Regions, resources, and economic geography: Sources of U.S. regional comparative advantage,

Size: px
Start display at page:

Download "Regions, resources, and economic geography: Sources of U.S. regional comparative advantage,"

Transcription

1 Regionl Science nd Urn Economics 29 (1999) 1 32 Regions, resources, nd economic geogrphy: Sources of U.S. regionl comprtive dvntge, Sukkoo Kim Wshington University in St. Louis nd NBER, Deprtment of Economics, Cmpus Box 1208, St. Louis, MO , USA Received 16 Decemer 1996; ccepted 1 Decemer 1997 Astrct This pper estimtes the Ryczynski eqution mtrix for the twenty two-digit U.S. mnufcturing industries for vrious yers etween 1880 nd As predicted y the stndrd generl equilirium theory of interregionl trde, the regression estimtes show tht consistent set of fctor endowments explins significnt mount of the geogrphic distriution of mnufcturing ctivities over time. Although these results do not rule out the importnce of incresing returns, they do suggest certin limits on how incresing returns ffect U.S. economic geogrphy Elsevier Science B.V. All rights reserved. Keywords: Ryczynski eqution mtrix; U.S. mnufcturing industries; Geogrphic distriution; U.S. economic geogrphy JEL clssifiction: F11; F12; R12; N7 1. Introduction One of the importnt fetures of industriliztion is the clustering of economic ctivities. Prior to industriliztion, production ws geogrphiclly dispersed due to the reltive intensive use of lnd in griculturl production. Industriliztion nd the growth of mnufcturing hve fundmentlly ltered the geogrphic ptterns of production s ctivities ecme concentrted in cities nd in regions. The cuses of the geogrphic concentrtion of industril ctivities ws systemticlly identified y Mrshll (1920). According to Mrshll, firms my choose to concentrte in / 99/ $ see front mtter 1999 Elsevier Science B.V. All rights reserved. PII: S (98)

2 2 S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32 given locle ecuse of informtion spillovers, vilility of specilized inputs, nd the pooling of the lor mrket for workers with specilized skills. These 1 forces re now known s the Mrshllin externlities or spillovers. However, it is lso importnt to note tht Mrshll identified nturl dvntges s one of the 2 chief cuses of geogrphic concentrtion. In recent rticle, Kim (1995) documents the long-run trends in U.S. regionl speciliztion nd industry locliztion to exmine which sources of geogrphic 3 concentrtion re most consistent with the dt. The dt revel some surprising nd interesting trends. Although the process of U.S. industriliztion etween the lte nineteenth nd the erly twentieth centuries coincided with drmtic increse in the geogrphic clustering of economic ctivities, the trend hs significntly reversed since the mid-twentieth century. Indeed, industries in the ggregte re less geogrphiclly concentrted thn they were during the mid-nineteenth century. The industry ptterns of locliztion re lso interesting. In generl, the dynmic trends nd cross-sectionl industry locliztion ptterns seem to e negtively correlted with mesures ssocited with high-tech industries. While the study of geogrphic concentrtion or regionl speciliztion is informtive nd extremely useful, these studies re suject to certin limits. The most serious prolem pertins to the lck of theoreticl justifiction for ny prticulr mesure of industry clustering. Consequently, it is difficult to decompose geogrphic concentrtion of industries into those cused y spillovers nd nturl dvntges. The recent work y Ellison nd Gleser (1997) ttempts to provide more theoreticlly motivted mesure of geogrphic concentrtion. An importnt feture of the Ellison Gleser mesure is tht it corrects for the 1 Recently, Krugmn (1991,) nd Dvid nd Rosenloom (1990) hve provided forml models of economic geogrphy sed on Mrshllin externlities. In urn economics, however, these types of models hve een in existence for some time. For exmple, see Muth (1963); Mills (1967, 1980); Henderson (1974, 1988); Fujit (1986, 1988); Berlint nd Wng (1993); Berlint nd Konishi (1994); Adel-Rhmn (1988); River-Btiz (1988), mong others. Also see Fujit nd Thisse (1996) for review of the literture. 2 Mrshll (1920, ) writes: Mny vrious cuses hve led to the locliztion of industries; ut the chief cuses hve een physicl conditions; such s the chrcter of the climte nd the soil, the existence of mines nd qurries in the neighorhood, or within esy ccess y lnd or wter. Thus metllic industries hve generlly een either ner mines or in plces where fuel ws chep. The iron industries in Englnd first sought those districts in which chrcol ws plentiful, nd fterwrds they went to neighorhood of collieries. Stffordshire mkes mny kinds of pottery, ll the mterils of which re imported from long distnce; ut she hs chep col nd excellent cly for mking the hevy sggrs or oxes in which the pottery is plced while eing fired. Strw pliting hs its chief home in Bedfordshire, where strw hs just the right proportion of silex to give strength without rittleness; nd Buckinghmshire eeches hve fforded the mteril for the Wycome chir mking. The Sheffield cutlery trde is due chiefly to the excellent grit of which its grindstones re mde. 3 Numerous studies hve exmined the phenomenon of geogrphic concentrtion. See Florence (1948); Hoover (1948); Perloff et l. (1960); Fuchs (1962); Krugmn (1991); Ellison nd Gleser (1997), nd Dumis et l. (1997), mong mny others. Most of these studies, however, cover reltively short term. Also see Kim (1998,).

3 S. Kim / Regionl Science nd Urn Economics 29 (1999) differences in the size of plnts nd for differences in the size of the geogrphic res. However, since the Ellison Gleser mesure is oservtionlly equivlent etween spillovers nd nturl dvntges, it still cnnot effectively distinguish etween these two sources of industry concentrtion. This pper ttempts to differentite etween the importnce of nturl dvntges nd spillovers over time y controlling for fctor endowments. More specificlly, this pper estimtes the Ryczynski eqution mtrix for the twenty two-digit U.S. mnufcturing industries for vrious yers etween 1880 nd Wheres the Heckscher Ohlin Vnek model of interregionl trde provides liner reltionship etween interregionl net exports nd fctor endowments, the Ryczynski theorem provides liner reltionship etween regionl production nd fctor endowments. Thus, in principle, the residul of the Ryczynski estimtes provides n upper ound estimte of the importnce of spillovers. The pper finds tht consistent set of fctor endowments, s predicted y the stndrd generl equilirium theory of interregionl trde, explins significnt mount of the geogrphic distriution of mnufcturing ctivities over time. However, the explntory power of fctor endowments declined slightly over time. Although the growth of the unexplined vrition my e ttriuted to the growing importnce of Mrshllin externlities or spillovers, this conclusion my not e wrrnted. Since the spillover effects re mesured s residul, it is difficult to scertin the exct cuses of this decline. The growth in the residul my e cused y the growing rndomness in the loction of mnufcturing ctivities s regionl differences in fctor endowments diminished over time or y the growth in the importnce of foreign trde in goods nd fctors. In ddition to reporting the Ryczynski regression estimtes for ech of the twenty 2-digit mnufcturing industries, the fctor intensities implied y the regression re compred ginst the independent estimtes of fctor intensities clculted from the ctul mounts of lor, cpitl, nd rw mterils used in mnufcturing. In generl, there is significnt correspondence etween the fctor intensities implied y the Ryczynski regressions nd the ctul fctor intensities. The pper is orgnized s follows. Section 2 presents forml frmework for the nlysis nd provides description of the dt sources. Section 3 presents the results of the Ryczynski regression estimtes. Section 4 exmines the chnges in the sources of U.S. regionl comprtive dvntge nd disdvntge over time. Section 5 concludes with summry. 2. Methodology nd dt This section provides the theoreticl frmework for the empiricl nlysis nd description of the dt used. The Heckscher Ohlin model of interregionl trde predicts tht region undnt in prticulr resource will produce nd export products which re reltively intensive in tht resource. However, due to the lck

