Softwood Lumber Products in the United States: Substitutes, Complements, or Unrelated?

Transcription

1 Forthcomng n Forest Scence Softwood Lumber Products n the Unted States: Substtutes, Complements, or Unrelated? Rao V. Nagubad, Daowe Zhang, Jeffrey P. Prestemon, and Davd N. Wear September 24, 2003 Rao Nagubad, Post-doctoral Fellow, and Daowe Zhang, Professor, School of Forestry & Wldlfe Scences, Auburn Unversty, AL ; Jeffrey Prestemon, Research Forester and Davd Wear, Project Leader, Economcs of Forest Protecton and Management, Southern Research Staton, USDA Forest Servce, P.O. Box 12254, Research Trangle Park, NC Zhang s the correspondng author and he can be reached at (334) , fax: (334) , and E-mal: zhangdw@auburn.edu Acknowledgments: Comments from two anonymous revewers and an Assocate Edtor of ths journal were greatly helpful. The fnancal support from USDA Forest Servce (Cooperatve Agreement SRS 01-CA ) s gratefully acknowledged.

2 Softwood Lumber Products n the Unted States: Substtutes, Complements, or Unrelated? ABSTRACT. Ths study addresses an mportant dmenson concernng the softwood lumber trade dspute between U.S. and Canada substtutablty among mported and domestcally produced speces. We employ the restrcted translog subcost functon approach to study ths ssue based on the monthly data of U.S. softwood products consumpton and prces between January1989 and July The results show that the spruce-pne-fr lumber speces group, manly mported from Canada, s largely unrelated to domestcally produced treated southern yellow pne, Douglas-fr, and other speces groups, but s a substtute to untreated southern yellow pne and engneered wood products. Furthermore, untreated southern yellow pne s facng more severe competton from structural panels rather than from the mported Canadan spruce-pne-fr group. Key words: Trade dspute, U.S.-Canada, southern pne, spruce-pne-fr, engneered wood products, translog subcost functon. 2

3 Softwood Lumber Products n the Unted States: Substtutes, Complements, or Unrelated? Introducton Among the trade conflcts between Canada and the Unted States, the softwood lumber dspute has been the most mportant n terms of trade volumes and values, complexty, procedure, poltczaton, and duraton (Zhang 1997, Gagne 1999). Softwood lumber s one of the largest commodtes produced by the forest ndustry n both countres. In 2001, Canada exported mllon cubc meters of softwood lumber to the U.S., accountng for 34 percent of total U.S. softwood lumber consumpton. Ths consttuted 85 percent of all softwood lumber exports from Canada and 93 percent of all softwood lumber mports nto the U.S. Wth the expraton of the Softwood Lumber Agreement that regned for fve years from 1996 to 2001, the softwood lumber trade confrontaton between U.S. and Canada has entered a new phase. In Aprl 2001, some U.S. lumber producers fled complants to the U.S. Department of Commerce, clamng that Canadan softwood lumber mports hurt them. A key argument used by these producers was that U.S. and Canadan softwood lumber are substtutes. Ther statstcal evdence was that the prce correlaton between spruce-pne-fr (an aggregate product class consstng of Pcea spp., Pnus spp., Abes spp.) (SPF, the man softwood lumber speces group mported from Canada) and southern yellow pne (manly, Pnus echnata Mll., P. taeda L., P. ellot Engelm.) (SYP, the largest sngle speces groups produced n the U.S.), was around 0.8 based on hstorcal data (Ragosta et al. 2000). More sophstcated emprcal analyses of lumber demand (e.g., Buongorno et al. 1988, Lewandrowsk et al. 1994) also lend credence to ths argument. In a latest development, on May 2, 2002, the U.S. Internatonal Trade Commsson voted unanmously that the U.S. lumber producers were threatened wth njury from lumber mported from Canada, clearng the way for new dutes. The dutes, averagng percent ad 3

4 valorem, are large, compared to dutes on other mported countres and zero tarffs on most other forest products from Canada. U.S. consumer groups and the Canadan lumber ndustry, n contrast, clam that Canadan lumber mports and softwood lumber produced n the U.S. are not substtutes and that Canadan mports meet U.S. consumer demand and do not njure U.S. lumber producers (ACAH 2002, NAHB 2002). The Natonal Assocaton of Home Bulders (NAHB) argues that the dfferent types of lumber for house framng are not nterchangeable and that one wood product cannot be substtuted for another wthout sgnfcantly harmng U.S. consumers by forcng them to spend more on a house. That s, because SPF from Canada s better suted for wall framng and SYP s best used for beams and josts, they may, n fact, even be complements. Hence, rasng tarffs on mported SPF would adversely affect demand for both SYP and SPF. Southern pne domnates outdoor wood constructon applcatons such as deckng and fencng, and t s used extensvely n large dmenson (tmber framng) applcatons. Whle SPF s the largest consumed lumber product n the U.S., t s also the least domestcally produced and largest mported lumber product (Fgure 1). On the other hand, SYP s largely domestcally produced. Canadan SPF s the preferred wood for wall framng due to ts workablty, strength, stablty and low densty. Indeed, both speces groups are used n the constructon ndustry, often n the same buldng. The purpose of ths paper s to look at the ssue of substtutablty and complementarty between varous softwood lumber products n U.S. consumer demand wth emprcal analyss based on economc theory. Insghts generated from ths study could help nform debates n the current U.S.-Canada softwood lumber trade dspute and mght have mplcatons on the 4

