Technological Spillovers The Argument. for Trade?*

Transcription

1 UNL, Draft Technologcal Spllovers The Argument for Trade?* Hans J. Czap Ph.D. student Department of Economcs; CBA Unversty of Nebraska, Lncoln Lncoln, NE Tel: Fax: Emal: *paper to be presented at the 2004 AAEA meetng n Denver Copyrght 2004 by Hans Czap. All rghts reserved. Readers may make verbatm copes of ths document for non-commercal purposes by any means, provded that ths copyrght notce appears on all such copes.

2 Technologcal Spllovers The Argument for Trade? Hans J. Czap Ths paper examnes the valdty of ant-trade arguments that are based on the absence of technologcal spllovers. Results of a pooled cross-secton tme-seres analyss for developng countres fal to support the exstence of technologcal spllovers. Fndngs of learnng-by-dong effects ndcate that protectonsm mght be benefcal under certan crcumstances. Key words: development, nternatonal trade, technologcal spllovers, protectonsm Economc growth s one of the man determnants for ncreases n human welfare. Whle developed countres enjoy hgh standards of lvng, the same cannot be sad of most developng countres, where large parts of the populaton lve at or below the poverty lne. Development polcy ams at mprovng standards of lvng n these countres, usng trade polcy as one of the tools to acheve ths goal. Whereas most recent contrbutons n the feld of nternatonal trade and development (see for example Romer 1990, Grossman and Helpman 1990) support the clam that trade n general augments growth, under certan crcumstances the argument can be made that protectonst trade polces mght actually be warranted for (e.g. Grossman and Helpman 1997, Romer 1990, Lewer and Van den Berg 2003). Further nsghts nto ths controversy are extremely mportant as they allow mproved polcy decsons. Deeper understandng of growth related ssues s partcularly mportant for less developed countres as they need economc growth most. Ths paper therefore explores, n contrast to prevous studes, the exstence of technologcal spllovers between sectors wthn developng countres and between developng countres and developed countres as a determnant of growth. Furthermore ths study dffers from prevous research n that the focus les on b-drectonal technologcal spllovers between the manufacturng sector 1

3 and the agrcultural sector n developng countres n addton to the analyss of technologcal spllovers between countres. Ths s of specal nterest for two reasons: Frst, the agrcultural sector represents a relatvely large part of most developng country economes; Second, growth promotng polces n these countres are usually targeted at the manufacturng sector due to the supposed lack of productvty mprovements n agrculture. A fndng of technologcal spllovers orgnatng from the agrcultural sector would have mportant ramfcatons for domestc economc polces as a hgh socal return to agrcultural actvty (compared to prvate return) mght justfy a shft n the target of growth promotng polces towards agrculture. The present paper s dvded nto three parts. The frst part of the paper addresses brefly the theoretcal mportance of the lack of technologcal spllovers for the justfcaton of protectonsm. The second part revews the lterature on technologcal spllovers. The thrd part, lastly, estmates the exstence and sgnfcance of nternatonal and sectoral technologcal spllovers for a set of Latn Amercan countres. Theoretcal Importance of Spllovers The assumptons of the ant trade models consdered n ths paper hnge on the lack of spllovers between sectors wthn countres and same sectors between countres. Technologcal spllovers result from the ncomplete approprablty of technologcal advances. Wth spllovers, productvty ncreases and technologcal progress n one frm or ndustry cause productvty ncreases n other sectors and therefore represent postve externaltes to the economy. To llustrate the testable hypotheses a smple model of specalzaton n lne wth Lewer and Van den Berg (2003) s presented. Smlar conclusons can be derved from more complex model as for example presented by Romer (1990), Grossman and Helpman (1997) and Lucas (1988). Consder a perfectly compettve economy wth an agrcultural and a manufacturng sector and two countres. Followng tradtonal reasonng, t s assumed Y = A F ( L, N ) for = 1,2 2