4 4 S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32 of systemtic U.S. interregionl trde dt, the Heckscher Ohlin model cnnot e directly estimted. This pper exploits the Ryczynski theorem which reltes regionl production with regionl fctor endowments. The Ryczynski theorem sttes tht, t constnt commodity prices, n increse in the supply of fctor will led to n increse in the production of the commodity tht uses tht fctor intensely nd reduction in the production of other commodities. The Ryczynski theorem nd the other three core theorems of the generl equilirium trde theory, the fctor price equliztion, the Stolper Smuelson, nd the Heckscher Ohlin theorems, re sed on the following ssumptions: the numer of goods nd numer of fctors re equl, fctors of production move costlessly within region ut re completely immoile cross regions, commodities re freely moile cross regions, oth commodity nd fctor mrkets cler competitively, regions hve ccess to identicl technologies, fctor endowments 4 re reltively similr, nd consumers hve identicl homothetic tstes. Given these ssumptions, the following liner reltionship etween regionl output nd regionl fctor supplies cn e derived: 21 Y 5 A V (1) where Y5n31 vector of outputs, mtrix A is n n3m fctor input intensities 5 or the Ryczynski mtrix, nd V is n m31 vector of endowments. The nlysis of this pper is sed on smple of U.S. sttes in 1880, 1900, 1967 nd The dt on vlue dded for the twenty two-digit mnufcturing industries re from the U.S. Census of Mnufctures nd Niemi (1974). Since the stndrd industril codes (SIC) were not developed until the midtwentieth century, the census dt for 1880 nd 1900 need to e ctegorized. The 1900 census dt come from Niemi (1974) who ctegorized the dt using the 1963 census definitions. The 1880 census dt nd the 1900 dt for the Mountin nd Pcific regions were ctegorized using the 1972 census definitions nd Niemi s product list. Tle 1 presents the men nd stndrd devition of sttes mnufcturing vlue dded for the twenty two-digit industries. 4 See Lemer (1984) nd Wong (1995) for more detil. Wong (1995) demonstrtes the vlidity of the Ryczynski theorem with fctor moility, nd lso derives conditions under which the Ryczynski theorem holds with incresing returns. 5 In the interntionl context, following the clssic work y Lemer (1984), scholrs estimte the Heckscher Ohlin Vnek model which provides liner reltionship etween net exports nd fctor endowments. There re three types of studies tht re sed on the Heckscher Ohlin Vnek eqution: 21 Ti5A (Vi2V W) where Ti is mtrix of net exports. Fctor content studies regress net exports on fctor intensities to infer fctor endowments (see Wright, 1990); cross country studies regress net exports on fctor endowments to infer fctor intensities (see Lemer, 1984); nd multifctor studies estimte fctor contents of trde nd fctor undnces nd test their correltion (Bowen et l., 1987; Trefler, 1995). Empiricl studies sed on the Ryczynski eqution cn e similrly ctegorized. This pper long with Richrdson nd Smith (1995) nd Hrrign (1995) re cross country or cross-stte studies while Grimes nd Prime (1993) nd Dvis et l. (1997) re regionl multifctor studies.

5 S. Kim / Regionl Science nd Urn Economics 29 (1999) Tle 1 Men nd stndrd devition of vlue dded in mnufcturing: U.S. sttes, Vrile Men Men Men Men (SD) (SD) (SD) (SD) 20 Food 5.2 (9.6) 15.8 (27.3) (639.5) (2484.2) 21 Tocco 1.1 (2.6) 3.6 (6.7) 40.1 (159.0) (1022.9) 22 Textiles 5.9 (14.6) 8.6 (19.7) (367.7) (1249.7) 23 Apprel 2.9 (8.5) 9.6 (29.5) (439.8) (1011.1) 24 Lumer nd Wood 4.2 (6.1) 8.0 (8.3) (139.0) (677.1) 25 Furniture nd Fixtures 1.1 (2.2) 2.2 (4.3) 83.3 (117.0) (603.6) 26 Pper 0.7 (1.8) 2.0 (4.6) (212.0) (969.5) 27 Printing nd Pulishing 1.6 (4.0) 6.1 (14.3) (547.5) (2777.1) 28 Chemicls 1.7 (3.5) 3.7 (7.1) (611.5) (3182.2) 29 Petroleum nd Col 0.1 (0.3) 1.9 (4.1) (275.0) (830.1) 30 Ruer nd Plstics 0.2 (0.6) 0.8 (2.4) (225.0) (1044.5) 31 Lether 3.3 (7.8) 4.3 (9.7) 49.2 (87.3) 55.1 (113.1) 32 Stone, Cly nd Glss 1.7 (3.1) 6.2 (12.2) (210.0) (765.1) 33 Primry Metl 2.9 (8.8) 8.3 (25.5) (718.9) (1369.2) 34 Fricted Metl 3.1 (5.4) 5.0 (9.3) (569.3) (1976.0) 35 Mchinery 2.8 (5.5) 9.2 (18.2) (867.9) (3050.8) 36 Electricl Mchinery 0.04 (0.1) (764.0) (2709.9) 37 Trnsporttion 1.6 (2.7) 5.3 (7.8) (1057.5) (4650.2) 38 Instruments 0.4 (0.8) 0.6 (1.5) (335.4) (2696.8) 39 Miscellneous 1.5 (3.9) 3.9 (9.3) 93.8 (163.8) (472.7) N Notes: Vlue dded is in million dollrs. Sources: U.S. Census of Mnufctures, 1880, 1900, 1967, nd Dt on fctor endowments re derived from vriety of sources. The dt on lor nd cpitl re from Census of Mnufctures: lor is the totl mnufcturing employees nd cpitl is the totl mount of gross deprecile ssets. While lnd is often used s fctor endowment in interntionl textooks, the mount of lnd is unlikely to serve s meningful fctor endowment. Consequently, this pper utilizes the production of vrious extrctive industries s proxy for lnd. 6 These industries include griculture, tocco, timer, petroleum, nd minerls. Dt on griculture nd tocco re from the Census of Agriculture. The 1880 dt on timer re from the Census of Agriculture, the 1900 dt re from specil reports on select industries reported in Tle 2 in the Census of Mnufctures, vol. 3, pt. 3, nd the 1967 nd 1987 dt re from the Sttisticl Astrct of the United Sttes. The 1880 nd 1900 dt on petroleum nd minerls re from the Census of 6 For exmple, in 1929, the percentge distriution of rw mterils used in mnufcturing supplied y griculture, mining, forestry, fishing, hunting nd trpping were 67.4, 27.6, 3.8, nd 1.3 percent, respectively (see Thompson, 1933).