5 compettveness of U.S. lumber ndustry and forest based economc development n the dfferent regons of the U.S. Earler studes conclude that the exchange rate nfluences the share of Canadan lumber n the U.S. lumber markets (Adams et al. 1986) and that Canadan lumber mports do not affect U.S. lumber prces (Buongorno et al. 1988). Other studes (e.g., Wear and Lee 1993, Zhang 2001) address the welfare consequences of varous trade restrctons on Canadan lumber. Studes examnng speces substtuton focus on tropcal and temperate sawlog mports n Japan (Vncent et al. 1991), tropcal log mports from dfferent world regons n Japan (Vncent et al. 1990), and domestc and mported ndustral woods n the 36 most mportant wood-mportng countres (Uusvuor and Kuuluvanen 2001). Hseu and Buongorno (1993) fnd sgnfcant elastcty of demand for Canadan lumber mports wth respect to U.S. domestc lumber prce, suggestng the possblty of substtuton. However, Hseu and Buongorno (1993) concentrate on substtuton among softwood lumber speces mported from Canada, not total U.S. consumpton, and ther data were pror to Lewandrowsk et al. (1994) examne the substtutablty among mported lumber speces and domestc lumber speces n the U.S. That study, also based on data from the 1970s and 1980s, does not dstngush between treated and untreated SYP and therefore does not address the perod after the emergence of large volumes of treated SYP (treated SYP producton ncreased ffteen-fold between 1977 and 1987, to a level where t essentally remans today). Nether study could have ncorporated nformaton related to hgher prces for naturally durable western speces such as redwood (Sequoa sempervrons [D. Don.] Endl.) (Olson et al. 1988, Berck and Bentley 1997) or the evoluton n buldng technologes that nvolve greater use of structural panels. Ths study, n contrast, s based on later data and ncludes all mportant 5

6 domestc and mported speces groups as well as structural panels and other engneered wood products. The rest of the paper s organzed as follows. In the next secton, the theory on whch our analyss s based and the emprcal specfcaton are descrbed. In the thrd and fourth sectons, data used n ths analyss and emprcal results are presented. The fnal secton draws some conclusons and polcy mplcatons. Methods In order to capture a full pcture of softwood products market, we nclude wood structural panels (WSP) whch nclude softwood plywood, OSB, and waferboard n our analyss. Followng Fuss (1977) and Uusvuor and Kuuluvanen (2001), we adopt a two-stage approach n our analyss. Frst, the aggregate producton functon of softwood utlzng ndustres (housng constructon and remodelng and repar ndustry) has several major nput categores, captal (K), labor (L), energy (E), softwood products (S), and other materals (M); each may consst of several components. Consderng that the softwood utlzng ndustres can choose among a number of softwood speces or products, S, = 1, 2,, N, and assumng a homothetc weakly separable producton technology, the underlyng producton functon can be wrtten as [ 1] Y = f [ S( S1, S 2,..., S N ), K, L, E, M ] where Y denotes the gross output of the softwood products utlzng ndustres, S, the total consumpton of softwood products, s an aggregate functon for softwood products. The correspondng cost functon wll be S 1 S2 [ 2] C = g[ P ( P, P S,..., P S N ), P K, P L, P where C denotes the total cost of the softwood utlzng ndustres, P S, P K, P L, P E, and P M are E, P M, Y ] respectve prces for nputs S, K, L, E, and M. Snce n equaton [2] P S, = 1, 2,, N, s the 6

7 prce per unt of softwood products, t s also the cost per unt of softwood products to the softwood utlzng ndustres. Second, assumng weak separablty, an unrestrcted subcost functon for the softwood product component of the producton functon can be wrtten as S 1 S2 [ 3] C = h( P, P S,..., P S N, Q) where C S s the total cost of softwood product component, S P ( =1, 2,, N) are prces for softwood products, and Q s aggregate softwood products consumpton by softwood utlzng ndustres. Ths approach of subcost functon permts us to study the structure of substtuton between varous softwood products ndependently of the other nputs. The least restrctve approach n the emprcal analyss s to assume a translog cost functon. It does not requre a pror restrctons on substtuton possbltes. The translog functon can be regarded as a quadratc approxmaton to the unspecfed true cost functon and s wrtten as a Taylor seres expanson to the second term of a twce-dfferentable analytc cost functon. Snce a subcost functon approach s used, ths s termed a translog subcost functon. After accountng for the effect of dfferent tme perods (explaned below) and addng perod dummy varables T 2 and T 3 to [3], ths functon takes the form: [4] ln C S = α j 0 j + α T2 ln P T S 2 + α ln P T3 S j T 3 + α + Q Q 1 ln Q + α 2 ln Q ln P S QQ (ln Q) + T2 2 + α ln P T ln P 2 S S + T3 T ln P 3 S where C S s the total cost of softwood products n the softwood utlzng ndustres; softwood product prces, = 1 to 6 (1 = SPF, 2 = SYP-untreated, 3 = SYP-treated, 4 = Douglas fr, 5 = WSP, and 6 = other speces); T 2 s from Sep 1994 to Mar 1996 and Apr to Jul 2001 (the Free Trade Perod), 0 otherwse; T 3 s 1 from Apr 1996 to Mar 2001 (the SLA perod), 0 S P are 7

8 otherwse; Q s the quantty of softwood products consumed by the softwood utlzng ndustry; and α and are coeffcents. Perod 1 (T 1 ) s from January 1989 to August 1994, durng whch the Memorandum of Understandng (MOU), the nterm duty, and the countervalng duty were n force. T 1 s omtted from equaton [4], and the effect of ths perod s captured by the ntercept. These three perods are dentfed to capture the effect of varous phases of US-Canada softwood lumber trade relatonshp on substtuton and demand for softwood lumber. Accordng to Shephard s lemma, cost-mnmzng behavor mples that the demand functons for the ndvdual softwood lumber speces, n terms of cost shares n the softwood lumber aggregate cost, can be expressed as S ln C S S j [ 5] S = m = α + j ln P + Q ln Q + T T + T ln P j = 1,2,..., N When [5] s estmated subject to constrants mposed by neoclasscal producton theory, the structure of substtuton and prce elastctes of demand for softwood products n the U.S. demand can be obtaned gven any set of relatve prces and total softwood product consumpton. Accordng to neoclasscal producton theory, lnear homogenety requres the followng restrctons: [6. a ] α = 1 [6. b] j j = j = 0 [ 6. c] j = j, j [ 6. d] = 0, = 0, = 0 = 1, Q T 2,..., N. 2 T3 These restrctons ensure that the cost functon s homogeneous of degree one n prces and that the Hessan matrx of the cost functon s symmetrc. Symmetry and lnear homogenety condtons ensure that the sum of factor shares n total cost adds up to unty. 8