4 that the manufacturng sector exhbts learnng by dong effects, whereas the agrcultural sector does not. 1 Both countres have the same sectoral producton functon (1) Where Y s the output n sector, A s the stock of techncal ablty or knowledge, L s the nput of land n sector, and N s the nput of labor n sector. The stock of techncal ablty s assumed to be a publc nput, thus the same for both sectors. The development of ths stock s, however, n ths model solely determned by producton n the manufacturng sector. Postve externaltes n terms of techncal progress do not only beneft the manufacturng sector, but also the producers n the agrcultural sector. The learnng by dong process s therefore characterzed by spllovers across sectors and dfferences between prvate and publc returns (see also Bernsten and Nadr (1988), and Rosegrant and Evenson (1992)). The rate of technologcal change can thus be modeled as 1 (2) A & = by Lastly, t s assumed that consumers are utlty maxmzers for any homothetc, ntertemporal utlty functon. Due to the perfectly compettve nature of the economy and the nonapproprablty of technologcal advances, each frm wll maxmze nstantaneous profts by equatng the margnal rate of transformaton between goods and ther prce rato. Under the further assumpton of no spllovers between countres t follows that output changes over tme as (3) 1 Y & = by F ( L, N ) + A( F (, ) (, ) L L N L& + F N L N N& ) mplyng a growth rate of output of 1 The reader should notce of course that the treatment of technologcal progress as a smple learnng by dong process s unsatsfactory, as we would expect that nnovatons are the result of ntentonal effort rather than a random byproduct of producton. The basc nsghts, however, are provded even n ths smple framework. 3

5 (4) g y 1 1 = bf ( L, N 1 F ) + L ( L, N ) L& + F N F ( L, N ) ( L, N ) N& n each sector. The assumpton of nonapproprablty of techncal progress by any ndvdual frm and the lack of spllovers between countres mples that from a socal perspectve the technology advancng sector s not producng at an optmal quantty of output. Socal welfare could be ncreased by subsdzng producton n the sector that produces postve externaltes. We thus deal wth a market falure. Followng standard neoclasscal trade theory, country A wll specalze n the sector that s resource ntensve n ts abundant factor, gven that producton functons are actually equal. Under the aforementoned assumpton of no nternatonal spllovers, ths mples that the country specalzng on the no-learnng sector wll experence a lower growth rate of ts output. 2 The laggng behnd of the no-learnng sector wll over tme ncrease the prce of ts output relatve to the prce of the output n the learnng sector. Ths wll cause the consumers to substtute for the now cheaper product and thus lead to a shft of producton towards the learnng product. Eventually both countres wll produce mostly the learnng product and thus experence smlar rates of growth. In ths scenaro statc gans from trade are gven, dynamc dsadvantages, however, wll outwegh these gans after some few years. Growth rates wll equalze eventually, but level changes wll have occurred durng the transton perod. In ths case an argument for protecton can be made. Protectng the learnng sector ncreases the proftablty of producng that product and leads producers to ncrease producton. The ncrease n output of learnng ntensve 2 That s as long as ( F L ( L, N ) L& + F N ( L, N ) N& ) does not dffer substantally across countres. F ( L, N ) However, even f there s a substantal dfference and the country specalzng n the no-learnng good grows faster than the other country, the former would stll be better off wth ncreased spllovers from the learnng ntensve sector. 4