6 6 S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32 Tle 2 Men nd stndrd devition of fctors of production: U.S. sttes, Vrile Men Men Men Men (SD) (SD) (SD) (SD) Lor 58.1 (106.7) (175.5) (475.4) (416.1) Cpitl 63.2 (110.8) (345.8) (5188.1) ( ) Agriculture 47.9 (49.7) 95.7 (91.0) (822.8) (1539.8) Tocco 10.1 (28.1) 1.1 (3.1) 26.8 (87.3) 36.9 (126.9) Timer 3.0 (5.2) 4397 (6844) (85574) (80484) Petroleum 1.5 (4.4) (1140.9) (6552.5) Minerls 3.1 (10.1) 12.7 (32.4) (196.5) (1222.5) N Notes: In 1880, lor is employees in thousnds; cpitl, griculture, timer, nd minerls re in million dollrs; tocco is in million pounds. In 1900, 1967, nd 1987, lor is employees in thousnds; cpitl, griculture, tocco, petroleum, nd minerls re in million dollrs; nd timer is in million ord feet. Sources: Census of Mnufctures, 1880, 1900, 1967, Census of Agriculture, 1880, 1900, 1967, Census of Mines nd Qurries, 1880, Census of Minerl Industries, 1967, Sttisticl Astrct of the United Sttes. See text. Mines nd Qurries while the 1967 nd 1987 dt re from the Census of Minerl 7 Industries. Tle 2 presents the men nd stndrd devitions of the sttes fctor endowments. The Ryczynski equtions re estimted using ordinry lest squres nd the estimtes re djusted for heteroscedsticity using White s (1980) procedure. For ech of the twenty 2-digit mnufcturing industries, vlue dded is regressed ginst seven fctor endowments for 1880, 1900, 1967 nd 1987: 7 Although dt on minerls y different types such s fuel, stone, chemicl, nd metl minerls re ville, the ctegories were ggregted to reduce the mount of potentilly spurious correltions. In 1880, fuel minerls consist of nthrcite nd ituminous col, stone minerls consist of slte, silicious, sndstones nd mrle nd lime, nd metl minerls consist of iron ore, led ore, zinc ore, copper ingots, nd other minor minerls. The 1900 dt on minerls were more detiled. The metl ctegory contins copper ore, iron ore, led ore, zinc ore, nd mngnese ore; the fuel ctegory contins nthrcite col, ituminous col nd nturl gs; stone ctegory contins cement, cly, limestones nd dolomites, mrle, sndstones nd qurtzites, siliceous crystlline rocks, slte, uhrstones nd millstones, corundum nd emery, crystlline qurtz, grnet, grindstones nd pulp stones, infusoril erth, tripoli, pumice, oilstones, whetstones, nd scythestones; the chemicl ctegory contins orx, fluorspr, gypsum, phosphte rock, sulphur nd pyrite, tlc nd sop stones, rytes, nd minerl pigments; miscellneous ctegory contins sestos, sphltum nd ituminous rock, uxite, feldspr, flint, fuller s erth, gold nd silver, grphite, lithium ore, mrl, mic, monzite, precious stones, silic snd, tungsten, urnium nd vndium, nd ll other minerls. For more detil dt description for the 1967 nd 1987 dt, see the Census of Minerl Industries.

7 S. Kim / Regionl Science nd Urn Economics 29 (1999) Y 5 1 LABOR 1 CAPITAL 1 AGRICULTURE TOBACCO 1 5 TIMBER 1 6 PETROLEUM 1 7 MINERALS 8 1. (19) The reder should e wre of numer of potentil prolems with the dt nlysis. First, due to dt limittions, lor nd cpitl endowments re totls for 9 mnufcturing only. Second, the vlue of products for resource endowments 10 (excluding lor nd cpitl) re suject to prolems of doule counting. However, since doule counting is expected to occur evenly cross ll sttes, it should not systemticlly is the results. Third, there re omitted vriles such s wter, wter power, nd climte. The omission of wter supply is likely to e prolemtic for some industries. For exmple, in 1975 the chemicls, primry metls nd pper industries utilized 19.4, 18.9 nd 8.9 illion gllons per dy wheres most other industries used pproximtely 5 illion gllons per dy (U.S. Wter Resource Council, 1978, p. 45.) Fourth, the ssumption of closed economy distorts the results for industries which import significnt mounts of resources or finl goods. Despite these complictions, however, the regression estimtes offer n useful explntion for the geogrphic distriution of U.S. mnufcturing 11 ctivities over time. 8 From technicl stndpoint, the mtrix A must e squre nd the reltive input intensities for goods must e different efore the mtrix cn e inverted. If the mtrix is squre, then from the estimtes of 21 A, fctor intensities cn e recovered y inverting the mtrix. In this pper, the numer of goods is greter thn the numer of fctors so tht the fctor intensities cnnot e recovered. In principle, it is lwys possile to choose n equl numer of goods nd fctors since there is n ritrry element in the ggregtion of industries nd fctors of production. The fct tht the numer of goods is greter thn the numer of fctors implies degree of indeterminteness equl to n2m, ut the indetermincy cn e resolved y hypothesizing smll interregionl trnsporttion cost. See Lemer (1984), (16 18). 9 The totl mnufcturing cpitl nd lor is chosen for historicl comprility. Since most of the vrition in cpitl nd lor endowments will e in mnufcturing, the prolem is not likely to e severe. 10 As noted y Perloff et l. (1960), Perhps the most serious wekness of gross vlue concept, over nd eyond prolems of enumertion, occurs in reltion to griculture where, for exmple, considerle dupliction of vlues is entiled. The gross vlue of griculturl products includes the vlue of crops fed to livestock nd lso the vlue of livestock sold to frmers. Thus the dupliction of vlues is greter where crops re rised nd fed on frms thn in res where csh crops like whet nd cotton re rised (p. 616). 11 On the other hnd, there re severl dvntges to estimting the Ryczynski eqution using U.S. regionl dt rther thn estimting the Heckscher Ohlin Vnek eqution using interntionl dt. First, some of the ssumptions of the model, tht regionl fctor endowments re in the sme cone of diversifiction, tht residents hve identicl homothetic tstes, nd tht firms hve ccess to identicl technologies, re more likely to e stisfied under U.S. regionl setting. Second, politicl nd institutionl rriers such s triffs nd trde lws which distort interntionl investigtions cn e neglected in U.S. regionl studies. Finlly, the U.S. Census Bureu provides excellent uniform dt on U.S. sttes tht is unlikely to e mtched y ny interntionl dt set.