9 The system of cost share equatons [5] contans all the nformaton needed to estmate the structure of substtuton and prce elastctes of demand. Allen partal elastctes of substtuton (AES) can be calculated as (Bnswanger 1974) 1 2 A + m - m σ = = 1,2,..., N m [ 7] 2 [ 8] σ A j = j + m m m m j j, j = 1,2,..., N; j The elastcty of substtuton measures the percent change n the nput use rato for a percent change n nput prce rato. It represents the ease wth whch a producton factor substtutes or complements for another n the producton process. A postve sgn ndcates substtutablty and a negatve sgn complementarty. Own and cross-prce elastctes of demand can be calculated as follows [ 9] 2 + m - m A η = mσ = = 1, 2,...,N m + m m A j j [ 10] ηj = m jσ j =, j = 1,2,..., N; j. m The own- and cross-prce elastctes of demand measure the percentage change n a factor use for a percent change n the own-prce or the prce of another good. Agan, a postve sgn ndcates substtutablty and a negatve sgn complementarty. Data Data on total softwood products consumpton n the U.S. were compled based on an dentty: consumpton = producton + mports exports. Monthly data on lumber producton were obtaned from the Amercan Forests and Paper Assocaton (AF&PA). However, AF&PA does not gve any speces break-down. The monthly shares of lumber producton by speces were 9

10 estmated by nstruments. 2 Import and export data for softwood lumber were taken from the U.S. Internatonal Trade Commsson s (USITC) Data Web ( From 1989 onward, the U.S. has used the Harmonzed Tarff Schedule (HTS) system of classfcaton. 3 Data pror to 1989 were not consdered due to changes n the classfcaton system n Intally, sx lumber speces groups, spruce-pne-fr (SPF), southern yellow pne (SYP), Douglas-fr, hem-fr, cedar, and an aggregate of other softwood speces, were determned on the bass of avalable nformaton from USDC classfcaton and USITC codes descrptons (Table 1). Southern yellow pne producton data were further dvded nto treated (SYP-R) and untreated (SYP-U) groups because these have dstnct uses and applcatons. In the absence of monthly data for treated SYP, annual and quarterly treated SYP producton statstcs were obtaned from Southern Forest Products Assocaton and monthly seres were constructed usng sutable procedures: a ftted regresson, extrapolaton, and apportonment. 4 Annual producton data for structural panels (plywood, OSB and wafer board) were obtaned from the Engneered Wood Assocaton, Tacoma, WA. Usng monthly proportons of producton from unpublshed monthly data for structural panels, monthly producton numbers were estmated. The monthly mport and export statstcs for structural panels were drawn from the USITC Data Web. Dmenson lumber prce seres from Random Lengths Yearbooks were used n the analyss 5 (Table 2). Sx groups of softwood products were constructed for the purpose of ths analyss SPF, SYP-U, SYP-R, DF, WSP, and other. Cedar, hem-fr groups, and other lumber were ncluded n the other group, and a volume weghted average prce for cedar, hem-fr and other speces was calculated and used n the analyss. Plywood and OSB were combned nto a WSP group, and a volume-weghted average prce for plywood and OSB was used n the 10

11 analyss. 6 Fgure 2 shows how the softwood products consumpton n the U.S. changed durng the years 1989, 1995, and Partcularly notceable s the fact that consumpton of SPF and WSP dramatcally ncreased, whle that of other speces and Douglas-fr declned. 7 Emprcal Results Most parameters of nterest can be obtaned from the sx cost share equatons [5] wthout consderng the total cost functon. But ths procedure neglects the nformaton contaned n the total cost functon, snce coeffcents α Q, α QQ, and coeffcents for perod dummes ( α T2, and α T 2 ) cannot be estmated wthout ncludng the total cost functon n the estmaton. The optmal procedure s to jontly estmate total cost functon and cost share equatons. Snce errors are contemporaneously correlated, maxmum-lkelhood estmates for seemngly unrelated regresson equatons (SURE) were obtaned. Because the cost shares should sum to one, to avod the problem of sngularty of covarance matrx, one cost share equaton (the other group) was dropped from estmaton. The remanng fve cost share equatons were estmated by normalzng the prces wth respect to the prce of the dropped equaton. Endogenety of Q and S P can be a problem n the model, as there may be a smultaneous equatons bas n the parameters estmated by ordnary least squares. The alternatve to ths s to use ether teratve Zellner-effcent estmaton (IZEF) or maxmum lkelhood estmaton (MLE) methods, whch are computatonally equvalent (Berndt and Chrstensen 1973, p.87-89). Snce a maxmum lkelhood estmaton procedure s used, the parameter estmates are also nvarant to the dropped equaton (Greene 1995, p ). The parameter coeffcents and ther standard errors for the dropped cost share equaton were estmated ndrectly usng the procedure outlned n Berndt (1991, p. 473). Estmated Cost Functon 11