6 goods wll beneft the growth rate of output and thus welfare through the ncrease of postve externaltes, e.g. spllovers between sectors. Relaxng the assumpton of no spllovers between countres, we arrve at a dfferent concluson. Now the change n knowledge does not depend only on the domestc producton of output, but also on the foregn producton, or. (5) A = b (Y 1A + Y 1B ). mplyng & & + C 1A 1A C C C C C C C C C C C (6) Y = b( Y + Y ) F ( L, N ) + A( F ( L, N ) L F ( L, N ) N ) and 1A 1A 1A 1B 1B 1B (7) g C = b( F ( L, N ) + F ( L, N )) y L ( F + C L ( L C, N C N C ) L& + F F C ( L C, N ( L ) & C C N C, N, C є [A,B] Analyzng ths last equaton clearly shows that the growth rate n each sector n each country s ndependent on who s producng the learnng ntensve good. It only matters how much s produced n total. Besde nternatonal spllovers also the degree of sectoral spllovers has to be consdered. Judgng from past experence t s not all that clear that manufacturng s ndeed the faster growng sector wth hgher spllovers. Data actually shows that n most countres the agrcultural sector experenced the hgher growth rates compared to the manufacturng sector. Grossman and Helpman (1997) developed a more complex model wth a two country, three sector economy wth varous specfcatons for the nternatonal economy. 3 Ther model allows a more detaled analyss of the results of alternatve specfcatons, among others the effect of technologcal spllovers. Dependng on the chosen specfcaton Grossman and Helpman (1997) are able to derve welfare enhancng as well as welfare decreasng effects of nternatonal trade wth spllovers 4. A careful analyss of C ) N& C ) 3 e.g. small country / large country case, smlar / dssmlar szed countres, Imtaton vs. Innovaton 4 The effect s n comparson to the autarky case. Baldwn and Forsld (1998) ponted out that ths mght actually be problematc as one should rather compare ncremental changes. 5

7 ther fndngs and results of other smlar models (see for example Aghon and Howtt 1999 or Rvera and Romer 1990), show, however, that wth nternatonal and/or sectoral knowledge spllovers the case of detrmental effects s rather specal and should not be used as a general argument aganst free trade. We can conclude therefore that spllovers are not necessary for trade to yeld postve effects, but they ncrease the probablty of ncreased growth rates. Ths argument can be used to reconsder ant-trade arguments for ther valdty. If spllovers do not exst protecton mght actually be called for n rare cases. If, however, knowledge spllovers do exst, t s dffcult to serously argue aganst a generally benefcal outcome of free trade. Emprcal evdence for spllovers Gven the mportance of spllovers (or lack thereof) for the justfcaton of protecton, t s necessary to nvestgate what evdence for spllovers has been found n the lterature. In a study by Bernsten and Nadr (1988) spllovers between ndustres are estmated. Whle sgnfcant spllovers between sectors were found, ther results ndcated that not all ndustres are equally lkely to cause spllovers. Studes by Jaffe and Trajtenberg (1998) and Jaffe et al. (1993) fnd that technology does not move to the same degree between countres and ndustres as t does wthn ts respectve country and ndustry. As ponted out by Eaton and Kortum (1999), however, foregn technology s stll very mportant, especally when consderng the amount of technology created abroad versus domestcally. A study conducted by Bernsten and Yan (1997) fnds that spllovers between Canadan and Japanese frms are nsgnfcant or of a small magntude, spllovers between US and Canada, however, are sgnfcant and farly large (Bernsten, 1996). In general they do fnd large dfferences between socal returns to R&D nvestments and prvate returns. Smlar results are obtaned by Nadr and Km (1996) n a study about nternatonal spllovers between major OECD countres. A paper by 6

8 Johnson and Evenson (1999) looks at the ssue of spllovers between sectors wthn and between countres. They use patent data and trace products used n varous sectors to ther orgn n other sectors. Ths provdes a mappng of spllovers between sectors. For developng countres t s arguable, however, that technology s more often than not an adaptaton or emulaton of an exstng product and thus does not receve a patent. Developng countres n general perform lttle R&D themselves, but results from Coe et al. (1997) suggest that there are sgnfcant spllovers from ndustralzed countres to developng countres. Summng up the nsghts from the spllover lterature, we can conclude that as of now there s no defnte knowledge of the true amount of spllovers between and wthn developed countres. Varous methods have been used to measure spllovers, but the proxes for technologcal progress and spllovers are less than perfect and results are nconclusve. Moreover, so far very few emprcal studes for developng countres have been conducted. Accordng to the author s best knowledge none of the studes that do target developng countres have analyzed ntra-sectoral spllovers between the agrcultural sector and the manufacturng sector and spllovers between the two sectors and ther counterpart n developed countres. Model Specfcaton Followng closely the work by Fulgnt and Perrn (1993), we can set up the general model wth the sector producton functons (8) y s = f(x s ; β s ) where x s s a vector of nputs x s = (x s1,, x sn ) n sector s, s є [A,M] wth A the agrcultural sector and M the manufacturng sector. β s represents a vector of all parameters n sector s. Some technology changng varables t k, k=1,2,,m determne the vector of parameters β s as (9) β s = G s (t s1,, t sm ) 7