8 8 S. Kim / Regionl Science nd Urn Economics 29 (1999) The estimtes of the Ryczynski equtions The Ryczynski estimtes presented in Tles 3 10 nd the fit of the regressions presented in Fig. 1 show tht fctor endowments explin significnt mount of the geogrphic vriility in U.S. mnufcturing production for most industries over time. However, the mount of vrition explined y the endow- 2 ments declined over time. The unweighted verge of the djusted-r for the twenty industries re 0.86 in 1880 nd 0.83, 0.78 nd 0.74 in 1900, 1967 nd 1987 respectively. One explntion for the fll in the explntory power of fctor endowments my e the growing importnce of spillovers. Or lterntively, the 2 fll in the djusted-r my e due to the greter rndomness in the loction of mnufcturing ctivities cused y regionl convergence in fctor endowments. Fctor endowments in U.S. regions hve ecome more similr over time s resources hve ecome incresingly more moile, nd technologicl innovtions hve fvored the development of sustitutes, recycling, nd less resource intensive methods of production. A closer exmintion of the Ryczynski regression estimtes y industries my provide etter clue s to why some industries re explined y spillovers or fctor endowments. If the residul of the Ryczynski estimtes re interpreted s the upper ound estimte on the importnce of spillovers, the results in Tles 3 10 suggest tht the importnce spillovers chnged over time for different industries. For exmple, in the lte nineteenth century, spillovers my hve plyed significnt role in the loction of the ruer nd plstics industry, ut the importnce fell over time. In the textiles, pper, lether nd trnsporttion industries, the regressions suggest tht spillovers my hve ecome significntly more importnt over time. However, the low explntory power of fctor endowments for these industries my e explined y fctors other thn spillovers. For the pper, lether, nd trnsporttion industries, the fll in the explntory power of domestic endowments my e due to the growing importtion of timer, hides nd utomoiles, wheres for the ruer industries, the opposite trend my reflect the sustitution of domestic synthetic rw mterils for imported nturl ruer. 2 Perhps, rther thn using the djusted-r, which my e ffected y some spurious correltions, the reliility of the Ryczynski regression estimtes is etter exmined y mtching the implied fctor intensities of the regression ginst the independent clcultions of fctor intensities. Tles 11 nd 12 present mnufcturing fctor intensities for ech of the twenty two-digit industries clculted using dt from the Census of Mnufctures for yers 1880, 1900, 1967, nd Since the Census of Mnufctures only provide informtion on lor, cpitl nd rw mterils consumed, fctor intensities re reported in cpitl lor nd mteril lor rtios. In generl, there is significnt correspondence etween the implied fctor intensities of the Ryczynski regressions nd the independent clcultions. For the

9 Tle 3 Ryczynski regression estimtes: mnufcturing 1880 Food Tocco Textiles Apprel Lumer Furniture Pper Printing Chemicl Petroleum Constnt (3.50) (2.07) (2.96) (2.46) (2.65) (4.60) (0.14) (2.22) (0.63) (2.94) Lor (3.06) (2.01) (3.04) (2.69) (5.83) (11.1) (16.9) (2.83) (4.63) (4.14) Cpitl (0.04) (0.35) (0.14) (0.74) (0.31) (0.13) (0.46) (0.81) (0.03) (1.22) Agriculture (4.46) (1.82) (3.54) (1.11) (3.89) (4.48) (3.51) (2.25) (0.64) (3.19) Tocco (0.92) (1.46) (0.50) (1.90) (0.46) (1.76) (0.66) (1.72) (1.83) (2.46) Timer (2.10) (0.79) (2.44) (1.35) (15.0) (2.34) (1.78) (2.01) (1.55) (4.77) Petroleum Minerls (2.02) (1.69) (0.63) (2.48) (3.00) (6.12) (13.7) (2.08) (2.18) (1.18) 2 Adj. R N Significnt t the 5 percent level. 5Significnt t the 1 percent level. Notes: See Tles 1 nd 2 for descriptive sttistics. The t-sttistics (solute vlue in prenthesis) re corrected for heteroscedsticity using White (1980) 3 procedure. Coefficients on fctor endowments were multiplied y 10. S. Kim / Regionl Science nd Urn Economics 29 (1999)

10 Tle 4 Ryczynski regression estimtes: Mnufcturing 1880 Ruer Lether Stone Primry F. Mch. Elec. Trns. Inst. Misc. Metl Metl Constnt (2.06) (1.13) (1.70) (0.16) (1.98) (1.86) (3.86) (2.12) (1.07) (0.42) Lor (3.26) (2.54) (4.92) (3.17) (3.94) (7.07) (10.9) (5.00) (8.92) (4.20) Cpitl (0.41) (0.72) (0.50) (1.16) (0.19) (0.38) (2.03) (0.20) (0.17) (1.07) Agriculture (3.18) (1.47) (3.53) (0.70) (3.33) (2.04) (2.59) (3.31) (0.79) (1.48) Tocco (0.68) (0.21) (1.45) (2.70) (0.18) (2.14) (2.26) (0.07) (0.34) (1.87) Timer (0.61) (0.25) (5.23) (3.09) (1.13) (4.67) (1.46) (1.94) (1.63) (0.39) Petroleum Minerls (2.40) (0.78) (3.46) (32.5) (0.26) (0.33) (5.43) (1.19) (2.62) (5.24) 2 Adj. R N Significnt t the 5 percent level. 5Significnt t the 1 percent level. Notes: See Tles 1 nd 2 for descriptive sttistics. The t-sttistics (solute vlue in prenthesis) re corrected for heteroscedsticity using White (1980) 3 procedure. Coefficients on fctor endowments were multiplied y S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32