12 Table 2 reports the estmated cost functon coeffcents and ther standard errors. The functon s lnear homogeneous n nput prces because of the restrctons mposed n [6.a] to [6.d]. A well-behaved cost functon s concave n the factor prces, and ts factor demand functons are strctly postve for postve nput prces and a postve output level. However, the translog cost functon does not satsfy these propertes globally, but can satsfy them locally. To ensure ths, the cost functon should be monotoncally ncreasng and strctly quas-concave n nput prces and level of output. We found that the monotoncty condton s satsfed at each monthly observaton. The quas-concavty condton s satsfed f the n x n matrx of substtuton elastctes s negatve sem-defnte (Baardsen 2000, p.539). We checked the 6 x 6 matrx of elastctes of substtuton usng both egenvalue and prncpal mnor methods and found t to be negatve sem-defnte. 8 Therefore, the estmated cost functon s well behaved and consstent wth economc theory. 9 Table 3 shows that 24 out of 40 drectly estmated coeffcents are dfferent from zero at 5% sgnfcance or better. The ntercept coeffcent for the second perod (coverng Free Trade perod), α T2 (-0.09), s negatve and sgnfcant at the 1% level, suggestng that the softwood costs to softwood utlzng ndustres decreased sgnfcantly n the free trade perod over the frst perod (MOU and dutes perod). The ntercept coeffcent for the thrd perod (coverng the SLA perod), α T3 (0.009), s postve but nsgnfcant, ndcatng that overall effect of the SLA was smlar to the ncreased levels of softwood costs experenced durng the tme of the MOU and nterm dutes. The nteracton terms between perod dummes and the softwood product prces show that SPF ( = -0.08) and SYP-R ( R = -0.04) consumpton declned sgnfcantly n the S T 2 second perod, and consumpton of structural panels ( T 2 E T 2 = 0.17) ncreased sgnfcantly. In the 12

13 thrd perod, SPF ( = 0.04) consumpton ncreased and Douglas-fr ( D = -0.01) S T 3 consumpton decreased. The effect of ncreased consumpton of dfferent softwood products on the total cost s captured by the coeffcents of nteracton terms between ther prces and total quantty of softwood consumpton (Q). The magntudes of the coeffcents for nteracton between softwood product prces and the total quantty consumed also ndcate the percentage change n the demand for the respectve softwood products when the total quantty of softwood consumpton ncreases by 1%, holdng other softwood product prces constant. Elastctes of Substtuton The Allen partal elastctes of substtuton and ther standard errors, calculated at the mean monthly levels of estmated cost shares usng equatons [7] and [8], are presented n Table 4. All own elastctes of substtuton (dagonals) have negatve sgns, ndcatng that a curvature condton for the total cost functon s satsfed. The results n the frst column of Table 4 show the relatonshp between the man mported softwood product group SPF, and the remanng softwood product groups. The SPF product group s unrelated to treated SYP (0.03), Douglas-fr (0.24), and other speces (0.31) groups as ndcated by nsgnfcant partal elastctes of substtuton. However, sgnfcant substtuton relatonshps exst between pars of SPF and untreated SYP, and SPF and engneered wood products. Specfcally, a one percent ncrease the rato of the SPF prce to that of untreated SYP would ncrease demand for untreated SYP by 1.46%, whle a smlar change wth respect to WSP would ncrease demand for WSP by 1.27%. The negatve elastcty of substtuton (-0.97, sgnfcant at the 10% level) between untreated and treated SYP suggests that they are complements. Lookng at the substtuton elastctes between untreated SYP and WSP (1.63), and between untreated SYP and SPF (1.46), T 3 13

14 t appears that untreated SYP s facng more competton from structural panels rather than from SPF. A test of equalty of substtuton elastctes between untreated SYP and SPF on the one hand, and untreated SYP and WSP on the other, usng a procedure for testng a lnear restrcton outlned n Greene (1993, pp.187) s rejected at the 10% sgnfcance level aganst the alternatve hypothess that the two elastctes of substtuton are not equal. 10 It s also notable that structural panels are provng to be the sgnfcant substtutes for untreated SYP, SPF, DF, and other lumber speces groups. Our results may not be drectly comparable to prevous studes because of dfferences n methodologes and tme perods used. These fndngs SPF s a sgnfcant substtute only to untreated SYP and engneered wood products, but unrelated to treated SYP, DF, and other speces comprsng about 71 percent of domestc softwood lumber producton or 76 percent of domestc softwood lumber consumpton (excludng SPF and engneered wood products) n the U.S. n 2001 dffer from Lewandrowsk et al. (1994) who fnd that the Canadan lumber s a substtute for softwood lumber n all regons (thus mplyng all speces) n the U.S. It s partcularly nterestng to note how the partal elastctes of substtuton between SPF (whch s manly mported from Canada) and the other lumber speces/products change over tme. Table 5 shows these partal elastctes of substtuton estmated usng average annual shares of varous lumber speces. The column labeled SPF*SYP-R gves the story of transformaton of the relatonshp between SPF and treated SYP from beng one of substtutes as ndcated by postve and sgnfcant partal elastctes of substtuton durng to beng unrelated or even moderate complements after Thus, although these speces groups were unrelated n the whole study perod, ther relatonshp changed from one of substtuton to complementarty over the study perod. 14

15 Prce Elastctes of Demand All of the own-prce elastctes of demand for softwood products, whch are shown on the dagonal of Table 6, have the correct sgn. Own-prce elastctes of demand for SPF, untreated SYP and other speces are less than one, whle those for treated SYP, DF and WSP exceeded one. These own-prce elastcty estmates are comparable wth Lewandrowsk et al. (1994) elastcty estmates of for Canadan lumber (manly SPF), and for southern pne (whch was manly untreated n the length of ther dataset). The own-prce elastcty of demand for structural panels (WSP) s comparable to the McKllop et al. (1980) estmate of for plywood, Spelter s (1984) estmates of 0.83 for plywood and to for structural partcleboard, and Adams and Haynes (1996) estmate of for OSB. The cross-prce elastctes of demand between all pars of softwood products are less than one. The Douglas fr and untreated SYP par has the hghest cross-prce elastcty of demand (0.97). If we look at the man softwood products ndvdually, a 1% ncrease n the prce of SPF lumber results n the largest ncrease n the quantty demanded of untreated SYP (0.40%), followed by WSP (0.35%), other thngs beng equal. Smlarly, a 1% ncrease n the prce of untreated SYP produces the largest ncrease n the quantty demanded of structural panels (0.26%), followed by SPF (0.24%), but also produces a largest decrease n the quantty demanded of treated SYP (-0.16%, sgnfcant at the 10% level). On the other hand, a one percent ncrease n the prce of structural panels would result n the ncrease n the quantty demanded of Douglas fr, untreated SYP, other speces, SPF, and treated SYP n that order. Conclusons and Implcatons We sought to determne whether SPF lumber manly mported from Canada s a substtute or a complement to softwood lumber produced n the U.S., wth a partcular focus on 15