9 The vector of parameters at any pont n tme s thus determned by the technology changng varables gven n each sector s. Productvty change due to changes n the technology changng varable s equvalent to the concept of elastcty of output wth respect to a change n some technology changng varable or for each sector (10) Ψ k y/ t k (t k /y). As specfcaton of the producton functon a Cobb-Douglas form s chosen: (11) C n y ( x ; β ) = A wth s s s s = 1 x βs s log A s s β = γ = α s0 s0 + + m k = 1 m k = 1 γ sk α t t sk sk sk + µ, + µ, s s0 k = 1,..., m = 1,..., n Ths producton functon exhbts varable elastctes of output wth respect to each nput. As specfed n Fulgnt and Perrn (1993) the µ 0 s a random varable dstrbuted ndependently of the x s and t s and the µ s are dstrbuted ndependently of the technology changng varables wth mean zero and a fnte postve sem-defnte covarance log matrx. y = The above specfcaton expressed n logs yelds s α s0 + α sktsk + γ s0 log xs + γ sktsk log xs + µ s log xs + µ s k = 1 = 1 = 1 k = 1 k = 1 (12) whch yelds for elastctes n (13) ψ sk = tsk γ sk log xs + α sk = 1 m n n m n 0 These equatons are used n the followng to estmate the mportance and exstence of spllovers n producton. 8

10 Methodology In order to capture the degree of nternatonal spllovers a group of countres wth farly smlar characterstcs was chosen. More specfcally t s nterestng to look at developng countres wth smlar clmate as the major developed countres, because t can be expected that agrcultural technology wll flow sgnfcantly only to countres wth smlar agrcultural condtons. Furthermore, countres wth armed conflcts were avoded n order to reduce the mpact of socal unrest on producton. A natural choce 5 for countres then was to pck Latn Amerca wth Brazl and ts southern and western neghbors, Uruguay, Paraguay, Argentna, Bolva, Peru and Chle. These countres of course dffer substantally from developed countres n ther geographcal features, but they are more or less as close as t gets as a sample wth the desred characterstcs. Data are stacked over countres, whch mples that estmates are for the regon as a whole. Estmates are obtaned usng pooled cross-secton tme-seres estmated separately for the manufacturng and the agrcultural sector. The followng two equatons (14a) log (14b) log A A0 + α Ak tk + γ A 0 log x A + γ Ak tk log k = 1 = 1 = 1 k = 1 y α x + µ M = M 0 + α Mk tk + γ M 0 log xm + γ Mk tk log k = 1 = 1 = 1 k = 1 y α x + µ = are thus estmated wth 3 technology changng varables and 3 tradtonal nputs for the agrcultural sector and 3 technology changng varables and 2 tradtonal nputs for the manufacturng sector. The estmated parameters are then used to calculate the elastctes as gven by (13). A M A0 M 0 5 Of course stll somewhat arbtrary 9