11 Tle 5 Ryczynski regression estimtes: Mnufcturing 1900 Food Tocco Textiles Apprel Lumer Furniture Pper Printing Chemicl Petroleum Constnt (1.99) (1.00) (2.18) (1.26) (0.33) (2.51) (0.28) (1.76) (0.69) (1.99) Lor (1.86) (0.46) (1.91) (0.92) (0.91) (1.39) (2.28) (0.27) (0.20) (1.60) Cpitl (3.66) (1.11) (1.42) (1.74) (0.56) (0.28) (0.13) (1.51) (0.75) (3.14) Agriculture (2.54) (0.94) (3.02) (0.81) (2.72) (2.37) (2.21) (2.63) (0.13) (1.29) Tocco (0.63) (4.71) (0.86) (0.58) (1.48) (1.32) (1.67) (1.38) (1.13) (0.30) Timer (0.60) (0.30) (0.40) (0.37) (3.31) (0.72) (0.15) (0.19) (0.65) (2.46) Petroleum (0.22) (0.06) (1.67) (0.06) (0.55) (0.65) (0.35) (0.29) (0.31) (2.67) Minerls (4.28) (0.88) (1.26) (2.98) (1.08) (1.46) (4.47) (2.90) (1.09) (2.38) 2 Adj. R N Significnt t the 5 percent level. 5Significnt t the 1 percent level. Notes: See Tles 1 nd 2 for descriptive sttistics. The t-sttistics (solute vlue in prenthesis) re corrected for heteroscedsticity using White (1980) 3 procedure. Coefficients on fctor endowments were multiplied y 10. S. Kim / Regionl Science nd Urn Economics 29 (1999)

12 Tle 6 Ryczynski regression estimtes: Mnufcturing 1900 Ruer Lether Stone Primry F. Mch. Elec. Trns. Inst. Misc. Metl Metl Constnt (1.75) (0.31) (1.04) (0.58) (1.33) (1.75) (1.42) (0.24) (0.68) Lor (1.70) (1.74) (0.54) (1.72) (0.25) (1.34) (0.63) (1.42) (0.38) Cpitl (1.30) (1.36) (1.95) (1.91) (1.22) (2.26) (0.34) (2.90) (2.37) Agriculture (2.40) (1.66) (1.15) (2.44) (1.42) (0.37) (4.08) (0.52) (1.28) Tocco (0.92) (0.89) (1.12) (1.26) (0.37) (0.68) (0.15) (0.58) (1.18) Timer (1.31) (0.72) (2.63) (1.95) (0.20) (2.55) (0.09) (3.11) (2.90) Petroleum (0.41) (0.98) (3.28) (1.11) (0.80) (1.40) (0.85) (1.14) (18.9) Minerls (0.88) (0.01) (1.04) (6.64) (1.63) (0.59) (7.16) (4.07) (6.84) 2 Adj. R N Significnt t the 5 percent level. 5Significnt t the 1 percent level. Notes: See Tles 1 nd 2 for descriptive sttistics. The t-sttistics (solute vlue in prenthesis) re corrected for heteroscedsticity using White (1980) 3 procedure. Coefficients on fctor endowments were multiplied y S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32

13 Tle 7 Ryczynski regression estimtes: Mnufcturing 1967 Food Tocco Textiles Apprel Lumer Furniture Pper Printing Chemicl Petroleum Constnt (2.70) (1.55) (1.99) (0.28) (0.44) (1.85) (2.79) (0.78) (1.01) (3.18) Lor (5.41) (1.14) (4.03) (3.01) (1.50) (3.63) (3.41) (3.63) (1.15) (1.98) Cpitl (1.75) (0.92) (1.67) (2.44) (1.93) (0.03) (0.15) (2.72) (1.97) (2.95) Agriculture (5.79) (0.23) (2.64) (1.76) (1.68) (2.10) (1.27) (0.01) (0.71) (1.79) Tocco (1.43) (11.1) (4.85) (0.74) (3.59) (4.15) (0.78) (3.46) (0.26) (1.16) Timer (2.01) (1.08) (0.57) (0.57) (13.6) (1.01) (1.21) (0.74) (2.47) (0.80) Petroleum (1.11) (0.80) (1.31) (2.21) (0.27) (0.26) (0.01) (2.12) (2.26) (2.71) Minerls (1.97) (1.13) (0.97) (0.98) (2.56) (0.79) (1.79) (0.80) (0.45) (0.67) 2 Adj. R N Significnt t the 5 percent level. 5Significnt t the 1 percent level. Notes: See Tles 1 nd 2 for descriptive sttistics. The t-sttistics (solute vlue in prenthesis) re corrected for heteroscedsticity using White (1980) 3 procedure. Coefficients on fctor endowments were multiplied y 10. S. Kim / Regionl Science nd Urn Economics 29 (1999)

14 Tle 8 Ryczynski regression estimtes: Mnufcturing 1967 Ruer Lether Stone Primry F. Mch. Elec. Trns. Inst. Misc. Metl Metl Constnt (0.61) (2.38) (0.87) (1.06) (1.09) (0.31) (2.05) (0.17) (0.19) (1.02) Lor (0.36) (4.97) (1.64) (5.98) (0.14) (0.95) (3.63) (0.36) (3.96) (6.25) Cpitl (1.50) (3.60) (3.21) (9.89) (2.36) (1.98) (0.79) (1.52) (3.31) (4.21) Agriculture (0.27) (2.38) (0.61) (1.76) (0.27) (1.56) (1.73) (0.07) (1.37) (1.70) Tocco (2.79) (4.45) (5.10) (3.26) (4.09) (2.85) (1.73) (2.12) (3.36) (3.89) Timer (1.86) (1.86) (1.84) (3.89) (1.91) (2.41) (0.18) (0.94) (0.70) (1.41) Petroleum (1.73) (1.75) (1.97) (6.40) (2.16) (2.55) (0.45) (1.02) (2.53) (2.72) Minerls (0.99) (0.60) (4.00) (1.33) (0.52) (0.88) (0.80) (1.42) (0.13) (0.21) 2 Adj. R N Significnt t the 5 percent level. 5Significnt t the 1 percent level. Notes: See Tles 1 nd 2 for descriptive sttistics. The t-sttistics (solute vlue in prenthesis) re corrected for heteroscedsticity using White (1980) 3 procedure. Coefficients on fctor endowments were multiplied y S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32