16 southern yellow pne, whch has emerged snce the 1980s as encompassng two dstnct products. Monthly U.S. consumpton data from January 1989 to July 2001 on sx groups of softwood products were analyzed by estmatng the restrcted translog cost functon and cost share equatons smultaneously usng a maxmum lkelhood procedure. The results show that the man mported Canadan speces group SPF s behavng lke an ndependent market and s largely unrelated to the domestcally produced treated southern yellow pne, Douglas-fr, and other speces groups. Interestngly, the relatonshp between SPF and treated SYP lumber has changed over tme from beng substtutes to beng unrelated or even complements. However, the SPF group competes wth the domestcally produced untreated southern yellow pne and structural panels and thus appears to be a sgnfcant substtute for these products n the U.S. In addton, untreated southern yellow pne s facng more severe competton from structural panels than from SPF mports from Canada. Jont analyss of three tme perods ndcates that the softwood costs to the U.S. softwood utlzng ndustres were sgnfcantly lower n the free trade perod than n the SLA perod and the perod when the MOU, nterm duty, and countervalng dutes were n force. The lower costs n free trade mght have happened because the larger volume coupled wth lower prces of Canadan SPF mports competed not only wth untreated SYP lumber, but also wth engneered wood products. Wth the mposton of the MOU, CVD and nterm dutes, and SLA, we found, hgher costs regned. The fndngs of Wear and Lee (1993), Lndsay et al. (2000), and Zhang (2001) that softwood lumber costs have ncreased for U.S. consumers support the results of ths analyss. The polcy mplcatons of ths study are four-fold. Frst, the mported softwood product (SPF) appears to be unrelated to some 71 percent of domestcally produced softwood lumber 16

17 whch ncludes treated SYP, Douglas-fr, and other speces n the U.S. SPF mports are fulfllng the ncreasng needs of softwood lumber for expandng U.S. housng constructon, remodelng and repar ndustry n the wake of restrctons on tmber harvestng on federal lands n PNW and of reduced supply from other regons n the U.S. The results support the contenton of U.S. consumer groups that Canadan lumber mports are specfcally demanded n the constructon of new homes and the upgrade and repars of exstng homes n Amerca. They also valdate the argument that, to some degree, Canadan mports are not substtutes for U.S. producton. On the other hand, Canadan mports are substtutes for untreated southern yellow pnes, mplyng that some domestc lumber producers are vulnerable to Canadan lumber mports. 11 Second, gven that SPF s a substtute for and drectly competes wth untreated southern yellow pne, t s not surprsng to see that major poltcal advocates for restrctng Canadan lumber mports are from the U.S. South. Even so, structural panels compete wth both untreated SYP and SPF, and untreated southern yellow pne s facng greater competton from these structural panels than from SPF. These facts were also noted by some consumer advocates who ndcated that the prncpal compettve threat to the use of SYP comes not from mports, but from engneered wood products such as wood I-josts and composte materals (Rayburn 2002). Although we dd not nclude wood I-josts and composte materals but ncluded only plywood and OSB n the wood structural panels group n the analyss, the results ndcate competton from the engneered wood products to untreated SYP, SPF, and the other speces product group, n that order. Perhaps, then, attractng more engneered wood products mlls to the U.S. South could help allevate the economc development problem assocated wth ncreasng softwood lumber mports. 17

18 Thrd, U.S. lumber producers efforts to seek a change n trade polcy as a means of obtanng a reducton n the share of Canadan mports and thereby ncreasng the share of domestcally produced lumber n the U.S. lumber consumpton have proved effectve n the wake of strong demand for housng n the U.S. n the 1990s. The U.S. softwood lumber producton ncrease of 14% from 1995 to 1999, roughly concdng wth a perod of the SLA, ndcates a trade polcy-nduced ncrease n the compettveness of the U.S. lumber ndustry. These ncreased shares surely have led to greater benefts enjoyed by producers of untreated southern pne lumber and manufacturers of engneered wood products. The Canadan share of U.S. softwood lumber consumpton, on the other hand, has been steady, and Canadan export of softwood lumber to the U.S. ncreased by 7.5%, n spte of the SLA. However, ths strategy of protectng the domestc sector could prove detrmental to secondary U.S. softwood manufacturers and to hardwood furnture manufacturers when ther products do not drectly compete wth Canadan lumber mports. Gven that demands for those products are postvely related to housng demand and gven that housng demand s negatvely affected by hgher nput prces, hgher Canadan spruce-pne-fr mport prces lead to lower demand for other products (e.g., Adams 2003). Fnally, a more recent tarff mposed on Canadan lumber mports (n May 2002) has not been successful n lftng lumber prces n the U.S. (although the counter factual absent the tarff, some domestc prces mght have been lower has not been proven). Further, Canada has taken the softwood lumber dspute to World Trade Organzaton and North Amercan Free Trade Agreement, leavng the ssue of future trade restrctons uncertan. Wth these knds of clouds and uncertanty stll loomng, and gven our new emprcal results showng only ncomplete competton between SPF and domestcally produced lumber, U.S. wood product producers may 18

19 proftably adopt an alternatve strategy. Ths would nclude cost cuttng, greater nvestment n research and development and n new wood engneerng technologes, and other nwardly focused nvestment approaches, yeldng more certan long run returns than those obtaned by head-to-head product competton and controversal short-run polcy fxes. In the near-term, trade actons aganst Canadan lumber could beneft some lumber producers and even enhance the proftablty of some n the U.S. lumber ndustry. However, the benefts enjoyed by the ndustry from such trade actons are lkely to be outweghed n economc terms by the costs of such actons on the aggregate U.S. economy. 19