11 Data Productvty growth s measured as the dfference of growth n output and growth of nputs. Whereas most other studes adjust ntermedate nputs for qualty changes, ths study refrans from dong so as a qualty ncrease n ntermedate nputs represents emboded technologcal change and should therefore be captured by the model. Mssng values n the data seres are extrapolated by takng the average yearly change and extendng forward and backward respectvely. Mssng data n the mddle are calculate by usng a geometrc average growth rate between the last data avalable before the gap and the frst data avalable after the gap. For sngle number gaps the average between precedng and succeedng value s taken. Consstent wth Fulgnt and Perrn (1993) and varous other studes (e.g. Hayam and Ruttan (1970), Kawagoe, Hayam and Ruttan (1985)), land, lvestock, machnery, fertlzer and labor are used as tradtonally measured physcal nputs n the agrcultural producton. Lvestock s captured n thousand lvestock unts wth converson factors for dfferent types of lvestock as proposed by Hayam and Ruttan (1970) (1.1 for camels and cameldes; 1.0 for buffaloes, horses and mules; 0.8 for cattle and asses; 0.2 for pgs; 0.1 for sheep and goats; 0.01 for poultry). Machnery s measured as the number of agrcultural tractors used. Fertlzer s the amount of fertlzer n metrc tons used n the agrcultural producton. In order to reduce the number of varables Lvestock, Machnery and Fertlzer are combned nto a captal stock ndex. The ndex (nput x 1 ) s calculated by summng the respectve quanttes weghted by ther cost share. Labor (nput x 2 ) s measured as thousands of actve partcpants n the agrcultural sector adjusted for qualty by multplyng the number of workers wth the lteracy rate n the respectve country, taken from the World Development Indcators (WDI). Land (nput x 3 ) s measured as the number of thousands of hectares of arable and permanent cropland and pastures, adjusted 10

12 for qualty by usng Peterson s (1987) nternatonal land qualty ndex 6. If not mentoned otherwse agrcultural data s obtaned from the FAO database. For the manufacturng sector only two nputs are consdered, captal stock and labor. Captal stock data s not readly avalable and thus has to be derved from the provded data. The constructon of the captal stock follows closely the method used by Easterly and Levne (2001). To be able to calculate the captal stock we assume that the country s n the steady state at the begnnng of our observaton perod. The captal stock K n perod t s gven by K t = (1-d)*K t-1 + I t-1, where d s the rate of deprecaton and I s nvestment. Usng the captal-output rato k nstead we can transform above equaton to k (1 d) * k + = 1+ g k = g + d, where s the nvestment-output rato and g the steady state growth rate of output. Followng Easterly and Levne (2001) we assume a deprecaton rate of The steady state growth rate s calculated usng the average growth rate of output n the respectve country from 1962 to 1972 and the average world growth rate calculated as The country growth rate s weghted by 0.25 whereas the world growth rate s weghted by The steady state value for the nvestment to output rato s calculated as the 10 year average of the nvestment-output rato, wth data obtaned from the WDI. Imprecse estmates of the captal stock wll be allevated over tme due to the deprecaton rate. Explct data on labor n the manufacturng sector s not avalable and s therefore approxmated by the dfference between the total labor force and the agrcultural labor force, gnorng labor n the servce ndustry. Smlarly to the adjustment of labor n the agrcultural sector manufacturng labor s adjusted for qualty usng the lteracy rate n the respectve country. The technology changng factors t consdered n ths study are spllovers from varous sources. Ths study attempts to dentfy the mportance of sectoral and nternatonal spllovers. Usually t s assumed that there s a connecton between R&D 6 Ths measure has been used frequently n prevous studes, e.g. Fulgnt and Perrn (1993), Frsvold and Ingram (1995), Lusg and Thrtle (1997). 11