15 Tle 9 Ryczynski regression estimtes: Mnufcturing 1987 Food Tocco Textiles Apprel Lumer Furniture Pper Printing Chemicl Petroleum Constnt (3.26) (0.38) (2.60) (0.26) (0.44) (1.08) (3.25) (0.47) (1.21) (5.63) Lor (4.82) (1.45) (0.59) (4.17) (0.61) (1.32) (0.06) (2.62) (0.49) (1.36) Cpitl (0.78) (1.05) (0.38) (2.90) (1.75) (0.54) (1.63) (1.90) (2.58) (0.06) Agriculture (8.14) (0.05) (2.74) (2.33) (0.57) (0.86) (2.32) (0.60) (1.56) (1.76) Tocco (2.71) (3.48) (5.15) (1.20) (1.67) (3.39) (0.39) (2.45) (1.14) (4.17) Timer (0.36) (0.74) (0.57) (0.63) (6.20) (0.31) (2.94) (1.39) (2.24) (2.30) Petroleum (1.61) (1.24) (0.80) (3.14) (1.36) (1.00) (1.25) (1.58) (1.60) (7.32) Minerls (2.30) (1.55) (1.79) (0.05) (0.55) (1.73) (1.11) (0.81) (0.40) (2.19) 2 Adj. R N Significnt t the 5 percent level. 5Significnt t the 1 percent level. Notes: See Tles 1 nd 2 for descriptive sttistics. The t-sttistics (solute vlue in prenthesis) re corrected for heteroscedsticity using White (1980) 3 procedure. Coefficients on fctor endowments were multiplied y 10. S. Kim / Regionl Science nd Urn Economics 29 (1999)

16 Tle 10 Ryczynski regression estimtes: Mnufcturing 1987 Ruer Lether Stone Primry F. Mch. Elec. Trns. Inst. Misc. Metl Metl Constnt (0.06) (2.29) (1.46) (0.07) (1.20) (1.05) (3.12) (1.30) (1.04) (1.14) Lor (0.38) (2.03) (3.26) (3.59) (0.49) (4.08) (4.84) (0.24) (5.47) (4.51) Cpitl (2.90) (1.32) (0.25) (4.67) (2.91) (0.56) (2.61) (0.77) (4.78) (2.70) Agriculture (0.87) (2.47) (0.06) (1.35) (0.91) (3.66) (3.16) (0.40) (0.67) (1.79) Tocco (1.45) (0.24) (1.41) (3.47) (5.28) (2.54) (0.32) (1.67) (2.57) (2.28) Timer (3.05) (1.52) (1.56) (1.98) (3.01) (2.14) (0.53) (0.05) (1.03) (1.51) Petroleum (2.75) (0.51) (0.81) (4.56) (2.99) (0.56) (2.74) (0.74) (4.52) (1.74) Minerls (0.82) (0.94) (2.92) (2.19) (0.98) (0.45) (1.82) (0.25) (0.30) (0.93) 2 Adj. R N Significnt t the 5 percent level. 5Significnt t the 1 percent level. Notes: See Tles 1 nd 2 for descriptive sttistics. The t-sttistics (solute vlue in prenthesis) re corrected for heteroscedsticity using White (1980) 3 procedure. Coefficients on fctor endowments were multiplied y S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32

17 S. Kim / Regionl Science nd Urn Economics 29 (1999) Fig. 1. The fit of the Ryczynski regression estimtes, lor intensive industries, the mtching is somewht miguous for the erlier two 12 periods ut is excellent for the ltter two periods. In 1967 nd 1987, lor coefficients on the instruments, pprel, printing nd pulishing, miscellneous, lether, nd electricl mchinery industries re the most significnt. These industries re lso significntly lor intensive ccording to Tles 11 nd 12 s they rnked the lowest in cpitl lor nd mterils lor rtios for those yers. There is lso resonle correltion etween the implied cpitl intensities from the regressions nd the independent estimtes of cpitl intensities from the census 12 The poor mtching of the implied fctor intensities of the Ryczynski regressions nd the ctul figures my e cused y the low vrition in fctor intensities in 1880 nd For exmple, the coefficient of vrition cross industries for cpitl lor nd mteril lor rtios for 1900 ws 0.40 nd 0.76 respectively wheres, for 1987, they were 1.23 nd 1.75 respectively.

18 Tle 11 Mnufcturing fctor intensities Cpitl/ Lor Mterils/ Lor Cpitl/ Lor Mterils/ Lor ($1000 per Employee) ($1000 per Employee) ($1000 per Employee) ($1000 per Employee) 20 Food Food Food Food Chemicls Petroleum Primry Metl Primry Metl Primry Metl Chemicls Chemicls Petroleum Petroleum Lether Petroleum Ruer Pper Primry Metl Pper Lether Electricl Ruer Ruer Chemicls Mchinery Pper Mchinery Lumer nd Wood Textiles Lumer nd Wood Instruments Pper Printing Electricl Textiles Fricted Metl Fricted Metl Miscellneous Fricted Metl Miscellneous Instruments Textiles Lumer nd Wood Apprel Lumer nd Wood Trnsporttion Printing Trnsporttion Miscellneous Tocco Lether Textiles Trnsporttion Fricted Metl Miscellneous Mchinery Ruer Apprel Trnsporttion Furniture Furniture Mchinery Stone, Cly nd Glss Tocco Lether Furniture Furniture Instruments Stone, Cly nd Glss Instruments Tocco Stone, Cly nd Glss Tocco Printing Apprel Printing Apprel Stone, Cly nd Glss Electricl 36 Electricl Sources: U.S. Census of Mnufctures, 1880, 1900, 1967, S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32

19 Tle 12 Mnufcturing fctor intensities Cpitl/ Lor Mterils/ Lor Cpitl/ Lor Mterils/ Lor ($1000 per Employee) ($1000 per Employee) ($1000 per Employee) ($1000 per Employee) 29 Petroleum Petroleum Petroleum Petroleum Chemicls Tocco Chemicls Tocco Primry Metl Food Tocco Food Pper Chemicls Pper Chemicls Stone, Cly nd Glss Trnsporttion Primry Metl Trnsporttion Food Primry Metl Stone, Cly nd Glss Primry Metl Tocco Pper Food Pper Ruer Textiles Trnsporttion Lumer nd Wood Fricted Metl Fricted Metl Ruer Textiles Textiles Ruer Mchinery Mchinery Trnsporttion Mchinery Electricl Stone, Cly nd Glss Lumer nd Wood Lumer nd Wood Instruments Ruer Mchinery Stone, Cly nd Glss Fricted Metl Fricted Metl Printing Electricl Textiles Electricl Instruments Instruments Lumer nd Wood Miscellneous Electricl Miscellneous Printing Instruments Miscellneous Furniture Miscellneous Lether Furniture Apprel Furniture Furniture Apprel Lether Lether Printing Lether Printing Apprel Apprel 29.6 Sources: U.S. Census of Mnufctures, 1880, 1900, 1967, S. Kim / Regionl Science nd Urn Economics 29 (1999)