20 End Notes: 1 We have chosen the Allen elastcty of substtuton for ts popularty and also because t can dstngush complementarty from substtutablty between nputs more frequently than other forms of elastcty of substtuton. Other measures of elastcty of substtuton nclude the Morshma elastcty of substtuton and shadow elastcty of substtuton (Chambers 1988). However, Fuss et al. (1978) argue that there s no omnpotent measure of elastcty of substtuton and the selecton of a partcular defnton should depend on the queston asked. 2 The monthly shares of lumber producton by speces were estmated by nstrumental varables, wth nstruments of housng starts, U.S. gross domestc product n 1996 real dollars, the U.S.- Canada exchange rate, speces-specfc lnear tme trends, and a lagged dependent varable (lagged share) as ndependent varables. The explanatory powers (R 2 s) of these equatons were: spruce-pne-fr (82%), southern yellow pne (93%), Douglas-fr (90%), hem-fr (80%), cedar (63%), and other speces (90%). Seasonalty n producton was ncorporated nto these estmates by usng monthly seasonalty varatons observed for all softwood lumber. The seasonal varatons were obtaned by estmatng an ordnary least squares equaton of total softwood lumber producton as a functon of the lagged dependent varable, monthly dummes, a lnear tme trend, housng starts, U.S. gross domestc product n 1996 n real dollars, and the U.S.- Canada exchange rate. The results obtaned usng these nstrumented and seasonalty-augmented shares were smlar to those that we found when we mputed monthly producton data for dfferent speces by smply assumng that the speces proporton was the same for all months n a partcular year. Addtonal detals of the nstrumented estmated equatons are not provded here to conserve space but can be obtaned from the authors upon request. 20

21 3 Under the HTS system, export and mport data pertanng to10 dgt HTS codes to belongng to varous softwood lumber speces were compled. Data were for the perod from January 1989 to July The quarterly treated SYP producton statstcs were avalable for the years 1993 to Snce the data showed apparent seasonalty, a regresson equaton, ncludng a tme trend and seasonal dummy varables, was ftted on the quarterly proportons of the treated SYP producton data. The ftted equaton was: *T *D *D *D 3, (adjusted R 2 =0.90) where T s tme n years startng 1993=1, D 1 to D 3 are dummy varables representng quarters 1 to 3, and the ntercept representng the quarter 4. The predcted quarterly proportons for the years 1989 to 1992 were determned by extrapolaton usng the predcted regresson equaton, and quarterly producton numbers for treated SYP were calculated from the annual treated SYP producton data. From these quarterly numbers, monthly producton data for treated SYP were estmated usng the monthly proportons of total SYP producton. Snce SYP lumber mports were a tny percent of total SYP lumber consumed, mports of treated SYP were assumed to be nl. The monthly export quanttes for treated lumber were taken from ITC Data Web. The monthly export numbers for treated lumber for all speces were reported under HTS speces code (1989 to 1996) and (1997 to 2001). Out of ths exported treated lumber, 85 percent was assumed to be of SYP treated lumber (GC&A 1990). 5 The results were dentcal wth nomnal or real prces. Augmented Dckey-Fuller tests (the t- test and the rho-test) were conducted by usng a general-to-specfc model selecton strategy, begnnng at 18 lagged dfference terms and successvely reduced n order untl the mnmum of the Akake Informaton Crteron was found (see Hall 1994). Both the t-test and the rho-test produced results that rejected the unt root null at 10 percent nomnal sgnfcance. Ths was true 21

22 for both deflated and undeflated prce seres. Addtonally, the more powerful test aganst a false unt root null, the rho-test, typcally rejected the unt root null at 5 percent nomnal sgnfcance. 6 The basc reported unts for plywood and OSB were MSF (thousand square feet) on a 3/8 bass. To ensure unformty wth the unts of lumber n MBF n the dependent varable n the total cost functon, plywood and OSB quanttes were converted nto MBF usng converson factor MSF MBF -1. For mport and export data whch were reported n cubc meters, the converson factors used were MBF m -3 for softwood lumber, and MSF m -3 of 3/8 bass for plywood and OSB. 7 These changes can be attrbuted to harvest reductons n the U.S. Pacfc Northwest, an ncreasng share of small dameter logs utlzed n mlls, and the contnued strength of the U.S. housng market. 8 The egenvalues were: 0.00, -2.11, -3.29, -4.78, -7.52, and , and the determnants of the prncpal mnors were: -2.28, 8.01, , , , and The results were hghly senstve to specfcaton. Other specfcatons were conducted, but none of them could satsfy the theoretcal propertes of the cost functon. Hence, the specfcaton n equaton [4] was retaned. 10 The calculated t-value s gven df = 149 (n-k = 151-2). 11 One revewer of ths paper asked that f SPF s unrelated to all these other lumber groups, how could SPF consumpton have ncreased at the same tme that U.S. output decreased. As output of U.S. softwood products declned (n the face of the strong demand descrbed) ther prces would go up. We have a three-part explanaton for ths phenomenon. Frst, there s no evdence that the producton or domestc consumpton n the U.S. have declned durng the study perod (1989 to 2001). In other words, an ncrease n SPF consumpton (and mports) can (and dd) occur 22

23 concurrently wth an ncrease n producton and consumpton of domestcally produced lumber, f the output effect on demand for the aggregate domnates the substtuton effect wthn the aggregate. Second, as mentoned, SPF s substtute for untreated SYP. If nvestment (and dsnvestment) n plants producng treated SYP s small or f there are no barrers to entry or ext n producng treated SYP from untreated SYP, producers could swtch back and forth between treated and untreated SYP. For example, f demand for treated SYP goes up due to an ncrease n housng starts, the prce of treated SYP wll go up. Ths means that, everythng else beng equal, the producton of untreated SYP wll go down. Less untreated SYP producton n that stuaton would therefore lead to hgher SPF mports. Thrd, ths study s based on margnal analyss. One cannot assume that a large (say 10-20%), abrupt decrease or ncrease n U.S. domestc lumber producton and consumpton would not have an mpact on SPF, snce we found that SPF mports are substtutes for 29% of domestcally produced lumber. 23