13 expendture n one sector and ts technologcal progress. The use of a producton functon wth output levels as the dependent varable necesstates the use of R&D stocks (or cumulatve knowledge stocks) as a determnng varable. Many studes (e.g. Coe and Helpman (1993), Nadr and Km (1996), Bernsten (1989)) have therefore used the R&D 7 stock as a frst proxy for technologcal progress. As much as t s desrable to use ths proxy for developed countres, t s arguable that for developng countres R&D expendture are of secondary mportance compared to learnng by dong effects as dscussed n the next paragraph. Furthermore, lack of R&D data for most of the countres n ths study would requre napproprately long extrapolaton from the few observatons avalable and thus put nto queston the nformatonal value of the proxy altogether. In ths study t s therefore decded to exclude R&D expendture and stocks as proxes for technologcal progress. Future studes wth more complete data sets mght consder ncludng ths varable however, snce even emulaton and adaptaton requres some amount of research. As already ndcated, R&D expendture s not the only means of mprovng technology. One mportant factor that lkely wll ncrease productvty wthn the sector pertans to the dea of a learnng curve. Ths most closely resembles the learnng by dong concepts proposed by varous endogenous trade theores. In the present study ths s modeled by usng a measure of cumulatve output lagged by two years as a proxy (t s1 ) for learnng by dong effects. The FAO database only has ndexes of output, but a value for output s needed n ths context. In order to obtan a seres for output values n the agrcultural sector therefore a base value for 1961 s taken from the USDA Economcs and Statstcs System database and then extended to the present date by multplyng wth the FAO ndexes. Furthermore, as also ponted out by Coe et al. (1997), Coe and Helpman (1993) and Keller (2001), nternatonal trade theory argues that knowledge wll be transferred through emboded technology n goods, drect knowledge transfer through foregn drect 7 Takng the R&D stock has the unfortunate mplcaton, however, that countres cannot easly skp technologcal steps n ther development. 12

14 nvestment (FDI) and some knowledge spllovers through general nteracton due to trade relatons. Lastly, there mght also be some technologcal spllovers through foregn ad. Data for the mports of ntermedate agrcultural goods are obtaned from the FAO database 8, whereas data for manufacturng are from the World Development Indcators. In order to capture the spllovers due to FDI, we smply use the amount of FDI (provded n form of a stock varable) n the country lagged by four years. Snce a suffcently long stock s not avalable nstead the stock s constructed by takng the percentage of FDI to GDP from the WDI database and multplyng t by the respectve GDP varable. Wth better data the FDI varable deally would have to be consdered wth respect to the destnaton sector or at least adjusted to the relatve weght of the respectve sectors n the economy. The four year lag s ntroduced due to the consderaton that FDI wll mport technology but presumably t wll take some tme for the newly ntroduced technology to spread wthn the country. For hgh technology ndustres the tme lag seems to be close to one year 9 (Tlten, 1971), whereas for sectors n general t takes about 4 years to mtate about 60% of patented nnovatons (Mansfeld, 1984). Consderng that most of developng countres do not posses a sgnfcant hgh tech ndustry, t s a reasonable approach to use a lag of 4 years. Data on foregn ad are obtaned from the WDI webpage and lagged four years as well. In order to keep the number of parameter manageable, the present study combnes these sources of nternatonal spllovers nto one ndex (t s2 ) by summng the cumulatve quanttes weghted by ther cost shares for each year. The value of the ndex s adjusted by the US productvty growth n the respectve sector. 10 Data for productvty ncreases n the agrcultural sector are taken from P. Pardey s webpage at the 9 Ths study s somewhat dated, but t s reasonable to assume that these results stll hold approxmately. 10 Tokgoz (2003) shows productvty ncreases n developed countres mght be lnked n varous ways to sales, competton and R&D n developng countres. For the sake of smplcty the present study keeps t as a workng hypothess nonetheless, snce the effect of one regon should not be sgnfcant for US productvty as a whole. 13

15 Unversty of Mnnesota, whereas data for productvty ncreases n the manufacturng sector are taken from the US Department of Labor - Bureau of Labor Statstcs. Besde nternatonal spllovers and learnng by dong effects, we also would expect that there s some degree of sectoral technologcal spllovers (t s3 ) wthn a country. Ths s measured by usng cumulatve manufacturng output and the nternatonal spllover ndex of the respectve other sector, both lagged by four years. Emprcal Results and Concluson The estmaton results for the parameters can be seen n the appendx. Gven the focus of ths paper on the mpact of the technology changng varables on output, only the results for the technologcal change elastctes of output and respectve t-ratos are presented n table1 for agrculture and table2 for manufacturng. It s strkng that only the elastcty wth respect to cumulatve output n the manufacturng sector s sgnfcant. (Table 1) Elastctes Agrculture Elastcty Asymptotc t-rato Ψ A cumulatve output Ψ A nternatonal spllover E E-04 Ψ A sectoral spllover E-05 14