20 20 S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32 of mnufctures. In 1900 the correltion is highest for the food, primry metl, chemicls, petroleum, mchinery nd instruments industries, which ll rnk in the top of the cpitl lor rtios, ut is rther low for the pprel nd stone, cly nd glss industries. In 1967 nd 1987 the correltion is the highest for the petroleum, chemicls, primry metl nd stone, cly nd glss industries which consistently rnk mong the highest cpitl lor rtios; the correltion is most disppointing for lumer nd wood which rnks in the ottom hlf of the cpitl lor rtios. Since independent estimtes of the fctor intensities for the extrctive resources griculture, tocco, timer, petroleum nd minerls re unville, the results of the regressions re compred to rw mteril intensities clculted from the Census of Mnufctures. In 1880 nd 1900 the most mteril intensive industries were food, petroleum, primry metl, lether nd ruer; in 1967 nd 1987, they were petroleum, tocco, food, chemicls, trnsporttion nd chemicls. According to the Ryczynski regressions, the respective extrctive resources were highly sttisticlly nd economiclly significnt for most of these cses. 4. Sources of U.S. regionl comprtive dvntge Tles show tht the sources of U.S. regionl comprtive dvntge Tle 13 Sources of U.S. regionl comprtive dvntge (elsticities t the mens in prenthesis) Food Agriculture (0.97), lor (0.65) 21 Tocco Lor (1.08), griculture (0.56) 22 Textiles Lor (1.45) 23 Apprel Lor (1.51) 24 Lumer nd Wood Timer (0.48), lor (0.44), griculture (0.23) 25 Furniture nd Fixtures Lor (0.93), griculture (0.61), timer (0.07) 26 Pper Lor (1.61), timer (0.06) 27 Printing nd Pulishing Lor (1.09), griculture (0.46) 28 Chemicls Lor (0.95), minerls (0.12) 29 Petroleum nd Col Lor (0.93), griculture (0.57) 30 Ruer nd Plstics Lor (1.71) 31 Lether Lor (1.64) 32 Stone, Cly nd Glss Lor (0.81), griculture (0.41), minerls (0.10) 33 Primry Metl Minerls (0.86), lor (0.33), tocco (0.03) 34 Fricted Metl Agriculture (0.97), lor (0.50) 35 Mchinery Lor (0.69), griculture (0.52) 36 Electricl Mchinery Lor (0.99), griculture (0.17) 37 Trnsporttion Lor (0.55), griculture (0.48), timer (0.12) 38 Instruments Lor (1.14) 39 Miscellneous Lor (1.46) 5Significnt t the 5 percent level. 5Significnt t the 1 percent level. Note: Fctor endowments re reported in this tle if they re sttisticlly significnt the 5 percent level.

21 S. Kim / Regionl Science nd Urn Economics 29 (1999) Tle 14 Sources of U.S. regionl comprtive dvntge (elsticities t the mens in prenthesis) Food Cpitl (1.68), griculture (0.70) 21 Tocco Tocco (0.21) 22 Textiles Lor (4.44) 23 Apprel Cpitl (4.06) 24 Lumer nd Wood Timer (0.31), griculture (0.25) 25 Furniture nd Fixtures Agriculture (0.37) 26 Pper Lor (1.58) 27 Printing nd Pulishing Agriculture (0.28) 28 Chemicls None 29 Petroleum nd Col Cpitl (2.63), petroleum (0.04) 30 Ruer nd Plstics Lor (6.19) 31 Lether Lor (4.91) 32 Stone, Cly nd Glss Cpitl (1.49), petroleum (0.07) 33 Primry Metl Cpitl (1.80), minerl (0.89) 34 Fricted Metl None 35 Mchinery Cpitl (0.73) 36 Electricl Mchinery (not ville) 37 Trnsporttion Agriculture (0.39), minerl (0.21) 38 Instruments Cpitl (3.35) 39 Miscellneous Cpitl (2.11) 5Significnt t the 5 percent level. 5Significnt t the 1 percent level. Note: Fctor endowments re reported in this tle if they re sttisticlly significnt the 5 percent level. chnged over time s technologicl dvnces in production nd trnsporttion 13 ltered fctor intensities nd fctor moility. In the lte nineteenth century, given the prevlence of smll scle mnufcturing nd the low moility of lor nd resources, the sources of comprtive dvntge in mnufcturing were lor nd resources. As mnufcturing ecme incresingly cpitl intensive through the turn of the twentieth century, cpitl lso ecme n importnt source of comprtive dvntge. In 1900, the dominnt sources of comprtive dvntge were cpitl nd resources. In the second hlf of the twentieth century, the differing comintions of lor, cpitl nd resources contriuted to explining the economic geogrphy of mnufcturing industries. However, s cpitl ecme 13 Since technologicl innovtions ffect fctor intensities in production s well s the nture nd moility of fctor endowments, these innovtions constntly chnge the optiml loction of mnufcturing ctivities from the perspectives of resources nd the Heckscher Ohlin model. Thus, to the extent the neoclssicl interregionl trde model explins the loction of mnufcturing ctivities, the dynmic ptterns of geogrphic concentrtion nd deconcentrting re then ccompnied y significnt closing nd opening of plnts in different loctions (see Dumis et l., 1997). Moreover, the sttic Ryczynski regression estimtes presented in this pper re in some sense lower ound estimtes of the importnce of resources since t ny given point in time plnts re in the process of relocting to new optiml regions.

22 22 S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32 Tle 15 Sources of U.S. regionl comprtive dvntge (elsticities t the mens in prenthesis) Food Lor (1.00), griculture (0.42), minerl (0.07), timer (0.04) 21 Tocco Tocco (1.12) 22 Textiles Lor (1.21), tocco (0.49) 23 Apprel Lor (4.03), petroleum (0.20) 24 Lumer nd Wood Timer (0.68), cpitl (0.17), griculture (0.05), tocco (0.02) 25 Furniture nd Fixtures Lor (0.89), tocco (0.20), griculture (0.13) 26 Pper Lor (0.77) 27 Printing nd Pulishing Lor (3.37), petroleum (0.12) 28 Chemicls Cpitl (0.63), petroleum (0.08) 29 Petroleum nd Col Cpitl (1.57), petroleum (0.51) 30 Ruer nd Plstics None 31 Lether Lor (3.02), petroleum (0.09) 32 Stone, Cly nd Glss Cpitl (0.72), minerl (0.12) 33 Primry Metl Cpitl (2.56) 34 Fricted Metl Cpitl (1.24) 35 Mchinery Cpitl (0.86) 36 Electricl Mchinery Lor (1.49), griculture (0.16) 37 Trnsporttion None 38 Instruments Lor (5.61), petroleum (0.22) 39 Miscellneous Lor (3.23), petroleum (0.10) 5Significnt t the 5 percent level. 5Significnt t the 1 percent level. Note: Fctor endowments re reported in this tle if they re sttisticlly significnt the 5 percent level. incresingly moile over the twentieth century, its importnce s source of regionl comprtive dvntge declined Lor nd cpitl Lor ws significnt source of comprtive dvntge for ll mnufcturing industries in For sixteen of these industries, lor ws economiclly the most significnt; for the reminder, it rnked second. Output elsticities with respect to lor were highest for ruer nd plstics, lether, pper, textiles, pprel nd miscellneous industries nd rnged from 1.71 to 1.45; they were the lowest for lumer nd wood nd primry metl industries t 0.44 nd 0.33, respectively. In 1900 the numer of products for which lor ws source of dvntge fell shrply to four, ut the output elsticities were higher in generl. Lor ws source of comprtive dvntge for ruer nd plstics, lether, textiles, nd pper. In 1967 nd 1987 the importnce of lor rose gin s it ws source of comprtive dvntge for ten industries in 1967 nd for nine industries