24 Lterature Cted Adams, D.M Market and resource mpacts of a Canadan lumber tarff. J. For. 101(2): Adams, D.M., and R.W. Haynes The 1993 tmber assessment market model: structure, projectons, and polcy smulatons. Gen. Tech. Rep. PNW-GTR-368. Portland, OR: U.S. Department of Agrculture, Forest Servce, Pacfc Northwest Research Staton. Adams, D.M., B.A. McCarl, and L. Homayounfarrokh The role of exchange rates n Canadan-Unted States lumber trade. For. Sc. 32: Amercan Consumers for Affordable Homes (ACAH) Home bulders, manufactured housng, lumber dealers and retal ndustry tell ITC they select lumber based on need not prce. Baardsen, S An econometrc analyss of Norwegan sawmllng based on mlllevel data. For. Sc. 46: Berck, P., and W.R. Bentley Hotellng s theory, enhancement, and the takng of the Redwood Natonal Park. Am. J. Agr. Econ. 79(2): Berndt, E.R The Practce of Econometrcs: Classc and Contemporary. Addson-Wesley, Readng, Mass. 702p. 24

25 Berndt, E.R., and L.R. Chrstensen The translog functon and the substtuton of equpment, structures, and labor n U.S. manufacturng J. Econometrcs. 1: Bnswanger, H.P A cost functon approach to the measurement of factor demand and elastctes of substtuton. Amer. J. Agr. Econ. 56: Buongorno, J., J.-P. Chavas, and J. Uusvour Exchange rates, Canadan lumber mports, and Unted States prces: A tme-seres analyss. Can. J. For. Res. 18: Chambers, R.G Appled Producton Analyss: A Dual Approach. New York: Cambrdge Unversty Press. 301p. Fuss, M.A The demand for energy n Canadan manufacturng: An example of the estmaton of producton structures wth many nputs. J. Econometrcs. 5(1): Fuss, M.A., D.L. McFadden, and Y. Mundlak A survey of functonal forms n the economc analyss of producton functons. P n Producton Economcs: A Dual Approach to Theory and Applcatons, Volume I: The Theory of Producton, Fuss, M., and D. McFadden (eds.). North Holland, Amsterdam. Gagne, G The Canada-U.S. softwood lumber dspute: An assessment after 15 years. J. World Trade. 33:

26 George Carter & Afflates (GC&A) Treated and plastc-composte lumber markets: 1990 Executve Summary. Greene, W.H Econometrc Analyss. New York: Macmllan Publshng Co., 791p. Greene, W LIMDEP User s Manual. Verson 7.0. Econometrc Software, Inc., New York. Hall, A Testng for a unt root n tme seres wth pretest data-based model selecton. J. Bus. Econ. Stat. 12(4): Hseu, J-S., and J. Buongorno Prce elastctes between speces n the demand for softwood lumber mports from Canada. Can. J. For. Res. 23: Lewandrowsk, J.K., M.K. Wohlgenant, and T.J. Grennes Fnshed product nventores and prce expectatons n the softwood lumber ndustry. Amer. J. Agr. Econ. 78: Lndsay, B., M.A. Groombrdge, and P. Loungan Nalng the Homeowner: The Economc Impact of Trade Protecton of the Softwood Lumber Industry. Center for Trade Polcy Analyss No. 11. CATO Insttute. Washngton, D.C. McKllop, W., T.W. Stuart, and P.J. Gessler Competton between wood products and substtute structural products: An econometrc analyss. For. Sc. 26:

27 Natonal Assocaton of Home Bulders (NAHB) Lumber tarffs wthout mert, bulders tell trade panel. March 23, Olson, K.W., R.L. Moomaw, and R.P. Thompson Redwood Natonal Park expanson: mpact on old-growth redwood stumpage prces. Land Econ. 64(3): Pndyck, R.S Interfuel substtuton and the ndustral demand for energy: An nternatonal comparson. Rev. Econ. and Stat. 61: Ragosta, J.A., H.L. Clark, C. Meacc, and G.I. Hume Canadan Governments should end Lumber Subsdes and Adopt Compettve Tmber Systems: Comments Submtted to the Offce of the Unted States Trade Representatve on Behalf of Coalton for Far Lumber Imports Pennsylvana Ave., NW., Washngton, DC. Random Lengths. Varous years. Forest Product Market Prces and Statstcs. Random Lengths Publcatons Inc., Eugene, OR. Rayburn, B Testmony before Senate Fnance Commttee. February 13, Spelter, H Prce elastctes for softwood plywood and structural partcleboard n the Unted States. Can. J. For. Res. 14:

28 Uusvuor, J., and J. Kuuluvanen Substtuton n global wood mports n the 1990s. Can. J. For. Res. 31: Vncent, J.R., A.K. Gandapur, and D.J. Brooks Speces substtuton and tropcal log mports by Japan. For. Sc. 36: Vncent, J.R., D.J. Brooks, and A.K. Gandapur Substtuton between tropcal and temperate sawlogs. For. Sc. 37: Wear, D.N., and K.J. Lee U.S. polcy and Canadan lumber: Effects of the 1986 memorandum of understandng. For. Sc. 39: Zhang, D The poltcal economy of U.S.-Canada trade n softwood lumber and newsprnt. Workng Paper, School of Forestry and Wldlfe Scences, Auburn Unversty, Auburn. Zhang, D Welfare mpacts of the 1996 U.S.-Canada softwood lumber (trade) agreement. Can. J. For. Res. 31:

29 20 16 Consumpton Imports Producton Bllon board feeet S-P-F SYP-U SYP-R DF WSP Other Fgure 1. Softwood products consumpton, producton and mports n the U.S.:

30 Bllon board feet S-P-F SYP-U SYP-R DF WSP Other Fgure 2. Softwood lumber products consumpton n the U.S.: 1989, 1995, and