16 (Table 2) Elastctes Manufacturng Elastcty Asymptotc t-rato Ψ M cumulatve output Ψ M nternatonal spllover E E-04 Ψ M sectoral spllover The results of the estmaton do not support the exstence of technologcal spllovers n developng countres. Furthermore, cumulatve output n the agrcultural sector does not seem to cause learnng by dong effects. To check for robustness of these fndngs a seemngly unrelated regresson (SUR) s used, estmated separately for the two consdered sectors. In lne wth the pooled cross-secton tme-seres approach the results of the SUR do not support the exstence of nternatonal or sectoral spllovers. In lght of these results protectonsm has to be consdered a vable tool for ncreasng domestc welfare under certan crcumstances. Further, the fndng of sgnfcant learnng-by-dong effects observed n the manufacturng sector supports the nfant ndustry argument for protecton. These conclusons have to be treated wth care, however, as only some possble channels of technologcal transfer are consdered. Indeed, prevous research and economc ntuton ndcate that some degree of spllovers should occur. Future research should consder possble endogenety of the varables. Furthermore, t s qute possble that the level of aggregaton s too large, as only some parts of an ndustry can be expected to receve technologcal spllovers. Ideally one would lke to consder only data of that ndustry or subsector. Gven the data restrctons, however, ths s not possble at the moment. Further econometrc refnements and dfferent estmaton approaches should therefore be consdered to ncrease the robustness of our results and shed some more lght on ths mportant ssue. 15

17 Lterature Aghon, P. and Howtt, P., Endogenous Growth Theory, MIT Press, Cambrdge Massachusetts, Baldwn, R.E. and Forsld, R, Incremental Trade Polcy and Endogenous Growth: A q-theory Approach, NBER Workng Paper 6477, Bernsten, J.I., Internatonal R&D Spllovers between Canada and the Unted States, Socal Rates of Return and Productvty Growth, The Canadan Journal of Economcs, Vol.29, 1996., The Structure of Canadan Inter-Industry R&D Spllovers, and the Rates of Return to R&D, The Journal of Industral Economcs, Vol. 37, Bernsten, J.I. and Nadr, M.I., Interndustry R&D Spllovers, Rates of Return, and Producton n Hgh-Tech Industres, The Amercan Economc Revew, Vol. 78, 1988., Research and Development and Intra-ndustry Spllovers: An Emprcal Applcaton of Dynamc Dualty, The Revew of Economc Studes, Vol. 56, Bernsten, J.I. and Yan, X., Internatonal R&D Spllovers between Canadan and Japanese Industres, The Canadan Journal of Economcs, Vol.30, Coe, D. T. and Helpman, E., Internatonal R&D Spllovers, NBER Workng Paper 4444, Coe, D. T., Helpman, E., and Hoffmaster, A. W., North-South R&D Spllover, The Economc Journal, Vol. 107,

18 Easterly, W. and Levne, R., It s Not Factor Accumulaton: Stylzed Facts and Growth Models, World Bank Economc Revew, 2001 Eaton, J. and Kortum, S., Internatonal Technology: Theory and Measurement, Internatonal Economc Revew, Vol. 40, Evenson, R.E. and Rosegrant, M.W., Agrcultural Productvty and Sources of Growth n South Asa, Amercan Journal of Agrcultural Economcs, Frsvold, G. and Ingram, K., Sources of Agrcultural Productvty Growth and Stagnaton n Sub-Saharan Afrca, Agrcultural Economcs 13, Fulgnt, L. E. and Perrn, R. K., Prces and Productvty n Agrculture, The Revew of Economcs and Statstcs, Vol. 75, Grossman, G. M., and Helpman, E., Innovaton and Growth n the Global Economy, The MIT Press, Cambrdge Massachusetts, 1997., Trade, Innovaton, and Growth, Amercan Economc Revew, Vol. 80, Papers and Proceedngs of the Hundred and Second Annual Meetng of the Amercan Economc Assocaton, Hayam, Y. and Ruttan, V.W., Agrcultural Productvty Dfferences Among Countres, Amercan Economc Revew 60, Jaffe, A. and Trajtenberg, M., Internatonal Knowledge Flows: Evdence from Patent Ctatons, NBER Workng Paper 6507, Jaffe, A., Trajtenberg, M., Henderson, R., Geographc Localzaton of Knowledge Spllovers as Evdenced by Patent Ctatons, NBER Workng Paper 3993,