23 S. Kim / Regionl Science nd Urn Economics 29 (1999) Tle 16 Sources of U.S. regionl comprtive dvntge (elsticities t the mens in prenthesis) Food Lor (0.90), griculture (0.33), minerl (0.07) 21 Tocco Tocco (1.72) 22 Textiles Tocco (0.59) 23 Apprel Lor (3.17), petroleum (0.19) 24 Lumer nd Wood Timer (0.63), cpitl (0.51) 25 Furniture nd Fixtures Tocco (0.16) 26 Pper Timer (0.11) 27 Printing nd Pulishing Lor (2.78) 28 Chemicls Cpitl (1.33) 29 Petroleum nd Col Petroleum (0.64), griculture (0.20), timer (0.18), minerls (0.15) 30 Ruer nd Plstics Cpitl (1.32) 31 Lether Lor (2.51) 32 Stone, Cly nd Glss Lor (0.89), minerls (0.13) 33 Primry Metl Cpitl (4.14), minerls (0.16) 34 Fricted Metl Cpitl (1.52) 35 Mchinery Lor (1.08), griculture (0.26) 36 Electricl Mchinery Lor (2.24), griculture (0.23), petroleum (0.15), minerls (0.07) 37 Trnsporttion None 38 Instruments lor (4.93), petroleum (0.34) 39 Miscellneous Lor (2.60), petroleum (0.09) 5Significnt t the 5 percent level. 5Significnt t the 1 percent level. Note: Fctor endowments re reported in this tle if they re sttisticlly significnt the 5 percent level. 14 in Industries for which output elsticity ws consistently greter thn one for oth yers re instruments, pprel, printing nd pulishing, miscellneous, lether, nd electricl mchinery. Cpitl ws not source of comprtive dvntge for ny mnufcturing industry in However, the importnce of cpitl rose in 1900 s it ws significnt for eight industries. By 1967 nd 1987, the numer settled down to seven nd five industries respectively. Industries for which cpitl ws source of comprtive dvntge for two or more yers re primry metl, petroleum, chemicls, stone, cly nd glss, fricted metl, mchinery, nd lumer nd wood Extrctive resources: Agriculture, forestry nd minerls Agriculturl products were sources of comprtive dvntge for food mnufcturing for ll yers. Food mnufcturing ws resource oriented ecuse it ws 14 Given the growing importnce of eduction, differentition of lor types y eduction nd skill levels hs grown over time. See Richrdson nd Smith (1995) for the estimtes of the Ryczynski regression mtrix using six different lor types for 1987.

24 24 S. Kim / Regionl Science nd Urn Economics 29 (1999) 1 32 intensive in rw mterils nd experienced significnt weight reduction in the mnufcturing process. For exmple, over eighty percent of inputs to met pcking cme from the griculturl sector nd s much s fifty percent of weight ws lost when livestock ws trnsformed into wholesle met. Other products such s flour milling, diry processing, nd cnning, preserving, nd freezing of fruits nd vegetles were lso resource oriented for similr resons. There were some exceptions to this rule, ut their contriution to the food sector ws reltively minor. Certin everge, kery nd confectionery products, especilly those whose qulity deteriorted rpidly with trnsporttion, tended to locte ner their mrkets rther thn resources. Agriculturl products were lso sources of comprtive dvntge for mny industries other thn food mnufcturing, ut their importnce diminished significntly over time. In 1880, griculture ws significnt for tocco, lumer nd wood, furniture nd fixtures, printing nd pulishing, petroleum nd col, stone, cly nd glss, fricted metls, electricl mchinery, nd trnsporttion industries. Given the low levels of scle economies in tocco mnufcturing, tocco lef ws grown widely in griculturl res in The geogrphic correltion of griculturl products nd industries such s lumer nd wood, furniture, fricted metl, mchinery, electric mchinery nd trnsporttion ws likely to e cused y the significnt use of wood in the mnufcturing of these products nd the reltive undnce of timer nd sesonl lor in the griculturl regions. By 1900, however, the griculturl products ecme less significnt source of comprtive dvntge in mnufcturing. Tocco lef production ws source of comprtive dvntge for tocco mnufcturing for ll yers fter 1900 nd its significnce rose over time. As the demnd for tocco products shifted from cigrs nd chewing tocco to cigrettes, tocco lef ecme significnt source of comprtive dvntge for 15 tocco mnufcturing. Since tocco lef production is concentrted in few 15 There were three importnt types of cigrette lef tht were grown in the United Sttes. Flue-cured tocco ws grown lrgely in North Crolin, South Crolin, Virgini nd Georgi. Burley tocco ws grown primrily in Kentucky nd Tennessee. Mrylnd tocco ws grown in tht stte. The Americn mnufcturers lended these lefs in different proportions to produce cigrettes. The lower grdes of Burley nd flue-cured lef were used to produce chewing tocco. The cigr lef, which is considerly different from the cigrette lef, ws grown primrily in Pennsylvni, Wisconsin, Connecticut, nd Msschusetts. Prior to 1880, when scle economies in the tocco industry were low, when cigrs nd chewing tocco were reltively more importnt thn cigrettes, nd when the Turkish lef held premium sttus, most regions produced tocco lef of some sort nd mnufctured tocco products. Thus, it is not surprising tht griculturl production rther thn rw tocco production ws correlted with tocco mnufcturing in However, s the demnd for cigrettes grew, tocco mnufcturing mechnized, scle economies rose, nd the industry ecme intensive in rw mterils. The supply of cigrette lefs ecme n importnt source of comprtive dvntge for tocco mnufcturing nd tocco production ecme incresingly clustered in the few cigrette lef growing sttes (see Nicholls, 1951).