31 Table 1. Softwood lumber speces groups dentfed by the U. S. Department of Commerce and the U. S. Internatonal Trade Commsson. Speces Group USDC Lsted Speces HTS Lsted Speces SPF spruce & fr SPF lodgepole pne lodgepole pne spruce stka spruce other spruce SYP southern pne southern yellow pne Douglas-fr Douglas-fr Douglas-fr Hem-fr hem-fr hem-fr whte fr fr other fr Cedar western red cedar western red cedar other cedar other cedar yellow cedar cedar Others ponderosa pne ponderosa pne other pne other pne redwood redwood eastern whte pne eastern whte pne other eastern softwoods eastern red pne western whte pne hemlock sugar pne larch other western softwoods pne other 31

32 Table 2. Softwood lumber products and the prces used n the analyss. Product SPF SYP-U SYP-R DF Hem-fr Cedars Other Plywood OSB Prce SPF, kln-dred, #1&2, Random, delvered to Great Lakes; Southern Pne (Eastsde), kln-dred, #2, Random, net f.o.b. mll; Southern Pne, treated, 2x4-12, #2, Random, net f.o.b. treatng plant; Douglas-fr, kln-dred, 2x4, Std&Btr, Random, net f.o.b. mll; Hem-fr (Inland-Spokane), kln-dred, 2x4, Std&Btr, Random, net f.o.b. mll; Western Red Cedar, green, 2x4, Std&Btr, Random, net f.o.b. mll; Fr & Larch, kln-dred, 2x4, Std&Btr, Random, net f.o.b. mll; Southern Plywood, (East) 23/32, Underlayment, C X-Band, T&G, net f.o.b. mll; OSB, Southeast, 23/32, T&G, net f.o.b. mll. 32

33 Table 3. Parameter estmates of translog subcost functon, U.S.: January1989 to July Parameter Estmate S.E. Parameter Estmate S.E. α ** α ** T 2 α T α Q α QQ α S α U α R ** α D * α E * ** SS ** SU ** SR ** SD ** SE ** UU ** UR ** UD ** UE RR ** RD ** RE DD ** DE EE ** ** SQ RQ DQ EQ T 2 S ** T 2 U T 2 R ** T 2 D T 2 E ** T 3 S ** T 3 U T 3 R ** T 3 D T 3 E Indrectly Estmated Coeffcents α UQ **, and * ndcate sgnfcance at the 1% and 5% levels, respectvely. S.E. s standard error. The O OO ** OS ** OU OR OD OE OQ T 2 O T 3 O subscrpts are: S = SPF, U = SYP-Untreated, R = SYP-Treated, D = Douglas fr, E = wood structural products, and O =other speces; Q = total quantty of softwood products; T 2 and T 3 are perod dummes. 33

34 Table 4. Allen partal elastctes of substtuton between softwood lumber products, U.S.: January 1989 to July SPF SYP-U SYP-R DF WSP Other SPF ** (0.80) SYP-U ** (0.092) ** (0.216) SYP-R (0.278) (0.552) ** (5.249) DF (0.149) (0.191) ** (1.941) ** (1.601) WSP ** (0.055) ** (0.083) (0.901) ** (0.605) ** (0.308) Other (0.175) (0.282) (1.928) (1.147) ** (0.512) * (0.971) **, * ndcate sgnfcance at the 1% and 5% levels, respectvely. Fgures n parentheses are standard errors: SE ( j ) SE( j ) / m m σ = (Bnswanger 1974, Pndyck 1979). j 34

35 Table 5. Allen partal elastctes of substtuton between SPF and other softwood lumber products: January 1989 July Year SPF*SPF SPF*SYP-U SPF*SYP-R SPF*DF SPF*WSP SPF*Other ** (0.082) ** (0.098) ** (0.160) (0.145) ** (0.065) * (0.169) ** (0.081) ** (0.098) * (0.176) * (0.134) ** (0.063) (0.173) ** (0.085) ** (0.104) ** (0.165) * (0.139) ** (0.065) (0.171) ** (0.083) ** (0.101) * (0.169) * (0.141) ** (0.067) * (0.165) ** (0.085) ** (0.104) ** (0.165) (0.154) ** (0.062) (0.171) ** (0.100) ** (0.111) (0.248) (0.164) ** (0.055) (0.195) ** (0.163) ** (0.147) * (0.745) (0.208) ** (0.051) (0.261) ** (0.079) ** (0.089) (0.488) (0.168) ** (0.048) (0.182) ** (0.064) ** (0.076) (0.595) (0.145) ** (0.050) * (0.167) ** (0.061) ** (0.069) * (0.961) ** (0.125) ** (0.054) * (0.161) ** (0.060) ** (0.069) (0.582) * (0.141) ** (0.055) * (0.155) ** (0.062) ** (0.071) (0.611) * (0.139) ** (0.055) * (0.155) ** ** * ** (0.103) (0.107) (0.739) (0.184) (0.047) (0.207) **, * ndcate sgnfcance at the 1% and 5% levels, respectvely. Fgures n parentheses are standard errors. 35

36 Table 6. Own and cross-prce elastctes of demand for softwood lumber products, U.S.: January 1989 to July % effect on For a 1% change n the prce of the quantty SPF SYP-U SYP-R DF WSP Other demanded of SPF ** ** ** (0.022) (0.015) (0.012) (0.014) (0.013) (0.035) SYP-U ** * ** (0.025) (0.035) (0.024) (0.018) (0.020) (0.056) SYP-R ** ** (0.076) (0.089) (0.234) (0.178) (0.211) (0.381) DF ** ** ** (0.040) (0.031) (0.084) (0.147) (0.142) (0.227) WSP ** ** ** ** ** (0.015) (0.013) (0.039) (0.056) (0.072) (0.101) Other ** * (0.048) (0.045) (0.083) (0.105) (0.120) (0.192) **, * ndcate sgnfcance at the 1% and 5% levels, respectvely. Fgures n parentheses are standard errors: η = (Bnswanger 1974, Pndyck 1979). SE ( j ) SE( j ) / m 36