19 Johnson, D.K.N. and Evenson, R.E., R&D Spllovers to Agrculture: Measurement and Applcaton, Contemporary Economc Polcy, Vol.17, Kawagoe, T., Hayam, Y. and Ruttan, V., The Intercountry Comparson of Agrcultural Producton Effcency, Amercan Journal of Agrcultural Economcs 67, Keller, W., Internatonal Technology Dffuson, NBER Workng Paper 8573, Lewer, J. and Van den Berg, H., Internatonal Trade: The Engne of Growth?, unpublshed manuscrpt, Lucas, R., On the Mechancs of Economcs Development, Journal of Monetary Economcs 22, Lusg, A. and Thrtle, C., Total Factor Productvty and The Effects of R&D n Afrcan Agrculture, Journal of Internatonal Development 9, Mansfeld, E., R&D and Innovaton: Some Emprcal Fndngs, n Grlches, Z. (ed.); R&D, Patents and Productvty; Unversty of Chcago Press, Nadr, M.I. and Km, S., Internatonal R&D Spllovers, Trade and Productvty n Major OECD Countres, NBER Workng Paper 5801, Peterson, W., Internatonal Land Qualty Indexes, Unversty of Mnnesota, Staff Paper, Rvera-Batz, L. and Romer, P.M., Economc Integraton and Endogenous Growth, Quarterly Journal of Economcs, Romer, P.M., Endogenous Technologcal Change, Journal of Poltcal Economy, Vol.98,

20 Tlten, J.E., Internatonal Dffuson of Technology: The Case of Semconductors, Brookngs Insttutons, Washngton D.C., Tokgoz, S., R&D Spllovers n Agrculture: Results from a Trade Model, Workng Paper 03-WP344, Center for Agrcultural and Rural Development, Iowa State Unversty, Databases Food and Agrcultural Assocaton (FAO); Internatonal Fnancal Statstcs, Pardey, P.; Unversty of Mnnesota, Statstcal Abstract of the unted states - The natonal Data Book, varous years USDA Economcs and Statstcs System - World Agrculture: Trends and Indcators (89024); US Department of Labor Bureau of Labor Statstcs, World Development Indcators; 19

21 Appendx Econometrc Output ASYMPTOTIC VARIABLE ESTIMATED STANDARD T-RATIO NAME COEFFICIENT ERROR Standard nputs - Agrculture Captal Stock (x1) E Labor (x2) E Land (x3) E Technology Changng Varables - Agrculture Cumulatve Output (t1) E E Internatonal Spllovers (t2) E E Sectoral Spllovers (t3) E E

22 Cross Effects and constant - Agrculture AX1T E E AX1T E E AX1T E E AX2T E E AX2T E E AX2T E E AX3T E E AX3T E E AX3T E E CONSTANT ASYMPTOTIC VARIABLE ESTIMATED STANDARD T-RATIO NAME COEFFICIENT ERROR Standard nputs - Manufacturng Captal Stock (x1) E Labor (x2) E

23 Technology Changng Varables - Manufacturng Cumulatve Output (t1) E E Internatonal Spllovers (t2) E E Sectoral Spllovers (t3) E E Cross Effects and Constant - Manufacturng X1T E E X1T E E X1T E E X2T E E X2T E E X2T E E CONSTANT